Journal of FST, Volume 01, Issue 01, July 2022 | 69

Fruits and Vegetables Disease Detection System Based on

Indications Using Machine Learning Approach: A Systematic
Review
Tajbia Karim1,Mariam Chowdhury2, Saima Murtuza2, Afrida Israt Jahan2, Afrina
Khatun3
Abstract
In agriculture science, automated and computerized methods increase the
country's growth, economy and productivity as it is highly dependent on the export
of fruits and vegetables. Nowadays, it is impossible to check the quality of fruits
and vegetables with bare hands as they are exported in a batch. In this world of
technology, artificial intelligence plays an essential role by introducing many
algorithms to detect diseases that hamper quality. This paper presents a detailed
review of which algorithm best detects diseases in fruits and vegetables. The paper
also includes details about pre-processing, segmentation, different algorithms for
detection, and image enhancement. An analysis of different algorithms proposed
by researchers for disease detection within fruits and vegetables was conducted.
From contemporary research works, we have come to know that there is not one
perfect method for detecting diseases of all fruits and vegetables. By careful
analysis, we have recommended which machine learning method might be suitable
for specific types of fruits and vegetables.
Keywords: Automated, Pre-processing, Segmentation, Fruits, Vegetables
1. Introduction
In this paper, we reviewed many papers that included diverse machine-learning
approaches to detect diseases in fruits and vegetables depending on the indications.
Bangladesh is an agricultural country, and this agriculture sector is crucial to the

1 Department of Information and Communication Technology, Bangladesh University of

Professionals, Dhaka, Bangladesh, Email:
2 Department of Information and Communication Technology, Bangladesh University of
Professionals, Dhaka, Bangladesh, Email:
2 Department of Information and Communication Technology, Bangladesh University of
Professionals, Dhaka, Bangladesh, Email:
2 Department of Information and Communication Technology, Bangladesh University of
Professionals, Dhaka, Bangladesh, Email:
3 Department of Information and Communication Technology, Bangladesh University of
Professionals, Dhaka, Bangladesh, Email:
70 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review
country's economy. The farmers of this country are highly dependent on
producing fruits and vegetables. However, sometimes this production is hindered
due to diseases caused by many pathogens. This is almost the same scenario for all
fruits and vegetables. Researchers worldwide tried to give solutions to this specific
problem described in their paper. By reviewing their works, we tried to analyze the
best way to detect diseases among fruits and vegetables using different machine-
learning approaches. As fruits and vegetables play a vital role in a country's
economy as they are exported abroad, our analysis will be helpful for the
researchers to find out which algorithm will be best for disease detection in fruits
and vegetables with the highest accuracy rate, which will be beneficial for
everyone, especially those who work in the farming sector and the industrial sector
as well. The motivation behind this paper was to reduce the loss in the agricultural
and food business due to fruit and vegetable diseases. As the weather in
Bangladesh is changing every day, it is affecting agricultural products greatly. So,
to reduce this loss, the researchers and other business personnel can follow this
paper to eliminate fruit and vegetable diseases to a certain level.
By analyzing thoroughly, we found out that in many fruit businesses (export &
import, juice companies, restaurants) these advanced techniques can be used to
distinguish between fresh and rotten fruits and vegetables. In future research, if the
researchers want to detect the maturity state of a fruit, it will become easy to find
out what disease can infect it with the help of a practical machine learning
approach. We compared datasets available in different papers and the approaches
they followed. To the best of our knowledge, we could not find any papers that
showed any comparison among the datasets and approaches. The comparison chart
in our review paper reflects which algorithm is better for specific cases.
2. Literature Review
Researchers have started to apply Machine Learning to detect diseases in fruits and
vegetables approximately since 2015 (Behera et al., 2018). The information about
the work of researchers is shown in Table 1. The first model that we reviewed here
was done in 2009 by (Qin et al., 2009). They proposed a model in which citrus
canker was detected on citrus using hyperspectral reflectance imaging with
spectral information divergence. Spectral Information Divergence (SID) was used
for detection, with an accuracy rate of 96.2%. Behera et al. (2018) proposed a
model using K-means clustering and Support Vector Machine (SVM) to detect
diseases of mandarin and oranges. They noticed the symptoms of the disease in
those plants by observing leaves, stems, and fruits. Also, they used image pre-
 Journal of FST, Volume 01, Issue 01, July 2022 | 71

processing techniques such as image enhancement and lab color transformation.

