See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/236189439

Machine vision system: A tool for quality inspection of food and agricultural
products

Article in Journal of Food Science and Technology -Mysore- · April 2012

DOI: 10.1007/s13197-011-0321-4 · Source: PubMed

CITATIONS READS

244 7,834

4 authors:

Krishna Patel Abhijit Kar

Post Graduate College, Ghazipur Indian Agricultural Research Institute
39 PUBLICATIONS 564 CITATIONS 48 PUBLICATIONS 1,725 CITATIONS

SEE PROFILE SEE PROFILE

Shyam Narayan Jha Mohammad Ali Khan

Indian Council of Agricultural Research Aligarh Muslim University
117 PUBLICATIONS 4,106 CITATIONS 55 PUBLICATIONS 1,837 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Mohammad Ali Khan on 11 June 2016.

The user has requested enhancement of the downloaded file.

J Food Sci Technol (March–April 2012) 49(2):123–141
DOI 10.1007/s13197-011-0321-4

REVIEW

Machine vision system: a tool for quality inspection of food

and agricultural products
Krishna Kumar Patel & A. Kar & S. N. Jha & M. A. Khan

Revised: 28 January 2011 / Accepted: 3 February 2011 / Published online: 9 April 2011
# Association of Food Scientists & Technologists (India) 2011

Abstract Quality inspection of food and agricultural Keywords Machine vision . Image processing . Image
produce are difficult and labor intensive. Simultaneously, analysis . Quality inspection . Food and agricultural
with increased expectations for food products of high products
quality and safety standards, the need for accurate, fast
and objective quality determination of these characteristics
in food products continues to grow. However, these Introduction
operations generally in India are manual which is costly
as well as unreliable because human decision in identifying Agriculture is Indian economy’s mainstay and it comprises
quality factors such as appearance, flavor, nutrient, texture, 18.5% of the GDP (gross domestic products). Due to the
etc., is inconsistent, subjective and slow. Machine vision advance of cultivation technology, the total cultivation
provides one alternative for an automated, non-destructive areas and yields for agricultural products have increased
and cost-effective technique to accomplish these require- rapidly in recent years, generating tremendous market
ments. This inspection approach based on image analysis values. In the last 2 years agriculture and its allied sectors
and processing has found a variety of different applications have registered note worthy growth rate of 4% as opposed
in the food industry. Considerable research has highlighted to the average annual growth rate of 2.5% during the 10th
its potential for the inspection and grading of fruits and 5 year plan (Anon 2009a). According to Anon (2009b)
vegetables, grain quality and characteristic examination and India’s fruits and vegetables production is registering year-
quality evaluation of other food products like bakery on-year growth and touching a new high. It has produced
products, pizza, cheese, and noodles etc. The objective of 197.54 millions tonnes (68.47 MT fruits and 129 MT
this paper is to provide in depth introduction of machine vegetables) fruits and vegetables (Anon 2009b), 27.72
vision system, its components and recent work reported on millions tonnes oilseed, 7.64 millions tonne fish and
food and agricultural produce. 556,280 millions numbers of eggs (Anon 2009c) in 2008–
09. However this is indicating vast potential for India to
emerge as a major exporter of agricultural produce, its share
K. K. Patel (*) : A. Kar
in global market is very low due to very high post harvest
Division of Post Harvest Technology, losses in handling and processing, mismanagement of
Indian Agricultural Research Institute, trades and procurements, lack of knowledge of preservation
New Delhi 110012, India and quick quality evaluation techniques.
e-mail: [email protected]
Futhermore, the ever-increasing population and the
S. N. Jha increased expectation of food products of high quality and
CIPHET, safety standards, there is a need for the growth of accurate,
Ludhiana, Punjab, India fast and objective quality determination of food and
agricultural products. However, ensuring product quality
M. A. Khan
AMU, is one of the most important and challenging tasks of the
Aligarh 202002, Uttar Pradesh, India industries before export of food and agricultural produce.
124 J Food Sci Technol (March–April 2012) 49(2):123–141

The Computer vision is a rapid, economic, consistent and experienced continued growth both in theory and applica-
objective inspection technique, which has expanded into tion. Sonka et al. (1999) reported that more than 1,000
many diverse industries. Its speed and accuracy satisfy ever papers are published each year in the expanding fields of
increasing production and quality requirements, hence computer vision and image processing.
aiding in the development of totally automated processes. Machine vision is an engineering technology that combines
This non-destructive method of inspection has found mechanics, optical instrumentation, electromagnetic sensing,
applications in the agricultural and food industry. digital video and image processing technology. As an
At the current stage, the quality has been assessed integrated mechanical-optical-electronic-software system,
traditionally by hand, inspecting the products individually machine vision has been widely used for examining,
or sampling large batches which is very time consuming monitoring, and controlling a very broad range of applica-
and inconsistent. The inspection is performed by personnel tions. It is the construction of explicit and meaningful
trained to detect defects, colors, sizes or strange features, descriptions of physical objects from images (Ballard and
and classify the product in its appropriate category. The Brown 1982) and it encloses the capturing, processing and
enormous variability that can present this type of products analysis of two-dimensional image (Timmermans 1998).
in terms of colours, textures or different types of defects, However, in another study by Sonka et al. (1999) noted that
hinders their classification, being machine vision, a valid it aims to duplicate the effect of human vision by
alternative to automate this task. Automation means every electronically perceiving and understanding an image, and
action that is needed to control a process at optimum provides suitably rapid, economic, consistent and objective
efficiency as controlled by a system that operates using assessment (Sun 2000). So, It can be say that the machine
instructions that have been programmed into it or response vision is the use of devices for optical, non-contact sensing
to some activities. Since automated systems are faster and to automatically receive and interpret the image of a real
more precise, the automatic qualitative inspections on food scene in order to obtain information and/or control machines
and agricultural products have been attracted much interest or process image (Zuech et al. 2000).
and reflected the progress of machine vision applications. Nevertheless we can say that the computer vision
Now, applications of these techniques have been widely technology not only provides a high level of flexibility
used for shape classification, defects detection, and quality and repeatability at a relatively low cost, but also, and more
grading and variety classification etc. importantly, it permits fairly high plant throughput without
For example, Lefebvre et al. (1994) have reported that compromising accuracy. Applications of these techniques
the modern commercial systems can classify the fruits have now expanded to various areas such as medical
based on external physical parameters and external defects diagnostic, automatic manufacturing and surveillance,
by using machine vision (non-invasive) system and remote sensing, technical diagnostics, autonomous vehicle,
Sapirstein (1995) has reported that the digital image robot guidance and in the agricultural and food industry,
analysis techniques have much capabilities to generate including the inspection of quality and grading of fruit and
precise descriptive data based on pictorial information vegetable. Some important applications of machine vision
which can contribute increased and more widespread use have been given in the Table 1 Singh et al. (2004).
in quality analysis. Latter on Kato (1997) determined many The food industry continues to be among the fast
physical characteristics of fruits and vegetables non- growing segments of machine vision systems (Gunasekaran
destructively and according to Aleixos et al. (1999) new 1996). It has also been used successfully in the analysis of
hardware architectures are making possible the implemen- grain characteristics and in the evaluation of foods such as
tation of these kinds of application in real-time so far. potato chips, meats, cheese and pizza. Crowe and Delwiche
Currently non-destructive quality evaluation techniques (1996a, b) have developed a machine vision system for
have gained momentum (Iwamoto et al. 1995; Jha and sorting and grading of fruits based on color and surface
Matsuoka 2000) and machine vision system provides a defects. Tao et al. (1991) and Tao (1996b) have developed a
means to perform entire tasks automatically (Diaz et al. high capacity color vision sorter based on optical properties
2000). These techniques, particularly for fruits and vegeta- such as reflectance to determine colour, shape, size, textural
bles, are quick and easy to use (Jha and Matsuoka 2004). feature (Majumdar and Jayas 2000a, b, c, d), volume (Wang
and Nguang 2007) and surface area (Eifert et al. 2006;
Thakur et al. 2007) of fruits and vegetables including
Machine vision system apples, peaches, tomatoes and citrus. Computer vision
systems provide suitably rapid, economic, consistent and
Machine/computer vision is a relatively young discipline objective assessment; they have been used increasingly in
with its origin traced back to the 1960 s (Baxes 1994). the food and agricultural industry for inspection and
Following an explosion of interest during the 1970s, it has evaluation purposes (Sun 2000). They have proved to be
J Food Sci Technol (March–April 2012) 49(2):123–141 125

Table 1 Summary of machine/computer vision applications for different areas

Application areas Purpose

Industrial automation and Process control, quality control, geometrical measurement

image processing Barcode and package label reading, object sorting
Parts identification on assembly lines, defect and fault inspection
Inspection of printed circuit boards and integrated circuits
Medical image analysis Tumor detection, measurement of size and shape of internal organs, blood cell count
X-ray inspection
Robotics Obstacle avoidance by recognition and interpretation of objects in a scene Collision avoidance,
machining monitoring
Hazard determination
Radar imaging Target detection and identification, guidance of helicopters and aircrafts in landing, guidance of
remote piloted vehicles (RPV), guiding missiles and satellites from visual cues
Food industry Sorting of vegetables and fruits, location of defects e.g. location of dark contaminants and insects
in cereals
Document analysis Handwritten character recognition, layout recognition, graphics recognition

Singh et al. (2004)

successful computer vision system for the objective commercial level. This is indicated by the sales of ASME
measurement and assessment of several agricultural prod- (Application Specific Machine Vision) systems into the
ucts (Timmermans 1998). Over the past decade advances in North American food market, which reached 65 million
hardware and software for digital image processing have dollars in 1995 (Locht et al. 1997). Graves and Batchelor
motivated several studies on the development of these (2003) summarized more than 20 machine vision applica-
systems to evaluate the quality of diverse and processed tions that were classified by tasks in the natural product
foods (Locht et al. 1997; Gerrard et al. 1996). Computer industry, more than 15 in manufacturing industry, and seven
vision has long been recognized as a potential technique for other machine vision tasks applied to various situations
the guidance or control of agricultural and food processes such as security and surveillance, medicine and health
(Tillett 1990). Therefore, over the past 20 years, extensive screening, military, and traffic control and monitoring.
studies have been carried out, thus generating many Gunasekaran (1996) reported that the food industry is
publications. now ranked among the top ten industries using machine
The majority of these studies focused on the application vision technology. This paper reviews the latest develop-
of computer vision to product quality inspection and ment of computer vision technology with respect to quality
grading. Traditionally, quality inspection of agricultural inspection in the agricultural and food industry.
and food products has been performed by human graders.
However, in most cases these manual inspections are time-
consuming and labour-intensive. Moreover the accuracy of Principle components of computer vision
the tests cannot be guaranteed (Park et al. 1996). By
contrast it has been found that computer vision inspection A computer vision system generally consists of five basic
of food products was more consistent, efficient and cost components: illumination, a camera, an image capture
effective (Lu et al. 2000; Tao et al. 1995a). Also with the board (frame grabber or digitizer), computer hardware and
advantages of superior speed and accuracy, computer vision software as shown in Fig. 1 (Wang & Sun 2002a).
has attracted a significant amount of research aimed at However, the mechanical design for a specified
replacing human inspection. Recent research has highlight- machine vision system usually is uniquely structured to
ed the possible application of vision systems in other areas suit the inspection of a particular product. For instance,
of agriculture, including the analysis of animal behaviour the conveyor belts in a poultry bone detection machine
(Sergeant et al. 1998), applications in the implementation of vision systems are usually flat, with no texture, and
precision farming and machine guidance (Tillett and Hague made of USDA approved plastic materials. While
1999), forestry (Krutz et al. 2000) and plant feature prototype machines for chicken defect and disease
measurement and growth analysis (Warren 1997). inspection in the chicken plants use hooks to hold the
Besides the progress in research, there is increasing birds when they are dangling from moving chains and
evidence of computer vision systems being adopted at passing through light beams (Chao et al. 2000). On the
126 J Food Sci Technol (March–April 2012) 49(2):123–141

