International Journal of Trend in Research and Development, Volume 5(1), ISSN: 2394-9333

www.ijtrd.com
Fruit Classification System Using Computer Vision:
A Review
1
Raja Sekar L, 2Ambika N, 3Divya V and 4Kowsalya T,
1
Assistant Professor, 2,3,4UG Scholar,
1,2,3,4
Department of EIE, Bannari Amman Institute of Technology, Sathyamangalam, Erode, India

Abstract— Automation in food processing comes into play to Using computer vision, image processing is used to automatic
increase productivity, quality and profitable growth of the detection and classification of plant disease from colour,
country. Fruit grading is an important process for producers texture and shape [5].Food quality can be improved by quality
which affects the fruits quality evaluation and export market. inspection using computer vision [6].
Although the grading and sorting can be done by the human,
II. BASIC STEPS OF IMAGE PROCESSING
but it is slow, labour intensive and tedious. Hence, there is a
need of an intelligent fruit grading system. In recent years, Step1: Image Acquisition: This is the first step of image
researchers had developed numerous algorithms for fruit processing in which camera is used for capturing fruits images
sorting using computer vision. Colour, textural and in digital form and store in any digital media.
morphological features are the most commonly used to
Step2: Image Pre-processing: This section removes noise,
identify the diseases, maturity and class of the fruits.
smoothen the image also perform resizing of images. RGB
Techniques include clustering and color based segmentation,
artificial neural network and different classifiers based images are converted to the grey images also contrast of image
classification of diseases. The main focus of our work is is increased at certain level.
obtaining the analysis of different fruit diseases detection Step3: Image Segmentation: Segmentation is used for
techniques. Subsequently, these features are used to train soft partitioning an image into various parts.
computing technique network. In this paper, use of image
processing in agriculture has been reviewed so as to provide an Step4: Feature Extraction: This section is used for obtaining
insight to the use of vision based systems by highlighting their features like color, texture and shape which reduce resources
advantages and disadvantages.. to describe large set of data before classification of image.

Keywords— Fruit diseases, SVM classifier, segmentation, Step5: Classification: This section analyzes numerical property
Feature extraction, Automation, Fruit classification, computer of image features and organize its data into categories. It use
vision, colour, morphological, texture. neural network which performs training and classification of
fruits diseases.
I. INTRODUCTION
III. COMPUTER VISION
In recent years, use of image processing has been increasing
day by day in different areas such as industrial image Computer vision is used to gather the information from the
processing, medical imaging, real time imaging, texture images which are captured from the real time world. It is a
classification, object recognition, etc. Image processing and field that includes methods for image acquisition, processing,
computer vision in agriculture is another fast growing research analyzing and understanding the images in order to gather
field. It is an important analysing tool for pre-harvest to post- symbolic and numerical information. Basically its aim is to
harvest of crops. It has lots of applications in agriculture. The duplicate the effect of human vision by electronically
cultivation of crops can be improved by the technological perceiving, understanding and classification of
support. Fruits and vegetables losses are caused by disease. images[7].Computer vision is widely used in post-harvest
Diseases are seen on the leaves and fruits of plant, therefore industries for quality inspection and grading of fruits and
disease detection plays an important role in cultivation of vegetables.
crops. Pathogens, fungi, microorganism, bacteria and viruses A .Computer Vision Setup
are types of fruit diseases also unhealthy environment is
responsible for diseases. There are lots of techniques to A computer vision system constitutes of five components :
identify diseases in fruits in its early stages. The old method of illumination, a camera, an image capture board also known as
disease detection in fruit is naked eye observation and it‘s not a frame grabber or digitizer, computer hardware and software
effective. Using digital method, the disease detection can [8].
efficient, accurately, time consuming is less, saves time. B. Colour Features
Different image processing techniques and lots of algorithms
have been developed by researchers with the help of Images are captured in RGB colour models. It is most common
MATLAB for accurate fruit disease identification. In land Colour model in image processing and based on primary
identification, image processing is used for identification of colours red(R), green (G), blue (B). Basically, for colour
land that will be suitable for agriculture [1]. In plant nitrogen features, every image is separated into red, green and blue
identification, image processing can be used for estimation of planes, respectively, and through these planes, mean, median,
plant nitrogen identification and chlorophyll identification [2]. standard deviation are calculated [9]. NTSC or YIQ colour
In pest control, image processing is a good tool for space consists of three components luminance (Y), which
identification of pest infected areas because it favors to build represents gray scale information, hue (I) and saturation (Q),
up the pest population [3] [4]. which carry information of a signal.

