You are on page 1of 12

©2010 International Journal of Computer Applications (0975 – 8887)

Volume 1 – No. 4



Narendra V G1 and Hareesh K S2

Sr. Lecturer Dept. Of CSE, Manipal Institute Of Technology, Manipal, Karnataka, India-576 104
Reader Dept. Of CSE, Manipal Institute Of Technology, Manipal, Karnataka, India-576 104

The paper presents the recent development and application of image analysis and computer vision systems in sorting and
grading of products in the field of agricultural and food. Basic concepts and technologies associated with computer vision,
a tool used in image analysis and automated sorting and grading is highlighted.

For the ever-increasing population, losses in handling and processing and the increased expectation of food products of
high quality and safety standards, there is need for the growth of accurate, fast and objective quality determination of the
characteristics of food and agricultural food products. Computer vision and image analysis, are non-destructive and cost-
effective technique for sorting and grading of agricultural and food products during handling processes and commercial
purposes. Different approaches based on image analysis and processing identified is related to variety of applications in
agricultural and food products.

Keywords: Quality, Sorting and Grading, Automation, Image Processing, Machine Vision;

1. INTRODUCTION obtain information or, to control machines or

Technological advancement is gradually finding processes [47].
applications in the agricultural and food products, in
response to one of the greatest challenges i.e. meeting In Timmermans [51] opinion, computer vision system
the need of the growing population. Efforts are being includes the capturing, processing and analyzing
geared up towards the replacement of human operator images to facilitate the objective and non-destructive
with automated systems, as human operations are assessment of visual quality characteristics in
inconsistent and less efficient. Automation means every agricultural and food products. The techniques used in
action that is needed to control a process at optimum image analysis include image acquisition, image pre-
efficiency as controlled by a system that operates using processing and image interpretation, leading to
instructions that have been programmed into it or quantification and classification of images and objects
response to some activities. Automated systems in most of interest within images.
cases are faster and more precise. However, there are
some basic infrastructures that must necessarily be in Harvesting and packing account for the major portion
place in automation [1]. of the effort and cost incurred by farmers producing
fresh fruits and vegetables to market. However the
The technology of image analysis is relatively young and processing and manufacturing sectors require product
its origin can be traced back to the 1960s. It has sorting and grading for commercial and production
experienced tremendous growth both in theory and in purposes. There is continuous growth in the
application. It has found application in areas such as development of mechanical harvesting system, and the
medical diagnostics, automated manufacturing, aerial need for automated inspection, as well as grading
surveillance, remote sensing and very recently in the systems so that the losses incurred during harvesting,
field of automated sorting and grading of agricultural production and marketing can be minimized. With
and food products. Computer vision is a novel these, the need arises to not only grow and harvest a
technology for acquiring and analyzing an image of a quality crop, but also to pack in a consistent and
real scene by computers and other devices in order to acceptable manner to gain or to maintain market share

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

as well as prepare materials, which suits processing vision instead of the conventional use of black and
operations. These cannot be achieved without sorting white cameras and geometric features for automatic
and grading. classification of tissue culture segments of potato
plantlets. It was found that instead of using only
Sorting and grading of an agricultural and food products geometric features combining it with colour
is accomplished based on appearance, texture, colour, significantly increased the separability of the classes in
shape and sizes. Manual sorting and grading are based the feature space.
on traditional visual quality inspection performed by
human operators, which is tedious, time-consuming, Real time detection of defects in fruits using a general
slow and non-consistent. A cost effective, consistent, hardware and image processing techniques was
superior speed and accurate sorting can be achieved with reported by Delwiche and Crowe [9]. Two fruits
machine vision assisted sorting and grading. Automated (apples and peaches) were tested at the rate of five
sorting and grading had undergone substantial growth in fruits/second to evaluate system performance. They
the field of agricultural and food, in the developed and developed an algorithm to acquire and analyze two
developing nations because of availability of combined near infrared (NIR) images of each fruit in
infrastructures. Computer application in agriculture and real time with a pipeline image processing system.
food industries have been applied in the areas of sorting, Their result showed that the system was capable of
grading of fresh products, detection of defects such as executing the sorting algorithm at a rate of 14
cracks, dark spots and bruises on fresh fruits and seeds. fruits/second, which was greater than conventional
The new technologies of image analysis and machine fruit conveying rates. In this submission, it was
vision have not been fully explored in the development observed that acquiring more than 2 images per fruit
of automated machine in agricultural and food and using more than 6 lines of structured illumination
industries. There is increasing evidence that machine per fruit would reduce the sorting errors slightly in the
vision is being adopted at commercial level [22]. case of the apples and greatly improve the system
performance with peaches. Algorithms were
The method used by the farmers and distributors to sort developed by Panigrahi and Misra [32] to measure
and grade agricultural and food products are through fractal-based feature and dimensions (length, width,
traditional quality inspection and handpicking which is area and perimeter) of ear corn images to discriminate
time-consuming, laborious and less efficient. Sun et al. among various shapes. The analysis showed that the
[48] observed that the basis of quality assessment is combination of the feature is able to discriminate the
often subjective with attributes such as appearance, shape differences while Carrion et al [8] described
smell, colour, texture and flavor frequently examined by sorting based on an unsupervised vision system.
human inspectors. Francis [13] found that human
perception could easily be fooled. It is pertinent to Raji [36] developed an algorithm for determining the
explore the possibilities of adopting faster systems, area of 2-dimensional objects by image analysis. This
which will save time and more accurate in sorting and can be adapted in detection of leave type as a form of
grading of agricultural and food products. One of such sorting to select the desired ones during harvesting.
reliable method is the automated computer vision system The discrimination of leave from weed using feature
for sorting and grading. identification by analysis was investigated and
reported by Tsheko [53].
Advances in computer technology have produced a surge
of interest in image analysis during the last decade and Bull [7] observed the image capture techniques, which
the potential of this technique for the guidance or control could be or are being used in the field of the
of agricultural and food processes have been recognized agricultural and food, to generate image that can be
[37]. Series of studies have been conducted in recent analyzed and used as feedbacks for an automated
years to investigate the application of computer vision systems. The appropriate technique to be adopted by
technology to sorting and grading of fresh agricultural each industry would depend on the property of the
and food products. Yam and Spyridon [60] used a simple sample to be monitored, the nature of the sample, its
digital imaging method for measuring and analyzing environments and other practical restrictions such as
colour of food surfaces and found that the method allows imaging time. With increased awareness and
measurements and analysis of the colour of food sophistication on the part of the consumers and the
surfaces that are adequate for food engineering research. expectation for improved quality in consumer
Payne and Shearer [33] developed a machine vision agricultural and food products which has increased the
algorithm for grading of fresh market produce according need for enhanced quality monitoring in field of
to colour and damage while Alchanatis et al [3], used a agricultural and food, there is a need to develop
neural network based classifier and colour machine various techniques of image analysis to meet the