Fig 1 (a) Principle components

of machine vision system (Pun-
dit et al. 2005) (b) Functional
block diagram of basic machine
vision system (Singh et al.
2004)

other hand, most conveyor belts for a modern apple detecting objects that are translucent, such as hatching eggs
packing line are roller conveyors made of special dark (Das and Evans 1992) or used to measure the geometric
colored rubber (Tao 1996a, b). dimensions of obscured object, such as measuring the shoot
The lighting system, a critical part of a controlled height or root diameters of pine to estimate the pine
machine vision system, must be carefully designed. The seedlings (Wilhoit et al. 1994). The side light scatter can be
ultimate purpose of lighting design is to provide a used to determine the cellular granularity (Jain et al. 1991).
consistent scene eliminate the appearance of variations, Structured light can be used to form a 3-D shape of apple
and yield appropriate, application-specific lighting. Proper surfaces for stem/calyx detection (Yang 1993). Moreover,
selection of lighting sources (incandescent, fluorescent, controlled lighting design sometimes acts as an active
halogen, Xenon, LED), lighting arrangements (backlight- sensing means that can “create” new information. Laser
ing, front lighting, side lighting, structured lighting, ring stripes of structured lights combined with X-ray imaging on
lighting), and lighting geometry (point lighting, diffuse deboned chicken meat inspection is a successful example of
lighting, collimated lighting) is the “key to value” (Zuech generating “extra” information in machine vision systems.
2004). Primary factors that influence the selection is Because of a controlled lighting system design, intelligent
whether the object under inspection is: 1) flat or curved; image processing technology applied to a machine vision
2) absorbing, transmissive or reflective; and 3) the nature of system is normally simplified and can achieve high
the feature to be imaged in comparison with the back- accuracy. Those algorithms tend to maximize the utilization
ground. For instance, backlighting is usually used for of pre-obtained object properties such as the appearance,
J Food Sci Technol (March–April 2012) 49(2):123–141 127

geometry, surface issues, shape, size, color, and positions, required, and the processing capability of the processor
as well as the effect of lighting sources. For instance, the board itself (Gunasekaran and Ding 1994).
rule-based decision-making method, which is quite effec-
tive under controlled condition (Bartlett et al. 1988), is
rarely used in the computer vision pattern recognition field Image processing and analysis
due to uncontrolled possibilities. In short, a major key to a
successful machine vision application is to start with a good The image processing and image analysis are the core of
contrast, repeatable image that is not affected by ambient computer vision with numerous algorithms and methods
light or the surroundings. A functional block diagram of available to achieve the required classification and measure-
basic machine vision system has been given in Fig. 1(b). ments (Krutz et al. 2000). However, image processing/
Image capturing is an application program that enables analysis involves a series of steps, which can be broadly
users to upload pictures from digital cameras or scanners divided into three levels: low level processing (image
which are either connected directly to the computer or to acquisition and pre-processing), intermediate level process-
the network. A system consisting of a high-pixel ing (image segmentation and image representation and
resolution CCD (charge coupled device) chip and description) and high level processing (recognition and
associated hardware is the most common method for interpretation) (Gunasegaram and Ding 1994; Sun 2000), as
generating digital images. However, because digital indicated in Fig. 2. The machine vision system, basically,
images are inherently monochrome, or black and white, comprises two main processes: (1) Image processing and
other hardware and software are needed to generate color (2) Image analysis (Zuech et al. 2000). Image processing is
images. An electronic system that can classify fruits by not meant for extracting information from the image but for
mass, by colour, by diameter, by bruising, shape and size removal of faults like noise, blurring of image etc. To
should be composed of a camera, a lens, a light source, a enhance and improve the acquired image for further
filter, a PC and image processing software (Sarkar and analysis, various processing algorithms are used. Suppose
Wolfe 1985; Hahn and Sanchez 2000). Lino et al. (2008) the image obtained becomes blurred due to motion of the
used electronic systems composed of CCD camera and PC object. This blurring needs to be removed by image
for image capturing in quality evaluation of tomatoes and processing tools, before extracting any information about
lemons. Recently Kanali et al. (1998) investigated the this object (Niku 2005). The various image processing
feasibility of using a charge simulation method (CSM) techniques used for image enhancement are Dilation (to
algorithm to process primary image features for three increase brightness of each pixel surrounded by neighbours
dimensional shapes recognition. However, only 2- with a higher intensity), Erosion (to reduce the brightness
dimensional (2D) data is needed for grading, classifica- of pixels surrounded by neighbours with a lower intensity),
tion, and analysis of most agricultural images. 3- dimen- Threshold (to convert the grey scale image into binary
sional may be needed for information on structure or image), Close (to remove dark spots isolated in bright
added details. Sonka et al. (1999) have developed 3-D regions and smoothes boundaries) and Open (to remove
vision technique from a series of 2-D images to derive a bright spots isolated in dark regions and smoothes
geometric description. The imaging unit consists of a high boundaries), etc. (Saeed et al. 2005). Tools used to improve
performance 12-bit digital CCD camera and spectral range image include lookup tables (converts grayscale values in
290–1,100 nm with the peak quantum efficiency of 60% at the source image into other grayscale values in the trans-
500 nm were used for quality evaluation of pickling formed image), spatial filters (improve the image quality by
cucumbers using hyperspectral reflectance and transmit- removing noise and smoothing, sharpening, and trans-
tance imaging by Ariana and Lu (2008). forming the image), grayscale morphology (extract and
The process of converting pictorial images into alter the structure of particles in an image), and frequency-
numerical form is called digitisation. In this process, an domain processing (remove unwanted frequency informa-
image is divided into a two dimensional grid of small tion), etc. (Anon 2008). In spite of above, the basic machine
regions containing picture elements defined as pixels by vision and image processing algorithms can be divided in
using a vision processor board called a digitiser or frame five major groups as: (1) Segmentation and algorithm
grabber. There are numerous types of analogue to digital development (2) Edge-detection techniques (3) Digital
converters (ADC) but for real time analyses a special morphology (4) Texture and (5) Thinning and skeletoniza-
type is required, this is known as a ‘flash’ ADC. Such tion algorithms (Russ 1992)
‘flash’ devices require only nanoseconds to produce a
result with 50–200 mega samples processed per second (1) Segmentation and algorithm development
(Davies 1997). Selection of the frame grabber is based on Segmentation is typically used to locate objects and
the camera output, spatial and grey level resolutions boundaries (lines curves etc.) in images. More pre-
128 J Food Sci Technol (March–April 2012) 49(2):123–141

Fig 2 Different levels in the

image processing process (Sun
2000)

cisely it is the process of assigning a label to every Leemans et al. (1998) to segment defects on golden
pixel in an image. There are four main methods or delicious apples and they reported that the proposed
algorithms for the selection of the global threshold: algorithm is effective in detecting various defects such
manual selection, isodata algorithm, objective func- as bruises, russet, scab, fungi or wounds.
tion, and histogram clustering (Zheng and Sun 2008). Since the algorithm is a set of well-defined rules or
There are many other thresholding based segmentation procedures for solving a problem in a finite number of
techniques such as, the minimum error technique steps. The algorithm developed for the surface defect
(Kittler and Illingworth 1986), the moment preserving detection mainly includes modules of image prepro-
technique (Tsai 1985), the window extension method cessing, defect segmentation, stem-calyx recognition,
(Hwang et al. 1997) and the fuzzy thresholding and defect area calculation and grading. Forbes and
technique (Tobias and Seara 2002). And some impor- Tattersfield (1999) developed a machine vision algo-
tant segmentation techniques are given in Fig. 3. rithm using neural networks and the algorithm was
Although, a large number of segmentation techni- tested on the estimation of pear fruit volume from two-
ques have been developed to date, no universal dimensional digital images. Another imaging algo-
method can perform with the ideal efficiency and rithm was developed by Hahn and Sanchez (2000) to
accuracy the infinity diversity of imagery (Bhanu et al. measure the volume of non-circular shaped agricultur-
1995). Therefore, it is expected that several techniques al produce, such as carrots. Using these methods,
will need to be combined in order to improve the Wang and Nguang (2007) and Sabliov et al. (2002)
segmentation results and increase the adaptability of successfully estimated the surface area and volume of
the methods. For instance, Hatem and Tan (2003) eggs, lemons, limes, peaches and tamarillos. Li et al.
developed an algorithm with an accuracy of 83% for (2002) have an algorithm for defect detection of apple.
the segmentation of cartilage and bone in images of (2) Edge-detection techniques
vertebrate by using the thresholding-based method Edge detection is an essential tool for machine
twice. vision and image processing. In image processing, an
Furthermore, Li and Wang (1999) have developed a edge is the boundary between an object and its
method based on reference image of apple to background. They represent the frontier for single
accomplish defect segmentation for curved fruits. An objects. Therefore, if the edges of object’s image can
algorithm using colour information was developed by be identified with precision, all the objects can be

