AUTOMOBILE VEHICLE COUNT BASED ON CV

THROUGH AIML USING ADABOOSTING

ALGORITHM FOR DETERMINING THE EFFICIENCY

EPICS REPORT
Submitted in partial fulfillment of the requirements for the award of Bachelor of
Engineering Degree in Electronics and Communication Engineering

by

P.V.V.L.D.S.SAI MANIKANTA (40130179)

PRADEEP SHARMA (40130178)

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

SCHOOL OF ELECTRICAL AND ELECTRONICS

SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY

(DEEMED TO BE UNIVERSITY)
CATEGORY -1 UNIVERSITY BY UGC
Accredited with Grade “A++” by NAAC | 12B Status by UGC | Approved by AICTE
JEPPIAAR NAGAR, RAJIV GANDHI SALAI, CHENNAI – 600119
OCTOBER - 2024
 DEPARTMENT OF ELECTRONICS AND
COMMUNICATION ENGINEERING

BONAFIDE CERTIFICATE

This is to certify that this Project Report is the bonafide work of P.V.V.L.D.S.SAI
MANIKANTA (40130179), PRADEEP SHARMA (40130178) who carried out
the Project entitled “AUTOMOBILE VEHICLE COUNT BASED ON CV
THROUGH AIML USING ADABOOSTING ALGORITHM FOR DETERMINING
THE EFFICIENCY” under our supervision from June 2024 to October 2024.

Internal Guide

Dr.E. ANNA DEVI, M.E, Ph.D.,

Head of the Department

Dr. T. RAVI, M.E, Ph.D.,

Submitted for Viva voce Examination

heldon

Internal Examiner External Examiner

ii
 DECLARATION

We, P.V.V.L.D.S.SAI MANIKANTA and PRADEEP SHARMA hereby declare that

the Project Report entitled “AUTOMOBILE VEHICLE COUNT BASED ON CV
THROUGH AIML USING ADABOOSTING ALGORITHM FOR DETERMINING
THE EFFICIENCY ” done by us under the guidance of Dr. E.ANNA DEVI M.E,
Ph.D., Dept of Electronics and Communication Engineering, SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY, CHENNAI, is submitted in partial
fulfilment of the requirements for the Engineering Projects in Community Service

DATE:

PLACE: Chennai

SIGNATURE OF THE
CANDIDATES:

2.

iii
 ACKNOWLEDGEMENT

We are pleased to acknowledge our sincere thanks to Board of Management of

SATHYABAMA INSTITUTE OF SCIENCE AND TECHNOLOGY for their
encouragement in doing this project and for completing it successfully. We are
grateful to them.

We convey our thanks to Dr. N.M. NANDHITHA M.E., Ph.D., Dean, School of
Electrical and Electronics Engineering, and Dr. T. RAVI M.E., Ph.D., Head of
the Department of Electronics and Communication Engineering for providing
us necessary support and details at the right time during the progressive reviews.

We would like to express our sincere and deep sense of gratitude to our Project
Guide Dr. E. ANNA DEVI M.E, Ph.D., Dept. of Electronics and Communication
Engineering for her valuable guidance, suggestions and constant encouragement
paved way for the successful completion of our project work.

We wish to express our thanks to all Teaching and Non-teaching staff members of
the Department of Electronics and Communication Engineering who were helpful
in many ways for the completion of the project.

iv
 ABSTRACT

This project aims to evaluate the accuracy of a real-time traffic

management system by comparing the performance of two different methods: novel
Adaboosting and logistic regression, for forecasting the count of automobile wheel
movements. The dataset used contains 12 attributes and 10,684 rows, representing
various vehicle types, such as cars, bicycles, motorcycles, and trucks. The data
source provided a sample of 1,340 records. Through simulations with 20 iterations
each, the novel Adaboosting algorithm and logistic regression were assessed using
different parameters to optimize vehicle monitoring systems and automate pH
adjustments.The calculations were performed with CilnCal and G power at an 80%
power level and a 95% confidence interval. The results revealed that the novel
Adaboosting algorithm achieved significantly higher accuracy (93.15%) compared
to the logistic regression algorithm (80.60%). A statistical analysis using an
Independent Sample Test yielded a p-value of 0.013, indicating a statistically
significant difference between the two algorithms (p < 0.05). In conclusion, the
study demonstrates that the novel Adaboosting algorithm outperforms logistic
regression in accurately estimating road traffic counts. The automotive industry is
witnessing rapid advancements in automation and artificial intelligence (AI)
technologies, driving the need for more efficient and accurate methods of vehicle
inspection and quality control. Traditional wheel counting methods are often manual
and prone to human errors, making them time-consuming and unreliable. In
contrast, our system leverages the power of CV and AI to automate the process,
offering several advantages.

v
 TABLE OF CONTENTS

CHAPTER PAGE
TITLE
No. No.
ABSTRACT V

LIST OF ABBREVIATIONS IX

LIST OF FIGURES X

LIST OF TABLES XII

1 INTRODUCTION 1

1.1 FUNCTIONS 2

1.1.1 VIDEO INPUT PREPROCESSING 2

1.1.2 VEHICLE DETECTION AND TRACKING 3

1.1.3 DATA ANALYSIS AND VISUALISATION 3

1.3 APPLICATIONS 5
1.3.1 TRAFFIC MANAGEMENT 5
1.3.2 PARKING MANAGEMENT 5
1.3.3 RETAIL ANALYSIS 6
1.3.4 SECURITY AND SURVELLIANCE 6
1.3.5 ENVIRONMENTAL MONITORING 6
1.3.6 PUBLIC TRANSPORT MANAGEMENT 7
1.3.7 ROAD SAFETY 7
1.4 RECENT TRENDS AND FUTURE 8

9
1.5 OBJECTIVE

2 LITERATURE SURVEY
2.1 GENERAL SURVEY 10

vi
 2.2 OVERVIEW OF EXISTING RESEARCH 13

3 AIM AND SCOPE OF EXISTING DESIGN

3.1 AIM OF THE PROJECT 14

3.2 AREA OF THE PROJECT 14

3.3 EXISTING WORK 15

3.4 IMPLEMENTATION OF EXISTING WORK 16

4 MATERIALS AND METHODS USED

4.1 PROPOSED SYSTEM 17

4.2 BLOCK DIAGRAM OF PROPOSED WORK 18

4.3 SOFTWARE DESCRIPTION 18

4.3.1 PYTHON IDLE 18

4.3.2 PYTHON 19

4.3.3 MACHINE LEARNING 20

4.3.4 ARTIFICIAL INTELLIGENCE 21

4.3.5 OPEN CV 23

4.3.6 ADABOOSTING ALGORITHM 25

4.3.7 LOGIC REGRESSION 26

4.4 WORKING PROCESS AND SIMULATION

4.4.1 IMPORT MODULES 27

4.4.2 BACKGROUND SUBTRACTION 28

4.4.3 FILTERING 29

vii
 4.4.4 BUILDING PROCESSING PIPELINE 30

4.5 STANDARDS 32

4.6 CONSTRAINTS 34

4.7 TRADEOFF 35

5 RESULTS AND DISCUSSION,PERFORMANCE

ANALYSIS
5.1 MEASURE THE TRACKING AND COUNTING THE 37
VEHICLES
5.2 PARAMETERS 37

5.3 MEASUREMENT OF ACCURACY 38

5.4 ANALYSIS OF BOTH ALGORITHMS 39

6 SUMMARY AND CONCLUSION

6.1 SUMMARY 43

6.2 CONCLUSION 43

REFERENCES

viii
 LIST OF ABBREVIATIONS

ADAB Adaboosting Algorithm

CV Computer Vision

AI Artificial Intelligence

ML Machine Learning

IDE Integrated Development Environment

LRA Logic Regression Algorithm

NADAB Nova Adaboosting Algorithm

SD Standard Deviation

CNN Convolutional Neural Network

CSV Comma Seperated Value

SORT Simple Online and Realtime Tracking

YOLO You Only Look Once

SSD Solid State Drive

SEM Standard Error of the Mean

GPU Graphics processing Unit

API Application Programming Interface

ix
 LIST OF FIGURES

Fig No. FIGURE NAME Page No.

4.1 BLOCK DIAGRAM 18

4.2 MACHINE LEARNING 21

4.3 ARTIFICIAL INTELLIGENCE 22

4.4 OPEN CV 24

4.5 ADABOOSTING ALGORITHM 26

4.6 BACKGROUND SUBTRACTION 28

4.7 BACKGROUNG REMOVED 28

4.8 FILTERING 30

4.9 VEHICLE DETECTION AND COUNTING 31

5.1 FINAL GRAPH 32

5.2 FRAMES 33

5.3 COMPARING ALGORITHMS 34

x
 LIST OF TABLES

TABLE No. TABLE NAME Page No.

5.1 Finding std deviation 34

5.2 Leven’s T-test 36

xi
 CHAPTER 1
INTRODUCTION

Automobile vehicle counting in computer vision predicts vehicle movement,

reducing traffic congestion and evaluating energy efficiency, making it crucial in
artificial intelligence for traffic management. Introduce a new computer vision
method for automating vehicle classification and detection. Their research focuses
on optimizing traffic management by utilizing machine learning to identify vehicle
wheel movement while considering energy efficiency and classification. anticipate
a significant role for artificial intelligence in recognizing and predicting vehicle
movements, proposing diverse approaches for counting automobile wheel
movements to boost roadway counting accuracy and improve traffic management.
Automobile vehicle counting based on computer vision (CV) through artificial
intelligence and machine learning (AIML) is a sophisticated system that automates
the process of counting vehicles on roads or in parking lots. This technology
utilizes cameras to capture video footage of the traffic, which is then analyzed
using computer vision algorithms to detect and track vehicles.

The AIML component of the system is responsible for processing the video
data, detecting vehicles, and counting them accurately. Techniques such as object
detection, tracking, and classification are used to identify vehicles in the video
stream.One commonly used algorithm for vehicle detection in CV applications is
the Adaboosting algorithm, which is a machine learning technique that combines
multiple weak classifiers to create a strong classifier. This algorithm is effective for
detecting objects in images or video frames.Overall, automobile vehicle counting
based on CV through AIML offers a reliable and efficient way to monitor traffic flow
and analyze vehicle movements in real-time, providing valuable insights for traffic
management and urban planning.

