Computer Aided Inspection System For Food Products Using Machine Vision - A Review

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

29

International Conference on Current Trends in Engineering and Technology, ICCTET13

Computer Aided Inspection System For Food


Products Using Machine Vision A Review
S.Janardhana
AIMS,
India.
[email protected]

Dr.J.Jaya
Akshaya College of
Engineering & Technology,
India.

Abstract The aim of this review paper is to emphasis


Computer Aided Inspection (CAI) system for the detection
of defects in food products. As the food products are being
consumed directly (processed / Unprocessed) by human
beings, defects in food products cause serious health
hazards which insists to develop an inspection system
which facilitates for food products of high quality and
safety standards. CAI system provides an alternative for an
automated, non-destructive and cost effective technique to
accomplish accurate, fast and objective quality
determination. This CAI system is based on image Analysis
and processing which can be adopted in food industry. The
various stages of CAI system consists of preprocessing,
enhancement, segmentation, feature extraction and
classification. Considerable research as proven its potential
for inspection and grading of fruits, meat, vegetables,
grain, bakery products, confessionary products and
processed food products. This paper reviews the recent
development and advances of CAI in the food industry. The
fundamental elements of the system and technology are
examined.
Keywords - CAI System, Non-Destructive Image Analysis
And Processing, Grading, Fruits, Vegetables, Grain,
Processed Food Products.
1. Introduction
Quality is defined as the expectation of consumer
rather than superior to other products of same category. In
todays scenario consumer plays a major role in deciding the
quality of products. It has become an at most important
factor to drive the quality in all aspects of the product right
from the Raw material to finished goods.
Especially in case of food industry, manufacturers
need to take extreme care of the finished goods as these
products are directly feed from farm to fork, most of the
products are consumed directly or indirectly at the end user
(either human beings / animals). Chances of food products
getting contaminated are very high as these manufacturing
processes undergoes different stages, where the Raw
material is being handled by various modes (Manual,
Machines). The increased awareness and sophistication of
consumers have created the expectation for improved quality
in consumer food products. This has increased a pressure in
industries to produce the cost effective high quality products
with consistency. This in turn has resulted to look for any
alternative mode of inspection system. (Farhad and Terry
(2002) [15]). In order to overcome the conventional
inspection methods automated inspection using machine
vision with the help of computer, camera and sensors have
played a major role. Automatic grading and sorting
incorporates machine vision technology encompassing the
image processing techniques.

IEEE 2013
IEEE 32107
July 3, 2013, Coimbatore, India.

K.J.Sabareesaan
Research scholar ,
Annamalai University,
India.

Ms. Jaina George


Assistant Professor,
Akshaya College of
Engineering & Technology,
India.
2. Components Of Computer Vision System

Fig.1 Components Of A Computer Vision System (Wang &


Sun, 2002a) [17]
The hardware configuration of computer-based
Computer vision systems is relatively standard Narendra VG,
Hareesh KS (2010) [16]. Typically, a computer vision system
consists of:
An illumination device, which illuminates the sample
under test.
A solid-state Charged Coupled Device array camera, to
acquire an image
A frame-grabber, to perform the A/D (analog-to-digital)
conversion of scan lines into picture elements or pixels
digitized in an N row by M column image
A personal computer or microprocessor system, to provide
disk storage of images and computational capability with
software and specific application programs.
Digital Image Processing can be classified into three
classifications

Noise Removal, Image


Sharpening,
Contrast Enhancement

Low Level
Image processing

Segmentation, Restoration,
Description

Middle Level
Image processing

Extraction, Recognition,
Interpretation

a.
b.

High Level
Image processing

Low level image processing ( noise removal, Image


sharpening, Contrast enhancement)
Middle level processing (Segmentation)

International Conference on Current Trends in Engineering and Technology, ICCTET13


c.

High level Image Processing (Analysis based on


output Of segmentation).