They were successful in detecting diseases with an accuracy of 90%. Then, Singh
et al. (2019) proposed a system for detecting diseases in Mangifera indica (mango)
by using Multilayer Convolutional Neural Network (MCNN). They used leaf for
detection and for pre-processing, contrast enhancement and image reshaping were
used. Their accuracy score was 97.13%. In the research of JJMCOE and J (2017),
diseases were discovered in Malus domestica (apple). They have used the K-means
clustering segmentation algorithm and Learning Vector Quantization Neural
Network (LVQNN) for detection. For pre-processing, they have used image
enhancement. They used 24 apple images that were real-time in nature, and they
achieved an accuracy of 95%. Ullah et al. (2019) proposed a model to detect fruit
disease by analyzing the leaves. They used contrast enhancement and image re-
scaling for pre-processing and Convolutional Neural Network (CNN) as a
detection algorithm. Their work had an accuracy rate of 97.33%. Ouyang et al.
(2012) made a model to detect strawberry plant disease. The Top-hat
transformation was used for pre-processing, which removes noise interference by
median filtering. Their detection algorithm was an image segmentation algorithm
with an accuracy rate of 99%. In the research of Pujari et al. (2013) grape, mango,
and pomegranate fruits were used to grade and classify the disease of those fruits
based on their statistical texture. Segmentation and Artificial Neural Networks
(ANN) were used for detection purposes, and the accuracy rate was 84.65% for
regular fruits and 76.6% for affected fruits. In 2019, (Doh et al.,2019) proposed a
model using both SVM and K-Nearest Neighbour (KNN) for disease detection in
citrus fruits and found an accuracy of 93.12% and 88.96% respectively.
In recent years, Yang et al. (2022) have achieved 99.70% accuracy in detecting
disease in apples using a sorting image recognition system. Before that, Zaki et al.
(2021) detected the disease in onion plants by observing young onion leaves. They
have used CNN for this detection, and for pre-processing techniques, they have
used image pre-processing. They achieved an accuracy of 77.05% for a data set of
batch size 8 and 85.47% for a dataset of 16 images. Khattak et al. (2021) made a
model which can detect infection in the citrus-based plant using their fruits and
leaves and found an accuracy score of 94.55%. They have used CNN for this
model. Another model where CNN is used as a detecting algorithm was proposed
by Chowdhury et al. (2021). They did their research on tomato plants and found
an accuracy score of 99.89%. Lastly, Mostafa et al. (2021) did their research on
the guava plant using Deep Convolutional Neural Network (DCNN) structure and
got an accuracy score of 97.74%.
72 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review

3. Fruits and Vegetable Disease Detection

Disease detection of fruits and vegetables using Machine learning algorithms
mainly involves five steps: Dataset collection, Data pre-processing, Segmentation,
Feature selection, and Classifier.
3.1. Dataset
The primary data in fruit and vegetable disease detection applications are image
data. Some of the main sources of collected data are google and various GitHub
repositories. Some researchers also collected images from the fruits and vegetables
available in the supermarket. Images acquired by different researchers are shown
in Table 2.
3.2. Pre-processing
Images are acquired by various techniques consisting of multiple noises that
worsen the aspect of an image. For this reason, the image without any pre-
processing cannot contribute the appropriate data. Pre-processing enhances the
image data, which overcomes reluctant misshaping and enlarges the image features
essential for the processing and building an appropriate image (which is degraded
form) than the original image for a specific application. The techniques used for
image pre-processing for machine learning algorithms are Contrast enhancement
and Rescaling.
3.2.1. Contrast Enhancement
The following equation is used to filter the contrast of any given image by
assigning a constant intensity value to the pixel of that image using the histogram
of an image. It is done with the help of the histogram equalization method.
) ( ( , ) )− (
× − 1)
( )
( ( , )) = ( ( × )− (
(1)
)