Fig 3 Typical segmentation

techniques: (a) thresholding (b)
edgebased segmentation and (c)
region-based segmentation (Sun
2000)
J Food Sci Technol (March–April 2012) 49(2):123–141 129

located and their properties such as area, perimeter, cordingly, there is no ideal method for measuring
shape, etc., can be calculated (Bolhouse 1997). It is textures. Nevertheless, a great number of methods
also the process of locating edge pixels and increasing have been developed, and these are categorized into
the contrast between the edges and the background (i. statistical, structural, transform-based, and model-
e. edge enhancement) in such a way that edges based methods (Zheng et al. 2006). These methods
become more visible. In addition, edge tracing is capture texture measurements in two different ways—
another terminology used by researcher, includes the by the variation of across pixels and their neighbour-
process of following the edges, usually collecting the ing pixels (Bharti et al. 2004).
edge pixels into a list (Parker 1997). (5) Thinning and skeletonization algorithms
Some of the well-known edge detectors that have An image skeleton is a powerful analog concept that
been widely used are the Sobel, Prewitt, Roberts, and may be employed for the analysis and description of
Kirsch detectors (Russ 1999). The first quantitative shapes in binary images. It plays a central role in the pre-
measurements of the performance of edge detectors, processing of image data (Gunasekaran 2008a, b). A
including the assessment of the optimal signal-to-noise comprehensive review on thinning methodologies has
ratio and the optimal locality, the maximum suppres- been presented by Lam et al. (1992). In general, a
sion of false response, were performed by Canny skeleton may be defined as a connected set of medial
(1986), who also proposed an edge detector taking lines along the limbs of a figure and skeletonization is a
into account all three of these measurements. The process to describe the global properties of an object
Canny edge detector was used in the food industry for and to reduce the original image into a more compact
boundary extraction of food products (Du and Sun representation. A basic method for skeletonization is
2004, 2006b). thinning (Parker 1994). Ni and Gunasekaran (1998)
(3) Digital morphology have applied a sequential thinning algorithm for
Digital morphology is a group of mathematical evaluating cheese shred morphology when they are
operations that can be applied to the set of pixels to touching and overlapping. Naccache and Shinghal
enhance or highlight specific aspects of the shape so that (1984) compared the results of 14 skeletonization
they can be counted or recognized (Parker 1997). In algorithm and Davies (1997) deduced the ideal shape
morphological processing, images are represented as after analysed the skeleton shape of object.
topographical surfaces on which the elevation of each Furthermore, image processing can be done inde-
point is assigned as the intensity value of the pendently by image acquisition software (Lichtenthaler
corresponding pixels (Vincent and Soille 1991). One et al. 1996). Mehl et al. (2004) have reported about,
such method was proposed by Du and Sun (2006a) to numerous, general and more specific image processing
segment pores in pork ham images. In other methods, softwares/tools in his literature. Bailey et al. (2004)
post processing is conducted to merge the over demonstrated an image processing approach which
segmented regions with similar image characteristics estimated the mass of agricultural products rapidly
together again. Such a method with a graphic algorithm and accurately. Koc (2007) determined the volume of
to determine the similarity of merging neighbouring watermelons and Rashidi et al. (2007) estimated
regions was developed by Navon et al. (2005). volume of kiwifruits using image processing. The
(4) Texture surface area and volume of asymmetric agricultural
Texture effectively describes the properties of products were measured by Sabliov et al. (2002) using
elements constituting the object surface, thus the an image processing algorithm. Image processing
texture measurements are believed to contain substan- algorithms are the basis for mmachine vision (Martin
tial information for the pattern recognition of objects and Tosunoglu) and the image algebra (Ritter and
(Amadasun and King 1989). The repetition of a Wilson 1996) forms a solid theoretical foundation to
pattern or patterns over a region is called texture. This implement computer vision and image processing
pattern may be repeated exactly, or as set or small algorithms. A quadratic discriminant model based on
variations. Texture has a conflictive random aspect: an algorithm was developed by Harrell (1991) for
the size, shape, color, and orientation of the elements consultation during on-line sorting.
of the pattern (textons) (Parker 1994). Although Image analysis uses digital data to display images
texture can be roughly defined as the combination of and to mathematically manipulate images to highlight
some innate image properties, including fineness, various characteristics. Areas of homogeneous color or
coarseness, smoothness, granulation, randomness, lin- edges between different color areas are examples of
eation, hummocky, etc., a strictly scientific definition useful image features. Often, filtering of the original
has still not been determined (Haralick 1979). Ac- data can help to eliminate image noise, such as shading
130 J Food Sci Technol (March–April 2012) 49(2):123–141

caused by uneven lighting. Then, statistical and necessary to take 2-D images vertically and horizontally to
mathematical techniques are applied to the data to obtain its roundness and thickness, respectively.
distinguish elements of the image (Ballard and Brown Recently Kanali et al. (1998) investigated the feasibility of
1982).Various operations and techniques are applied to using a charge simulation method (CSM) algorithm to
the processed image to extract the desired information. process primary image features for three-dimensional shape
Among these operations and techniques are: Object recognition. The required features were transferred to a retina
recognition; Feature extraction; Analysis of position, model identical to the prototype artificial retina and were
size, orientation, etc. these terms being self explanatory compressed using the CSM by computing output signals at
in nature. A first order statistics which are based on work cells located in the retina. An overall classification rate
statistical properties of the gray level histogram of an of 94% was obtained when the prototype artificial retina
image region has been given in Tables 2 and 3. discriminated between distinct shapes of oranges for the 100
data sets tested. Gunasekaran and Ding (1999) obtained 3-D
images of fat globules in cheddar cheese from 2-D images.
3-D technique This enabled the in situ 3-D evaluation of fat globule
characteristics so as the process parameters and fat levels
In general, only 2-dimensional (2D) data are needed for may be changed to achieve the required textural qualities.
grading, classification, and analysis of most agricultural
images. However, in many applications 3-dimensional
image analysis maybe needed as information on structure Applications
or added detail is required. A 3-D vision technique has been
developed to derive a geometric description from a series of Assessment of fruits
2-D images (Sonka et al. 1999). In practice this technique
might be useful for food inspection. For example, when Computer vision has been widely used for the inspection and
studying the shape features of a piece of bakery, it is grading of fruits. It offers the potential to automate manual

Table 2 Summary of machine vision applications for food and agricultural produce

Product type Product Application Accuracy Ref.

Fruits Apple Defects detection ——— Paulus and Schrevens (1999)

Classification 78% Steinmetz et al. (1999)
Grading 95% Yang (1994)
Oranges Classification 93% Ruiz et al. (1996)
Quality Evaluation 87% Kondo (1995)
Strawberries Sorting 94–98% Nagata et al. (1997)
98–100% Bato et al. (2000)
Oil Palm Sorting — Alfantni et al. (2008)
Papayas Classification 94% Riyadi et al. (2007)
Nuts Classification 95% Pearson and Toyofuku (2000)
Pomegranate Sorting (Aril) 90% Blasco et al. (2009)
Vegetables Potatoes Classification 84% Wooten et al. (2000)
Chilli Classification 87% & 96.3% Hahn and Sanchez (2000)
Lemon Classification and sorting —— Lino et al. (2008)
Cereals Wheat Classification 94% Zayas et al. (1996)
Weed identification ———— Zhang and Chaisattapagon (1995)
Disease infection 97% Ruan et al. (1997)
Corn Size 73–90% Paulus et al. (1997)
Whiteness ———— Xie and Paulsen (1997)
Whole and broken kernel 91–94% Ni et al. (1997a, b)
Grading 80–96% Steenhoek and Precetti (2000)
Rice Grading 91% Wan et al. (2000)
Wheat, barley, oats, rye Classification 98, 97, 100 and 91 Majumdar et al. (1997)
J Food Sci Technol (March–April 2012) 49(2):123–141 131

Table 3 First order statistics which are based on statistical properties of the gray level histogram of an image region

Texture Features Expression Measure of Texture

P
Average Gray f1 ¼ gHðgÞ A measure of average intensity
g ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r
Level P
Standard f2 ¼ g  f 1 Þ2 HðgÞ A measure of average contrast
g
Deviation
1
Smoothness f3 ¼ 1  Measures the relative smoothness of the intensity in a region f3 is 0 for a region
ð1þf 22 Þ
of constant intensity and approaches 1 for regions with large excursions in the
values of its intensity levels.
P
Skewness f4 ¼ ðg  f 1 Þ3 H ðg Þ Measures the skewness of an histogram. This measure is 0 for symmetric histograms,
g
positive by histograms skewed to the right and negative for histograms skewed to the left
P
Uniformity f5 ¼ H 2 ðg Þ Measures uniformity. This measure is maximum when all gray levels are equal (maximal
g
uniform) and decreases from there.
P
Entropy f6 ¼  HðgÞlog2 HðgÞ A measure of randomness
g

g-symbolizes the possible values of intensity and H(g) the percentage of pixels with intensity value g, respectively (Bountris et al. 2005)

grading practices and thus to standardize techniques and sorted the apple as accepted or rejected. The accuracy
eliminate tedious inspection tasks. Kanali et al. (1998) reached 82.5%, which is poor compared with European
reported that the automated inspection of produce using standards (Heinemann et al. 1995). Other studies involving
machine vision not only results in labour savings, but can ‘Golden Delicious’ apples were performed for the purpose of
also improve inspection objectivity. classification into yellow or green groups using the HSI (hue,
Computer vision has been used for such tasks as shape saturation, intensity) colour system method (Tao et al.
classification, defects detection, quality grading and variety 1995a). The results show that an accuracy of over 90%
classification of the apple. Paulus and Schrevens (1999) was achieved for the 120 samples tested.
developed an image processing algorithm based on Fourier Steinmetz et al. (1999) investigated sensor fusion for the
expansion to characterize objectively the apple shape so as to purpose of sugar content prediction in apples by combining
identify different phenotypes. Experimentation by Paulus et image analysis and near-infrared spectrophotometic sen-
al. (1997) also used Fourier analysis of apple peripheries as a sors. The repeatability of the classification technique was
quality inspection/classification technique. This methodology improved when the two sensors were combined giving a
gave insight into the way in which external product features value of 78% for the 72 test samples. An online system
affect the human perception of quality. The research found with the use of a robotic device (Molto et al. 1997) resulted
that as the classification involved more product properties in a running time of 3.5 s per fruit for the technique.
and became more complex, the error of human classification Ruiz et al. (1996) studied three image analysis methods
increased. Leemans et al. (1998) investigated the defect to solve the problem of long stems attached to mechanically
segmentation of ‘Golden Delicious’ apples using machine harvested oranges. The techniques include colour segmen-
vision. The proposed algorithm was found to be effective in tation based on linear discriminant analysis, contour
detecting various defects such as bruises, russet, scab, fungi curvature analysis and a thinning process which involves
or wounds. In similar studies Yang (1996) assessed the iterating until the stem becomes a skeleton. It was found
feasibility of using computer vision for the identification of that these techniques were able to determine the presence or
apple stems and calyxes which required automatic grading absence of a stem with certainty. A study by Kondo (1995)
and coring. Back propagation neural networks were used to investigated the quality evaluation, i.e. by the correlation of
classify each patch as stem/calyx or patch-like blemish. appearance with sweetness, of Iyokan oranges using image
Earlier studies proposed the use of a ‘flooding’ algorithm to processing and reported that the method could effectively
segment patch-like defects (russet patch, bruise, and also predict the sweetness of the oranges.
stalk or calyx area) (Yang 1994). It was found that this Nagata et al. (1997) investigated the use of computer
method of feature identification is applicable to other types vision to sort fresh strawberries, based on size and shape.
of produce with uniform skin colour. This technique was In their study they developed a sorting system with an
improved by Yang and Marchant (1995), who applied a accuracy of 94–98% into three grades based on shape and
‘snake’ algorithm to closely surround the defects. To five grades on size. However, another automatic straw-
discriminate russet in ‘Golden Delicious’ apples a global berry sorting system was developed by Bato et al. (2000),
approach was used and the mean hue on the apples was had average shape and size accuracies of 98 and 100%,
computed (Heinemann et al. 1995). A discriminant function respectively.
132 J Food Sci Technol (March–April 2012) 49(2):123–141