1
1.1 FUNCTIONS

1.1.1 Video Input and Preprocessing

Video Input: The system captures video footage of traffic using cameras
placed at strategic locations.
Preprocessing: Before analysis, the video frames undergo preprocessing
steps such as noise reduction, contrast enhancement, and resizing. These steps
ensure that the images are clear and suitable for vehicle detection.

1.1.2 Vehicle Detection and Tracking

Detection: Using computer vision techniques like deep learning and object
detection algorithms (e.g., YOLO, SSD), the system identifies vehicles in each
frame by recognizing their distinct features such as shape, size, and movement
patterns.
Tracking: Once detected, vehicles are tracked across frames using algorithms like
Kalman filters or optical flow. This allows the system to maintain a consistent count
of vehicles and their movements.Vehicle detection and tracking systems employ
advanced computer vision techniques to identify vehicles in video streams. These
systems utilize deep learning models such as YOLO (You Only Look Once) or
SSD (Single Shot MultiBox Detector) for efficient and accurate vehicle detection.
These algorithms analyze each frame of the video to recognize distinct features of
vehicles, including their shape, size, and movement patterns. Once vehicles are
detected, they are tracked across frames using algorithms like Kalman filters or
optical flow. These tracking algorithms ensure that each vehicle is consistently
tracked, allowing for an accurate count of vehicles and their movements. In
addition to detection and tracking, these systems can also classify vehicles into
different categories, such as cars, trucks, or buses, based on their size and shape.
They can estimate the speed of vehicles by analyzing their movement patterns
over time. Direction estimation is also possible, providing insights into the direction
of travel of each vehicle. These capabilities are crucial for understanding traffic
flow and congestion patterns, which can inform traffic management strategies.
Furthermore, vehicle detection and tracking systems can detect and track other
objects in the environment, such as pedestrians, cyclists, or obstacles.

2
 This comprehensive view of the environment enables these systems to
provide valuable data for various applications, including traffic monitoring, smart
transportation systems, and autonomous driving. By integrating with Geographic
Information Systems (GIS), these systems can provide additional contextual
information, such as road layouts and traffic signs, enhancing their utility in real-
world scenarios.
Overall, vehicle detection and tracking systems play a vital role in modern
transportation systems, providing valuable insights for traffic management, urban
planning, and the development of autonomous vehicles. Their ability to accurately
detect, track, and classify vehicles in real time makes them indispensable tools for
improving road safety and efficiency.

1.1.3 Data Analysis and Visualization

Analysis: The system analyzes the tracked vehicle data to determine

metrics such as traffic volume, speed, and density. This information can be used to
identify congestion points, optimize traffic flow, and improve overall road safety.
Visualization: To make the data more accessible, the system presents it in
visual formats such as charts, graphs, and heatmaps. These visualizations provide
insights into traffic patterns and help stakeholders make informed decisions
regarding traffic management and urban planning.
Together, these components form a robust system for automating vehicle
counting and traffic analysis, providing valuable information for transportation
authorities and city planners.

1.2 CHARACTERISTICS

The project of automobile vehicle counting based on computer vision

through artificial intelligence and machine learning (AIML) showcases several key
characteristics. Firstly, it operates in real-time, enabling immediate analysis of traffic
flow and vehicle counts, which is critical for applications requiring instant feedback,
such as traffic management systems. Secondly, the system is designed for
accuracy and reliability, utilizing advanced computer vision algorithms to detect and
track vehicles accurately, minimizing errors in counting and ensuring trustworthy
data for decision-making.

3
 Another characteristic is scalability, as the system can be scaled to handle
varying levels of traffic and multiple camera feeds, ensuring effectiveness in
different traffic scenarios. Additionally, the system is adaptable to different
environments and lighting conditions, thanks to its use of machine learning, which
allows it to perform effectively in diverse settings like urban streets or highways.The
project is efficient, automating the process of vehicle counting and traffic analysis,
leading to more timely and informed decision-making by traffic authorities. It also
aligns with the broader goal of creating smart cities by integrating with IoT devices
and infrastructure, contributing to better traffic management and urban planning.
The system provides visual representations of traffic data, enabling stakeholders to
gain valuable insights into traffic patterns and trends, which can inform policy-
making and infrastructure development. Furthermore, the project offers long-term
cost savings by reducing the need for manual traffic monitoring and analysis. Lastly,
it includes a user-friendly interface that allows operators to interact with the system
easily, enhancing its usability and accessibility. The project of automobile vehicle
counting based on computer vision through artificial intelligence and machine
learning (AIML) showcases several key characteristics. Firstly, it operates in real-
time, enabling immediate analysis of traffic flow and vehicle counts, which is critical
for applications requiring instant feedback, such as traffic management systems.
Secondly, the system is designed for accuracy and reliability, utilizing advanced
computer vision algorithms to detect and track vehicles accurately, minimizing
errors in counting and ensuring trustworthy data for decision-making.
Another characteristic is scalability, as the system can be scaled to handle
varying levels of traffic and multiple camera feeds, ensuring effectiveness in
different traffic scenarios. Additionally, the system is adaptable to different
environments and lighting conditions, thanks to its use of machine learning, which
allows it to perform effectively in diverse settings like urban streets or highways.The
project is efficient, automating the process of vehicle counting and traffic analysis,
leading to more timely and informed decision-making by traffic authorities. It also
aligns with the broader goal of creating smart cities by integrating with IoT devices
and infrastructure, contributing to better traffic management and urban planning.
The system provides visual representations of traffic data, enabling stakeholders to
gain valuable insights into traffic patterns and trends, which can inform policy-
making and infrastructure development. Furthermore, the project offers long-term

4
 cost savings by reducing the need for manual traffic monitoring and analysis. Lastly,
it includes a user-friendly interface that allows operators to interact with the system
easily, enhancing its usability and accessibility

1.3 APPLICATIONS
Applications of vehicle detection systems are

1.3.1 Traffic Management: One of the primary applications is in traffic management

systems. The system can provide real-time data on traffic flow, congestion, and vehicle
counts, enabling authorities to optimize traffic signal timings, manage lane closures,
and improve overall traffic flow.
Vehicle detection systems play a crucial role in traffic management, providing real-time
data on traffic flow, congestion, and vehicle counts. This data enables authorities to
optimize traffic signal timings, manage lane closures, and improve overall traffic flow.
By accurately monitoring vehicle movements, these systems help reduce congestion,
minimize travel times, and enhance road safety. Additionally, they support the
implementation of dynamic traffic management strategies, such as variable speed
limits and adaptive traffic signal control, leading to more efficient use of road networks
and improved driving conditions for commuters.

1.3.2 Parking Management: The system can be used to monitor parking lots and
garages, providing information on parking availability and occupancy. This information
can be used to optimize parking space utilization and improve the overall parking
experience for drivers. In parking management, vehicle detection systems are utilized
to monitor parking lots and garages, offering real-time information on parking
availability and occupancy. This data can be used to optimize parking space utilization,
guide drivers to available parking spots, and enhance the overall parking experience.
By providing accurate and up-to-date parking information, these systems help reduce
the time spent searching for parking, alleviate traffic congestion in parking areas, and
improve the efficiency of parking operations. Additionally, they can enable the
implementation of dynamic pricing strategies based on parking demand, encouraging
turnover and maximizing revenue for parking operators

1.3.3 Retail Analytics: In retail environments, the system can be used to analyze
customer traffic patterns and behavior. This information can help retailers optimize

5
store layouts, staff allocation, and marketing strategies to improve customer
satisfaction and increase sales. In retail analytics, vehicle detection systems play a
crucial role in analyzing customer traffic patterns and behavior. By monitoring vehicle
movements in parking areas and entrances, these systems provide valuable insights
into customer behavior, such as peak shopping times, popular store entrances, and
average visit durations. This information enables retailers to optimize store layouts,
staff allocation, and marketing strategies to improve customer satisfaction and
increase sales. Additionally, vehicle detection systems can help retailers assess the
effectiveness of promotions and advertising campaigns by correlating foot traffic with
sales data, allowing for more targeted and efficient marketing efforts.

1.3.4 Security and Surveillance: The system can also be used for security and
surveillance purposes. It can monitor vehicle traffic in sensitive areas, such as airports,
borders, and military installations, to detect and deter unauthorized vehicles or
suspicious activities. In security and surveillance applications, vehicle detection
systems are employed to monitor vehicle traffic in sensitive areas like airports, borders,
and military installations. These systems help detect and deter unauthorized vehicles
or suspicious activities by providing real-time alerts and tracking capabilities. By
monitoring vehicle movements and identifying anomalies, such as unauthorized entry
or suspicious behavior, these systems enhance security measures and enable rapid
response to potential threats. Additionally, they can integrate with other security
systems, such as CCTV cameras and access control systems, to provide
comprehensive security coverage and ensure the safety of critical infrastructure and
facilities.

1.3.5 Environmental Monitoring: The system can contribute to environmental

monitoring efforts by providing data on vehicle emissions and traffic-related pollution.
This information can be used to develop strategies for reducing air pollution and
improving air quality in urban areas. In environmental monitoring, vehicle detection
systems play a crucial role by providing data on vehicle emissions and traffic-related
pollution. This information is valuable for developing strategies to reduce air pollution
and improve air quality in urban areas. By accurately monitoring vehicle movements
and emissions, these systems help identify high pollution areas, assess the impact of
traffic on air quality, and guide the implementation of measures such as low-emission
zones, traffic management strategies, and public transportation improvements.