Here, ( ) = cumulative frequency of the gray level

( ) = minimum value of cumulative distribution function
( )( ( , )) = intensity of the current pixel,
R and C = product of several pixels in rows and columns
L = number of intensities
 Journal of FST, Volume 01, Issue 01, July 2022 | 73

3.2.2. Rescaling
The images have been collected from different sources and captured with different
devices. So, the images are not of the same size, therefore the images need to be
resized.
3.3. Segmentation
Image segmentation is required to separate a digital image into distinct areas
(Bhargava and Bansal 2021).
3.3.1. K-Means Clustering
K-means clustering is a type of vector quantization that divides n numbers of data
into k number of clusters. Every observation is assigned to the cluster with the
closest mean which acts as the prototype for the cluster. As a result, Voronoi cells
are created in the data space (Vertica, 2022). Within-cluster variances are
minimized using K-means clustering (squared Euclidean distances). Regular
Euclidean distances, on the other hand, are not the more difficult Weber problem:
the mean minimizes squared errors, but only the geometric median reduces
Euclidean distances. Better Euclidean solutions can be produced, for example, by
employing k-medians and k-medoids. The supervised KNN classifier, which is
commonly confused with k-means due to its name, is connected to the
unsupervised k-means algorithm. Using the cluster generated by k-means, the 1-
nearest neighbor classifier is used to categorize incoming data into existing
clusters.
3.3.2. Thresholding
Thresholding is an image segmentation technique that applies to the change of the
pixels of an image to make it easier to analyze. It turns into a binary image from a
color or grayscale image using thresholding. While the user can or should pick the
threshold T manually in some circumstances, the user often prefers the threshold
to be established automatically by an algorithm. The threshold should be the "best"
threshold in certain circumstances, separating the brighter foreground items from
the darker background objects. The threshold for picture intensity (image
brightness) can be preset manually or automatically by some applications. Pixels
whose bit values are zero turn black, whereas those with a bit value larger than
zero become white (a bit value of one).
74 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review
3.4. Feature Selection
Feature selection is the process of selecting the most important features to input in
machine learning algorithms. Feature selection techniques are implemented to
reduce the number of input variables by eliminating redundant or irrelevant
features and narrowing down the set of features to those most relevant to the
machine learning model. Some of the most common feature selection methods are:
3.4.1. Global Color Histogram (GCH)
GCH is a common term in machine learning. Color is regarded as one of the most
important forms of visual representation. Changes to an image's size, rotation, or
translation into different forms do not affect the image's color. Color space, color
reduction, and the color feature extraction procedure are all factors to consider
when extracting color characteristics from a picture (Han et al., 2022).

3.4.2. Color Coherence Vector (CCV)

CCV works by ranging each pixel as either coherent or incoherent. For each color,
the number of coherent versus incoherent pixels is stored in a CCV. By separating
pixels from incoherent pixels, it is possible to make finer distinctions than color
histograms. They're all partially connected hidden layers, with the output layer
being the fully connected layer at the end. The size of the input image is mirrored
in the output shape (Pass et al., 1997).