Fruit Shape is one of the most important quality element model to determine the firmness of pears was
parameters for evaluation by customer’s preference. Addi- developed by Dewulf et al. (1999). The application of
tionally, misshaped fruits are generally rejected according computer vision technology to detect pear bruising was
to sorting standards of fruit. This Rashidi et al. (2007) was studied by Zhang and Deng (1999). Results from the
carried out to determine quantitative classification of experiments confirmed that different bruised areas can be
algorithm for fruit shape in kiwifruit (Actinidia deliciosa) precisely detected with most relative errors controlled to
while Riyadi et al. (2007) classified papaya size according within 10%. However, all above discussed quality evaluat-
to combinations of the area, mean diameter and perimeter ing and other commercial sorting machines that are
features and they studied the uniqueness of the extracted currently available for commodities (cherries, nuts, rice
features. The proposed technique showed the ability to etc.) are not capable of handling and sorting pomegranate
perform papaya size classification with more than 94% arils for different reasons, thus making it necessary to build
accuracy. In another study by Alfantni et al. (2008) specific equipment. This work Blasco et al. (2009)
developed an automated grading system for oil palm describes the development of a computer vision-based
bunches using the RGB colour model to distinguish machine to inspect the raw material coming from the
between the three different categories of oil palm fruit extraction process and classifies it in four categories. The
bunches. In their study, result showed that the ripeness of machine is capable of detecting and removing unwanted
fruit bunch could be differentiated between different material and sorting the arils by colour.
categories of fruit bunches based on RGB intensity.
Pearson and Slaughter (1996) developed a machine vision Assessment of vegetables
system for the detection of early split lesion on the hull of
pistachio nuts and reported that the developed system Computer vision has been shown to be a viable approach
classified early split nuts with 100% success and normal to inspection and grading of vegetables (Shearer and
nuts with 99% accuracy. In other research a multi-structure Payne 1990a, b). Heinemann et al. (1994) assessed the
neural network (MSNN) classifier was proposed and applied quality features of the common white Agaricus bisporus
to classify four varieties (classes) of pistachio nuts by mushroom using image analysis in order to inspect and
Ghazanfari et al. (1996). An automated machine vision grade the mushrooms by an automated system. Their study
system was developed to identify and remove pistachio nuts showed that the disagreement between human inspectors
with closed shells from processing streams (Pearson and ranged from 14% to 36% while by the vision system it
Toyofuku, 2000). The system included a novel material ranged from 8% to 56%. Reed et al. (1995) found that
handling system to feed nuts to line scan cameras without computer vision could be combined with harvester
tumbling. The classification accuracy of this machine vision technology to select and pick mushrooms based on size.
system for separating open shell from closed shell nuts was Computer vision has also been applied to objective
approximately 95%, similar to mechanical devices. measurement of the developmental stage of mushrooms
Nielsen et al. (1998) developed a technique to correlate (Van Loon 1996). This study found that cap opening of
the attributes of size, colour, shape and abnormalities, mushrooms correlated the best with the stage of develop-
obtained from tomato images, with the inner quality of the ment except for tightly closed mushrooms. Other research
tomato samples. They applied fuzzy sets into their study. described the development of computer vision techniques
Recently, chaos theory was introduced into this area for the detection, selection, and tracking of mushrooms
(Morimoto et al. 2000). In this study tomato fruit shape prior to harvest (Williams and Heinemann 1998).
was quantitatively evaluated using an attractor, fractal From the spectral analyses on the colour of different
dimension and neural networks. The results showed that a mushroom diseases Vızhanyo and Tillett (1998) concluded
combination of these three elements offers more reliable that the colour of the developed, senescent mushroom
and more sophisticated classification. Computer vision has differs from any browning caused by diseases allowing
also been used in the assessment of tomato seedling quality earlier detection of infected specimens. Similar research
as a classification technique to ensure only good quality developed a method, involving a series of complex colour
seedlings were transplanted (Ling and Ruzhitsky 1996). operations, to distinguish the diseased regions of mush-
The classification process adopted an adaptive thresholding rooms from naturally senescing mushrooms (Vızhanyo and
technique, the Oust method. Felfoldi 2000). Intensity normalisation and image transfor-
As consumer awareness and sophistication increases the mation techniques were applied in order to enhance colour
importance of objective measurement of quality is ever differences in true-colour images of diseased mushrooms.
increasing. In a study by Miller and Delwiche (1989) the The method identified all of the diseased spots as ‘diseased’
maturity of market peaches was evaluated by colour and none of the healthy, senescent mushroom parts were
analysis and the combined image processing with a finite detected as ‘diseased’.
J Food Sci Technol (March–April 2012) 49(2):123–141 133

Potatoes have many possible shapes which need to be Morrow et al. (1990) presented the techniques of vision
graded for sale into uniform classes for different markets. inspection of mushrooms, apples and potatoes for size,
This created difficulties for shape separation. A Fourier shape and colour. The use of computer vision for the
analysis based shape separation method for grading of location of stem/root joint in carrot has also been assessed
potatoes using machine vision for automated inspection (Batchelor and Searcy 1989). Feature extraction and pattern
was developed by Tao et al. (1995b). A shape separator recognition techniques were developed by Howarth and
based on harmonics of the transform was defined. Its Searcy (1992) to characterize and classify carrots for
accuracy of separation was 89% for 120 potato samples, in forking, surface defects, curvature and brokenness. The
agreement with manual grading. Earlier, Lefebvre et al. rate of misclassification was reported to be below 15% for
(1993) studied the use of computer vision for locating the the 250 samples examined.
position of pulp extraction automatically for the purpose of
further analysis on the extracted sample. An image
acquisition system was also constructed for mounting on a For grain classification and quality evaluation
sweetpotato harvester for the purpose of yield and grade
monitoring (Wooten et al. 2000). It was found that culls Wheat
were differentiated from saleable sweet potatoes with
classification rates as high as 84%. Grain quality attributes are very important for all users and
Chilli is a variety grown extensively consumed by especially the milling and baking industries. Computer vision
almost all the population. It has a high processing demand has been used in grain quality inspection for many years. An
and proper sorting is required before filling or canning. A early study by Zayas et al. (1989) used machine vision to
sorter that, Hahn and Sanchez (2000) classifies chilli by identify different varieties of wheat and to discriminate
three different width sizes, by means of a photodiode wheat from non-wheat components. In later research Zayas
scanner, was built they reported that the accuracy on the et al. (1996) found that wheat classification methods could
necrosis detection and width classification was of 96.3 and be improved by combining morphometry (computer vision
87%, respectively. analysis) and hardness analysis. Hard and soft recognition
Color and size are the most important features for accurate rates of 94% were achieved for the 17 varieties examined.
classification and sorting of citrus (Khojastehnazhand et al. Twenty-three morphological features were used for the
2010) and studies of the reflectance properties of citrus fruit discriminant analysis of different cereal grains using machine
(Gaffney 1973) can be considered as the bases for subsequent vision (Majumdar et al. 1997). Classification accuracies of
work on image analysis for citrus inspection. These studies 98, 91, 97,100 and 91% were recorded for CWRS (Canada
determined the visible and near-infrared wavelengths at Western Red Spring) wheat, CWAD (Canada Western
which greater contrast between the peel and major defects Amber Durum) wheat, barley, oats and rye, respectively.
can be achieved (Molto and Blasco 2008). A simple The relationship between colour and texture features of
technique was proposed by Cerruto et al. (1996), which wheat samples to scab infection rate was studied using a
segment blemishes in oranges using histograms of the three neural network method (Ruan et al. 1997). It was found that
components of the pixel in HIS (Hue, Saturation, Intensity) the infection rates estimated by the system followed the
color space. To estimate the maturity of citrus, Ying et al. actual ones with a correlation coefficient of 0.97 with human
(2004) used a dynamic threshold in the blue component to panel assessment and maximum and mean absolute errors of
segment between fruit and background. They then used 5 and 2%, respectively. In this study machine vision-neural
neural networks to distinguish between mature and immature network based technique proved superior to the human
fruit. However, in another study, Khojastehnazhand et al. panel. Image analysis has also been used to classify dockage
(2010) developed an image processing technique for estimat- components for CWRS (Canada Western Red Spring) wheat
ing citrus fruits physical attributes including diameters, and other cereals (Nair et al. 1997). Morphology, colour and
volume, mass and surface area using image processing morphology/colour models were evaluated for classifying the
technique (Khojastehnazhand et al. 2009) and they reported dockage components. Mean accuracies of 89 and 96% for
that this image processing procedure can be readily applied to the morphology model and 71 and 75% for the colour model
other axi-symmetric agricultural products such as eggs, pearl, were achieved when tested on the test and training data sets,
pepper, carrot, limes and onions. The main advantage of these respectively.
techniques is their robustness when dealing with changes of
color. Corn
Some other earlier studies of computer vision associated
with vegetable grading and inspection include colour and In order to preserve corn quality it is important to obtain
defect sorting of bell peppers (Shearer and Payne 1990a, b). physical properties and assess mechanical damage so as to
134 J Food Sci Technol (March–April 2012) 49(2):123–141

design optimum handling and storage equipment. Measure- over a two-crop season period and developed an online
ments of kernel length, width and projected area indepen- classification system based on neural classifier.
dent of kernel orientation have been performed using
machine vision (Ni et al. 1997a). While, Steenhoek and Processed food products
Precetti (2000) performed a study to evaluate the concept of
two-dimensional image analysis for classification of maize Visual features such as colour and size indicate the quality
kernels according to size category and they reported that the of many prepared consumer foods. Sun (2000) investigated
classification accuracy of both machine vision and screen this in research on pizza in which pizza topping percentage
systems was above 96% for round-hole analysis. However, and distribution were extracted from pizza images. Then it
sizing accuracy for flatness was less than 80%. was found that the new region-based segmentation tech-
Ng et al. (1997) developed a machine vision algorithm nique could effectively group pixels of the same topping
for corn kernel mechanical and mould damage measure- together and the topping exposure percentage can be easily
ment, which demonstrated a standard deviation less than determined with accuracy 90%.
5% of the mean value. They found that this method was To avoid the missguideness of quality assessment by visual
more consistent than other methods available. The auto- based human perception, computer vision has been widely
matic inspection of corn kernels was also performed by Ni used in the assessment of confectionary products so far.
et al. (1997b) using machine vision and they showed Davidson et al. (2001) measured the physical features of
classification rates of 91 and 94% for whole and broken chocolate chip biscuits, including size, shape baked dough
kernels respectively for whole and broken kernel identifi- colour, and fraction of top surface area that was chocolate
cation during their on-line tests study. The whiteness of chip using image analysis. Four fuzzy models were
corn has been measured by an on-line computer vision developed to predict consumer ratings based on three of
approach by Liu and Paulsen (1997) and they were this the features. A prototype-automated system for visual
technique easy to perform with a speed of 3 kernels/sec. In inspection of muffins was developed by Abdullah et al.
other studies Xie and Paulsen (1997) used machine vision (2000) and they reported that it was able to correctly classify
to detect and quantify tetrazolium staining in corn kernels. 96% of pre graded and 79% of ungraded muffins with an
The tetrazolium-machine vision algorithm was used to accuracy of greater than 88%. Now, the machine vision
predict heat damage in corn due to drying air temperature system has also been used in the assessment of quality of
and initial moisture content. Conclusively it could be say crumb grain in bread and cake products (Sapirstein 1995).
that the machine vision system is playing a versatile role in
quality evaluation corn so for. Cheese