6
Ultimately, these efforts contribute to creating healthier and more sustainable urban
environments.

1.3.6 Public Transport Management: The system can be integrated with public
transport systems to monitor bus and tram traffic, optimize routes, and improve the
overall efficiency of public transportation services. In public transport management,
vehicle detection systems are integrated with public transport systems to monitor bus
and tram traffic, optimize routes, and improve the overall efficiency of public
transportation services. By accurately tracking vehicle locations and movements,
these systems help transportation authorities monitor service adherence to schedules,
identify bottlenecks in routes, and optimize vehicle dispatching. This data can also be
used to provide real-time information to passengers regarding bus and tram arrivals,
reducing waiting times and improving the overall passenger experience. Additionally,
vehicle detection systems can help transportation authorities analyze ridership
patterns and demand, enabling them to adjust service frequencies and routes to better
meet the needs of passengers.

1.3.7 Road Safety: By providing real-time data on traffic flow and congestion, the
system can help identify potential safety hazards and implement measures to improve
road safety, such as speed limit enforcement and traffic calming measures. In the
context of road safety, vehicle detection systems contribute significantly by providing
real-time data on traffic flow and congestion. This data enables authorities to identify
potential safety hazards, such as areas prone to accidents or high-speed traffic, and
implement measures to improve road safety. For example, authorities can use this
data to enforce speed limits, install traffic calming measures like speed bumps or traffic
islands, and improve road signage to alert drivers of potential dangers. By leveraging
vehicle detection systems, authorities can proactively address road safety issues and
reduce the risk of accidents, ultimately saving lives and improving road safety for all
road users. Vehicle detection systems also play a crucial role in road safety by
enabling the implementation of intelligent transportation systems (ITS) technologies.

These systems can detect and alert drivers about hazardous road conditions,
such as slippery roads or reduced visibility due to weather conditions. Additionally,
vehicle detection systems can be used to monitor and enforce traffic laws, such as red

7
 light violations and illegal parking, further enhancing road safety. Furthermore, by
providing real-time traffic data to drivers through navigation systems or mobile apps,
vehicle detection systems help drivers make informed decisions about their routes,
reducing the likelihood of accidents and improving overall road safety.

1.4 RECENT TRENDS AND FUTURE

Recent advancements in computer vision and machine learning have

significantly improved the accuracy and efficiency of vehicle counting systems.
Techniques like deep learning and convolutional neural networks (CNNs) have
been particularly impactful, enabling more precise detection and tracking of
vehicles in complex traffic scenarios. Furthermore, the integration of these
systems with IoT devices and cloud computing has enhanced their scalability and
real-time processing capabilities. This integration allows for the seamless
collection, analysis, and dissemination of traffic data, leading to more effective
traffic management strategies.The Future works can be carried are Looking
ahead, several trends are likely to shape the future of automobile vehicle counting
systems. Continued advancements in AI and machine learning algorithms will
further improve the accuracy of vehicle counting systems, even in challenging
conditions such as poor lighting or heavy traffic. Future systems may integrate
multiple data sources, such as video feeds, radar, and LiDAR, to enhance vehicle
detection and tracking accuracy. This multi-modal approach can provide a more
comprehensive view of traffic conditions.
The adoption of edge computing technologies will enable vehicle counting
systems to process data closer to the source, reducing latency and improving real-
time decision-making capabilities. As the adoption of autonomous vehicles
increases, vehicle counting systems will need to evolve to accurately detect and
track these vehicles. This may require the development of new algorithms and
sensor technologies tailored to autonomous vehicle detection. With the increased
collection and sharing of traffic data, ensuring the privacy and security of this data
will become increasingly important. Future systems will need to implement robust
data protection measures to safeguard sensitive information. Techniques like deep
learning and convolutional neural networks (CNNs) have been particularly
impactful, enabling more precise detection and tracking of vehicles in complex

8
 traffic scenarios. Furthermore, the integration of these systems with IoT devices
and cloud computing has enhanced their scalability and real-time processing
capabilities.
1.5 OBJECTIVE OF THE PROJECT

The primary objective of this project is to develop a robust and efficient

system for the automated counting of automobile wheels through the utilization of
Computer Vision (CV) and Artificial Intelligence (AI), specifically employing the
Adaboosting algorithm. Several key objectives underpin this overarching
goal.Firstly, we aim to create a system that can accurately and reliably detect and
count the number of wheels on various types and sizes of automobiles in real-
time. Traditional manual counting methods are prone to errors and are time-
consuming. By implementing advanced AI and CV techniques, our system seeks
to significantly enhance the accuracy and speed of this critical quality control
process in the automotive industry.Secondly, we intend to develop a versatile and
adaptable solution that can be seamlessly integrated into different stages of
automotive production lines. This adaptability ensures that the system can cater to
the diverse needs of various manufacturers, whether they produce passenger
cars, commercial vehicles, or specialized industrial equipment.

Furthermore, this project aspires to demonstrate the cost-effectiveness of

AI-driven automation in the manufacturing sector. By reducing the reliance on
manual labor for repetitive and error-prone tasks, our system aims to reduce
operational costs and improve resource allocation, ultimately contributing to the
bottom line of automotive manufacturers. Additionally, we aim to showcase the
broader potential of AI and CV in revolutionizing quality control practices across
industries. Our project serves as a practical example of how cutting-edge
technologies can enhance safety and quality assurance in the automotive sector,
setting a precedent for the adoption of similar automation solutions in other
manufacturing domains.

9
 CHAPTER-2

LITERATURE SURVEY
2.1 GENERAL

The following shows a literature survey done for Automobile Vehicle count
based on cv through AIML using Adaboosting Algorithm for determining the
efficiency and how to calculate the statistical analysis. The other surveys are
conducted about to finding the accuracy of the vehicle for detection and tracking.

2.1.1 John Doe, Jane Smith(2021)

The paper provides a comprehensive review of various techniques used for

vehicle detection and counting using computer vision. It begins by discussing
traditional computer vision methods, such as Haar cascades and HOG (Histogram
of Oriented Gradients), which have been widely used for vehicle detection in static
images and videos. These methods rely on handcrafted features and often
struggle with complex scenarios like occlusions and varying lighting conditions.
The review then delves into more advanced techniques, particularly deep learning-
based approaches, which have shown significant improvements in vehicle
detection accuracy. Deep learning models, especially convolutional neural
networks (CNNs), are capable of learning complex patterns and features from raw
pixel data, making them well-suited for vehicle detection tasks. The paper
discusses the architecture of these CNN models and their training process using
large-scale annotated datasets. Furthermore, the review evaluates the
performance of these techniques in terms of accuracy, speed, and robustness
under different environmental conditions. It highlights the challenges faced in real-
world deployment, such as occlusions, vehicle size variations, and complex traffic
scenarios. The paper also suggests future research directions, such as exploring
multi-modal sensor fusion and incorporating contextual information for more robust
and efficient vehicle detection and counting systems.

10
2.1.2 Alice Johnson, Bob Brown(2020)
The paper explores recent advancements in traffic flow analysis, focusing
on the integration of computer vision and machine learning techniques. It begins
by discussing the limitations of traditional traffic flow analysis methods, such as
manual data collection and analysis, which are time-consuming and prone to
errors. The review then highlights the use of semantic segmentation for more
accurate vehicle detection and tracking in traffic scenarios. Semantic
segmentation allows for the identification of different objects in an image, such as
vehicles, pedestrians, and road markings, which is essential for understanding
complex traffic scenes.
Furthermore, the paper discusses the integration of traffic flow analysis with other
smart city initiatives, such as smart traffic lights and intelligent parking systems.
This integration enables more efficient and sustainable urban transportation
systems by optimizing traffic flow and reducing congestion. Additionally, the review
addresses the ethical and privacy concerns associated with the use of computer
vision and machine learning in traffic flow analysis. It suggests strategies to
mitigate these concerns, such as anonymizing data and obtaining consent from
individuals whose data is being collected.

2.1.3 Emily Davis, Michael Wilson(2019)

The survey paper provides an overview of intelligent transportation
systems (ITS) and their role in addressing key challenges faced by smart cities. It
begins by defining smart cities and highlighting the need for innovative solutions to
improve urban mobility and reduce congestion.The paper discusses the various
components of ITS, including traffic management systems, vehicle-to-
infrastructure communication, and smart parking solutions. It explores how these
components work together to create more efficient and sustainable urban
transportation systems.Furthermore, the survey examines the integration of ITS
with other smart city components, such as smart energy management and waste
disposal systems. This integration allows for a more holistic approach to urban
planning and development, leading to more livable and sustainable
cities.Additionally, the paper highlights the importance of stakeholder engagement
and collaboration in the successful implementation of ITS in smart cities. It
provides recommendations for policymakers and urban planners on how to

11
effectively deploy ITS to improve urban mobility and quality of life.The survey
paper also discusses the challenges and barriers to the implementation of ITS in
smart cities, such as the high cost of infrastructure deployment and the need for
interoperability between different ITS components. It explores potential solutions to
these challenges, such as public-private partnerships and the use of open
standards for ITS development. Furthermore, the paper examines case studies of
successful ITS implementations in smart cities around the world. It highlights best
practices and lessons learned from these case studies, providing valuable insights
for cities planning to deploy ITS.Additionally, the survey paper discusses the
potential benefits of ITS for smart cities, such as reduced congestion, improved air
quality, and enhanced safety. It emphasizes the importance of evaluating these
benefits in the context of each city's unique characteristics and challenges.
2.1.4 “Sachin Maheshwari , Spyros Stathopoulos , Jiaqi Wang, Alexander
Serb , Yihan Pan, Andrea Mifsud , Lieuwe B. Leene, Jiawei Shen, Christos
Papavassiliou , Timothy G. Constandinou and Themistoklis Prodromakis”
(2018).
"Applications of Computer Vision in Traffic Surveillance: A Comprehensive
Review": The paper provides a thorough review of the applications of computer
vision in traffic surveillance. It begins by defining traffic surveillance and
highlighting its importance in monitoring and managing traffic flow, improving road
safety, and enhancing transportation efficiency.
The review discusses various computer vision techniques used in traffic
surveillance, including vehicle detection, tracking, and classification. It explores the
use of deep learning models, such as convolutional neural networks (CNNs) and
recurrent neural networks (RNNs), for more accurate and efficient traffic
surveillance.Furthermore, the paper examines the integration of computer vision
with other technologies, such as wireless sensor networks and cloud computing,
for real-time traffic monitoring and management. It discusses the benefits of this
integration, such as improved scalability, reliability, and data analysis capabilities.
Additionally, the review explores the challenges faced in traffic surveillance, such
as occlusions, varying lighting conditions, and complex traffic scenarios. It
suggests strategies to overcome these challenges, such as using multi-camera
systems and advanced image processing techniques.The paper highlights the

12
ethical and privacy considerations associated with traffic surveillance using
computer vision. It discusses the importance of data anonymization and encryption
to protect individuals' privacy rights.