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

demand of the growing population. Quality is the sum of • A high-resolution colour monitor, which aids in
these attributes that can lead to the production of visualizing images and the effects of various image
products acceptable to the consumer when they are analysis routines.
In the food industries in developing countries, there is 3.1. Illumination
growing need for automation due to the fact that the Computer Vision Systems are affected by the level and
labour intensive manual processes are not efficient, quality of illumination as with the human eye. The
accurate and effective. In the developed countries on the performance of the illumination system greatly
other hand Gunasekaran [14] observed that the food influences the quality of image and plays an important
industry is now ranked among the top 10 industries using role in the overall efficiency and accuracy of the
this technology. system. Illumination systems are the light sources. The
light focuses on the materials (especially when used).
This study therefore sets out to review the state and level Lighting type, location and colour quality play an
of application of image processing and machine vision in important role in bringing out a clear image of the
automated sorting and grading with its application and object. Lighting arrangements are grouped into front-
steps to be taken in improving and developing the or back-lighting [15]. Front lighting serve as
emerging technology beyond its present state to illumination focusing on the object for better detection
agricultural and food products sorting and grading. of external surface features of the product while back-
lighting is used for enhancing the background of the
object. Light sources used include incandescent lamps,
2. TECHNIQUES IN IMAGE PROCESSING fluorescent lamps, lasers, X-ray tubes and infra-red
2.1. Equipment and Techniques lamps.
A computer vision system consists of two basic
components which are image acquisition: illumination 3.2. Image Acquisition
and image capture device (camera) and image analysis: Image capturing devices or sensors are used to view
an image capture board (frame grabber or digitizer) and and generate images of the samples. Some of the
analysis software. A typical laboratory set for image devices or sensors used in generating images include
processing is as shown in Figure 1. scanners, ultrasound, X-ray and near infrared
spectroscopy. However, in machine vision, image
sensors used are the solid state charged coupled device
(CCD) (i.e. camera) technology with some
applications using thermionic tube devices. Recent
technology has seen the adoption of digital camera,
which eliminates the additional component required to
convert images taken by photographic and CCD
cameras or other sensors to readable format by
computer processors. Images captured or taken by
digital camera maintain the features of the images with
Figure 1: Components of a computer vision system [56] . little noise due to its variable resolution.

2.2. Computer Vision System 4. IMAGE PRE-PROCESSING

The hardware configuration of computer-based This refers to the initial processing of the raw image.
Computer vision systems is relatively standard. The images captured or taken are transferred onto a
Typically, a computer vision system consists of: computer and are converted to digital images. Digital
• An illumination device, which illuminates the sample images though displayed on the screen as pictures, are
under test digits, which are readable by the computer and are
• A solid-state CCD array camera, to acquire an image converted to tiny dots or picture elements representing
• A frame-grabber, to perform the A/D (analog-to- the real objects. In some cases preprocessing is done to
digital) conversion of scan lines into picture elements or improve the image quality by suppressing undesired
pixels digitized in a N row by M column image distortions referred to as “noise” or by the
• A personal computer or microprocessor system, to enhancement of important features of interest.
provide disk storage of images and computational
capability with vendor-supplied software and specific The images or pictures are transformed into computer
application programs digital readable format (i.e. digitized) if a digital
camera did not take them by the image board digitizer.

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

The digitized format is then transferred and used as the sending signals to a device to take a feature based
input data by the image processing software to carry out decision. This process is useful in colour (maturity)
the necessary processes. Each or a combination of the and feature based (defect and damages detection)
digits represent the feature a small portion of the image sorting.
called picture element (pixel). Objects are described as
black and white pictures which are represented by digits
ranging from 0 to 255 where 0 is black and 255 is white.

Each pixel in coloured pictures is represented by 3 digits

representing 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
using the matrix theory does the analysis in image Original Image Threshold image
processing. Further details can be obtained in Tsheko Figure 3. Thresholding
[53], Raji [36], Raji et al [37] and any text or reports on
image processing and computer machine vision. Edge based segmentation relies on detection by edge-
to-edge operators, which detect discontinuities in grey
level, the pixel, colour, texture etc. Edge detection is
useful in shape and size sorting. An example of edge
detection result is shown in Figure 4. The application
of this process was reported by Raji et al [37] who
demonstrated the detection of defects in the shape of
bread and biscuit samples on a processing line.

Figure 2: Different levels in the image processing

Figure 4: Edge based segmentation
Image acquisition and image pre-processing are
categorized as low-level processing while the
intermediate and the high level processing stages as Region based segmentation involves the grouping
classified by Sun [43] (Figure 2) are further processing together and extraction of similar pixels to form
stages required when integrating the preprocessing regions representing single objects within the image.
stages into an application device such as a sorter and (Figures 5). In this process the other regions are
grader. deleted leaving only the feature of interest.