Rice and lentils The evaluation of the functional properties of cheese is

assessed to ensure the necessary quality is achieved,
As rice is one of the leading food crops of the world its especially for specialized applications such as consumer
quality evaluation is of importance to ensure it remains food toppings or ingredients. Wang and Sun (2001)
appealing to consumers. Liu et al. (1997) developed a developed a computer vision method to evaluate the
digital image analysis method for measuring the degree of melting and browning of cheese. This novel non-contact
milling of rice and they obtained the coefficient of method was employed to analyze the characteristics of
determination of R2 =0.98 for the 680 samples. Wan et al. cheddar and mozzarella cheeses during cooking and the
(2000) employed three online classification methods for results showed that the method provided an objective and
rice quality inspection: range selection, neural network and easy approach for analyzing cheese functional properties
hybrid algorithms. The highest recorded online classifica- (Wang and Sun 2002a, b). Ni and Gunasekaran (1998)
tion accuracy was around 91% at a rate of over 1,200 developed an image-processing algorithm to recognize
kernels/min. The range selection method achieved this individual cheese shred and automatically measure the
accuracy but required time-consuming and complicated shred length. It was found that the algorithm recognized
adjustment. In another study, milled rice from a laboratory shreds well, even when they were overlapping. It was also
mill and a commercial-scale mill was evaluated for head reported that the shred length measurement errors were as
rice yield and percentage whole kernels, using a shaker low as 0.2% with a high of 10% in the worst case.
table and a machine-vision system called the Grain Check
(Lloyd et al. 2000). However, a flat bed scanner was used Noodles
in machine vision system developed by Shahin and Symons
(2001) for colour grading of lentils. In their study, they The important of noodles in the Asian diet is very
scanned and analyzed different grades of large green lentils significant, as currently they account for 30–40% of most
J Food Sci Technol (March–April 2012) 49(2):123–141 135

countries’ wheat consumption (Miskelly 1993). The major- grading. Based on reflectance characteristics, they discrim-
ity of Asian noodle manufacturers noodle products to be inated between fat and lean in the Longisimus muscle and
derived from white seed-coated wheat as compared to red generated binary muscle images. In a more recent study
seed-coat material. Initial research had used image analysis Gerrard et al. (1996) examined the degrees of marbling and
to characterize the white wheat products (Hatcher and colour in 60 steaks and they predicted the lean colour (R2 =
Symons 200b). As noodles appearance and color are the 0.86) and marbling scores (R2 = 0.84). Image texture
consumer’s initial quality parameters, there is a preference analysis has also been used in the assessment of beef
for manufacturers to use high quality, low yield patent tenderness (Li et al. 1997a, b). Statistic regression and
flours and free from various contaminants of the flour (i.e. neural network were performed to compare the image
bran and germ), as they result in brighter noodles. The features and sensory scores for beef tenderness and it was
standard methodology to assess bran contamination has found that the texture features considerably contributed to
been ash content, the use of auto fluorescence by different the beef tenderness. In another study by Lu et al. (2000,
wheat seed tissue (Munck et al. 1979). Hatcher (2001) 1997), evaluation of pork quality has also been investigated
demonstrated that the image analysis could be effectively and recommended neural network modeling as an effective
used to quantify, measure, and discriminate varietal differ- tool for evaluating fresh pork colour.
ences in white seed coated wheat varieties in preparation of Furthermore, Gray-scale intensity, Fourier power spec-
yellow alkaline noodles prepared from high-quality patent trum, and fractal analyses were used as a basis for
flour. separating tumorous, bruised and skin torn chicken
carcasses from normal carcasses (Park et al. 1996). A
Potato chips and French fries neural network classifier used performed with 91% accura-
cy for the required separation based on spectral images
The images of commercial potato chips were evaluated for scanned at both 542 and 700 nm wavelengths. In a further
various colour (Brosnan and Sun 2004) and textural study Park and Chen (2001) found that a linear discriminant
features to characterize and classify the appearance (Pedre- model was able to identify unwholesome chicken carcasses
schi et al. 2004) and to model the quality preferences of a with classification accuracy of 95.6% while a quadratic
group of consumers. The most frequently used color model model (97% accuracy) was better to identify wholesome
is the RGB model, in which each sensor captures the carcasses. Daley et al. (1994) analyzed chicken carcasses
intensity of the light in the red (R), green (G) and blue (B) for systemic defects using global colour histograms based
spectra respectively, for food quality evaluation (Du and on a neural network classifier.
Sun 2004). However, the features derived from the image However, the correct assessment of meat quality, to fulfill
texture contained better information than colour features to the consumer’s needs, is crucial element within the meat
discriminate both the quality categories of chips and industry. Although there are several factors that affect the
consumers preferences. perception of taste, tenderness is considered the most
Pedreschi et al. (2006) recently designed and imple- important characteristic. Cortez and Portelinha (2006) pre-
mented a computer vision system to measure representa- sented, a feature selection procedure, based on a Sensitivity
tively and precisely the color of highly heterogeneous food Analysis, is combined with a Support Vector Machine, in
materials, such as potato chips. Leon et al. (2006) order to predict lamb meat tenderness. This real-world
developed a computation color conversion procedure that problem is defined in terms of two difficult regression tasks,
allows the obtaining of digital images from the RGB by modeling objective (e.g. Warner-Bratzler Shear force) and
images by testing five models: linear, quadratic, gamma, subjective (e.g. human taste panel) measurements. In both
direct and neural network. After the evaluation of the cases, the proposed solution is competitive when compared
performance of the models, the neural network model was with other neural (e.g. Multilayer Perceptron) and Multiple
found to perform the best. In another study, a stepwise Regression approaches.
logistic regression model was developed by Mendoza et al.
(2007) and they reported that that it was able to explain
86.2% of the preferences variability when classified into Advantages and disadvantages
acceptable and non-acceptable chips.
Machine vision system is seen as an easy and quick way to
Meat and meat products acquire data that would be otherwise difficult to obtain
manually (Lefebvre et al. 1993). Since, the capabilities of
Visually discernible characteristics are routinely used in the digital image analysis technology to generate precise
quality assessment of meat. McDonald and Chen (1990) descriptive data on pictorial information have contributed
pioneered early work in the area of image based beef to its more widespread and increased use (Sapirstein 1995).
136 J Food Sci Technol (March–April 2012) 49(2):123–141

Quality control in combination with the increasing automa- Furthermore, Gunasekaran and Ding (1993) also had agreed
tion in all fields of production has led to the increased that machine vision technique’s popularity in the food
demand for automatic and objective evaluation of different industries is growing constantly and they also pointed out
products. Sistler (1991) confirm that computer vision meets that its development so for at a robust level and
these criteria and states that the technique provides a quick competitively priced sensing too (Yin and Panigrahi
and objective means for measuring visual features of 1997). Moreover, all above advantages, there are more
products. In agreement it found that a computer vision other advantages of machine vision system to different
system with an automatic handling mechanism could agriculture sectors, are summarized bellow in Table 2.
perform inspections objectively and reduce tedious human An ambiguity of computer vision is that its results are
involvement (Morrow et al. 1990). In other study, Gerrard influenced by the quality of the captured images. Often due
et al. (1996) also recognized that machine image technol- to the unstructured nature of typical agricultural settings
ogy provides a rapid, alternative means for measuring and biological variation of plants within them, object
quality consistently. identification in these applications is considerably more
Another benefit of machine vision systems is the difficult. Also if the research or operation in being
non-destructive and undisturbing manner in which conducted in dim or night conditions artificial lighting is
information could be attained (Zayas et al. 1996), needed. Some advantages and disadvantages of computer
making inspection unique with the potential to assist vision to different sectors of the agricultural and horticul-
humans involving visually intensive work (Tao et al. tural industries have been summarized in Table 4.
1995b). Tarbell and Reid (1991) noted that an attractive So, according to above findings, computer vision has
feature of a machine vision system is that it can be used to been widely used for the sorting and grading of fruits and
create a permanent record of any measurement at any vegetable. It offers potential to automate manual grading
point in time. Hence archived images can be recalled to practices and thus to standardize techniques and eliminate
look at attributes that were missed or previously not of tedious inspection tasks. Kanali et al. (1998) reported that
interest. the automated inspection of produce using machine vision
Human grader inspection and grading of produce is often not only results in labour savings, but can also improve
a labour intensive, tedious, repetitive and subjective task inspection objectivity.
(Park et al. 1996). In addition to its costs, this method is
variable and decisions are not always consistent between
inspectors or from day to day (Tao et al. 1995a; Heinemann Conclusion
et al. 1994). In contrast Lu et al. (2000) had found computer
vision techniques adoptable, consistent, efficient and cost The paper on machine vision system reviews the recent
effective for food products. Hence computer vision could developments in computer vision for the agricultural and
be used widely in agricultural and horticulture to automate food industry. Machine vision systems have been used
many labour intensive processes (Gunasekaran 2001). increasingly in industry for inspection and evaluation

Table 4 Benefits and drawbacks of machine vision

Reference

Advantages
Generation of precise descriptive data Sapirstein (1995)
Quick and objective Li et al. (1997a, b)
Reducing tedious human involvement Ni et al. (1997b)
Consistent, efficient and cost effective Lu et al. (2000)
Automating many labour intensive process Gunasekaran (2001)
Easy and quick, consistent Gerrard et al. (1996)
Non-destructive and undisturbing Tao et al. (1995a); Zayas et al. (1996)
Robust and competitively priced sensing technique Gunasekaran and Ding (1993)
Permanent record, allowing further analysis later Tarbell and Reid (1991)
Disadvantages
Object identification being considerably more difficult in unstructured scenes Shearer and Holmes (1990)
Artificial lighting needed for dim or dark conditions Stone and Kranzler (1992)

Brosnan and Sun (2004)