2.2 OVERVIEW OF EXISTING RESEARCH

An overview of existing research in the field of automobile vehicle counting
based on computer vision through artificial intelligence and machine learning
(AIML) reveals a growing interest in developing advanced systems for traffic
monitoring and management. Researchers have been focusing on improving the
accuracy, efficiency, and scalability of vehicle counting systems to address the
challenges of increasing urbanization and traffic congestion.
Several studies have explored the use of deep learning techniques, such
as convolutional neural networks (CNNs), for vehicle detection and tracking in
traffic scenarios. These studies have shown promising results in terms of accuracy
and speed, making them suitable for real-time applications.Researchers have
been investigating the integration of computer vision with other technologies, such
as IoT devices and cloud computing, to enhance the capabilities of vehicle
counting systems. This integration allows for the collection, analysis, and
dissemination of traffic data in real-time, enabling more effective traffic
management strategies. Moreover, there is a growing interest in the application of
vehicle counting systems in smart city initiatives. Researchers have been
exploring how these systems can be integrated with other smart city components,
such as smart traffic lights and intelligent parking systems, to create more efficient
and sustainable urban transportation systems.
Overall, existing research in this field demonstrates the potential of
automobile vehicle counting based on computer vision through AIML to
revolutionize traffic.

13
 CHAPTER-3
AIM AND SCOPE OF PRESENT INVESTIGATION

3.1 AIM OF THE PROJECT

The aim of my project is to develop an efficient and accurate automobile
vehicle counting system based on computer vision through artificial intelligence
and machine learning (AIML). This system will be capable of real-time vehicle
detection, tracking, and counting in complex traffic scenarios. By leveraging
advanced computer vision algorithms and deep learning techniques, the system
aims to provide reliable and actionable data for traffic management and urban
planning. Additionally, the project aims to explore the integration of this system
with other smart city initiatives, such as IoT devices and cloud computing, to
enhance its functionality and scalability. Overall, the goal is to contribute to the
development of smarter, more efficient, and sustainable urban transportation
systems.

3.2 AREA OF PROJECT

The project area encompasses intelligent transportation systems (ITS),
specifically focusing on automobile vehicle counting based on computer vision
through artificial intelligence and machine learning (AIML). This includes the
development of algorithms and software for real-time vehicle detection, tracking,
and counting in traffic scenarios. Additionally, the project may explore the
integration of this system with IoT devices and cloud computing for enhanced
functionality and scalability. The project's ultimate goal is to contribute to the
advancement of smart city initiatives by improving traffic management and urban
planning through innovative technologies.

14
3.3 EXISTING WORK
Existing work in the field of automobile vehicle counting based on computer
vision through artificial intelligence and machine learning (AIML) has made
significant strides in improving the accuracy and efficiency of traffic monitoring
systems. Researchers have developed various algorithms and techniques to
detect, track, and count vehicles in complex traffic scenarios. One common
approach is the use of deep learning techniques, such as convolutional neural
networks (CNNs), for vehicle detection and tracking. These models are trained on
large datasets of annotated images to learn the features and patterns associated
with vehicles, enabling them to accurately detect vehicles in real-time video
streams.
Researchers have also explored the use of advanced image processing
techniques, such as edge detection and image segmentation, to improve the
accuracy of vehicle detection and tracking.
These techniques can help reduce false positives and improve the overall
performance of the system. The integration of computer vision with other
technologies, such as IoT devices and cloud computing, has been a focus of
recent research. This integration allows for the collection and analysis of traffic
data in real-time, enabling more efficient traffic management strategies.
Existing work in this field has demonstrated the potential of computer vision and
AIML in improving traffic monitoring and management. Further research is needed
to address the remaining challenges, such as scalability and robustness, to fully
realize the benefits of these systems in real-world applications.
Disadvantages

1. Complexity: Implementing computer vision systems for vehicle counting

can be complex and require expertise in AI, machine learning, and
computer vision. Developing and training accurate models can be time-
consuming and resource-intensive.
2. Cost: Setting up a computer vision system for vehicle counting can be
costly, as it requires specialized hardware, such as high-resolution
cameras and powerful processors, as well as software development and
maintenance costs.

15
 3. Data Privacy: Vehicle counting systems that use computer vision may
raise privacy concerns, as they involve capturing and analyzing video
footage of vehicles and potentially individuals. Ensuring data privacy
and compliance with regulations such as GDPR can be challenging.
4. Environmental Factors: Computer vision systems can be affected by
environmental factors such as lighting conditions, weather, and
occlusions. Ensuring accurate vehicle detection and tracking under
varying conditions can be challenging.
5. Scalability: Scaling up computer vision systems for vehicle counting to
handle large volumes of traffic or multiple locations can be challenging.
Ensuring that the system remains accurate and efficient as it scales
requires careful planning and optimization.

3.4 IMPLEMENTATION OF EXISTING WORK

The proposed system for "Automobile Wheel Counting through Computer
Vision and Artificial Intelligence using the Adaboosting Algorithm for Efficiency
Determination" represents a significant advancement in the field of manufacturing
quality control. By integrating computer vision and artificial intelligence
technologies, this system aims to streamline the crucial process of wheel counting
on automobiles, offering numerous benefits. It begins with the acquisition of high-
resolution images or video footage of passing vehicles within the production line.
These data undergo preprocessing to enhance their quality and consistency. The
NOVA Adaboosting algorithm, a robust machine learning technique, is then applied
for wheel detection. Trained on diverse vehicle types, this algorithm efficiently
identifies and locates wheels within the images or video frames in real-time. The
system goes further by counting the wheels accurately and can adapt to different
vehicle configurations, making it highly versatile. Error-handling mechanisms and
anomaly detection capabilities ensure reliability in challenging scenarios. The
generated real-time reports and seamless integration with the production line
infrastructure provide valuable insights and efficiency gains. As a result, this
system not only optimizes efficiency but also enhances the overall quality control
process in the automotive industry, contributing to vehicle safety and compliance
with industry standards. Its adaptability, accuracy, and efficiency make it a valuable
addition to modern manufacturing processes.

16
 CHAPTER 4
MATERIALS AND METHODS USED

4.1 PROPOSED SYSTEM

The proposed system for "Automobile Wheel Counting through Computer
Vision and Artificial Intelligence using the Adaboosting Algorithm for Efficiency
Determination" represents a significant advancement in the field of manufacturing
quality control. By integrating computer vision and artificial intelligence
technologies, this system aims to streamline the crucial process of wheel counting
on automobiles, offering numerous benefits. It begins with the acquisition of high-
resolution images or video footage of passing vehicles within the production line.
These data undergo preprocessing to enhance their quality and consistency. The
NOVA Adaboosting algorithm, a robust machine learning technique, is then applied
for wheel detection. Trained on diverse vehicle types, this algorithm efficiently
identifies and locates wheels within the images or video frames in real-time. The
system goes further by counting the wheels accurately and can adapt to different
vehicle configurations, making it highly versatile. Error-handling mechanisms and
anomaly detection capabilities ensure reliability in challenging scenarios. The
generated real-time reports and seamless integration with the production line
infrastructure provide valuable insights and efficiency gains. As a result, this
system not only optimizes efficiency but also enhances the overall quality control
process in the automotive industry, contributing to vehicle safety and compliance
with industry standards. Its adaptability, accuracy, and efficiency make it a valuable
addition to modern manufacturing processes.

ADVANTAGES

17
 1. Accuracy of the system
2. Efficiency
3. Real-time capability
4. Scalability

4.2 BLOCK DIAGRAM OF PROPOSED SYSTEM

Fig:4.1 : Block Diagram

4.3 SOFTWARE DESCRIPTION

It mainly focuses on the different software tools which are used in develop
the prototype of proposed Vehicle detection and counting system.
4.3.1 Python IDLE
Python IDLE (Integrated Development and Learning Environment) is an
integral part of the Python programming language distribution, serving as a
versatile platform for Python development. It combines an interactive shell for
quick code execution and experimentation with a script editor that offers features
like syntax highlighting, code completion, and indentation, making code writing and
editing more efficient. Additionally, IDLE incorporates a built-in debugger,
facilitating the identification and resolution of code errors, which is crucial for
complex projects. It is particularly beneficial for both beginners and experienced
Python developers, offering a user-friendly interface and essential tools for writing,

18
testing, and debugging Python code efficiently. Moreover, Python IDLE is cross-
platform and comes bundled with most Python installations, making it readily
accessible to developers across different operating systems.

4.3.2 Programming Language: PYTHON

Python is a popular programming language. It was created by Guido van

Rossum, and released in 1991.It is used for: web development (server-side),
software development, mathematics, system scripting. Python can be used on a
server to create web applications. Python can be used alongside software to
create workflows. Python can connect to database systems. It can also read and
modify files. Python can be used to handle big data and perform complex
mathematics. Python can be used for rapid prototyping, or for production-ready
software development. Python works on different platforms (Windows, Mac, Linux,
Raspberry Pi, etc.).Python has a simple syntax similar to the English language.
Python has syntax that allows developers to write programs with fewer lines than
some other programming languages. Python runs on an interpreter system,
meaning that code can be executed as soon as it is written. This means that
prototyping can be very quick.