The intermediate-level processing involves image High level processing deals with recognition and
segmentation, image representation and image interpretation, typically using statistical classifiers or
description. Image segmentation is a process of cutting, multiplayer neural networks of region of interest.
adding and feature analysis of images aimed at dividing These steps provide information necessary for the
an image into regions that have a strong correlation with process or machine control for quality sorting and
objects or areas of interest using the principle of matrix grading.
analysis. Segmentation can be achieved by the following
techniques: thresholding, edge based segmentation and
region based segmentation. Thresholding is used in
characterizing image regions based on constant
reflectivity or light absorption of their surface. This
shows that regions with same features are characterized
and extracted together. Figure 3 shows a thresholding
process where only the dark region is of interest, the
other regions are converted to the background colour in
the threshed image before further processing such as

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

Figure 5. Region based segmentation segmentation techniques were found to be inadequate

An interaction of all these levels and knowledge for this application developed a new segmentation
database are very important and essential for more algorithm. It was found that the new region-based
precise decision-making and is seen as an integral part of segmentation technique could effectively group pixels
the image processing process. of the same topping together. As the result, topping
exposure percentage can be easily determined. The
These theories when applied to images of products taken study reported that the accuracy of measuring the
can be used to extract features, which are needed for the topping percentage by the new algorithm reached
necessary processes. Generally, edge detection to 90%.
determine shape and feature extraction to determine
differences in colour is useful in sorting and harvesting. The evaluation of the functional properties of cheese is
assessed to ensure the necessary quality is achieved,
5. APPLICATIONS IN AGRICULTURAL AND especially for specialized applications such as
FOOD PROCESSING OPERATIONS consumer food toppings or ingredients. Wang and Sun
Computer vision systems are being used increasingly in [55] developed a computer vision method to evaluate
the field of agricultural and food products for quality the melting and browning of cheese. This novel non-
assurance purpose. The system offers the generation of contact method was employed to analyze the
precise descriptive data and reduction of tedious human characteristics of cheddar and mozzarella cheeses
involvement. Computer vision system has proven during cooking and the results showed that the method
successful for the objective, online measurement of provided an objective and easy approach for analyzing
several agricultural and food products with application cheese functional properties [56,57]. Ni and
ranging from routine inspection to the computer vision Gunasekaran [29] developed an image-processing
system guided robotic control [48]. Some of the areas algorithm to recognize individual cheese shred and
where the techniques have been applied in agricultural automatically measure the shred length. It was found
and food processing which need to be developed further that the algorithm recognized shreds well, even when
for commercial purposes include. they were overlapping. It was also reported that the
shred length measurement errors were as low as 0.2%
5.1 Bakery Products with a high of 10% in the worst case.
Appearance of baked products is an important quality
attribute which influences the visual perceptions of The application of this method is therefore a promising
customers and hence potential demands of the products. approach to solving quality control inspection in the
The appearance of the internal and external features bakery products. This will improve the quality of
contributes to the overall impression of the products baked products, where there are no standard sizes,
quality. Computer vision system has been used to shape and texture. The required system though may
measure characteristics such as colour, size and shape add to the cost of production but the added advantage
with a view to sorting them to products with same will far outweigh the initial cost involved.
characteristics before packing.
5.2. Fruits
Raji et al [37] developed a programme in FORTRAN External qualities i.e. sizes, shapes and colour are
using the principle of edge detection in image analysis to considered of paramount importance in the marketing
determine the edge of sliced breads and biscuits (round and sale of fruit. Presence of blemishes influences
and rectangular) with a view to detecting defects consumer perceptions and therefore determines the
(breakage). Scott [41] described a system, which level of acceptability prior to purchase. Computer
measures the defects in baked loaves of bread, by vision system has been used for the automated
analyzing its weight and slope of the top. The internal inspection and grading of fruit to increase product
structure (crumb grain) of bread and cake was also throughput.
examined by machine vision [40]. Dos Mohammed et al
[10] also developed a system for the automated visual The number of fruits (ripe and unripe) on a tree has
inspection of muffins. been counted by image analysis prior to harvesting.
This process involved the development of fruit
Visual features such as colour and size indicate the location algorithms [24]. Images were taken from a
quality of many prepared consumer foods. Sun [43] distance of about 150cm and all the fruit on the tree
investigated this in research on pizza in which pizza placed within the vision field of a camera were
topping percentage and distribution were extracted from counted. The visible fruits were considered to be those
pizza images. Combining three algorithms used to that the human eye can distinguish on the monitor.
segment many different types of pizzas as the traditional Algorithms based on the red/green relation and on

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

threshold achieved the highest detection percentages on shown to be a viable means of meeting these increased
citrus fruit. requirements for the vegetables field.