A. Image Acquisition
Image acquisition is the process of capturing the
images, Image capture devices or sensors are used to view
and reproduce images of the sample, these devices includes
scanners, ultrasound, X-ray and near infrared spectroscopy.
In machine vision most commonly used is CCD (Charge
coupled device) i.e. camera. However recent developments
in technologies have seen an adoption of digital camera
which in turn reduces the additional component required to
convert images taken by photographic and CCD cameras.
Images captured by all of the above devices maintain the
features of the images with little noise due to its variable
resolution.
B. Image Pre-Processing
This mainly refers to initial processing of raw image.
The image captured through sensors are transferred into
computer, these are converted to digital image. Digital
images are digits which are readable by computer and are
converted to tiny dots or picture elements representing the
real objects. In some cases pre-processing are done to
improve the image quality by removing the undesired
distortions referred as noise.
Each or a combination of the digits of the image in
digital form represents a small portion of the image called
picture element (Pixel). Objects are described as black and
white picture which are represented by digits ranging from 0
to 255 where 0 is black and 255 is white. Each pixel in
colored image is represented by 3 digits RGB (Red, Green,
Blue) components with each being ( 0 to 255) darkest to
lightest RGB. An arrangement of these digits in row
column format gives a representation of the image. With this
arrangement image analysis can be done using matrix theory
and other mathematical techniques.
C. Segmentation
Image segmentation is a process of cutting adding and
feature analysis of images aimed at dividing an image into
regions that have a strong co-relation with objects or area of
interest using the principal of matrix analysis. Segmentation
can be achieved following techniques Thresholding, Edge
based
segmentation,
Region based
segmentation.
Thresholding is a process where only the dark region is of
interest, the other regions are converted to the back ground
colors in the threshed image before further processing. This
process is useful in color (Maturity) and feature based
(Defect and Damage detection) sorting. Edge based
segmentation relies on detection by edge to edge operators,
which detects discontinuities in gray level, the pixel color,
texture, etc. Edge detection is useful in shape and size
sorting. Region based segmentation involves the grouping
together and extraction of similar pixels to form region
representing single objects with in the image. In this process
the other regions are deleted leaving only the feature of
interest.
D. Feature Extraction / Selection
Feature extraction is a key step in most pattern
analysis tasks. There are several general guidelines to

IEEE 2013
IEEE 32107
July 3, 2013, Coimbatore, India.

30

describe about feature extraction. Firstly is discrimination:


Features of pattern in different classes should have
significantly different values. Secondly is reliability: Feature
should have similar values for pattern of same class. Thirdly
is independence: Features should not be strongly co-related
to each other. Finally is Optimality: Some redundant features
should be detected. Among a number of approaches for the
task, the principal component analysis is the most widely
used approach.
High level processing deals with recognition and
interpretation of images using statistical classifiers or
multilayer neural network of region of interest. Digital
morphology is a group of mathematical operations that can
be applied to the set of pixels to enhance or high light
specific aspects of the shape so that they can be counted and
recognized.
These steps provide information necessary for the process or
machine control for quality sorting and grading. An
interaction of all these levels and knowledge database are
very important and essential for more precise decisionmaking and is seen as an integral part of the image
processing process. These theories when applied to images
of products taken can be used to extract features, which are
needed for the necessary processes. Generally, edge
detection to determine shape and feature extraction to
determine differences in color is useful in identification and
classification
4. Computer Vision For Grading And Sorting Of Fruits
And Vegetables
The application of computer vision for automatic
quality control of agriculture products through image
processing has many potential tasks, The variability in the
agriculture products make it very difficult and challenging to
adopt the existing industrial algorithms to this domain. There
are many processes in agriculture where decisions are made
based on the appearance of the product. Applications for
grading the fruit by its quality, size or ripeness are based on
its appearance, as well as a decision on whether it is healthy
or diseased. Humans are easily able to perform intensive
tasks like harvesting and pruning using basically the visual
sensory mechanism. This suggests that a system based on a
visual sensor should be able to emulate the interpretation
process of the human visual recognition system.
According to Juan Pablo Mercol et.al, the main
advantage of decision trees and classification rules over
neural networks are their simplicity and interpretation of the
obtained classification rules. The main advantages of
decision trees and classification rules over neural networks
are their simplicity and interpretation of the obtained
classification rules. Although most of the algorithms produce
good results (higher than 80%), the SMO neural network and
the LMT decision tree are the ones which, in the experiments
done, produce the models with the highest accuracy (86%).
Most of the errors produced by the SMO algorithm are
intermediate oranges misclassified as good ones (4 errors)
while the LMT misclassifies 4 defective oranges as
intermediate. Also, both misclassify 6 intermediate oranges
as defective. However, based on the parsimony principle, the
best algorithm is the classification rule 1R, because it
achieves rather good accuracy (83.9%), and only generates
three classification rules which use only one attribute (the
contrast of the Value component of the HSV color space).