IJTRD | Jan - Feb 2018

Available [email protected] 22
 International Journal of Trend in Research and Development, Volume 5(1), ISSN: 2394-9333
www.ijtrd.com
This colour model is used in television in United State. YCbCr methods are derived by analysing the spatial distribution of
colour space is used in digital video. Here, Y contains pixel gray level values such as Gray level co-occurrence
luminance information, Cb contains colour information matrix (GLCM),run-length and auto covariance function [13].
between blue component and reference value and Cr contains
E. Application of Computer Vision in Fruit Grading System
colour information between blue component and reference
value [10].HSV colour space is used by people to select Sorting of agriculture products automatically is more efficient
colours from a colour wheel or palette. In this colour space, H as compared to the current manual system which is very slow,
means hue that refers to tint; S means saturation that represents tedious, labour intensive and error prone. However, there is a
shade; V means value that refers to tone. HSI colour space need of an automatic sorting system which can identify the
means hue, saturation and intensity. HSI is the best tool for agricultural products based on their characteristics. Computer
developing image processing algorithm based on colour that vision has been widely used in fruit grading system. Grading is
are natural and perceived by humans.Image acquisition method based upon shape, size, colour, intensity and texture of the
is capturing of digital images, after that the paper deals image fruits.
pre processing techniques which includes feature extraction.
Savakar had graded five different types of fruit images(Apple,
Three feature vectors namely: morphology, texture and color
are used for feature extraction. Image components for Chickoo, Orange, Mango and Sweet Lemon). Total5000
boundaries are extracted using morphology. Various visual sample images had been captured, i.e. 1000 images of
eachtype of fruit. The algorithm had been developed by
patterns are described by texture feature. RGB color space is
extracting 18colours and 27 texture features. The colour
converted to HSI color space in color feature extraction also
features were calculated by separating RGB (Red, Green and
histogram of image is computed. ANN neural network and
back propagation algorithm is used for classification diseases Blue)components. The RGB image was then converted into
in network. Lastly fruit grading is determined by the HIS model and its components were separated. Mean, variance
and ranges were calculated for each RGB and HSI component
calculation of weight and spread of disease on the fruit[36].
individually. Texture features were calculated using Gray
Then color based segmentation techniques such as clustering,
LevelCo-occurrence Matrices (GLCM). The study revealed
YCbCr, RGB, La*b, HSV are used. In feature extraction three
that classification of chickoo, apple, sweet lemon, orange and
features are extracted. Morphology, color, texture features are
helpful for the classification purpose. In texture feature mango were 94%, 93%, 93%, 92%, 92% respectively [14].
extraction, Gabor filter is used and in morphology feature Deepa proposed a method to evaluate the extracted features
used for grading and classification of defected and non-
extraction, boundary of image is obtained. The eroded images
defected fruits. The image database included 200 mosambi
are subtracted from original image to extract shape vector from
fruits. Shape features, intensity features and texture features
healthy fruit image. Using minimum distance classifier
were calculated. This database was then classified based on
(MDC), diseased and non-diseased fruit are classified after
training and testing of images[44]. PNN and the result showed that shape, intensity and texture
features gave 100%, 92%, 96%identification rate respectively
C. Morphological Features [15]. Mustafa et al. presented a novel approach for fruit
grading system. In this paper five fruits (apples, bananas,
The most commonly used features for classification of fruits is
carrots, mangoes and oranges) were analysed. Shape and
morphological features, i.e. shape and size. Size features are
colour features were extracted from fruit sample images.
physical dimensional measurements that tell about the
Morphological features were used to distinguish between
appearance of an object. Area, perimeter, major and minor axis
almost similar shapes and sizes such as apple and orange or
lengths, and aspect ratio are commonly used as morphological
bananas and carrot. Colour features were used to remove
features. Morphological features are widely used in automatic
misclassification between apple and orange or banana and
sorting purpose in industries [9]. The area is a scalar quantity
carrot and increase the accuracy to 79-90% [9]. Khojastehnazh
that actual number of pixels in the region. Perimeter is a scalar
and et al. proposed an algorithm for sorting and classification
quantity and it is the distance around the boundary of the
of lemon fruits based upon the colour and size in Visual Basic
region. Major axis length is scalar quantity that is specifying
6. Volume of sample image had been calculated and RGB
the length of major axis (in pixels) of the major axis of the
images were converted into HSI images. HSI values were
ellipse that has the same normalized second central moment as
estimated and these data were stored in a database. During the
the region. Minor axis length is also scalar that is specifying
sorting stage, calculated volume and colour are compared with
the length of minor axis (in pixels) of the minor axis of the
the saved information in the database. The overall system gave
ellipse that has the same normalized second central moment as
94.04% accuracy [16]. Jack man et al. proposed a system for
the region. Shape features are measured by roundness (4π ×
food quality assessment based on computer vision. In this
Area/Perimeter2), aspect ratio(Major Axis/Minor Axis) and
paper, to extract surface texture feature of food, three
compactness (Perimeter2/Area)[12].