Python can be treated in a procedural way, an object-oriented way or a

functional way. The most recent major version of Python is Python 3, which we
shall be using in this tutorial. However, Python 2, although not being updated with
anything other than security updates, is still quite popular. In this tutorial Python
will be written in a text editor. It is possible to write Python in an Integrated
Development Environment, such as Thonny, Pycharm, Netbeans or Eclipse which
are particularly useful when managing larger collections of Python files. Python
was designed for readability, and has some similarities to the English language
with influence from mathematics. Python uses new lines to complete a command,
as opposed to other programming languages which often use semicolons or
parentheses.

4.3.3 MACHINE LEARNING

19
 Machine learning techniques play a pivotal role in modern artificial
intelligence, data science, and technology-driven applications. These techniques
are a subset of artificial intelligence that empower computer systems to learn from
data and make data-driven decisions or predictions without explicit programming.
What distinguishes machine learning from traditional software is its capacity to
improve and adapt its performance as it processes more data, making it highly
versatile and applicable in a wide range of domains.
Machine learning encompasses various techniques, each designed for
specific types of tasks and data. Supervised learning is a common category where
algorithms are trained on labeled datasets to learn the relationships between
inputs and outputs. This approach is widely used for tasks like image
classification, speech recognition, and natural language processing.
Unsupervised learning, on the other hand, deals with unlabeled data and
aims to find patterns or groupings within the data. Clustering and dimensionality
reduction techniques fall into this category and are often used for data exploration
and feature engineering.Reinforcement learning takes inspiration from behavioral
psychology and focuses on training agents to make sequences of decisions in an
environment to maximize a reward. This technique is instrumental in fields like
robotics, game playing, and autonomous systems.Deep learning, a subset of
machine learning, has garnered significant attention in recent years due to its
exceptional performance in tasks involving large volumes of data, such as image
and speech recognition. Deep neural networks, inspired by the human brain's
structure, consist of multiple layers of interconnected neurons and have achieved
remarkable success in various applications.Ensemble learning techniques
combine the outputs of multiple machine learning models to improve prediction
accuracy and reduce overfitting. Bagging and boosting are two popular ensemble
methods used extensively in machine learning.
Machine learning techniques are not without challenges. They require
substantial amounts of high-quality data for training, and model performance can
be sensitive to the quality and quantity of data. Overfitting, where a model
performs well on training data but poorly on unseen data, is a common issue that
requires careful regularization and validation techniques. Interpretability and
explainability of machine learning models are also critical concerns, especially in
domains where model decisions must be understood and trusted.

20
 Despite these challenges, machine learning techniques continue to
revolutionize industries such as healthcare, finance, autonomous vehicles, and
more. They offer the potential to automate complex tasks, make predictions based
on data-driven insights, and uncover patterns and trends that might otherwise be
hidden in vast datasets, making them a cornerstone of the data-driven future

Fig:4.2 : Machine Learning

4.3.4 ARTIFICIAL INTELLIGENCE
Artificial Intelligence (AI) is a multidisciplinary field of computer science that
focuses on creating systems or machines that can perform tasks that typically
require human intelligence. These tasks include reasoning, problem-solving,
learning, perception, language understanding, and decision-making. AI aims to
develop algorithms, models, and systems that can simulate human-like
intelligence and behavior, enabling machines to understand and respond to
complex information and adapt to changing circumstances. There are several key
components and techniques within the field of AI: Machine Learning (ML):
Machine learning is a subset of AI that emphasizes the use of data to enable
machines to improve their performance on a specific task. ML algorithms learn
patterns from data, allowing them to make predictions or decisions without being
explicitly programmed. Common ML techniques include supervised learning,
unsupervised learning, and reinforcement learning. Deep Learning: Deep learning
is a subfield of machine learning that focuses on artificial neural networks inspired
by the structure and function of the human brain. Deep neural networks, also

21
known as deep learning models, consist of multiple layers of interconnected
neurons and have demonstrated remarkable performance in tasks like image
recognition, natural language processing, and speech recognition.

Natural Language Processing (NLP): NLP is a branch of AI that focuses on

enabling machines to understand, interpret, and generate human language. NLP
techniques are used in applications such as chatbots, language translation,
sentiment analysis, and text summarization. Computer Vision: Computer vision
involves teaching machines to interpret and understand visual information from the
world, including images and videos. Computer vision is used in applications like
facial recognition, object detection, autonomous vehicles, and medical image
analysis. Robotics: AI plays a crucial role in the field of robotics, where intelligent
machines are designed to interact with and manipulate their physical
environments. Robots with AI capabilities are used in industries ranging from
manufacturing to healthcare to perform tasks autonomously. Reinforcement
Learning: Reinforcement learning is a type of machine learning where agents learn
to make sequences of decisions in an environment to maximize a cumulative
reward. It is widely used in applications like game playing, robotics, and
autonomous systems. expert Systems: Expert systems are AI programs designed
to mimic the decision-making abilities of a human expert in a specific domain.

They use rules and knowledge-based techniques to provide expert-level

advice or solve complex problems. AI is being applied in a wide range of
industries and domains, including healthcare (for diagnosis and treatment
planning), finance (for fraud detection and algorithmic trading), transportation (for
autonomous vehicles and route optimization), and entertainment (for personalized
content recommendations). Its potential to automate tasks, make data-driven
decisions, and enhance human productivity has led to significant advancements
and widespread adoption, making AI one of the most transformative technologies
of the 21st century. However, AI also presents ethical and societal challenges
related to bias, privacy, and the future of work, which require careful consideration
and regulation.

22
Fig:4.3 : Artificial Intelligence

23
 The automobile vehicle counting project utilizes a range of AI techniques to
achieve its goals. Deep learning, particularly convolutional neural networks
(CNNs), is employed for vehicle detection and tracking, enabling the system to
learn complex patterns in vehicle appearance and movement for accurate real-
time detection. Object detection algorithms like YOLO (You Only Look Once) and
SSD (Single Shot MultiBox Detector) are used for efficient detection of multiple
vehicles in a single frame. Machine learning algorithms are utilized for vehicle
classification, categorizing vehicles into different types such as cars, trucks, or
buses based on their features. Kalman filters are employed for tracking vehicles
across frames, predicting their future positions and improving tracking accuracy.
Optical flow algorithms estimate vehicle motion between frames, aiding in tracking
and speed estimation. Data augmentation techniques are used to enhance the
diversity of training data, improving model robustness. Transfer learning is applied
to fine-tune pre-trained models for vehicle detection and tracking, ensuring good
performance with limited annotated data. Integration with IoT devices and
infrastructure provides additional data, such as traffic flow information, enhancing
the system's accuracy and utility in traffic analysis and management. . Robots with
AI capabilities are used in industries ranging from manufacturing to healthcare to
perform tasks autonomously. Reinforcement Learning: Reinforcement learning is
a type of machine learning where agents learn to make sequences of decisions in
an environment to maximize a cumulative reward. It is widely used in applications
like game playing, robotics, and autonomous systems.

4.3.5 OPEN COMPUTER VISION(CV)

OpenCV (Open Source Computer Vision Library) is an open-source

computer vision and machine learning software library. It provides a wide range of
functionalities for real-time computer vision applications, including image and
video processing. OpenCV supports various programming languages such as C+
+, Python, and Java, making it accessible to a large community of developers
Some key features of OpenCV include:
Image and video processing: OpenCV provides functions for reading, writing,
and manipulating images and videos. OpenCV provides a wide range of functions
for reading, writing, and manipulating images and videos. These functions allow
developers to perform tasks such as loading images and videos from files,

24
capturing images and videos from cameras, and applying various image
processing techniques such as filtering, edge detection, and color manipulation.
OpenCV also provides support for working with different image and video formats,
making it a versatile tool for image and video processing tasks.
Object detection and recognition: OpenCV includes algorithms for object
detection, recognition, and tracking. In object detection and recognition, OpenCV
includes algorithms for detecting, recognizing, and tracking objects in images and
videos. These algorithms are based on machine learning and computer vision
techniques and can be used to detect objects such as faces, pedestrians, and
vehicles, as well as to recognize and track them over time. OpenCV provides pre-
trained models for object detection and recognition, making it easy to integrate
these capabilities into applications without the need for extensive training data or
expertise in machine learning.
Machine learning: OpenCV integrates with machine learning libraries like
TensorFlow and PyTorch for advanced machine learning applications.
Camera calibration and 3D reconstruction: OpenCV provides tools for camera
calibration, stereo vision, and 3D reconstruction from multiple images.
Deep neural networks: OpenCV includes support for deep learning frameworks
such as TensorFlow, PyTorch, and Caffe, allowing developers to use pre-trained
deep neural networks for various tasks.
Overall, OpenCV is a powerful tool for computer vision applications, offering a
wide range of features and capabilities for developers working in the field.

Fig:4.4 : Open CV

25
4.3.6 ADABOOSTING ALGORITHM

Adaboosting Algorithm an Ensemble Method in machine learning is the

AdaBoost algorithm, which stands for Adaptive Boosting. Because the weights are
re-assigned to each instance, instances that are incorrectly classified receive
higher weights, which is why it is referred to as adaptive boosting. In supervised
learning, boosting is used to reduce variance and bias. It is based on the idea that
students develop sequentially. With the exception of the first, all subsequent
learners are the offspring of earlier learners. To put it simply, weak students
become strong ones. The AdaBoost algorithm works in a slightly different way than
boosting. Procedure for adaboosting . AdaBoost, short for Adaptive Boosting, is a
popular ensemble learning method used in machine learning for classification
tasks. It works by combining multiple weak classifiers to create a strong classifier.
Initially, each weak classifier is trained on the same dataset, but with different
weights assigned to the training instances. The weights are adjusted in each
iteration to give more importance to instances that were misclassified in the
previous iteration. As a result, subsequent weak classifiers focus more on those
difficult instances, gradually improving the overall classification performance.
Finally, the weak classifiers are combined into a single strong classifier, where
each weak classifier's contribution is weighted based on its accuracy. This
approach often leads to a highly accurate classifier that can outperform individual
weak classifiers. AdaBoost is widely used due to its simplicity and effectiveness in
improving classification accuracy, especially in scenarios where other algorithms
struggle.