International Conference on Current Trends in Engineering and Technology, ICCTET13


Because of this, the processing speed is higher than if all the
features are extracted and processed by the classifier. In the
opposite side, the algorithm with the worst accuracy is the
J48 classification tree (76.3%), which misclassifies one good
orange as defective. Despite that the classification accuracy
of the RBF network is 83.9%, it incorrectly classifies 1 good
orange as defective, and 2 defective as good ones, making
this algorithm a bad choice.
According to Nagle et.al (2012) in Non-destructive
mango quality assessment using image processing, Based on
the findings of this study, it is strongly recommended that
better performance cameras and webcams are required for
the defects detection. Furthermore, higher intensity UV
illuminants and suitable wavelength need to be investigated.
In order to detect anthracnose infection at the earliest stage
of disease development, complete microbiological and
photochemical analyses are required. The developed
computer vision system shows encouraging potential for the
application in automatic sorting and grading of mango fruits
for the export industry in Thailand. However, the CVS which
was built for the purpose of this study cannot inspect the
entire fruit surface. Therefore an additional feature capable
of capturing images of the entire fruit has to be developed.
According to Wamiq M. Ahmed et.al(2013),
Techniques for automated detection and classification of
microorganisms can be a great help in preventing outbreaks
and maintaining the safety of the nations food supply. A set
of 1000 scatter patterns representing ten different bacterial
strains was used. Zernike and Chebyshev moments as well as
Haralick texture features were computed from the available
light-scatter patterns. The most promising features were first
selected using Fishers discriminant analysis, and
subsequently a support-vector-machine classifier with a
linear kernel was used. With extensive testing, we were able
to identify a small subset of features that produced the
desired results in terms of classification accuracy and
execution speed. The use of distributed computing for
scatter-pattern analysis, feature extraction, and selection
provides a feasible mechanism for large-scale deployment of
a light scatterbased approach to bacterial classification.
According to Hasan M. Velioglu, image processing
techniques has been widely used in determination of shape,
color, texture and size of agricultural products. A manual
measurement technique has been replaced by image
processing according to the advantages such as accurate and
consistent results of this novel technique. These systems are
flexible in application and they can be used in process lines
instead of human inspection. The present study showed that
the images of beans under UV light can be used in separation
of healthy and damaged ones automatically with using
suitable algorithm. However there is still need to use
different imaging techniques such as multispectral or
hyperspectral imaging to reach more accurate results.
According to Abraham Gastelum-Barrios et.al, recent
developments in agricultural technology have led to the
demand for a new era of automated, noninvasive methods
that leave the crop intact and do not interfere with its natural
growth. The methods used are invasive-destructive, invasive
non-destructive and noninvasive. Recent developments in
agricultural technology have led to the demand for a new era
of automated, noninvasive methods that leave the crop intact
and do not interfere with its natural growth.

IEEE 2013
IEEE 32107
July 3, 2013, Coimbatore, India.