approaches had been suggested which were pixel co-
D. Texture Features occurrence, Run length and Difference Histogram methods.
Fourier Transform was also proposed as a method to extract
Texture represents the surface appearance and distribution of
texture feature in term of convolution of sinusoidal waves but
elements. It is an important feature in machine vision and
it had a limitation that it is used only where a small number of
predicts surface in form of contrast, roughness, orientation,
frequencies can reproduce the surface image.
entropy, etc. A wide variety of techniques have been proposed
for describing image texture. In model based approach, a set of Another classical approach Wavelet Transform was also
parameters is derived from variation of pixel elements that is proposed for texture analysis. An artificial intelligence was
used to define an image model such as Gaussian Markov provided to computer to analysis the texture which was
Random Field (GMRF), fractional Brownian motion (fBm). In efficient and robust [17]. Liminget al. presented a system for
GMRF, conditional probability of a certain pixel depends upon automatic grading of strawberry. In this RGB image was
the value of the neighbouring pixels, while fBm exploits the changed into L*a*b* colour model. The size was calculated
self-similarity of texture at varying scales. Statistical based using the major axis length and colour features were extracted

IJTRD | Jan - Feb 2018

Available [email protected] 23
 International Journal of Trend in Research and Development, Volume 5(1), ISSN: 2394-9333
www.ijtrd.com
from the dominant colour model on a* channel. K means fuzzy (IF–THEN) rules. A fuzzy rule is the basic unit to gain
clustering method was used for classification purpose and it knowledge in fuzzy systems.
gave 90% accuracy for shape classification and 88.8%
Like a conventional rule in artificial intelligence, fuzzy system
accuracy for colour grading [18].Alavi proposed a system for
has two components: an ‗if‘ part and a ‗then‘ part which are
grading of Mozafati dates. Total100 fruits were taken and three
also known as antecedent and consequent, respectively [19].
quality parameters were calculated such as quantity of juice,
Fuzzy is easy to train and requires less number of samples.
size and freshness. These fruits were graded using both fuzzy
However, it gives a lower accuracy rate as compared to any
interference system and human experts for comparison and
other classifier[30].Kavdir et al. proposed a method of apple
graded using fuzzy showed 86% conformity result as
grading in colour,size and defects of apples are extracted.
compared to human experts [19].Suresha et al presented an
These features were gathered and evaluated using fuzzy
automatic grading of apples with the help of support vector
system and this gave 89% accuracy in classification [24]. Date
machines (SVM). In this, apple images were captured into
fruits were graded using Fuzzy by extracting some features
RGB colour model and threshold based segmentation was used
like quality of juice, size and freshness and it gave 86%
to extract the region of interest from the background. The RGB
accuracy [19]. Rokunuzzman et al. presented an algorithm to
colour model was then converted into HSV model and average
classify tomatoes and it gave 84%accuracy rate [29].
red and green colour components were determined for
classification. This classifier gave 100% accuracy in grading B. Artificial Neural Network (ANN)
[20]. Leemans et al. proposed apples grading method and two
ANN is massively parallel distributed information processing
types of apple were used-Golden Delicious and Jona gold.
system that is made up of artificial neurons has certain
Features were extracted from shape, colour, texture and stem
position. This method for apple external quality grading gave performance characteristics resembling biological neurons of
72% accuracy for Golden Delicious and 78% Jona gold apples. the human brain [31]. A neural network is characterized by its
architecture that represents in which pattern it is connected
The grading of healthy fruits was better and an error rate
between nodes and which method is used to determine the
decreases to 5 and 10%, respectively [21].Nozari et al.