The adoption can also be extended to the peasants Plantlet segments of potato were subcultures for
through tractor hiring bodies or by forming cooperatives classification by colour machine vision system by
to acquire such devices, which only need to be attached Alchanatis et al., [3]. In related development, two
to a tractor. The problem of unplanned fruit trees, which algorithms were developed for analyzing digital binary
may make machine movement very difficult, will also images and estimating the location of stem root joints
need to be addressed. The high losses incurred presently in processing carrots [5]. Both algorithms were
on fruits in the market comes as a result of the damages capable of estimating the stem/root location with a
and bruises as well as potential damage region inflicted standard deviation of 5mm. Also Haworth and Searcy
on the fruits falling from the trees due to the method of [17] classified carrots on surface defects, curvature
harvesting of shaking the trees or the branches. The and brokenness. A line scans images with discrete
adoption can also be extended to the peasants through Fourier transform was developed for the classification
tractor hiring bodies or by forming cooperatives to of broccoli heads for assessing its maturity. For the
acquire such devices, which only need to be attached to a 160 observations from each of three broccolis
tractor. The problem of unplanned fruit trees, which may cultivates, an accuracy of 85% was achieved for
make machine movement very difficult, will also need to multiple cultivars.
be addressed. The high losses incurred presently on
fruits in the market comes as a result of the damages and Mushrooms‟ discolouration is undesirable in
bruises as well as potential damage region inflicted on mushroom houses and it reduces market value. The
the fruits falling from the trees due to the method of colour and shape of the cap is the most important
harvesting of shaking the trees or the branches. consideration of fresh mushrooms. Felfodi and
Vizhanyo [11] used mushroom images recorded by a
Strawberry appearance and fruit quality are dependent machine vision system to recognize and identify
on a number of pre- and post- harvest factors, hence discolouration caused by bacterial disease. The
variation occurs, necessitating the need for sorting. method identified all the diseased spot as „diseased‟
Nagata et al. [26] investigated the use of computer vision and none of the healthy mushrooms parts were
to sort fresh strawberries, based on size and shape. The detected as „diseased‟. Reed et al [38] used camera-
experimental results show that the developed system was based technology to select mushroom by size for
able to sort the 600 strawberries tested with an accuracy picking by a mushroom harvester.
of 94-98% into three grades based on shape and five
grades on size. Another automatic strawberry sorting Potatoes have many possible shapes, which need to be
system was developed by Bato et al. [6]. Average shape graded for sale into uniform classes for different
and size accuracies of 98 and 100%, respectively, were markets. This created difficulties for shape separation.
obtained regardless of the fruit orientation angle with A Fourier analysis based shape separation method for
judgement time within 1.18 s. grading of potatoes using machine vision for
automated inspection was developed by Tao et al.
Prior to export, papayas are subjected to inspection for [50]. A shape separator based on harmonics of the
the purpose of quality control and grading. For size transform was defined. Its accuracy of separation was
grading, the fruit is weighted manually hence the 89% for 120 potato samples, in agreement with
practice is tedious, time consuming and labour intensive. manual grading. Earlier, Lefebvre et al. [19] studied
Therefore, a computer vision system for papaya size the use of computer vision for locating the position of
grading using shape characteristic analysis. The shape pulp extraction automatically for the purpose of
characteristics consisting of area, mean diameter and further analysis on the extracted sample. An image
perimeter were extracted from the papaya images. acquisition system was also constructed for mounting
Slamet Riyadi et al. [43] classified according to on a sweet potato harvester for the purpose of yield
combinations of the three features to study the and grade monitoring [58]. It was found that culls
uniqueness of the extracted features. The proposed were differentiated from saleable sweet potatoes with
technique showed the ability to perform papaya size classification rates as high as 84%.
classification with more than 94% accuracy.
5.3. Vegetables Chilli is a variety grown extensively consumed by
The need to be responsive to market demands places a almost all the population. It has a high processing
greater emphasis on quality assessment resulting in the demand and proper sorting is required before filling or
greater need for improved and more accurate grading canning. A sorter that, Federico Hahn [12] classifies
and sorting practices. Computer vision system has chilli by three different width sizes was built. The

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

conveyor used baby suckers to align each chilli during Computer vision has been used in grain quality
sensing. Chilli width was determined by means of a inspection for many years. An early study by Zayas et
photodiode scanner, which detected the incoming al. [61] used machine vision to identify different
radiation sent by a laser line generator. Chilies varieties of wheat and to discriminate wheat from non-
presenting necrosis were detected with a radiometer and wheat components.
removed to increase product quality. Horizontal and In later research Zayas et al. [62] found that wheat
vertical widths were measured for 200 chilies. The classification methods could be improved by
accuracy on the necrosis detection and width combining morphometry (computer vision analysis)
classification was of 96.3 and 87%, respectively. On-line and hardness analysis. Hard and soft recognition rates
necrosis measurements were 85% accurate when only of 94% were achieved for the seventeen varieties
the relative reflectance at 550nm was used. examined. Twenty-three morphological features were
used for the discriminant analysis of different cereal
Vegetable quality is frequently referred to size, shape, grains using machine vision [23]. Classification
mass, firmness, colour and bruises from which fruits can accuracies of 98, 91, 97,100 and 91% were recorded
be classified and sorted. However, technological by for CWRS (Canada Western Red Spring) wheat,
small and middle producers implementation to assess CWAD (Canada Western Amber Durum) wheat,
this quality is unfeasible, due to high costs of software, barley, oats and rye, respectively. 25 kernels per image
equipment as well as operational costs. Based on these were captured from a total of 6000 for each grain type
considerations, the Antonio Carlos Loureiro Lino et al. examined.
[4] research is to evaluate a new open software that The relationship between colour and texture features
enables the classification system by recognizing fruit of wheat samples to scab infection rate was studied
shape, volume, colour and possibly bruises at a unique using a neural network method [39]. It was found that
glance. The software named ImageJ, compatible with the infection rates estimated by the system followed
Windows, Linux and MAC/OS, is quite popular in the actual ones with a correlation coefficient of 0.97
medical research and practices, and offers algorithms to with human panel assessment and maximum and mean
obtain the above-mentioned parameters. The software absolute errors of 5 and 2%, respectively. In this study
allows calculation of volume, area, averages, border machine vision-neural network based technique
detection, image improvement and morphological proved superior to the human panel. Image analysis
operations in a variety of image archive formats. has also been used to classify dockage components for
CWRS (Canada Western Red Spring) wheat and other
Some other earlier studies of computer vision associated cereals [27]. Morphology, colour and morphology-
with vegetable grading and inspection include colour and colour models were evaluated for classifying the
defect sorting of bell peppers [42]. Morrow et al. [25] dockage components. Mean accuracies of 89 and 96%
presented the techniques of vision inspection of for the morphology model and 71 and 75% for the
mushrooms, apples and potatoes for size, shape and colour model were achieved when tested on the test
colour. The use of computer vision for the location of and training data sets, respectively. Overall 6000
stem/root joint in carrot has also been assessed [5]. kernels for each grain type were analyzed. Machine
Feature extraction and pattern recognition techniques vision was used to identify weeds commonly found in
were developed by Howarth and Searcy [17] to wheat fields in experimentation by Zhang and
characterize and classify carrots for forking, surface Chaisattapagon [63]. Five shape parameters were used
defects, curvature and brokenness. The rate of in leaf shape studies and were found effective in
misclassification was reported to be below 15% for the distinguishing broadleaf weed species such as
250 samples examined. More recently sweet onions were pigweed, thistle and kochia from wheat.
line scanned for internal defects using X-ray imaging
[52]. An overall accuracy of 90% was achieved when In order to preserve corn quality it is important to
spatial and transform features were evaluated for product obtain physical properties and assess mechanical
classification. damage so as to design optimum handling and storage
equipment. Measurements of kernel length, width and
Image analysis and machine vision in general from the projected area independent of kernel orientation have
foregoing can be said to offer a fast and reliable process been performed using machine vision [30]. The
in product sorting and grading, separation and detection algorithm accuracy was between 0.86 and 0.89
of some other facilities. measured by the correlation coefficient between
predicted results and actual sieving for a 500 g sample.
5.4. Grains The processing time of the size-grading program was
Grain quality attributes are very important for all users reported as being between 0.66 and 0.74 s per kernel.
and especially the milling and baking industries. Steenhoek and Precetti [46] performed a study to