31

According to Mercado-Luna et al., the overall goal of


precision agriculture is to make cultural operations more
efficient, to reduce environmental impact and to enhance
crop quality and yields.
N. Razmjooy et al.(2012), presented work on a potato
defect detection combining with size sorting system using
the machine vision. This work also focus on the mathematics
methods used in automation with a particular emphasis on
the issues associated with designing, implementing and using
classification algorithms to solve equations. Experimental
results show that support vector machines have very high
accuracy and speed between classifiers for defect detection.
The best algorithm method in classification has an accuracy
of about 95% and the size grading section has an accuracy of
about 96.86% for the samples in these experiments.
B. Obara(2007), presented a work concerned with the
development and application of an effective automatic
algorithm of image analysis in order to detect grain
boundaries on microscope images of Redziny dolostone and
Wisniowka quartzite. Method is based on an image preprocessing procedure that is focused on colour system
transformation, followed by rock grain boundaries
segmentation using the image analysis methods for each
colour image. Newly-developed method may facilitate
petrographical and stereological studies of rock structures.
The proposed method has one main disadvantage. The
difficulty is associated with inadequate colour information
for the correct recognition of grain boundaries.
Peggy Walter et.al, presented a work on the analysis
of the effects of thermal history and presence of additives on
fat bloom in chocolate. NMR relaxometry experiments, MRI
experiments, DSC analysis. All samples, even those with
BOB(1,3-Dibehenoyl, 2-oleoylglycerol),
bloomed after
completion of the experiment, indicating that heating
chocolate samples to 50C and allowing them to slowly cool
down afterward was not the best method to achieve a good
tempering.
Tom Pearson et.al, have proposed Methodology to
introduce the signal processing community to the challenges
that arise in food inspection. Food inspection for acrylamide
detection involves the detection of this well-known
neurotoxin in French fries and fried potato chips using image
processing. In particular, it is expected that food inspection
will require more automated equipment to perform these
tasks, for which machine learning and classification
algorithms with automated feature selection are good
candidates.
RuchiTaneja et al, presents the recent developments
in computer vision system in the field of agricultural and
food products. Computer vision systems have been used
increasingly in industry for inspection and evaluation
purposes as they can provide rapid, economic, hygienic,
consistent and objective assessment. However, difficulties
still exist, evident from the relatively slow commercial
uptake of computer vision technology in all sector from the
relatively slow commercial uptake of computer vision
technology in all sectors. However, techniques such as 3D
and color vision willensure computer vision development
continues to meet the accuracy and quality requirements
needed in this highly competitive and changing industry.
Chetna V. Maheshwari et al, proposes parametric
superiority method for quality assessment which is nondestructive and cost-effective technique. provides one

International Conference on Current Trends in Engineering and Technology, ICCTET13


automated method for counting the number of Oryza sativa L
(rice seeds) from the foreign elements with a high degree of
quality and then quantify the same by introducing Q-curves
for quantification and assessment of the rice seeds. The edge
can be detected by any of template based edge detector but
Shen-Castan Infinite symmetric exponential filter based edge
detector is an optimal edge detector like canny edge detector
which gives optimal filtered image.
Kit L. Yam et al presented a simple method that uses
a combination of digital camera, computer, and graphics
software to measure and analyze the surface color of food
products. Photoshop is shown to be able to analyze the color
for food samples, although the software was not originally
designed for this purpose. However, it is already one of the
most powerful software for color analysis, and the
manufacturer and users are regularly making enhancements.
Narendra V G et al, presents the recent development
and application of image analysis and computer vision
system in quality evaluation of products in the field of
agricultural and food. Computer vision systems have been
used increasingly in industry for inspection and quality
evaluation purposes as they can provide rapid, economic,
hygienic, consistent and objective assessment. However,
difficulties still exist, evident from the relatively slow
commercial uptake of computer vision technology in all
sectors. Even though adequately efficient and accurate
algorithms have been produced, processing speeds still fail
to meet modern manufacturing requirements.
5. Conclusion
This paper reviews the recent development in
computer vision for agriculture and food industry. Despite
the numerous studies developed in the area of sorting,
grading and quality inspection of agriculture and food
products, there is still no standardization method which
could be proposed for the assessment of quality of different
types of food. The inspection system is itself a costly affair
due to the diversity of products and their characteristics and
also there is a demand for specific knowledge of the system
in order to perform adjustments which are adequate to the
problem. Even though adequately efficient and accurate
algorithms have been produced processing speeds still fail to
meet modern manufacturing requirements. Many food
manufacturing industry prefer human grading and sorting
due to the high investment and high skilled work force for
executing the job, there is still scope to explore and develop
an cost effective computer machine vision system for quality
inspection of agriculture and food products.