connection weights and the activation function. Feed forward
presented an algorithm for grading of Mozafatidates. Total 100
hierarchical architecture is the most commonly used ANN.
fruits were taken and classified based on length, width and
thickness. These fruits were graded using both ANFIS and A typical three-layered feed-forward neural network is made
human experts for comparison and ANFIS showed 93.5% of a multiple elements called nodes, and connection pathways
conformity result as compared to human experts [22]. Razak that link them. When an input signal is received at a neuron,
etal. presented automatic grading of mango using fuzzy then neuron processes it, and sends an output signal to other
analysis. In this size, colour and skin features were extracted. interconnected neurons [31].ANN is suitable to analyse
Size of mango was determined by calculating area of sample complex problems. However, it has some demerits such as it
image. Then RGB component was extracted from image and needs large training set and it is a time consuming process.
mean of three colour components was detected. For shape ANN gave high accuracy for training set but inadequate
analysis, edge detection algorithm had been implemented. interpolation for testing data [32].Mustafa et al. proposed a
Fuzzy inference rules were applied for mango grading into method to determine the size and ripeness of banana. Shape
different classes and it gave 80% overall accuracy [23].Kavdir and colour features were extracted. Then ANN was used for
et al. proposed a method of apple grading in which some classification and it gave accuracy of79-90% [9].
quality features were extracted such as colour, size and defects Rokunuzzman et al. presented an algorithm to classify
of apples. These features were gathered and evaluated using tomatoes ANN also and it gave 87.5% accuracy rate which
the fuzzy system and this gave 89% accuracy in classification was more than the accuracy rate given by Fuzzy logic[29].
[24]. Zhang proposed that a new method to differentiate apple Alipasandi et al. proposed a method in which peach was
stem-end/calyx from true defect according to their depth classified and it gave 99.3% accuracy rate [33]. K-means
information on apple surface. The projector was designed to clustering algorithm is used for labeling each pixel in the
generate NIR structured light based on position encoding. The image also authors used SURF (Speed up Robust Feature)
NIR structured light was used to sense the depth information algorithm for extracting features such as color, structure,
of apple surface, then the stem-end/calyx region was texture and morphology. SURF algorithm uses the blob
identified. It gave an overall 95.24% detection rate [25]. detector and local descriptors.
IV. CLASSIFIER MODELS Blob analysis consists of refinement, extraction and analysis
This section includes the most popular classifier models that for detecting the scope of interest of given images. Artificial
include fuzzy logic, artificial neural network, support vector neural network (ANN) is applied for pattern matching and
classification of diseases [40].
machine and adaptive network-based fuzzy interference
system. C. Adaptive Neural Fuzzy Interference System
A. Fuzzy Logic ANFIS was first introduced by Jang which is capable of
approximating any real and continuous function on a compact
Zadeh proposed the theory of fuzzy sets. This theory
introduced making of the membership function operations over set to any degree of accuracy. ANFIS is a fusion of artificial
the range of real numbers [0, 1]. To calculate logic neural network and fuzzy interference system. In ANFIS, a
fuzzy inference system (FIS) has been constructed, then
membership functions, new operations were proposed and
membership function parameters are tuned using either a back
showed to be areas on able tool to generalize classic logic.
propagation algorithm alone or in combination with a least
Fuzzy logic is nonparametric classification procedure which
can deduce nonlinear classification between input and output, squares method. This adjustment allows to FIS to train from
and maintain flexibility in decision making even on complex the extracted data. Normally, an ANFIS is made up of the
membership layer, the fuzzification layer, the normalization
problems. Fuzzy systems provide the means of translating the
layer, the defuzzification layer and the fifth layer, the output
expert knowledge of humans about the process in terms of
layer [34]. ANFIS is represented faster than ANN. However,
IJTRD | Jan - Feb 2018
Available [email protected] 24
 International Journal of Trend in Research and Development, Volume 5(1), ISSN: 2394-9333
www.ijtrd.com
when the trained parameters are applied to checking data, total Mongolia,‖ in IEEE International Geoscience & Remote
error of ANFIS is larger than ANN Nozari et al. presented an Sensing Symposium (IGRASS), pp.3194-3197, 2010.
algorithm for grading of Mozafatidates and classified these [2] V. K. Tewari, A. K. Arudra, S. P. Kumar, V. Pandey, N.
based on weight, length, width and thickness. These fruits S. Chandel,―Estimation of plant nitrogen content using
were graded using both ANFIS and human experts for digital image processing,‖ in International Commission of
comparison and ANFIS showed 93.5%accuracy as compared Agricultural and Bio systems Engineering, vol. 15(2), pp.
to human experts [22]. 78-86, July 2013.