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

evaluate the concept of two-dimensional image analysis structure or added detail is required. A 3-D vision
for classification of maize kernels according to size technique has been developed to derive a geometric
category. A total of 320 maize kernels were categorized description from a series of 2-D images [44]. In
into one of 16 size categories based on degree of practice this technique might be useful for food
roundness and flatness. Classification accuracy of both inspection. For example, when studying the shape
machine vision and screen systems was above 96% for features of a piece of bakery, it is necessary to take 2-
round-hole analysis. However, sizing accuracy for D images vertically and horizontally to obtain its
flatness was less than 80%. roundness and thickness, respectively.

Ng et al. [28] developed a machine vision algorithm for Kanali et al. [18] investigated the feasibility of using a
corn kernel mechanical and mould damage charge simulation method (CSM) algorithm to process
measurement, which demonstrated a standard deviation primary image features for three dimensional shape
less than 5% of the mean value of the 250 grains recognition. The required features were transferred to
examined. They found that this method was more a retinamodel identical to the prototype artificial retina
consistent than other methods available. The automatic and were compressed using the CSM by computing
inspection of 600 corn kernels was also performed by Ni output signals at work cells located in the retina. An
et al. [31] using machine vision. For whole and broken overall classification rate of 94% was obtained when
kernel identification on-line tests had successful the prototype artificial retina discriminated between
classification rates of 91 and 94% for whole and broken distinct shapes of oranges for the 100 data sets tested.
kernels, respectively. Gunasekaran and Ding [16] obtained 3-D images of fat
globules in cheddar cheese from 2-D images. This
The whiteness of corn has been measured by an on-line enabled the in situ 3-D evaluation of fat globule
computer vision approach by Liu and Paulsen [20]. For characteristics so as the process parameters and fat
the 63 samples (50-80 kernels per sample) tested the levels may be changed to achieve the required textural
technique was found to be easy to perform with a speed qualities.
of 3 kernels per s. In other studies Xie and Paulsen [59]
used machine vision to detect and quantify tetrazolium 7. PROBLEMS OF AUTOMATED SORTING
staining in corn kernels. The tetrazolium-machine vision AND GRADING
algorithm was used to predict heat damage in corn due to The major problem of automated sorting and grading
drying air temperature and initial moisture content. is one of socio-economic effects, which tend to reduce
employment when the number of operators required in
As rice is one of the leading food crops of the world its the processing line is reduced. It is not suitable in
quality evaluation is of importance to ensure it remains processes where manual skill is necessary or
appealing to consumers. Liu et al. [20] developed a economically more attractive. It requires higher initial
digital image analysis method for measuring the degree and maintenance cost and there may be the need for a
of milling of rice. They compared the method with precise understanding of the process for programming
conventional chemical analysis and obtained a to achieve the required product quality Equipment
coefficient of determination of R2 =/0.9819 for the 680 associated problem involves radiance on accurate
samples tested. Wan et al. [54] employed three online sensors to precisely measure process condition and the
classification methods for rice quality inspection:- increased risk, delays and cost if the automatic system
namely range selection, neural network and hybrid fails. The farm layout and very low production level
algorithms. The highest recorded online classification which may make it uneconomically viable has also
accuracy was around 91% at a rate of over 1200 been highlighted.
kernels/min. The range selection method achieved this
accuracy but required time-consuming and complicated 8. PROSPECTS OF AUTOMATED SORTING
adjustment. In another study, milled rice from a AND GRADING
laboratory mill and a commercial-scale mill was The adoption of this emerging technology by first
evaluated for head rice yield and percentage whole putting more effort into researches on the appropriate
kernels, using a shaker table and a machine-vision methods and ways of application will be of immense
system called the GrainCheck [21]. benefit to this country. Some of the other associated
benefits include increased production rates (e.g.
6. 3-D technique through optimization of equipment utilization), more
In general, only 2-dimensional (2D) data are needed for efficient operation, production of more consistent
grading, classification, and analysis of most agricultural product quality, greater product stability and safety.
images. However, in many applications 3-dimensional
image analysis maybe needed as information on