[1]

[2]

References
Naglea, Marcus, Kiatkamjon Intania, Busarakorn
Mahayotheeb, Vicha Sardsudc, Joachim Mllera,
Non-Destructive Mango Quality Assessment Using
Image Processing: Inexpensive Innovation For
The Fruit Handling Industry, September 19-21,
2012 , Conference On International Research On
Food Security, Natural Resource.
Wamiq M. Ahmed, Bulent Bayraktar, Arun K.
Bhunia, E. Daniel Hirleman, J. Paul Robinson, And
Bartek
Rajwa
Classification
Of Bacterial
Contamination Using Image Processing And
Distributed Computing , , IEEE Journal Of

IEEE 2013
IEEE 32107
July 3, 2013, Coimbatore, India.

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

32

Biomedical And Health Informatics, Vol. 17, No. 1,


January 2013.
Juan Pablo Mercol, Juliana Gambini, Juan Miguel
Santos, Automatic Classification Of Oranges Using
Image Processing And Data Mining Techniques.
Hasan M. Velioglu, Ozgur Saglam , Evaluation of
Insect Damage on Beans using Image Processing
Technology.
Abraham Gastelum-Barrios, Rafael A. BorquezLopez, Enrique Rico-Garcia, Manuel Toledano-Ayala
and Genaro M. Soto-Zarazua, Tomato
quality
evaluation with image processing: A
review,
African Journal of Agricultural Research
Vol.
6(14), pp. 3333-3339, 18 July, 2011.
Mercado-Luna A, Rico-Garca E, Lara-Herrera A,
Soto-Zaraza G, Ocampo-Velzquez R, GuevaraGonzles R, Herrera-Ruiz G, Torres-Pacheco I,
Nitrogen
determination
on
tomato
(Lycopersiconesculentum Mill.) seedlings by color
image analysis (RGB), , Afr. J. Biotech., 9(33):
5326-5332, 2010.
Navid Razmjooya, B. Somayeh Mousavib, F.
Soleymani, A real-time mathematical computer
method for potato inspection using machine vision,
Computers and Mathematics with
Applications 63
(2012) 268279.
B. Obara, A new algorithm using image colour
system transformation for rock grain segmentation,
Mineralogy and Petrology (2007)
91: 271285.
Peggy Walter and Paul Cornillon, Influence of
Thermal Conditions and Presence of Additives on Fat
Bloom in Chocolate, Paper no. J9577 in JAOCS 78,
927932 (September 2001).
Tom Pearson, A. Enis Cetin, Ahmed H. Tewfik, and
Vural Gkmen, An Overview of SignalProcessing
for Food Inspection, IEEE Signal Processing
Magazine , May 2007.
RuchiTaneja, Gagandeepsingh Gill, Reman, Quality
Evaluation of Agricultural and Food Products Using
EVision-A Review, International Journal
Of
Research In Management, Science & Technology
Vol.1, No. 1.
Chetna V. Maheshwari, Quality Assessment Of
Oryza Sativa Ssp Indica (Rice) Using Computer
Vision, International Journal of Innovative Research
in Computer and Communication
Engineering
Vol. 1, Issue 4, June 2013.
Kit L. Yam, Spyridon E. Papadakis, A simple digital
imaging method for measuring and analyzing color
of food surfaces, Journal of Food Engineering 61
(2004) 137142.
Narendra V G, Hareesh K S, Quality Inspection and
Grading of Agricultural and Food Products by
Computer Vision- A Review, International Journal
of Computer Applications (0975 8887) Volume 2
No.1, May 2010.
Farhad, N. and S. Terry, Performance analysis and
optimization of shape recognition and classification
using ANN, Robot. Comput. Integ. Manuf., 18(34): 177-185, 2002.
Narendra VG, Hareesh KS Prospects of computer
vision automated grading and sorting systems in
agricultural and food
products for
quality

International Conference on Current Trends in Engineering and Technology, ICCTET13

[17]

evaluation, (2010) Int J Comp App 1:


0975
8887.
Wang, H.-H., & Sun, D.-W.,Correlation between
cheese meltability determined with a computer vision
method and with Arnott and Schreiber, Journal of
Food Science (2002a),
67(2),745749.

IEEE 2013
IEEE 32107
July 3, 2013, Coimbatore, India.

33

You might also like