[3] M. Krishna, Jabert G., ―Pest Control in Agriculture
D. Support Vector Machine
Plantation using Image Processing,‖ in IOSR Journal of
In machine learning, support vector machines are supervised Electronic and Communication Engineering(IOSR-
learning systems based on the statistical learning theory for JECE), vol.6(4), pp. 68-74, June 2013.
data classification and pattern recognition in classification and [4] G. Bhandane, S. Sharma, V.B. Nerkar, ―Early Pest
regression analysis problems. Identification inAgriculture Crop using Image Processing
,‖ in International Journal of Electrical, Electronics and
This method is used for classification of both linear and
Computer Engineering, vol.2(2), pp. 77-82,2008.
nonlinear data. In support vector machine model, samples are [5] J. K. Patil, R. Kumar, ―Advances in Image Processing for
presented as points in space, and mapped such that the samples Detection of Plant Diseases,‖ in Journal of Advanced Bio
of the distinct classes are separated by a clear boundary which
informatics Applications and Research ISSN, vol2(2),
is as wide as possible. In this method, an optimal boundary of
pp.135-141, June 2011.
two data sets in a vector space is known as a hyper plane,
[6] T. Bronsnan, D. W. Sun, ―Improving quality inspection of
which is obtained independent of the probabilistic distribution food products using computer vision- a review,‖ in
of training vectors in the set. Originally ,SVM was introduced Journal of food engineering, vol. 61,pp. 3-16, 2004.
for the problems related to two classes only, but we can extend
[7] Da-Wen Sun, ―Computer Vision- an Objective, Rapid and
it for multi-class problem using near one plot against one class
Non-Contact Quality Evaluation Tool for the Food
or one against all strategies. If the space between two data sets
Industry,‖ In Journal of Food Engineering, vol.61, pp.1-2,
is not linearly separable i.e. there is no separating hyper plane
2014.
to distinguish the classes, a kernel function may be used. The [8] H. H. Wang, D. W. Sun, ―Evaluation of the Functional
kernel function evaluates the relationships within the data and Properties of Cheddar Cheese using a Computer Vision
makes complex divisions in the space [20]. SVM has the
Method‖, in Journal of Food Engineering, vol. 49(1),
highest accuracy rate as compared to other techniques
pp.47–51, 2002.
[35].Suresha et al. proposed an Apple classifier using SVM
[9] N. B. A. Mustafa, K. Arumugam, S. K. Ahmed,Z. A. M.
and it gave 100% accuracy when only colour features are
Sharrif,― Classification of Fruits using Probabilistic
compared [20].Zheng et al. presented an algorithm of mango Neural Networks-Improvement using Colour Features‖,
grading. Fractal dimension and L*a*b* colour model are used IEEE International Conference on TENCON, 2011.
for grading purposes. Using SVM, fractal dimension and
[10] R. C. Gonzalez and R. E. Woods, Digital Image
colour gave 85.19%and 88.89% accuracy respectively [28]. K-
Processing. Prentice Hall,2nd ed., 2002.
means based segmentation is used to detect region of interest
[11] M. W. Schwarz, W. B. Cowan, J. C. Beatty, ―An
to select infected part only in the image. After that features are
Experimental Comparison of RGB, YIQ, LAB, HSV, and
extracted from the segmented portion of the given images for Opponent Colour Models‖, in ACM Transactions on
training and to store in database. Then with the help of support Graphics, vol. 6(2), April 1987.
vector machine and feature database, the image will be
[12] A. Hazra, K. Deb, S. Kundu, P. Hazra, ―Shape Oriented
classified using feature derived from segmented part of the
Feature Selection for Tomato Plant Identification‖, in
image. Due to the learning involved, this system detects type
International Journal of Computer Applications
of diseases with greater accuracy also the system suggest way Technology and Research, vol 2(4), pp. 449 - 454, 2013.
to prevent the occurrence of the disease and helps farmers by [13] O.S. Al-Kadi, ―Combined Statistical and Model based
identifying the problem in the fruit[43].
Texture Features for Improved Image Classification‖, in
CONCLUSION Advances in Medical, Signal and Information Processing,
MEDSIP, 4th International Conference, pp 1-4,2008.