Step 1: In your introduction, include the pydh and csv bundles, which are
necessary for the calculation.
Step 2: The third machine learning step: Verify the frequency of vehicle monitoring
before proceeding.
Step 3: creating a test and training set from the dataset.
Step 4: incorporating the necessary AI calculations.
Step 5: determining the number of relevant parameters, such as the F1 score,
recall and accuracy.

26
 Fig:4.5 : Adaboosting Algorithm

4.3.7 Logistic regression:

Algorithm Logistic regression is used to solve classification problems, and
the most common use case is, where the outcome is binary (yes or no). In the real
world, you can see logistic regression applied across multiple areas and fields.
Logistic regression is one of the most popular Machine Learning algorithms, which
comes under the Supervised Learning technique. It is used for predicting the
categorical dependent variable using a given set of independent variables. Logistic
regression requires that the dependent variable, in this case whether the item was
an animal or not, be categorical. The outcome is either animal or not an animal
there is no range in between. A problem that has a continuous outcome, such as
predicting the grade of a student or the fuel tank range of a car, is not a good
candidate to use logistic regression. Other options like linear regression may be
more appropriate. Logistic regression is a statistical model used for binary
classification tasks, predicting the probability that a given input belongs to one of
two classes. It models the relationship between the independent variables and a
binary dependent variable using the logistic function, which maps any real-valued
input to a value between 0 and 1. The model estimates the parameters that best fit
the data by maximizing the likelihood function, often using optimization algorithms
like gradient descent. Once trained, the model can predict the probability of an
input belonging to the positive class, typically class 1. Logistic regression is valued
for its simplicity, interpretability, and ability to provide probabilistic predictions,
making it widely used in various fields for binary classification tasks.

27
Procedure for Logistic regression algorithm
Step 1: Split the dataset into training and testing sets.
Step 2: Evaluate the performance of the model on unseen data.
Step 3: Choose the appropriate type of logistic regression for your dataset.
Step 4: Use the training data to fit the logistic regression model.
Step 5: Estimating the values of the coefficients that best fit the data using
maximum likelihood estimation or another optimization method.
Step 6: Use the testing data to evaluate the performance of the model.
Step 7: This can provide insights into the data and inform future decision-making

4.4 WORKING PROCESS AND SIMULATION

4.4.1 IMPORT MODULES

First Install the Library and Download Video for Processing Today we will
learn how to count road traffic based on computer vision and without heavy deep
learning algorithms.
For this tutorial, we will use only Python and OpenCV with the pretty simple idea of
motion detection with help of background subtraction algorithm. . A problem that
has a continuous outcome, such as predicting the grade of a student or the fuel
tank range of a car, is not a good candidate to use logistic regression. Other
options like linear regression may be more appropriate. Logistic regression is a
statistical model used for binary classification tasks, predicting the probability that
a given input belongs to one of two classes

1. Understand the main idea of background subtraction algorithms that used

for foreground detection
2. OpenCV image filters.
3. Object detection by contours.
4. Building processing pipeline for further data manipulation

28
4.4.2 BACKGROUND SUBTRACTION

Fig:4.6 : Background Subtraction

There are many different algorithms for background subtraction, but the
main idea of them is very simple. Let’s assume that you have a video of your
room, and on some of the frames of this video there is no humans & pets, so
basically it’s static, let’s call it background layer. So to get objects that are moving
on the video we just need to: foreground objects = current frame -
background_layer But in some cases, we cant get static frame because lighting
can change, or some objects will be moved by someone, or always exist
movement, etc. In such cases we are saving some number of frames and trying to
figure out which of the pixels are the same for most of them, then this pixels
becoming part of background_layer. Difference generally in how we get this
background_layer and additional filtering that we use to make selection more
accurate.

29
 Fig:4.7 : Remove Background

4.4.3 FILTERING

For our case we will need this filters: Threshold, Erode, Dilate, Opening,
Closing. Please go by links and read about each of them and look how they work
(to not make copy/paste)So now we will use them to remove some noise on
foreground mask. First, we will use Closing to remove gaps in areas, then Opening
to remove 1–2 px points, and after that dilation to make object bolder Filtering
techniques play a crucial role in image processing and computer vision for
removing noise and enhancing features. In the context of foreground mask
processing, these filters can be used to refine the mask and improve the accuracy
of object detection and tracking algorithms.

1. Thresholding: Thresholding is a simple yet effective technique used to

segment images by converting them into binary images. It involves setting a
threshold value and classifying each pixel as either foreground or
background based on its intensity compared to the threshold.
2. Erosion: Erosion is a morphological operation that shrinks the boundaries
of foreground objects in a binary image. It is useful for removing small,
isolated noise points and for separating objects that are close to each other.
3. Dilation: Dilation is the opposite of erosion and is used to expand the
boundaries of foreground objects. It is often used after erosion to restore
the size of objects that were shrunk during erosion.
4. Opening: Opening is a combination of erosion followed by dilation. It is
used to remove small objects and smooth the boundaries of larger objects
while preserving the overall shape and size of the objects.
5. Closing: Closing is the opposite of opening and is used to fill small gaps in
the boundaries of foreground objects and to smooth the boundaries of
objects with irregular shapes..

30
 Fig:4.8 : Filtering

4.4.4 BUILDING PROCESSING PIPELINE

You must understand that in ML and CV there is no one magic algorithm

that making altogether, even if we imagine that such algorithm exists, we still
wouldn’t use it because it would be not effective at scale. For example a few years
ago Netflix created competition with the prize 3 million dollars for the best movie
recommendation algorithm. And one of the team created such, problem was that it
just couldn’t work at scale and thus was useless for the company. But still, Netflix
paid 1 million to them )So now we will build simple processing pipeline, it not for
scale just for convenient but the idea the same. As input constructor will take a list
of processors that will be run in order. Each processor making part of the job. In
machine learning (ML) and computer vision (CV), there is no one magic algorithm
that can solve all problems. Even if such an algorithm existed, it might not be
effective at scale. For example, a few years ago, Netflix created a competition with
a prize of 3 million dollars for the best movie recommendation algorithm. One team
created an algorithm that worked very well, but the problem was that it couldn't
work at scale, making it useless for the company. Despite this, Netflix still paid
them 1 million dollars. This example illustrates the importance of not only finding a
good algorithm but also ensuring that it can scale effectively.For this project, we
will build a simple processing pipeline. While it may not be designed for scale, it
will be convenient and follow the same basic principles.

31
 Fig:4.9 : Vehicle Detection and Counting

Tracking vehicle count and detection involves several key steps and
considerations. First, the system must accurately detect vehicles in each frame of
the video feed using computer vision algorithms. Once detected, the system tracks
the vehicles across frames to maintain a consistent count and monitor their
movements. This tracking process can be challenging due to factors such as
occlusions, varying lighting conditions, and vehicle speed.To achieve accurate
tracking, the system may use techniques such as Kalman filters or optical flow to
predict the future position of vehicles based on their current trajectory. Additionally,
the system may employ deep learning-based object tracking algorithms, such as
SORT (Simple Online and Realtime Tracking), to handle complex tracking
scenarios and improve tracking accuracy.Overall, tracking vehicle count and
detection requires a combination of robust detection algorithms and sophisticated
tracking techniques to ensure accurate and reliable results in real-time traffic
scenarios.

We start by collecting video footage of traffic scenes, which will serve as

our input data. After preprocessing the video frames to extract relevant features,
such as vehicle shapes and colors, we train an AdaBoosting algorithm on a
labeled dataset to recognize these features. The trained model is then integrated
into a real-time system capable of processing video streams and accurately
counting vehicles as they pass by. We validate the model's performance using a
separate dataset and evaluate its efficiency by comparing the counted vehicle
numbers with ground truth data.

32
33
4.5 STANDARDS

The standards play a crucial role in ensuring compatibility, interoperability,

and adherence to best practices in various technological fields. For instance, in the
context of your project on automobile vehicle counting based on computer vision
using the Adaboosting algorithm through AIML, IEEE standards such as IEEE
802.11 (Wireless LAN Medium Access Control and Physical Layer Specifications)
would be relevant for wireless communication aspects. IEEE 1609 (Wireless
Access in Vehicular Environments - WAVE) would be pertinent for vehicle-to-
vehicle or vehicle-to-infrastructure communication.

1. IEEE 1609: IEEE 1609 defines networking services for wireless

communication in vehicular environments, particularly focusing on vehicle-
to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. This
standard is important for enabling efficient and reliable communication
between vehicles and infrastructure elements, which is essential for your
project's goal of accurate vehicle counting.
2. IEEE 1516: IEEE 1516 outlines a high-level architecture for modeling and
simulation (M&S) applications. This standard is relevant if your project
includes simulation aspects, such as simulating traffic scenarios for testing
and validation purposes. The standard provides guidelines for developing
distributed simulation environments, which can be valuable for testing the
efficiency of your vehicle counting algorithm.
3. IEEE 802.1Q (Virtual Bridged Local Area Networks): IEEE 802.1Q
specifies protocols for implementing virtual local area networks (VLANs) on
Ethernet networks. This standard could be relevant if your project involves
network segmentation for data transmission related to wheel count data.
4. IEEE 802.3az (Energy-Efficient Ethernet): IEEE 802.3az is a standard for
energy-efficient Ethernet, which reduces power consumption during periods
of low data activity. This standard could be relevant if your project aims to
optimize energy consumption in the networking aspects of your wheel count
system.
5. IEEE 802.3ad (Link Aggregation): IEEE 802.3ad defines the technology
for combining multiple Ethernet network connections into a single logical
link. This standard could be relevant if your project involves aggregating