3.4.3. SVM

SVM is a collection of supervised learning methods for classification, regression,

and detecting outliers. The goal of the SVM algorithm is to find the best line or
decision boundary for categorizing n-dimensional space into classes in the future
so that new data points can be easily placed in the appropriate category. The best
decision boundary is known as a hyperplane. Making a straight line between two
classes is how a simple SVM works (McGregor and M, 2020). There are two types
of SVM. Linear SVM is used for linear regression and classification problems.
Kernel SVM is used for non-linear data. It can fit a hyperplane instead of a two-
dimensional space with more features which has more flexibility for non-linear
data.
 Journal of FST, Volume 01, Issue 01, July 2022 | 75

3.4.4. Local Binary Pattern (LBP)

LBP is a robust method for image feature extraction in image processing. LBP
reflects the correlation among pixels within a local area (usually a 3×3 area) which
represents the local information (Bingham et al., 2015). By applying LBP, texture
pattern probability can be precise in a histogram. LBP values need to be
determined for all the image pixels (Prakasa, 2016).

3.4.5. Gray level co-occurrence matrix (GLCM)

It is a typical texture-based feature extraction model. The GLCM determines the

relationship between pixels by performing an operation according to the second-
order statistics in the images (Öztürk and Akdemir, 2018).

3.4.6. LVQNN
LVQNN is a superior classification technique for digital pictures. In general, it's a
good idea to prepare data for LVQNN the same way it is done for KNN.

Table-1: Comparison of various classification methods for disease detection of

vegetables and fruits
Fruit/Vegetabl Affecte Pre-
Work Algorithm Accuracy
e d Part processing
Mandarin Leaf Image K-means
Behera
enhancement clustering
et al. 90%
Stem Lab color
(2018) Orange SVM
Fruit transformation
77.05%
and
Zaki et
Young Image pre- 85.47% for
al. Onion CNN
leaf processing a batch size
(2021)
of 8 and 16
dataset
Contrast
Singh et
enhancement
al. Mango Leaf MCNN 97.13%
Rescaling of
(2019)
the image data
 76 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review

K-Means
JJMCO clustering
Image More than
E and J Apple Fruit segmentation
enhancement 95%
(2017) algorithm
LVQNN
Contrast
Apple
enhancement
Bell paper
Blueberry
Cherry(sour)
Corn(maize)
Ullah et Grape
al. Orange Leaf CNN 97.33%
(2019) Peach Image resizing
Potato
Raspberry
Soybean
Strawberry
Squash
Tomato
Top hat
transform
Ouyang Image
Median
et al. Strawberry Fruit segmentation 99%
filtering for
(2012) algorithm
removing noise
interference
Grape 84.65%
Mango and 76.6%
Pujari et
ANN for healthy
al. Fruit None
classifier and
(2013) Pomegranate
diseased
fruits
Qin et SID method
al. Citrus Fruit None for 96.2%
(2009) classification
 Journal of FST, Volume 01, Issue 01, July 2022 | 77

Keras image
Khattak
Fruit pre-processing 94.55%
et al.
Citrus and using data CNN
(2021)
leaves generator class
and API
Grayscale
processing
Sorting
Yang et Binarization
image
al. Apple Fruit Enhancement 99.70%
recognition
(2022) Processing
system
Feature
extraction
Image SVM
ANN
segmentation (93.12%)
Doh et
SVM
al. Citrus Fruit
Feature Phenotyping ANN
(2019)
extraction K-means (88.96%)
clustering
Resizing and
Chowdh
Normalizing
ury et
Tomato Leaves Leaf CNN 99.89%
al.
segmentation
(2021)
Augmentation
Image DCNN
acquisition structures (
AlexNet,
Mostafa
SqueezeNet,
et al. Guava Plant 97.74%
GoogLeNet,
(2021) Illumination
ResNet-50,
and ResNet-
101)
78 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review

Table-2: Features of the datasets obtained by different researchers

Total Number of
Work Fruit/Vegetable Image Source
Image Data
Apple Internet
Bell paper
Blueberry
Cherry(sour)
Corn(maize)
Grape
Ullah et al. Orange
2,17,204
(2019) Peach Smartphone
Potato
Raspberry
Soybean
Strawberry
Squash
Tomato
Real-time image
Singh et al.
Mango 2200 GitHub
(2019)
repository
Zaki et al. (2021) Onion 1000 Real-time image
JJMCOE and J
Apple 24 Real-time image
(2017)
Khattak et al.
(2021) Citrus 2293 Different dataset