J Food Sci Technol (March–April 2012) 49(2):123–141 137

purposes as they can provide rapid, economic, hygienic, functions for strawberry (Akihime variety). J Jpn Soc Agric
Machinery 62(2):101–110
consistent and objective assessment. However, there are
Baxes GA (1994) Digital image processing principles and applica-
some difficulties still exist in the adaptation of machine tions. Wiley, New York
vision system, evident from the relatively slow commercial Bhanu B, Lee S, Ming J (1995) Adaptive image segmentation using a
uptake of machine vision technology in all sectors. In spite genetic algorithm. IEEE Trans Syst Man Cybern 25:1264–1274
Bharti MH, Liu JJ, MacGregor JF (2004) Image texture analysis:
of this, with the advent of machine vision technology, a
methods and comparisions. Chemometr Intell Lab Syst 72:57–71
revolution has come in the field of automation. Automating Blasco J, Cubero S, Gomez-Sanchis J, Mira P, Molto E (2009) Development
an operation in a manufacturing plant requires a high of a machine for the automatic sorting of pomegranate (Punica
degree of pre-installation systems engineering and post- granatum) arils based on computer vision. J Food Eng 90:27–34
Bolhouse V (1997) Fundamentals of machine vision, robotic
installation process integration. The use of machine vision industries Assn. Proceedings of coastal sediments '87. ASCE,
technology in manufacturing can be as simple as producing New York, pp 468–483
an inspection quality report or as complex as total process Bountris P, Farantatos E, Apostolou N (2005) Advanced image
automation. At the end, we can say that machine vision is a analysis tools development for the early stage bronchial cancer
detection. World Acad Sci Engi Technol 9:152
powerful tool of automation that includes both image
Brosnan T, Sun DW (2004) Improving quality inspection of food
processing and image analysis tools. products by computer vision––a review and grading of agricul-
tural and food products by computer vision systems –a review. J
Food Eng 61:3–16
References Canny J (1986) A computational approach to edge detection. IEEE
Trans Pattern Anal Mach Intelligency 8(6):679–698
Cerruto E, Failla S, Schillaci G (1996) Identification of blemishes on
Abdullah MZ, Abdul-Aziz S, Dos-Mohamed AM (2000) Quality oranges. International conference on agricultural engineering,
inspection of bakery products using color based machine vision AgEng’96, Madrid, Spain, Eur Agric Eng paper no. 96 F-017,
system. J Food Qual 23:39–50 Sept, 23–26
Aleixos N, Blasco J, Molto E (1999) Design of a vision system for Chao K, Park B, Chen YR, Hruschka WR, Wheaton FW (2000)
real-time inspection of oranges. Pattern Recognit Image Anal Design of a dual-camera system for poultry carcasses inspection.
1:387–394 Appl Eng Agric 16(5):581–587
Alfantni MSM, Shariff ARM, Shafri HZM, Saaed OMB, Eshanta OM Cortez P, Portelinha M (2006) Lamb meat quality assessment by
(2008) Oil palm fruit bunch grading system using red, green and support vector machines, neural processing letters, Springer, In
blue digital number. J Appl Sci 8(8):1444–1452 press, ISSN: 1370–4621
Amadasun M, King R (1989) Textural features corresponding to Crowe TG, Delwiche MJ (1996a) Real-time defect detection in fruit-
textural properties. IEEE Trans Syst Man Cybern 19:1264–1274 Part I: design concepts and development of prototype hardware.
Anon (2008) Image analysis and processing. National instruments Trans ASAE 39(6):2299–2308
(accessed on 12 Jan. 2011), http://zone.ni.com/devzone/cda/tut/p/ Crowe TG, Delwiche MJ (1996b) Real-time defect detection in fruit—
id/3470 Part II: an algorithm and performance of a prototype system.
Anon (2009a) Indian-Agriculture sector at the Cusp of a Revolution. Trans ASAE 39(6):2309–2317
(Accessed date 30 Dec2010). http://fairsnexhibition.wordpress.com/ Daley W, Carey R, Thomson C (1994) Real-time colour grading and
2009/09/19/indian-agriculture-sector-at-the-cusp-of-a-revolution/ defect detection of food products. Optics in agriculture, forestry
Anon (2009b) Indian horticulture data base 2009. National Horticul- and biological processing SPIE (International Society for Optical
ture Board (accessed date 26 Oct 2010) http://nhb.gov.in/ Engineering). Int Soc Opt Eng 2345:403–411
database-2009.pdf Das K, Evans MD (1992) Detecting fertility of hatching eggs using
Anon (2009c) All India area, production and yield of food grains machine vision. II. Neural network classifiers. Trans ASAE 35
along with coverage irrigation. Agric Stat Glance 2009 (accessed (6):2035–2041
date 26 Oct 2010) http://planningcommission.nic.in/sectors/ Davidson VJ, Ryks J, Chu T (2001) Fuzzy models to predict
agri_html/ASGlance2009.pdf consumer ratings for biscuits based on digital features. IEEE
Ariana DP, Lu R (2008) Quality evaluation of pickling cucumbers using Trans Fuzzy Syst 9(1):62–67
hyperspectral reflectance and transmittance imaging: Part I. Devel- Davies ER (1997) Machine vision, 2nd edn. Academic, New York
opment of a prototype. Sens Instrumen Food Qual 2:144–151 Dewulf W, Jancsok P, Nicolai B, De Roeck G, Briassoulis D (1999)
Bailey DG, Mercer KA, Plaw C, Ball R, Barraclough H (2004) High Determining the firmness of a pear using finite element model
speed weight estimation by image analysis. In: Mukhopadhyay analysis. J Agric Eng Res 74(3):217–224
SC, Browne RF, Gupta GS (eds.), Proceedings of the 2004 New Diaz R, Faus G, Blasco M, Blasco J, Molto E (2000) The application
Zealand National Conference on Non- destructive Testing. of a fast algorithm for the classification of olives by machine
Palmerston North, New Zealand, July 27–29, p89–96 vision. Food Res Int 33:305–309
Ballard DA, Brown CM (1982) Computer vision. Prentice-Hall, Du CJ, Sun DW (2004) Shape extraction and classification of pizza
Eaglewoood Cliffs base using computer vision. J Food Eng 64(4):489–496
Bartlett SL, Besl PJ, Cole CL, Jain R, Mukherjee D, Skifstad KD Du CJ, Sun DW (2006a) Automatic measurement of pores and
(1988) Automatic solder joint inspection. Trans IEEE Pattern porosity in pork ham and their correlations with processing time,
Anal Mach Intell 1(10):31–43 water content and texture. Meat Sci 72(2):294–302
Batchelor MM, Searcy SW (1989) Computer vision determination of Du C-J, Sun D-W (2006b) Estimating the surface area and volume of
stem-root joint on processing carrots. J Agric Eng Res 43:259–269 ellipsoidal ham using computer vision. J Food Eng 73(3):260–268
Bato PM, Nagata M, Cao QX, Hiyoshi K, Kitahara T (2000) Study on Eifert JD, Sanglay GC, Lee DJ, Sumner SS, Pierson MD (2006)
sorting system for strawberry using machine vision (part 2): Prediction of raw produce surface area from weight measure-
development of sorting system with direction and judgement ment. J Food Eng 74:552–556
138 J Food Sci Technol (March–April 2012) 49(2):123–141

Forbes KA, Tattersfield GM (1999) Estimating fruit volume from Jha SN, Matsuoka T (2004) Nondestructive determination of acid brix
digital images. Africon IEEE 1:107–112 ratio of tomato juice using near infrared spectroscopy. Int J Food
Gaffney JJ (1973) Reflectance properties of citrus fruit. Trans ASAE Sci Technol 39:425–430
16(2):310–314 Kanali C, Murase H, Honami N (1998) Three-dimensional shape
Gerrard DE, Gao X, Tan J (1996) Beef marbling and colour score recognition using a chargesimulation method to process image
determination by image processing. J Food Sci 61(1):145–148 features. J Agric Eng Res 70:195–208
Ghazanfari A, Irudayaraj J, Kusalik A (1996) Grading pistachio Kato K (1997) Electrical density sorting and estimation of soluble
nuts using a neural network approach. Trans ASAE 39 solids content of watermelon. J Agric Eng Res 67:161–170
(6):2319–2324 Khojastehnazhand M, Omid M, Tabatabaeefar A (2009) Determina-
Graves M, Batchelor BG (2003) Machine vision for the inspection of tion of orange volume and surface area using image processing
natural products. Springer-Verlag Pub, ISBN 9781852335250, technique. Int Agrophys 23:237–242
p1– 472 Khojastehnazhand M, Omid M, Tabatabaeefar A (2010) Development
Gunasekaran S (1996) Computer vision technology for food quality of a lemon sorting system based on color and size. Afr J Plant Sci
assurance. Trends Food Sci Technol 7(8):245–256 4(4):122–127
Gunasekaran S (2001) Non-destructive food evaluation techniques to Kittler J, Illingworth J (1986) Minimum error thresholding. Pattern
analyse properties and quality. food science and technology Recognit 19(1):41–47
series. Marcel Dekker, New York, p 105 Koc AB (2007) Determination of watermelon volume using ellipsoid
Gunasekaran S (2008a) Quality evaluation of cheese. Computer vision approximation and image processing. J Postharvest Biol Technol
technology for food quality evaluation-edited by Da-Wen Sun, 45:366–371
Director, Food Refrigeration and Computerised Food Technology Kondo N (1995) Quality evaluation of orange fruit using neural
Rresearch Group, National University of Ireland, Dublin, Ireland. networks. In: Food Processing Automation IV Proceedings of the
ISBN: 978-0-12-373642-0, p457 FPAC Conference. ASAE, 2950 Niles Road, St. Joseph, MI
Gunasekaran S (2008b) Quality evaluation of cheese. Coputer vision 49085–9659, USA
technology-edited by Da-Wen Sun p457 Krutz GW, Gibson HG, Cassens DL, Zhang M (2000) Colour vision
Gunasekaran S, Ding K (1993) Using computer vision for food quality in forest and wood engineering. Landwards 55(1):2–9
evaluation. Food Technol 6:151–154 Lam L, Lee SW, Suen CY (1992) Thinning methodologies-a
Gunasekaran S, Ding K (1994) Using computer vision for food quality comprehensive survey. IEEE Trans Pattern Anal Mach Intell 14
evaluation. Food Technol. 48:151–154 (9):869–885
Gunasekaran S, Ding K (1999) Three-dimensional characteristics of Leemans V, Magein H, Destain MF (1998) Defects segmentation on
fat globules in cheddar cheese. J Dairy Sci 82:1890–1896 ‘Golden Delicious’ apples by using colour machine vision.
Hahn F, Sanchez S (2000) Carrot volume evaluation using imaging Comput Electron Agric 20:117–130
algorithms. J Agric Eng Res 75:243–249 Lefebvre M, Gil DS, Brunet E, Natonek C, Baur P, Gugerli PT (1993)
Haralick RM (1979) Statistical and structural approaches to texture. Computer vision and agricultural robotics for disease control: the
Proceeding of the IEEE 67:786–804 potato operation. Comput Electron Agric 9:85–102
Harrell RC (1991) Processing of color images with bayesian Lefebvre M, Zimmerman T, Baur C, Gugerli P, Pun T (1994) Potato
discriminate analysis. Ist international seminar on use of on- operation: automatic detection of potato diseases. Proc SPIE
machine vision systems for the agricultural and bio- 2345:2–9
industries, CEMAGREF Montpellier, France, 3–6 September, Li Q, Wang M (1999) Study on high-speed apple surface defect
p 11–20 segment algorithm based on computer vision. Proceedings of
Hatcher DW (2001) In: Owens G (ed) Asian noodle processing. In International Conference on Agricultural Engineering (99-ICAE),
cereal processing technology. CRS Press, Wood head publishing Beijing, People’s Republic of China, 14–17 Dec, p27-31
Ltd, Cambridge, pp 131–157 Li J, Tan J, Martz FA (1997) Predicting beef tenderness from
Hatem I, Tan J (2003) Cartilage and bone segmentation in vertebra image texture features. In 1997 ASAE Annual International
images. Trans ASAE 46(5):1429–1434 Meeting, Paper No. 973124. St. Joseph, Michigan, USA:
Heinemann PH, Hughes R, Morrow CT, Sommer HJ, Beelman RB, ASAE
Wuest PJ (1994) Grading of mushrooms using a machine vision Li J, Tan J, Martz FA (1997) Predicting beef tenderness from image
system. Trans ASAE 37(5):1671–1677 texture features. In: 1997 ASAE Annual International Meeting
Heinemann PH, Varghese ZA, Morrow CT, Sommer HJ III, Technical Papers, Paper No. 973124, ASAE, 2950 Niles Road,
Crassweller RM (1995) Machine vision inspection of “Golden St. Joseph, MI 49085–9659, USA
Delicious” apples. Appl Eng Agric 11(6):901–906 Li B, Perabekam S, Liu G, Yin M, Song S, Larson A (2002)
Howarth MS, Searcy SW (1992) Inspection of fresh carrots by Experimental and bioinformatics comparison of gene expression
machine vision. In: Food Processing Automation II Proceedings between T cells from TIL of liver cancer and T cells from
of the 1992 Conference. ASAE, 2950 Niles Road, St. Joseph, MI UniGene. J Gastroenterol 37(4):275–282
49085–9659, USA Lichtenthaler HK, Lang M, Sowinska M, Heisel F, Miehe JA (1996)
Hwang H, Park B, Nguyen M, Chen Y-R (1997) Hybrid image Detection of vegetation stress via a new high resolution
processing for robust extraction of lean tissue on beef cut surface. fluorescence imaging system. J Plant Physiol 148:599–612
Comput Electron Agric 17(3):281–294 Ling PP, Ruzhitsky VN (1996) Machine vision techniques for
Iwamoto M, Kawano S, Yukihiro O (1995) An overview of research measuring the canopy of tomato seedling. J Agric Eng Res
and development of near infrared spectroscopy in Japan. J Near 65:85–95
Infrared Spectrosc 3:179–189 Lino ACL, Sanches J, Fabbro IMD (2008) Image processing
Jain NC, Paape MJ, Miller RH (1991) Use of flow cytometry for techniques for lemons and tomatoes classification. Bragantia
determination of differential leukocyte counts in bovine blood. campinas 67(3):785–789
Am J Vet Res 52(4):630–636 Liu J, Paulsen MR (1997) Corn whiteness measurement and
Jha SN, Matsuoka T (2000) Non-destructive techniques for quality classification using machine vision. In: 1997 ASAE Annual
evaluation of intact fruits and vegetables. Food Sci Technol Res 6 International Meeting Technical Papers, Paper No. 973045,
(4):248–251 ASAE, 2950 Niles Road, St. Joseph, MI 49085–9659, USA
J Food Sci Technol (March–April 2012) 49(2):123–141 139