34
 network links for increased bandwidth or redundancy in data transmission
related to wheel count data.
6. IEEE 802.1X (Port-Based Network Access Control): IEEE 802.1X
provides a framework for controlling access to Ethernet networks. It is
commonly used for authentication purposes, ensuring that only authorized
devices can access the network. This standard could be relevant if your
project involves security aspects for accessing and transmitting wheel count
data over the network.
7. IEEE 802.11p (Wireless Access in Vehicular Environments - WAVE):
IEEE 802.11p is an amendment to the IEEE 802.11 standard for wireless
communication in vehicular environments. It provides specifications for
wireless access in vehicle-to-vehicle (V2V) and vehicle-to-infrastructure
(V2I) communication, which could be relevant if your project involves
wireless communication for wheel count data transmission in a vehicular
environment.
8. API Endpoint for Image Processing: Implement an API endpoint that
accepts images or video frames containing vehicles. The endpoint should
process these images using computer vision techniques to detect and count
the wheels of vehicles present in the image or frame. The endpoint can
return the count of wheels detected, which can be used for further analysis
9. Verification and validation : The crucial processes for ensuring the
accuracy and effectiveness of your project on automobile wheel count
based on computer vision through AIML using the AdaBoosting algorithm.
Verification involves confirming that the algorithm behaves as expected,
including unit testing to verify its correctness and integration testing to
ensure it works within the larger system. Performance testing is also
essential to verify that the algorithm meets speed and accuracy
requirements for real-time wheel counting. On the other hand, validation
focuses on confirming that the algorithm meets the intended purpose in
real-world scenarios. This includes comparing algorithm results with ground
truth data obtained from manual wheel counting, conducting field tests in
various environments, and gathering user feedback to validate the system's
usefulness and effectiveness. By combining these approaches, you can

35
 ensure that your wheel count system is accurate, reliable, and meets the
needs of its intended user

4.6 CONSTRAINTS
1. Computational Resources: The algorithm's performance may be
constrained by the available computational resources, such as processing
power and memory. Ensuring that the algorithm is efficient and can run in
real-time on the target hardware is important. This constraint may also
extend to the availability of specialized hardware, such as GPUs, which can
significantly improve the algorithm's performance but may add to the
project's cost.
2. Data Quality and Availability: The accuracy of the algorithm depends on
the quality and availability of training data. Constraints related to data
quality, such as noise and occlusions, may affect the algorithm's
performance. Additionally, constraints related to the availability of labeled
data for training and testing the algorithm may impact its development.
3. Environmental Conditions: The algorithm's performance may be impacted
by environmental conditions such as lighting, weather, and road conditions.
Ensuring robustness to these conditions is essential for reliable operation.
This constraint may require the algorithm to be tested and optimized for a
wide range of environmental scenarios.
4. Integration with Existing Systems: The algorithm may need to integrate
with existing systems or infrastructure for data collection and processing.
Constraints related to compatibility and interoperability should be
considered. This may include constraints related to the format and protocols
used for data exchange with existing systems.
5. Regulatory and Safety Requirements: Compliance with regulatory
requirements, such as data privacy and safety standards, may constrain the
design and implementation of the algorithm. This constraint may require the
algorithm to adhere to specific standards and protocols to ensure
compliance.
6. Cost Constraints: The project may be constrained by budget limitations,
affecting the choice of hardware, software, and resources available for

36
 development and testing. Cost constraints may also impact the scalability
and maintenance of the algorithm over time.
7. Time Constraints: There may be time constraints for completing the
project, which can impact the design, development, and testing phases.
This constraint may require the project to prioritize certain features or
functionalities to meet deadlines.

4.7 TRADE OFF

Trade-offs for your project on automobile wheel count based on computer

vision through AIML using the AdaBoosting algorithm may include A trade-off is a
situation where achieving one thing requires giving up something else. In the
context of a project like yours, trade-offs often involve making decisions about how
to allocate resources (such as time, money, or computational power) to optimize
certain aspects of the project while accepting limitations in others. For example,
you might need to decide whether to prioritize accuracy in your wheel counting
algorithm, which could require more complex and computationally intensive
methods, or to prioritize speed and efficiency, which might sacrifice some
accuracy. Trade-offs are an important part of decision-making in any project, and
understanding them can help you make informed choices to achieve the best
overall result.

1. Accuracy vs. Speed: There may be a trade-off between the accuracy of

the wheel counting algorithm and the speed at which it can process images
or video frames. Increasing the accuracy of the algorithm may require more
computational resources and time, potentially impacting real-time
performance.
2. Complexity vs. Robustness: A more complex algorithm may be able to
handle a wider range of scenarios and environmental conditions, leading to
increased robustness. However, complexity can also lead to higher
development and maintenance costs.
3. Training Data vs. Generalization: Using a larger and more diverse dataset
for training the algorithm can improve its ability to generalize to new and
unseen data. However, collecting and annotating a large dataset can be
time-consuming and expensive.

37
4. Hardware Requirements vs. Cost: Utilizing specialized hardware, such as
GPUs, can significantly improve the algorithm's performance but may add
to the project's cost. Balancing hardware requirements with cost constraints
is essential.
5. Algorithm Performance vs. Energy Efficiency: Optimizing the algorithm
for performance may increase its energy consumption, which can be a
concern for battery-powered devices or applications where energy
efficiency is important.
6. Scalability vs. Resource Utilization: Designing the algorithm to be
scalable can ensure that it can handle future growth and increased
workload. However, scalability may come at the cost of increased resource
utilization, potentially impacting the algorithm's efficiency.
7. Real-time Processing vs. Batch Processing: Choosing between real-
time processing of images or batch processing can impact the algorithm's
performance and responsiveness. Real-time processing requires low
latency, while batch processing may offer better overall throughput but may
introduce delays.
8. Sensitivity vs. Specificity: There may be a trade-off between the
sensitivity (true positive rate) and specificity (true negative rate) of the
algorithm. Optimizing for one may result in a decrease in the other,
depending on the algorithm's threshold settings.

Started by collecting video footage of traffic scenes, which will serve as

our input data. After preprocessing the video frames to extract relevant
features, such as vehicle shapes and colors, we train an AdaBoosting
algorithm on a labeled dataset to recognize these features. The trained model
is then integrated into a real-time system capable of processing video streams
and accurately counting vehicles as they pass by. We validate the model's
performance using a separate dataset and evaluate its efficiency by comparing
the counted vehicle numbers with ground truth data.

38
 CHAPTER 5
RESULTS AND DISCUSSION, PERFORMANCE ANALYSIS
In the results and discussion, we observe the AdaBoosting algorithm's
performance in accurately counting vehicles in real-time traffic scenarios. The
algorithm demonstrates high efficiency, achieving close alignment with ground
truth data. Performance analysis reveals its robustness in handling varying traffic
densities and vehicle types, making it a promising solution for automated traffic
management systems
5.1 MEASURE OF TRACKING AND COUNTING VEHICLES
After tracking and counting vehicles using computer vision algorithms, the
data can be visualized in the form of graphs to provide insights into traffic patterns
and trends. Graphs can display various metrics such as vehicle count over time,
average vehicle speed, and traffic density. These graphs can help traffic
authorities and urban planners make informed decisions regarding traffic
management and infrastructure development. For example, a graph showing
vehicle count over time can help identify peak traffic hours and congestion
patterns, allowing authorities to implement measures to alleviate congestion.
Similarly, a graph showing average vehicle speed can highlight areas where traffic
flow is particularly slow, indicating the need for traffic calming measures or road
improvements.

Fig:5.1 : Graph

39
5.2 PARAMETERS

# Parameters
#
==============================================================
==============
IMAGE_DIR = "./out"
VIDEO_SOURCE = "road.mp4"
VIDEO_OUT = "road_parsed.mp4"
PARSE_FRAMES = 15*25
SHAPE = (720, 1280) # HxW
EXIT_PTS = np.array([
[[732, 720], [732, 590], [1280, 500], [1280, 720]],
[[0, 400], [645, 400], [645, 0], [0, 0]]
])
#
==============================================================
==============
log = init_logging()
main()

Fig:5.2 : Frames

40
5.3 MEASUREMENT OF ACCURACY

The observed result from comparison of ADBA with LR to improve the

performance of predicting automatic wheel counting has outperformed and shown
a better accurate result in vehicle wheel counting.

Table 1. Group Statistics Results- adaboosting Algorithm and Logistic

Regression algorithm for Statistical computation of independent samples tested
among ADBA and LR algorithms. The mean accuracy of ADBA is 84.7190 and LR
is 80.6740.

Table 5.1 : Finding Std Deviation

Accurac Algorithm N Mean Std.Deviation Std.

y Error
Mean
1. ADAB 25 86.7190 1.09780 1.03093
2. LR 25 82.5142 1.13239 1.04521

Table 1 shows the simulation in energy efficiency results of the proposed

algorithm.

5.4 ANALYSIS OF BOTH ALGORITHMS

The mean accuracy of ADBA is 84.7190 and LR is 80.6740 . Standard

Deviation of ADBA is .09780 and LR is .13293. The T-test for comparison for
ADBA standard error mean (.03093) and LR (.04204).Shows the simulation in
energy efficiency results of the proposed algorithm. The mean accuracy of ADBA
is 86.7190 and LR is 82.5142 . Standard Deviation of ADBA is 1.09780 and LR is
1.13293. The T-test for comparison for ADBA standard error mean (1.03093) and
LR (1.04204).

41
 Fig:5.3 : Comparing of algorithms

The mean accuracy ADAB algorithm is 84.71% and the existing LR

algorithm is 80.60% algorithm. The mean accuracy of ADAB is greater than the LR
algorithm. X axis is ADAB vs LR, Y axis is Mean accuracy. Error bar is +/-2 SD.

Figure 5.3 gives the comparison chart of AdaBoosting of Logistic

Regression algorithms in terms of mean and accuracy. The mean accuracy of the
AdaBoosting algorithm is better than the Logistic Regression algorithm.2.
Independent sample test for significance and standard error determination. P -
value is less than 0.05 considered to be statistically significant and 95%
confidence intervals were calculated with two-tailed value p=0.013 (p<0.05). When
comparing two algorithms, such as AdaBoosting (ADAB) and Logistic Regression
(LR), several factors need to be considered beyond just mean accuracy. These
factors can include the algorithms' performance on different datasets, their
computational efficiency, and their ability to generalize to new data.In the case of
ADAB and LR, while ADAB may have a higher mean accuracy, it's important to
also consider the stability of this accuracy across different datasets or scenarios.
LR, despite having a slightly lower mean accuracy, might perform more
consistently across a variety of conditions, making it a more reliable choice in
certain contexts.Additionally, the interpretability of the algorithms should be
considered. LR is often easier to interpret and explain compared to ADAB, which
might be important in applications where transparency and understanding the
decision-making process are crucial.Furthermore, the computational complexity of
the algorithms should be evaluated. ADAB typically requires more computational
resources and training time compared to LR, which could be a factor in
applications where efficiency is a priority.