Chowdhury et al. PlantVillage

(2021) Tomato leaves 18,161 tomato leaf
images

4. Comparative Analysis
After reviewing the works of multiple researchers, we can summarize that different
algorithms work best on different fruits and vegetables. In Figure 1, we can get a
brief idea about the accuracy rate for detecting disease in mango plants by using
different algorithms.
 Journal of FST, Volume 01, Issue 01, July 2022 | 79

Accuracy Vs Algorithm
120.00%
97.13% 96%
100.00%
80%
80.00%

60.00%

40.00%

20.00%

0.00%
MCNN SVM, K-means DCNN
clustering

Figure 1: Accuracy vs algorithms for Mango disease detection

We have collected the data from the work of multiple researchers. Singh et al.
(2019) used MCNN and found an accuracy score of 97.13%. Mia et al. (2020)
proposed a model that used SVM and K-means clustering as detection algorithms
and the accuracy rate was 80%. Ashok et al. (2021) used DCNN and found an
accuracy of 96%. So, we can understand that MCNN has the highest score for
accuracy in detecting disease in mango plants. By analyzing the comparison, we
observed that MCNN shows exceptionally high accuracy in image recognition
problems. It can also detect necessary features without any human guidance. On
the other hand, for SVM and K-means clustering with an 80% accuracy rate, we
observed if the dataset is large then SVM is not suitable as well as for K-means
clustering, it is generally prone to biases.
We can conclude that MCNN is the best among the algorithms we analyzed due to
its high accuracy rate and other features.
80 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review

Accuracy Vs Algorithm
98.00% 97.18% 96.90%
97.00%
96.00%
95.00%
94.00%
93%
93.00%
92.00%
91.00%
90.00%
CNN LBP K-means clustering
segmentation
algorithm

Figure 2: Accuracy vs algorithms for Apple disease detection

Figure 2 shows a bar chart of accuracy Vs different algorithms for detecting
diseases in Apple. Khan et al. (2021) used CNN and found an accuracy score of
97.18%. 18%. Dubey et al. (2012) used LBP to complete the detection and found
an accuracy of 93%. In the research of JJMCOE and J (2017), they used the K-
means clustering segmentation algorithm and LVQNN and found an accuracy of
96.6%. So, we can say that the CNN algorithm has the best accuracy rate for
detecting apple diseases. By analyzing the comparison of figure 2, we observed an
accuracy rate of 97.18% for CNN. From the earlier comparison, we know that it
gives a high accuracy rate in image recognition which is one of the notable features
of CNN. In the figure, the LBP shows an accuracy rate of 93% which is the lowest
among the other two algorithms. One of the disadvantages is its level of
recognition is still low.
We can conclude that CNN is the best among the algorithms we analyzed due to
its high accuracy rate and other features.
 Journal of FST, Volume 01, Issue 01, July 2022 | 81

Accuracy vs Algorithm
98.00% 96.20%
96.00% 94.55%
94.00% 93.12%
92.00%
90.00%
88.00% 86.96%
86.00%
84.00%
82.00%
SID method for CNN ANN SVM
classification