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

With the above in mind, the fruit and vegetable been neglected, and hence it needs more focused and
processing and marketing industries stand to gain from detailed study.
this emerging technology. This is because the losses
incurred during the harvesting season on fruit are REFERENCES
enormous. This results from large number of products to [1].A. Raji and A. Alamutu, “Prospects of Computer
be handled and sold at the same time. These products Vision Automated Sorting Systems in Agricultural
consist of ripe and unripe fruits. The introduction of Process Operations in Nigeria”, Agricultural
automated sorting and grading will encourage the sorting Engineering International: the CIGR Journal of
of the unripe (which can be kept for a relatively longer Scientific Research and Development, Vol. VII,
period) from the ripe, which are to be sold immediately. Invited Overview, February 2005.
Presently the practice at the fruit market in to sell [2].Ahmada, U., Kondo, N., Montaa, M., and H.
baskets of fruits containing both ripe and unripe. These Muraseh, “Machine Vision Based Quality Evaluation
products are found to get spoilt before they get to the of Iyokan Orange Fruit using Neural Networks”,
final destination. Computers and Electronics in Agriculture, 29(1-2),
pp.135-147, 2000.
One of the major hindrances to the introduction of this [3].Alchanatis V., Peleg, K., and M. Ziv,
technique is the level of production, which remains at “Classification of Tissue Culture Segments by Colour
the peasant level. Most of the products found in the Machine Vision”, Journal of Agricultural Engineering
markets are owned by a number of individuals with each Research, 55, pp. 299-311,1993.
controlling not more than 4 to 5 baskets, which will be [4].Antonio Carlos Loureiro Lino; Juliana Sanches;
too small for the adoption of an automated technique. Inacio Maria Dal Fabbro, “Image processing
However, encouraging wholesalers and retailers for techniques for lemons and tomatoes classification”,
groups and cooperatives in marketing as practiced in the Journal of Food and Engineering, 83(4), pp. 433–
village level for maize shelling will be a way out. 440, 2008.
The adaptation of computer vision system for quality [5].Batchelor, M.M., and S.W. Searcy, “Computer
evaluation of processed foods is the area for the greatest Vision Determination of Stem/Root Joint on
potential uptake of this technology, as analysis can be Processing Carrots”, Journal of Agricultural
based on a standard requirement in already automated Engineering Research, 43, pp. 259-269, 1989.
controlled conditions. More complex systems are needed [6].Bato, P.M., Nagata, M., Cao, Q.X., Hiyoshi, K.,
for the automated grading and sorting of fresh produce Kitahara, T., ”Study on sorting system for strawberry
because of the greater range in variability of quality and using machine vision (part 2): development of sorting
also as produce orientation may influence results. With system with direction and judgement functions for
the idea of precision and more environmental friendly strawberry (Akihime variety)”, Journal of the
agriculture becoming more realistic the potential for Japanese Society of Agricultural Machinery, 62(2),
computer vision system in this area is immense with the pp. 101-110, 2000.
need in field crop monitoring, assessment and guidance [7].Bull, C.R., “A Review of Sensing Techniques
systems. However, techniques such as 3D and colour which could be used to Generate Images of
vision will ensure computer vision development Agricultural and Food Materials”, Computers and
continues to meet the accuracy and quality requirements Electronics in Agriculture, 8, pp. 1-29, 1992.
needed in this highly competitive and changing industry. [8].Carrion, J.A., Torregrosa, E.O., and E. Molto,
“First Result of an Automatic Citrus Sorting Machine
9.CONCLUSIONS Based on an Unsupervised Vision System”,
The paper presents the recent developments in computer Proceedings of the InternationalConference on
vision system in the field of agricultural and food Agricultural Engineering, Oslo, August 1998.
products. Computer vision systems have been used [9].Delwiche, M.J., and T.G. Crowe, “Real-Time
increasingly in industry for inspection and evaluation Defect in Food-Part II: An Algorithm and
purposes as they can provide rapid, economic, hygienic, Performance of a Prototype System”, Transactions of
consistent and objective assessment. However, the ASAE, 39(6), pp. 2309-2317, 1996.
difficulties still exist, evident from the relatively slow [10].Dos Mohammed, A.M., Abdullah, Z.M., and A.S.
commercial uptake of computer vision technology in all Aziz, “Quality Inspection of Bakery Products using a
sectors. Even though adequately efficient and accurate Colour-Based Machine Vision System”, Journal of
algorithms have been produced, processing speeds still Food Quality, 23(1), pp. 39-50, 2000.
fail to meet modern manufacturing requirements. With [11].Felfodi, J., and T. Vizhanyo, “Enhancing Colour
few exceptions, research in this field has dealt with trials Differences in Images of Diseased Mushrooms”,
on a laboratory scales thus the area of mechatronics has Computers and Electronics in Agriculture, 26, pp.187-
198, 2000.