In this paper, different image processing based classification [14] D. Savakar, ―Identification and Classification of Bulk
techniques for fruit grading and sorting system is reviewed. Fruits Images using Artificial Neural Networks,‖ in
Machine vision based fruit grading systems are capable of International Journal of Engineering and Innovative
replacing labour work for inspection of fruit grading. Different Technology (IJEIT), vol. 1(3), March 2012.
researchers used algorithms for image segmentation, feature [15] P. Deepa, ―A Comparative Analysis of Feature Extraction
extraction, training and classification of fruit disease.Out of Methods for Fruit Grading Classifications,‖ in
morphological, colour and texture feature, morphological gave International journal of emerging technologies in
highest accuracy rate. In colour model, HIS (Hue, Saturation, computational and applied sciences
Intensity) colour model is commonly used for grading because (IJETCAS)vol.13(138), 2013.
it is related to human perception. In machine learning [16] M. Khojastehnazhand, M. Omid, A. Tabatabaeefar,
techniques, SVM (Support Vector Machine) gave highest ―Development of Lemon Sorting System based on Colour
accuracy, but ANFIS (Adaptive Neuro Fuzzy Interference and Size,‖ in African Journal of Plant Science vol. 4(4),
System) showed the best result out of these techniques. pp. 122-127, April 2010.
Further, Fuzzy gave lowest accuracy rate result, but it is easy [17] P. Jackman, Da-Wen Sun, ―Recent Advances in Image
to implement. Processing using Image Texture Features for Food
References Quality assessment,‖ trends in Food Science &
Technology vol. 29(1), pp. 35-43, 2013.
[1] B. Erdenee, T. Ryutaro, G. Tana, ―Particular Agricultural
Land Cover Classification Case Study of Tsagaannuur,
IJTRD | Jan - Feb 2018
Available [email protected] 25
 International Journal of Trend in Research and Development, Volume 5(1), ISSN: 2394-9333
www.ijtrd.com
[18] Xu Liming and Zhao Yanchao, "Automated strawberry [34] J. S. R. Jang.,"ANFIS: Adaptive-Network-based Fuzzy
grading system based on image processing," in Inference Systems," IEEE Transactions on Systems, Man,
Computers and Electronics in Agriculture, vol. 71, no. and Cybernetics, vol. 23(3), pp. 665-685, May 1993.
Supplement 1, pp. S32-S39, April 2010. [35] H. Zhibin, W. Xiaohu, L. Hu, D. Jun , ―A Comparative
[19] N. Alavi, "Date Grading using rule based Fuzzy Study Of Artificial Neural Network, Adaptive Neuro
Interference System,‖ in Journal of Agriculture Fuzzy inference System and Support Vector Machine for
Technology, vol.8 (4), pp. 1243-1254, 2012. Forecasting River Flow in the Semiarid Mountain
[20] M. Suresha, N.A. Shilpa, B. Soumaya, ―Apples Grading Region‖, in Journal of Hydrology, vol.509, pp. 379–
based on SVM classifier,‖ in International Journal of 386,2014.
Computer Applications on National Conference on [36] Anand Singh Jalal, Shiv Ram Dubey ―Detection and
Advanced Computing and Communication-NCACC, Classification of Apple Fruit Diseases Using Complete
April2012. Local Binary Patterns‖ IEEE Third International
[21] V. Leemans, H. Mageinb, M.F. Destain, ―On-line Fruit Conference on Computer and Communication
Grading according to their External Quality using Technology, pp. 978-0-7695-4872, 2012.
Machine Vision‖, Journal of Automation and Emerging [37] Monika Jhuria, Rushikesh borse, Ashwani Kumar ―Image
Technologies, Belgium. Biosystems Engineering, pp. Processing for Smart Farming: Detection of Disease and
397–404, 2002. Fruit Grading‖ Proceeding of the IEEE Second
[22] V. Nozari, M. Mazlomzadeh, , ―Date Grading based on International Conference on Image Information
some Physical properties,‖ in Journal of Agriculture Processing, pp. 978 -1-4673-6101, 2013.
Technology, vol. 9(7), pp. 1703-1713, 2013. [38] Mrunmayee Dhakate,Ingole A.B. ―Diagnosis of
[23] T. R. B. Razak, M.B. Othman, M. N. A. Bakar, K. A. Pomegranate Plant Diseases using Neural Network‖ IEEE
Ahmed, A. R.Mansor, ―Mango Grading using Fuzzy pp. 978-1-46738564, 2015.