Liu W, Tao Y, Siebenmorgen TJ, Chen H (1997) Digital image understanding cereal processing. In: Inglett GE, Muk L (eds) In
analysis method for rapid measurement of rice degree of milling. cereals for food and beverages. Academic, New york, pp 27–40
In: 1997 ASAE Annual International Meeting Technical Papers, Naccache NJ, Shinghal R (1984) An investigation into the skeletoni-
Paper No. 973028, ASAE, 2950 Niles Road, St. Joseph, MI zation approach of Hilditch. Pattern Recognit 17(3):279–284
49085–9659, USA Nagata M, Cao Q, Bato PM, Shrestha BP, Kinoshita O (1997) Basic
Lloyd BJ, Cnossen AG, Siebenmorgen TJ (2000) Evaluation of two study on strawberry sorting system in Japan. In: 1997 ASAE
methods for separating head rice from brokens for head rice yield Annual International Meeting Technical Papers, Paper No.
determination. In: 2000 ASAE Annual International Meeting, 973095, ASAE, 2950 Niles Road, St. Joseph, Michigan 49085–
Paper No. 006074 ASAE, 2950 Niles Road, St. Joseph, MI 9659, USA
49085–9659, USA Nair M, Jayas DS, Bulley NR (1997) Dockage identification in wheat
Locht P, Thomsen K, Mikkelsen P (1997) Full color image analysis as using machine vision. ASAE Annual International Meeting
a tool for quality control and process development in the food Technical Papers, Paper No. 973043, ASAE, 2950 Niles Road,
industry. Paper No. 973006, ASAE, 2950 Niles Road, St. Joseph, St. Joseph, Michigan 49085–9659, USA
MI 49085–9659, USA Navon E, Miller O, Averbuch A (2005) Image segmentation based on
Loon VPCC (1996) Het bepalen van het ontwikkelingsstadium bij adoptive local thresholds. Image Vis Comput 23(1):69–85
dechampignon met computer beeldanalyse. Champignoncultuur Ng HF, Wilcke WF, Morey RV, Lang JP (1997) Machine vision
40(9):347–353 evaluation of corn kernel mechanical and mould damage. In:
Lu J, Tan J, Gerrard DE (1997). Pork quality evaluation by image 1997 ASAE Annual International Meeting Technical Papers,
processing. In: 1997 ASAE Annual International Meeting Paper No. 973047, ASAE, 2950 Niles Road, St. Joseph,
Technical Papers, Paper No. 973125, ASAE, 2950 Niles Road, Michigan 49085–9659, USA
St. Joseph, MI 49085–9659, USA Ni H, Gunasekaran S (1998) A computer vision method for
Lu J, Tan J, Shatadal P, Gerrard DE (2000) Evaluation of pork color determining the length of cheese shreds. Artif Intell Rev 12(1–
by using computer vision. Meat Sci 56:57–60 3):27–37
Majumdar S, Jayas DS (2000a) Classification of cereal grains using Ni B, Paulsen MR, Liao K, Reid JF (1997a) Design of an automated
machine vision: I. Morphology models. Trans ASAE 43 corn kernel inspection system for machine vision. Trans ASAE
(6):1669–1675 40(2):491–497
Majumdar S, Jayas DS (2000b) Classification of cereal grains using Ni B, Paulsen MR, Reid JF (1997b). Size grading of corn kernels with
machine vision: II. Color models. Trans ASAE 43(6):1677–1680 machine vision. In: 1997 ASAE Annual International Meeting
Majumdar S, Jayas DS (2000c) Classification of cereal grains using Technical Papers, Paper No. 973046, ASAE, 2950 Niles Road,
machine vision: III. Texture models. Trans ASAE 43(6):1681–1687 St. Joseph, Michigan 49085–9659, USA
Majumdar S, Jayas DS (2000d) Classification of cereal grains using Nielsen HM, Paul W, Munack A, Tantau HJ (1998) Modelling image
machine vision: IV. Morphology, color, and texture models. processing parameters and consumer aspects for tomato quality
Trans ASAE 43(6):1689–1694 grading. In: Mathematical and Control Application in Agriculture
Majumdar S, Jayas DS, Bulley NR (1997) Classification of cereal and Horticulture. Proceedings of the Third IFAC Workshop,
grains using machine vision, part 1: morphological features. In: Pergamon-Elsevier, Oxford, UK
1997 ASAE Annual International Meeting Technical Papers, Niku SB (2005) Introduction to robotics—analysis, systems, applica-
Paper No. 973101, ASAE, 2950 Niles Road, St. Joseph, MI tions 248. Prentice Hall of India Pvt. Ltd., New Delhi
49085–9659, USA Park B, Chen YR (2001) Co-occurrence matrix texture features of
McDonald T, Chen YR (1990) Separating connected muscle tissues in multi spectral images on poultry carcasses. J Agric Eng Res 78
images of beef carcass ribeyes. Trans ASAE 33(6):2059–2065 (2):127–139
Mehl PM, Chen YR, Kim MS, Chan DE (2004) Development of Park B, Chen YR, Nguyen M, Hwang H (1996) Characterising
hyperspectral imaging technique for the detection of apple multispectral images of tumorous, bruised, skin-torn, and
surface defects and contaminations. J Food Eng 61:67–81 wholesome poultry carcasses. Trans ASAE 39(5):1933–1941
Mendoza F, Dejmek P, Aguilera JM (2007) Colour and image texture Parker JR (1994) Practical computer vision using C. Wiley, New York, p 40
analysis in classification of commercial potato chips. Food Res Parker JR (1997) Algorithms for image processing and computer
Int 40:1146–1154 vision. Wiley, p 155–160
Miller MK, Delwiche MJ (1989) A colour vision system for peach Paulus I, Schrevens E (1999) Shape characterisation of new apple
grading. Trans ASAE 32(4):1484–1490 cultivars by Fourier expansion of digital images. J Agric Eng Res
Miskelly DM (1993) Noodles-a new look at an old food. Food Aust 72(2):113–118
45:496–500 Paulus I, De Busschers R, Schrevens E (1997) Use of image analysis
Molto E, Blasco J (2008) Quality evaluation of citrus fruits. Computer to investigate human quality classification of apples. J Agric Eng
vision technology for food quality evaluation, p 247 Res 68(4):341–353
Molto E, Blasco J, Steinmetz V, Bourley A, Navarron F, Pertotto G Pearson TC, Slaughter DC (1996) Machine vision detection of early
(1997) A robotics solution for automatic handling, inspection and split pistachio nuts. Trans ASAE 39(3):1203–1207
packing of fruits and vegetables. Proceedings of the International Pearson T, Toyofuku N (2000) Automated sorting of pistachio nuts
Workshop on Robotics and Automated Machinery for Bio- with closed shells. Appl Eng Agric 16(1):91–94
productions. BioRobotics’97, Grandia, Valencia, Spain Pedreschi F, Mery D, Mendoza F, Anguiera JM (2004) Classification
Morimoto T, Takeuchi T, Miyata H, Hashimoto Y (2000) Pattern of potato chips using pattern recognition. J Food Sci 69:264–270
recognition of fruit shape based on the concept of chaos and Pedreschi F, Leon J, Mery D, Moyano P (2006) Development of a
neural networks. Comput Electron Agric 26(2):171–186 computer vision system to measure the color of potato chips.
Morrow CT, Heinemann PH, Sommer HJ, Tao Y, Varghese Z (1990) Food Res Int 39:1092–1098
Automate inspection of potatoes, apples, and mushrooms. In Rashidi M, Seyfi K, Gholami M (2007) Determination of kiwifruit
Proceedings of the International Advanced Robotics Programme, volume using image processing. J Agric Biol Sci 2:17–22
Avignon, France: 179–188 Reed JN, Crook S, He W (1995) Harvesting mushrooms by robot. In:
Munck L, Feil C, Gibbons GC (1979) Analysis of botanical Elliott TJ (ed) Science and cultivation of edible Fungi. Balkema,
components in cereals and cereals products: a new way of Rotterdam, pp 385–391
140 J Food Sci Technol (March–April 2012) 49(2):123–141