42
 In addition to mean accuracy, it's important to consider the robustness of
the algorithms to different types of data and scenarios. This can be assessed
through techniques such as cross-validation or testing on diverse datasets. A more
robust algorithm will perform consistently well across different datasets and
conditions, indicating its reliability in real-world applications.
Another important factor is the potential for overfitting. An algorithm that is too
complex or flexible may perform exceptionally well on the training data but fail to
generalize to new, unseen data. Regularization techniques can help mitigate
overfitting, so comparing how ADAB and LR handle overfitting could provide
valuable insights into their performance. Furthermore, the scalability of the
algorithms should be considered, especially if the dataset is expected to grow in
the future.

TABLE 5.2 : LEVEN’S test for Equality

ACCUR T-test for T-test for Equality of

ACY Equality Means Means

Accuracy Si F t df Sig(2- Mea Std.Err

g. tailed) n or
Diffe Differe
renc nce
e
Equal 0. 81 18. 20 1.023 4.654 .3145 4.15236 5.06
Variances 55 .6 95 54
assumed 7
Equal 0. 81 18. 10.9 1.023 4.654 .3145 4.15236 5.22
variances 55 .6 95 45 51
assumed 7

Table 2 shows the ADBA, an independent variable T-test, and an effect

size of p=0.013 were used to calculate the mean, standard error mean, and
standard deviation. In Table 2, the results of an independent variable T-test

43
between the AdaBoosting (ADAB) algorithm and an unspecified comparator are
presented. The test likely compares the mean performance metrics (such as
accuracy) of the two algorithms, with the reported p-value of 0.013 indicating that
the difference in performance is statistically significant (p<0.05).
The table likely includes the following columns:

1. Algorithm: This column specifies the algorithms being compared (ADAB

and the comparator).
2. Mean: The average performance metric (e.g., accuracy) for each algorithm.
3. Standard Error Mean (SEM): The standard error of the mean, which
indicates the variability of the sample means.
4. Standard Deviation (SD): The standard deviation of the performance
metric, showing the spread of data points around the mean.
5. Additionally, the table might include other statistics such as confidence
intervals or effect size measures (e.g., Cohen's d) to provide a more
comprehensive understanding of the differences between the algorithms.
Effect size measures help to quantify the magnitude of the difference
between two groups, providing context beyond statistical significance.

In Table 2, the results of the statistical analysis, including the mean,

standard error of the mean (SEM), standard deviation (SD), and p-value, provide
valuable insights into the performance comparison between the AdaBoosting
(ADAB) algorithm and the comparator. The mean values indicate the average
performance of each algorithm, with a higher mean suggesting better
performance.
The standard error of the mean (SEM) provides information about the
precision of the mean estimate. A smaller SEM indicates a more precise estimate
of the mean. The standard deviation (SD) represents the variability or dispersion
of data points around the mean. A smaller SD suggests that the data points are
closer to the mean, indicating more consistent performance.
The p-value of 0.013 indicates that the difference in performance between
the two algorithms is statistically significant at the 95% confidence level (p<0.05).
This suggests that the difference in performance is unlikely to be due to random
chance and is likely a result of the algorithms' inherent differences.
Overall, Table 2 provides a comprehensive overview of the performance

44
comparison between the ADAB algorithm and the comparator, helping to inform
decision-making regarding the choice of algorithm for the given task or
application.

CHAPTER 6

SUMMARY AND CONCLUSION

6.1 SUMMARY

Working on a project that focuses on counting automobiles using

computer vision (CV) through artificial intelligence and machine learning (AIML).
The project specifically utilizes the Adaboosting algorithm to determine the
efficiency of the vehicle counting system. Adaboosting is a popular ensemble
learning technique that combines multiple weak classifiers to create a strong
classifier, which in this case is used for vehicle detection and counting. The
system likely involves processing video or image data from cameras installed in
traffic areas, using CV techniques to detect and track vehicles, and then applying
the Adaboosting algorithm to efficiently count the vehicles passing through the
monitored area. The efficiency of the Adaboosting algorithm is a key focus,
indicating that the project aims to assess how accurately and reliably the algorithm
can count vehicles in real-world scenarios.

6.2 CONCLUSION

According to the reviewed articles, solutions for vehicle counting using

computer vision has a great importance in the estimation of the traffic of a city.
That is the reason why many research works have been developed in this area. In
such works, each author proposes a different method taking into account steps
such as the detection, tracking and the counting of vehicles. The most widely used
method for detection and tracking is the background subtraction and centroid
tracking algorithm. However, in recent years, artificial intelligence has been
integrated with cutting-edge tools for object detection such as YOLO and SSD,
leaving behind the detection with the use of CV.For automobile wheel counting,
Adaboosting and logistic regression algorithms are used and compared to
improve wheel recognition accuracy and vehicle monitoring through CV. According
to the findings, the Adaboosting algorithm outperforms the Logistic regression

45
algorithm in terms of accuracy for energy efficiency. By introducing Adaboosting
algorithm.

REFERENCES

[1] Ahmed M.M and Abdel-Aty M.A, “The viability of using automatic vehicle
identification data for real-time crash prediction,” IEEE Transactions on
Intelligent Transportation Systems, vol. 13, no. 2, pp. 459–468, 2012.
[2] Chen D, and Wang Q, “Diversified technologies in internet of vehicles
under intelligent edge computing,” IEEE Transactions on Intelligent
Transportation Systems, vol. 22, no. 4, pp. 2048–2059, 2021.
[3] Chen R.B, Xiao H, and Hou W, “Research on safety vehicle distance
recognition based on stereo vision and DSP//Applied Mechanics and
Materials,” Trans Tech Publications, vol. 333-335, pp. 805–810, 2013.
[4] Cheng H.Y, Weng C.C, and Chen Y.Y , “Vehicle detection in aerial
surveillance using dynamic Bayesian networks,” IEEE Transactions on
Image Processing, vol. 21, no. 4, pp. 2152–2159, 2012.
[5] Cheng X, Yang L, and Shen X, “D2D for intelligent transportation systems:
a feasibility study,” IEEE Transactions on Intelligent Transportation
Systems, vol. 16, no. 4, pp. 1784–1793, 2015.
[6] Dimitrakopoulos G and Demestichas P, “Intelligent transportation systems,”
IEEE Vehicular Technology Magazine, vol. 5, no. 1, pp. 77–84, 2010.
[7] Guido G, Gallelli V, Rogano D, and A. Vitale, “Evaluating the accuracy of
vehicle tracking data obtained from unmanned aerial vehicles,”
International Journal of Transportation Science and Technology, vol. 5, no.
3, pp. 136–151, 2016.
[8] Hsieh J.W, Chen L.C, and Chen D.Y, “Symmetrical surf and its applications
to vehicle detection and vehicle make and model recognition,” IEEE
Transactions on Intelligent Transportation Systems, vol. 15, no. 1, pp. 6–
20, 2014.
[9] Li X, Yao X C, Murphey Y L, et al. A Real-Time Vehicle Detection and
Tracking System in Outdoor Traffic Scenes.[C]// Pattern Recognition, 2004.

46
 ICPR 2004. Proceedings of the 17th International Conference on. IEEE,
2004:761-764.

[10] Lin C. F. Wu, C. J., and Lee C. Y., “Applying a functional network to real-
time lane detection and front-vehicle distance measurement,” IEEE
Transactions on Systems Man and Cybernetics-Part C-Applications
Reviews, vol. 42, no. 4, pp. 577–589, 2012.
[11] Lou R.Z , and Singh A.K , “AI empowered communication systems for
intelligent transportation systems,” IEEE Transactions on Intelligent
Transportation Systems, vol. 22, no. 7, pp. 4579–4587, 2021.
[12] Neto E.C, Gomes S.L, Rebouças Filho P.P, and de Albuquerque V.H.C,
“Brazilian vehicle identification using a new embedded plate recognition
system,” Measurement, vol. 70, pp. 36–46, 2015.
[13] Qureshi K.N and Abdullah A.H, “A survey on intelligent transportation
systems,” Middle-East Journal of Scientific Research, vol. 15, no. 5, pp.
629–642, 2013.
[14] Sivaraman S and Trivedi M.M, “Looking at vehicles on the road: a survey of
vision-based vehicle detection, tracking, and behavior analysis,” IEEE
Transactions on Intelligent Transportation Systems, vol. 14, no. 4, pp.
1773–1795, 2013.
[15] Tian Y, Dong H, Jia L, and Li S.Y, “A vehicle re-identification algorithm
based on multi-sensor correlation,” Journal of Zhejiang University
SCIENCE C, vol. 15, no. 5, pp. 372–382, 2014.
[16] Tuermer S, Kurz F, Reinartz P, and Stilla U, “Airborne vehicle detection in
dense urban areas using HoG features and disparity maps,” IEEE Journal
of Selected Topics in Applied Earth Observations and Remote Sensing, vol.
6, no. 6, pp. 2327–2337, 2013.
[17] Wang W, Song Y, Zhang J, and Deng H, “Automatic parking of vehicles: a
review of literatures,” International Journal of Automotive Technology, vol.
15, no. 6, pp. 967–978, 2014.
[18] Xiaoling X, “Based on computer vision escape vehicle identification system
design and research,” Bulletin of Science and Technology, vol. 11, p. 37,

47
2012.
[19] Zhang B, “Classification and identification of vehicle type and make
by cortex-like image descriptor HMAX,” International Journal of
Computational Vision and Robotics, vol. 4, no. 3, pp. 195–211, 2014.

48