Image Analysis,‖ in International Conference on [39] Ridhuna Rajan Nair, Swapnal Subhash Adsul, Namrata
Agricultural, Environment and Biological Vitthal Khabale,Vrushali Sanjay Kawade ― Analysis and
Sciences(ICAEBS),2012. Detection of Infected Fruit Part Using Improved k-means
[24] I. Kavdir, D.E. Guyer, ―Apple Grading using Fuzzy Clustering and Segmentation Techniques‖ IOSR Journal
Logic,‖ in Journal of Agric Turk, vol. 27, pp. 375-382, of Computer Engineering (IOSR-JCE), pp. 37-41, 2015.
2003. [40] Ashwini Awate, Damini Deshmankar, Prof. Samadhan
[25] C. Zhang , L. Chen, W. Huang , Z. Guo, Q. Wang, ―Apple Sonavane ―Fruit Disease Detection using Color, Texture
Stem-end/Calyx Identification Using a Speckle-array Analysis and ANN‖ IEEE International Conference on
Encoding Pattern,‖ in ISCP proceeding, 2012. Green Computing and Internet of Things (ICGCIoT), pp.
[26] S. A. Khoje, S. K. Bodhe , A. Adsul, ―Automated Skin 978-1-4673-7910, 2015.
Defect Identification System for Fruit Grading Based on [41] Pujitha N, Swathi C, Kanchana V ―Detection Of External
Discrete Curvelet Transform‖ in International Journal of Defects On Mango‖ International Journal of Applied
Engineering and Technology(IJET), vol. 5(4), 2013. Engineering Research ISSN 0973-4562 Volume 11,
[27] M. Pourjafar, M. R. Benam, S. M. Mazloumzadeh, Number 7, pp. 4763-4769 2016.
―Using Adaptive Neuro-Fuzzy Interference System for [42] Bhavini J. Samajpati, Sheshang D. Degadwala ―Hybrid
Classify Date Fruits,‖ in Journal of Agriculture Approach for Apple Fruit Diseases Detection and
Technology, vol. 9(5), pp.1309-1318, 2013. Classification Using Random Forest Classifier‖ IEEE
[28] H. Zheng, H. Lu, "A least-squares support vector machine International Conference on Communication and Signal
(LS-SVM)based on fractal analysis and CIELab Processing, pp. 978-50900396,2016.
parameters for the detection of browning degree on [43] Sherlin Varughese, Nayana Shinde, Swapnali Yadav,
mango (Mangiferaindica L.)," in Computers and Jignesh Sisodia ―Learning-Based Fruit Disease Detection
Electronics in Agriculture, vol. 83, pp. 47-51, January Using Image Processing‖ International Journal of
2012. Innovative and Emerging Research in Engineering
[29] M. Rokunuzzaman, H. P.W. Jayasuriya, "Development of Volume 3, Issue 2, p-ISSN: 23945494,2016.
a Low Cost Machine Vision System for Sorting of [44] Khot.S.T, Patil Supriya, Mule Gitanjali, Labade Vidya
Tomatoes‖, in AgricEngInt: CIGRJournal, vol.15 (1), ―Pomegranate Disease Detection Using Image Processing
March 2013. Techniques‖ International Journal of Advanced Research
[30] H. Atmaca, B. Cetisli, H. S. Yavuz, "The Comparison of in Electrical, Electronics and Instrumentation Engineering
Fuzzy Inference Systems and Neural Network Vol. 5, Issue 4, p-ISSN: 2320-3765, 2016.
Approaches with ANFIS Method for Fuel Consumption
Data‖ in Second International Conference on Electrical
and Electronics Engineering ELECO, Bursa-Türkiye,
Nov. 2001.
[31] S. Haykin, Neural Networks: A Comprehensive
Foundation, 2nd Edition, Prentice-Hall, 1999.
[32] Y. Huang, ―Advances in Artificial Neural Networks-
Methodological Development and Application‖, in
algorithms ISSSN 1999-4893, vol.2, pp.973-1007, 2009.
[33] A. Alipasandi, H. Ghaffari and S. Z. Alibeyglu,
―Classification of Three Varieties of Peach Fruit Using
Artificial Neural Network Assisted with Image Processing
Techniques‖, in International Journal of Agronomy and
Plant Production, vol. 4(9), pp. 2179-2186, 2013.

IJTRD | Jan - Feb 2018

Available [email protected] 26