Ritter GX, Wilson JN (1996) Handbook of computer vision Tao Y, Deck SH, Morrow CT, Heinemann PH, Sommer HJ, Cole RH
algorithms in image algebra. CRC Press (1991) Rule-supervised automated machine vision inspection of
Riyadi S, Rahni AAA, Rahni M, Mustafa MM, Hussain A (2007) produce. St Joseph MI, ASAE Paper 91–7003, 19 pages
Shape characteristics analysis for papaya size classification, The Tao Y, Heinemann PH, Varghese Z, Morrow CT, Sommer HJ III
5th Student Conference on Research and Development (1995a) Machine vision for colour inspection of potatoes and
(SCOReD), 11–12 Dec apples. Trans ASAE 38(5):1555–1561
Ruan R, Ning S, Ning A, Jones R, Chen PL (1997) Estimation of Tao Y, Morrow CT, Heinemann PH, Sommer HJ (1995b) Fourier
scabby wheat incident rate using machine vision and neural based separation techniques for shape grading of potatoes using
network. In: 1997 ASAE Annual International Meeting Technical machine vision. Trans ASAE 38(3):949–957
Papers, Paper No. 973042, ASAE, 2950 Niles Road, St. Joseph, Tarbell KA, Reid JF (1991) A computer vision system for characterising
MI 49085–9659, USA corn growth and development. Trans ASAE 34(5):2245–2249
Ruiz LA, Molto E, Juste F, Pla F, Valiente R (1996) Location and Thakur M, Quan C, Tay CJ (2007) Surface profiling using fringe
characterization of the stem-calyx area on oranges by computer projection technique based on Lua effect. Opt Laser Technol
vision. J Agric Eng Res 64(3):165–172 39:453–459
Russ JC (1992) The image processing handbook, 2nd edn. CRC Press. Tillett RD (1990) Image analysis for agricultural processes. Silsoe
Ch. 5, p283–346 Research Institute. Division Note DN 1585.
Russ JC (1999) The image processing handbook, 3rd edn. CRC, Boca Tillett ND, Hague T (1999) Computer vision based hoe guidance for
Raton cereals-an initial trial. J Agric Eng Res 74(3):225–236
Sabliov CM, Boldor D, Keener KM, Farkas BE (2002) Image Timmermans AJM (1998) Computer vision system for on-line sorting
processing method to determine surface area and volume of of pot plants based on learning techniques. Acta Hort 421:91–98
axisymmetric agricultural products. Int J Food Prop 5:641–653 Tobias OJ, Seara R (2002) Image segmentation by histogram thresh-
Saeed AEM, Ahmed BM, Abdelkarim EI, Ibrahim KEE, Mohamed olding using fuzzy sets. IIEEE Trans Image Process 11
OSA (2005) New anti cestodal drugs: Efficacy, toxicity and (12):1457–1465
structure activity relationship of some derivatives of the open Tsai WH (1985) Moment-preserving thresholding: a new approach.
lactam form of praziquantel, J Sci Technol 6:109–119 Comput Vis Graph Image Process 29(3):377–393
Sapirstein HD (1995) Quality control in commercial baking: Machine Vincent L, Soille P (1991) Watersheds in digital spaces: an efficient
vision inspection of crumb grain in bread and cake products. In: algorithm based on immersion simulations. IEEE Trans Pattern
Food processing automation IV Proceedings of the FPAC Anal Mach Intell 13(6):583–598
Conference. ASAE, 2950 Niles Road, St. Joseph, Michigan Vızhanyo T, Felfoldi J (2000) Enhancing colour differences in images
49085–9659, USA of diseased mushrooms. Comput Electron Agric 26:187–198
Sarkar N, Wolfe RR (1985) Feature extraction techniques for sorting Vizhanyo T, Tillett RD (1998) Analysis of mushroom spectral
tomatoes by computer vision. Trans ASAE 28(3):970–979 characteristics. Proceedings of AgEng’98 International Confer-
Sergeant D, Boyle R, Forbes M (1998) Computer visual tracking of ence on Agricultural Engineering, Oslo, 24–27 August, Paper
poultry. Comput Electron Agric 21:1–18 No. 98-F08019981-7
Shahin MA, Symons SJ (2001) A machine vision system for grading Wan YN, Lin CM, Chiou JF (2000) Adaptive classification method
lentils. Canadian Grain Commission, Grain Research Laboratory, for an automatic grain quality inspection system using machine
1404-303 Main Street, Winnipeg, Manitoba, Canada R3C 3G8. vision and neural network. In: 2000 ASAE Annual International
Publication Number GRL # 809, 43: 7.7–714 Meeting, Paper No. 003094 ASAE, 2950 Niles Road, St. Joseph,
Shearer SA, Holmes RG (1990) Plant identification using colour co- MI 49085–9659, USA
occurrence matrices. Trans ASAE 33(6):2037–2044 Wang TY, Nguang SK (2007) Low cost sensor for volume and surface
Shearer SA, Payne FA (1990a) Color and defect sorting of bell computation of axi-symmetric agricultural products. J Food Eng
peppers using machine vision. Eng Res 63:229–236 79:870–877
Shearer SA, Payne FA (1990b) Color and defect sorting of bell Wang H-H, Sun D-W (2001) Evaluation of the functional properties of
peppers using machine vision. Trans ASAE 33(6):2045–2050 cheddar cheese using computer vision method. J Food Eng 49
Singh M, Kaur G, Kaur S (2004) Machine vision: an upcoming realm (1):47–51
Everyman’s science 7 Oct to 7 Nov 42(4): 186–189 Wang H-H, Sun D-W (2002a) Melting characteristics of cheese:
Sistler FE (1991) Machine vision techniques for grading and sorting analysis of effects of cooking conditions using computer vision
agricultural products. Postharvest News Inf 2(2):81–84 techniques. J Food Eng 52(3):279–284
Sonka M, Hlavac V, Boyle R (1999) Image processing, analysis, and Wang H-H, Sun D-W (2002b) Correlation between cheese meltability
machine vision. Chapter 14. PWS publishing, California determined with a computer vision method and with Arnott and
Steenhoek L, Precetti C (2000) Vision sizing of seed corn, In: 2000 Schreiber. J Food Sci 67(2):745–749
ASAE Annual International Meeting, Paper No. 003095 ASAE, Warren DE (1997) Image analysis research at NIAB (National
2950 Niles Road, St. Joseph, MI 49085–9659, USA Institute of Agricultural Botany): chrysanthemum leaf shape.
Steinmetz V, Roger JM, Molto E, Blasco J (1999) On-line fusion of Plant Var Seeds 10(1):59–61
colour camera and strawberry using machine vision (part 2): Wilhoit JH, Auburn AL, Kutz LJ, Fly DE, South DB (1994) PC-based
development of sorting system with direction and judgement multiple camera machine vision systems for pine seedling
functions for strawberry (Akihime variety). J Jpn Soc Agri measurements. Appl Eng Agric 10(6):841–847
Machinery 62(2):101–110 Williams AL, Heinemann PH (1998) Detection, tracking, and
Stone ML, Kranzler GA (1992) Image based ground velocity selection of mushrooms for harvest using image analysis. In
measurement. Trans ASAE 35(5):1729–1734 ASAE Annual International Meeting, Orlando, FL, USA, 12–16
Sun D-W (2000) Inspecting pizza topping percentage and distribution July 1998. ASAE Paper no. 983041
by a computer vision method. J Food Eng 44(4):245–249 Wooten JR, White JG, Thomasson JA, Thompson PG (2000)
Tao Y (1996a) Spherical transform of fruit images for on-line defect Yield and quality monitor for sweetpotato with machine
extraction of mass objects. Opt Eng 35(2):334–350 vision. In: 2000 ASAE Annual International Meeting, Paper
Tao Y (1996b) Methods and apparatus for sorting objects including No. 001123, ASAE, 2950 Niles Road, St. Joseph, MI 49085–
stable color transformation. US Patent 5:533–628 9659, USA.
J Food Sci Technol (March–April 2012) 49(2):123–141 141

Xie W, Paulsen MR (1997) Machine vision detection of tetrazolium in Zayas I, Pomeranz Y, Lai FS (1989) Discrimination of wheat and
corn. In: 1997 ASAE Annual International Meeting Technical nonwheat components in grain samples by image analysis. Cereal
Papers, Paper No. 973044, ASAE, 2950 Niles Road, St. Joseph, Chem 66(3):233–237
MI 49085–9659, USA Zayas IY, Martin CR, Steele JL, Katsevich A (1996) Wheat
Yang Q (1993) Finding stalks and calyx of apples using structured classification using image analysis and crush force parameters.
lighting. Comput Electron Agri 8:31–42 Trans ASAE 39(6):199–2204
Yang Q (1994) An approach to apple surface feature detection by Zhang N, Chaisattapagon C (1995) Effective criteria for weed
machine vision. Comput Electron Agric 11:249–264 identification in wheat fields using machine vision. Trans ASAE
Yang Q (1996) Apple stem and calyx identification with machine 38(3):965–974
vision. J Agric Eng Res 63(3):229–236 Zhang T, Deng JZ (1999) Application of computer vision to detection
Yang Q, Marchant JA (1995) Accurate blemish detection with active bruising on pears. Trans Chin Soc Agric Eng 15(1):205–209
contour models. Comput Electron Agric 14(1):77–89 Zheng C, Sun D-W (2008) Image segmentation techniques. Computer
Yin H, Panigrahi S (1997) Image processing techniques for vision technology for food quality evaluation, p 42
internal texture evaluation of French fries. In: 1997 ASAE Zheng C, Sun D-W, Zheng L (2006) Recent development of image
Annual International Meeting Technical Papers, Paper No. texture for evaluation of food qualities – a review. Trends Food
973127, ASAE, 2950 Niles Road, St. Joseph, MI 49085– Sci Tecnol 17:113–128
9659, USA Zuech N (2000) Understanding and Applying Machine Vision, 2nd
Ying YB, Xu ZG, Fu XP, Liu YD (2004) Nondestructive maturity eds, Vision Systems International Yardley, Pennsylvania, Marcel
detection of Citrus with computer vision. In Proceedings of Dekker, Inc., p1–336
the Society of Photo-optical Instrumentation Engineers (SPIE), Zuech N (2004) Machine vision and lighting. Machine vision on line,
5271, Monitoring Food Safety, Agriculture, and plant Health, http://www.machinevisiononline.org/public/articles/archivedetails.
p 97–107 cfm?id=1430, (accessed on 5 Nov. 2010)

View publication stats