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

[12].Federico Hahn,”Automatic Jalapeno Chilli Grading [25].Morrow, C.T., Heinemann, P.H., Sommer, H.J.,
by Width”, Biosystems Engineering, 83(4), pp. 433–440, Tao, Y., Varghese, Z., “Automate inspection of
2002. potatoes, apples, and mushrooms”, In Proceedings of
[13].Francis, F.J., “Colour Quality Evaluation of the International Advanced Robotics Programme,
Horticultural crops”, HortScience, 15(1), pp. 14-15, Avignon, France, pp. 179-188, 1990.
1980. [26].Nagata, M., Cao, Q., Bato, P.M., Shrestha, B.P.,
[14].Gunasekaran, S., “Computer Vision Technology for Kinoshita, O., “Basic study on strawberry sorting
Food Quality Assurance”, Trends in Food Science and system in Japan”, In: 1997 ASAE Annual International
Technolog, 7(8), pp. 245-256, 1996. Meeting Technical Papers, Paper No. 973095, ASAE,
[15].Gunasekaran, S., “Non-Destructive Food 2950 Niles Road, St. Joseph, Michigan 49085-9659,
Evaluation Techniques to Analyze Properties and USA,1997.
Quality”, Food Science and Technology Vol. 105, [27].Nair, M., Jayas, D.S., Bulley, N.R., “Dockage
Marcel Decker, New York, 2001. identification in wheat using machine vision”, 1997
[16].Gunasekaran, S., Ding, K., “Three-dimensional ASAE Annual International Meeting Technical Papers,
characteristics of fat globules in cheddar cheese”, Paper No. 973043, ASAE, 2950 Niles Road, St.
Journal of Dairy Science 82, pp.1890-1896, 1999. Joseph, Michigan 49085-9659, USA,1997.
[17].Howarth, M.S., and S.W. Searcy, “Inspection of [28].Ng, H.F., Wilcke, W.F., Morey, R.V., Lang, J.P.,
Fresh Carrots by Machine Vision”, In Food Processing “Machine vision evaluation of corn kernel mechanical
Automation II Proceedings of the 1992 ASAE and mould damage”, In: 1997 ASAE Annual
Conference, St. Joseph, Michigan, USA, 1992. International Meeting Technical Papers, Paper No.
[18].Kanali, C., Murase, H., Honami, N., ”Three- 973047, ASAE, 2950 Niles Road, St. Joseph,
dimensional shape recognition using a chargesimulation Michigan 4 9085-9659, USA, 1997.
method to process image features”, Journal of [29].Ni, H., Gunasekaran, S., “A computer vision
Agricultural Engineering Research, 70, pp. 195-208, system for determining quality of cheese shreds”, In:
1998. Food Processing Automation IV Proceedings of the
[19].Lefebvre, M., Gil, S., Brunet, D., Natonek, E., Baur, FPAC Conference. ASAE, 2950 Niles Road, St.
C., Gugeril, P., Pun, T., “Computer vision and Joseph, Michigan 49085-9659, USA, 1995.
agricultural robotics for disease control: the potato [30].Ni, B., Paulsen, M.R., Reid, J.F., “Size grading of
operation”, Computers and Electronics in Agriculture 9, corn kernels with machine vision”, In: 1997 ASAE
pp. 85-102, 1993. Annual International Meeting Technical Papers, Paper
[20].Liu, J., Paulsen, M.R., “Corn whiteness No. 973046, ASAE, 2950 Niles Road, St.Joseph,
measurement and classification using machine vision”, Michigan 49085-9659, USA, 1997a.
In:1997 ASAE Annual International Meeting, Technical [31].Ni, B., Paulsen, M.R., Liao, K., Reid, J.F.,
Papers, Paper No. 973045, ASAE, 2950 Niles Road, St. “Design of an automated corn kernel inspection
Joseph, MI 49085-9659, USA, 1997. system for machine vision”, Transactions of the ASAE,
[21].Lloyd, B.J., Cnossen, A.G., Siebenmorgen, T.J., 40(2), pp. 491-497, 1997b.
“Evaluation of two methods for separating head rice [32].Panigrahi, S., and M.K. Misra, “Feature
from brokens for head rice yield determination”, In: Extraction Techniques for Corn Germplasm by Colour
2000 ASAE Annual International Meeting, Paper No. Computer Vision”, In 1990 ASAE International
006074 ASAE, 2950 Niles Road, St. Joseph, MI 49085- Summer Meeting, paper No 90-7050, St. Joseph,
9659, USA, 2000. Michigan, USA ,1990.
[22].Locht, P., Thomas, K., and P. Mikkelsen, “Full [33].Payne, F.A., and S.A. Shearer, “Colour and
Colour Image Analysis as a Tool for Quality Control and Defect Sorting of Bell Peppers using Machine Vision”,
Process Development in the Food Industry”, In 1997 Transactions of the ASAE, 33(6), pp.2045-2050, 1990.
ASAE AnnualInternational Meeting, Paper No 973006, [34].Pearson, T., and N. Toyofuku, “Automated
St. Joseph, Michigan, USA,1997. Sorting of Pistachio Nuts with Closed Shells”, Applied
[23].Majumdar, S., Jayas, D.S., Bulley, N.R., Engineering in Agriculture, 16(1), pp.91-94, 2000.
“Classification of cereal grains using machine vision, [35].Qui, W., and Shearer, S.A., “Maturity Assessment
part 1: Morphological features”, In: 1997 ASAE Annual of Broccoli using the Discourse Fourier Transform”,
International Meeting Technical Papers, Paper No. Transaction of the ASAE, 35(6), pp. 2057-2062, 1992.
973101, ASAE, 2950 Niles Road, St. Joseph, MI 49085- [36].Raji, A.O., “Discrete Element Modeling of the
9659, USA, 1997. Deformation of Bulk Agricultural Particulates”,
[24].Molto, E., Pla, F., and F. Juste, “Vision Systems for Unpublished Ph.D Thesis, University of Newcastle-
the Location of Citrus Fruit in a Tree Canopy”, Journal upon-Tyne, 1999.
of Agricultural Engineering Research, 52, pp. 101-110, [37].Raji, A.O., Fagboun, A.A., and M.K. Dania, “An
1992. Approach to Detecting Defects in Food Products”,

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

Proceedings of the First International Conference of the [51].Timmermans, A.J.M., “Computer Vision System
Nigerian Institution of Agricultural Engineers, pp.36-39, for Online Sorting of Pot Plants Based on Learning
2000. Techniques”, Acta Horticulturae, 421, pp. 91-98,
[38].Reed, J.N., Crook, S., and W. He., “Harvesting 1998.
Mushrooms by Robot,. In Science and cultivation of [52].Tollner, E.W., Shahin, M.A., Maw, B.W.,
edible fungi, pp . 385-391, Rotterdam: Elliott, Balkema, Gilaitis, R.D., Summer, D.R., “Classification of onions
1995. based on internal defects using imaging processing
[39].Ruan, R., Ning, S., Ning, A., Jones, R., Chen, P.L., and neural network techniques”, In: 1999 ASEA
“Estimation of scabby wheat incident rate using machine International Meeting, Toronto, Onteroi, Paper no.
vision and neural network”, In: 1997 ASAE Annual 993165, ASAF, 2950 Niles Road, St. Joseph, MI
International Meeting Technical Papers, Paper No. 49085-9659, USA, 1999.
973042, ASAE, 2950 Niles Road, St. Joseph, MI 49085- [53].Tsheko, R., “Image Analysis by Machine
9659, USA, 1997. Vision”, Ph.D Thesis, University of Newcastleupon-
[40].Sapirstein, H.D.,”Quality Control in Commercial Tyne, 1998.
Baking: Machine Processing”, Automation IV [54].Wan, Y.N., Lin, C.M., Chiou, J.F., “Adaptive
Proceedings of the ASAE FPAC Conference, St. Joseph, classification method for an automatic grain quality
Michigan, USA, 1995. inspection system using machine vision and neural
[41].Scott, A., “Automated Continuous Online network”, In: 2000 ASAE Annual International
Inspection, Detection and Rejection”, Food Technology Meeting, Paper No. 003094 ASAE, 2950 Niles Road,
Europe, 1(4), pp. 86-88, 1994. St. Joseph, MI 49085-9659, USA, 2000.
[42].Shearer, S.A., Payne, F.A., “Colour and defect [55].Wang, H.-H., Sun, D.-W., “Evaluation of the
sorting of bell peppers using machine vision”, functional properties of cheddar cheese using a
Transactions of the ASAE, 33 (6), pp. 2045-2050, 1990. computer vision method”, Journal of Food
[43].Slamet Riyadi, Ashrani A. Abd, Rahni Mohd, Engineering, 49(1), pp. 47-51, 2001.
Marzuki Mustafa, and Aini Hussain, Member, “Shape [56].Wang, H.-H., Sun, D.-W., “Melting
Characteristics Analysis for Papaya Size Classification”, characteristics of cheese: analysis of effects of cooking
The 5th Student Conference on Research and conditions using computer vision techniques”, Journal
Development-SCOReD 2007. of Food Engineering, 52(3), pp. 279-284, 2002a.
[44].Sonka, M, Hlavac, V, Boyle, R., “Image [57].Wang, H.-H., Sun, D.-W., “Correlation between
Processing, Analysis, and Machine Vision”, PWS cheese meltability determined with a computer vision
publishing, California, USA, 1999. method and with Arnott and Schreiber”, Journal of
[45].Steinmertz, V., Rogr, J.M., Molto, E., and J. Blasco, Food Science, 67(2), pp. 745-749, 2002b.
“Online Fusion of Colour Camera and [58].Wooten, J.R., White, J.G., Thomasson, J.A.,
Spectrophotometer for Sugar Content Prediction of Thompson, P.G., “Yield and quality monitor for sweet
Apples”, Journal of Agricultural Engineering Research, potato with machine vision”, In: 2000 ASAE Annual
73, pp. 207-216, 1999. International Meeting, Paper No. 001123, ASAE,
[46].Steenhoek, L., Precetti, C., “Vision sizing of seed 2950 Niles Road, St. Joseph, MI 49085-9659, USA,
corn”, In: 2000 ASAE Annual International Meeting, 2000.
Paper No. 003095 ASAE, 2950 Niles Road, St. Joseph, [59].Xie, W., Paulsen, M.R., “Machine vision
MI 49085-9659, USA, 2000. detection of tetrazolium in corn”, In: 1997 ASAE
[47].Sun, D.W., “Computer Vision: An Objective, Annual International Meeting Technical Papers, Paper
Rapid and Non-Contact Quality Evaluation Tool for the No. 973044, ASAE, 2950 Niles Road, St. Joseph, MI
Food Industry”, Journal of Food Engineering, 61, pp.1- 49085-9659, USA, 1997.
2, 2003. [60].Yam, K.L., and E.P. Spyridon , “A Simple Digital
[48].Sun, D.W., and T. Brosnan, “Improving Quality Imaging Method for Measuring and Analyzing Colour
Inspection of Food Products by Computer Vision: A of Food Surfaces”, Journal of Food Engineering, 61,
Review”, Journal of Food Engineering, 61, pp. 3-16, pp. 137-142, 2003.
2003. [61].Zayas, I., Pomeranz, Y., Lai, F.S.,
[49].Tao, Y., and Z. Wen, “An Adaptive Spherical “Discrimination of wheat and nonwheat components
Image Transform for Characterizing Corn Growth and in grain samples by image analysis”, Cereal
Development”, Transactions of the ASAE, 34(5), pp. Chemistry, 66(3), pp. 233-237, 1989.
2245-2249, 1999. [62].Zayas, I.Y., Martin, C.R., Steele, J.L., Katsevich,
[50].Tao, Y., Morrow, C.T., Heinemann, P.H., Sommer, A., “Wheat classification using image analysis and
H.J., ”Fourier based separation techniques for shape crush force parameters”, Transactions of the ASAE 39
grading of potatoes using machine vision”, Transactions (6), pp. 2199-2204, 1996.
of the ASAE, 38(3), pp. 949-957, 1995b.

©2010 International Journal of Computer Applications (0975 – 8887)
Volume 1 – No. 4

[63].Zhang, N., Chaisattapagon, C., “Effective criteria

for weed identification in wheat fields using machine
vision”, Transactions of the ASAE, 38(3), 965-974,