Figure 3: Accuracy vs algorithms for Citrus disease detection

Figure 3 represents a bar chart of different machine learning algorithms along with
their accuracy rate to detect diseases in citrus fruit and leaves. From the above
chart, we distinguished that the SID method has the highest accuracy rate. One of
the major reasons for this value is that the SID method is an image-based
classification procedure that is used to discriminate citrus diseases from other
confounding diseases. On the other hand, SVM has the lowest accuracy rate among
all other methods because it is sensitive to noisy attributes and requires a large
search time.
We can conclude that for citrus disease detection SID method for classification is
most suitable.
5. Conclusion
This paper highlights detailed descriptions of different algorithms for disease
detection within fruits and vegetables along with applied pre-processing factors.
The most significant features of agricultural by-products are their size, texture, and
color. After reviewing and studying many research papers we came to a position
where we can say which algorithm is best and will give high accuracy in detecting
diseases in fruits and vegetables. More precise results can be obtained if
researchers review papers with a large number of datasets which will allow them
82 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review
to acquire higher accuracy rates and will help them to understand which machine
learning algorithm is appropriate for detecting diseases in any kind of agro-based
products. We can conclude that this comparative study of different machine
learning algorithms can be applied to early-stage detection of commonly available
fruits and vegetables preventing massive economic loss with an increase in GDP
rate as a result of which any country can progress in doing business with agro-
based products without any complication.
References
Ashok, V., & Vinod, D. S. (2021). A novel fusion of deep learning and android
application for real-time mango fruits disease detection. In Intelligent
System Design. Springer, Singapore, pp. 781-79.
Behera, S. K., Jena, L., Rath, A. K., & Sethy, P. K. (2018, April). Disease
classification and grading of orange using machine learning and fuzzy
logic. In 2018 International Conference on Communication and Signal
Processing (ICCSP). IEEE, pp. 0678-0682.
Bhargava, A., & Bansal, A. (2021). Fruits and vegetables quality evaluation using
computer vision: A review. Journal of King Saud University-Computer
and Information Sciences, 33(3), pp. 243-257.
Bingham, E., Kaski, S., Laaksonen, J., & Lampinen, J. (Eds.). (2015). Advances
in independent component analysis and learning machines. Academic
Press.
Chowdhury, M. E., Rahman, T., Khandakar, A., Ayari, M. A., Khan, A. U., Khan,
M. S., ... & Ali, S. H. M. (2021). Automatic and reliable leaf disease
detection using deep learning techniques. AgriEngineering, 3(2), pp. 294-
312.
Doh, B., Zhang, D., Shen, Y., Hussain, F., Doh, R. F., & Ayepah, K. (2019,
September). Automatic citrus fruit disease detection by phenotyping using
machine learning. In 2019 25th International Conference on Automation
and Computing (ICAC). IEEE, pp. 1-5.
Dubey, S. R., & Jalal, A. S. (2012, November). Detection and classification of
apple fruit diseases using complete local binary patterns. In 2012 Third
International Conference on Computer and Communication Technology.
IEEE, pp. 346-351.
 Journal of FST, Volume 01, Issue 01, July 2022 | 83

Han, Ju & Ma, Kai-Kuang. (2002). Fuzzy color histogram and its use in color
image retrieval. IEEE transactions on image processing: a publication of
the IEEE Signal Processing Society, 11, pp. 944-52.
JJMCOE, J. (2017). Detection and Classification of Apple Fruit Diseases using K-
means clustering and Learning Vector Quantization Neural Network.
Khan, A. I., Quadri, S. M. K., & Banday, S. (2020). Deep learning for apple
diseases: classification and identification. arXiv preprint
arXiv:2007.02980.
Khattak, A., Asghar, M. U., Batool, U., Asghar, M. Z., Ullah, H., Al-Rakhami, M.,
& Gumaei, A. (2021). Automatic detection of citrus fruit and leaves
diseases using deep neural network model. IEEE Access, 9, pp. 112942-
112954.
McGregor, M. (2020, July 02). SVM machine learning tutorial – what is the
support vector machine algorithm, explained with code examples.
Retrieved June 23, 2022, from https://www.freecodecamp.org/news/svm-
machine-learning-tutorial-what-is-the-support-vector-machine-
algorithm-explained-with-code-examples/ (Last accessed 21 July 2022)
Mia, M., Roy, S., Das, S. K., & Rahman, M. (2020). Mango leaf disease
recognition using neural network and support vector machine. Iran Journal
of Computer Science, 3(3), pp.185-193.
Mostafa, A. M., Kumar, S. A., Meraj, T., Rauf, H. T., Alnuaim, A. A., &
Alkhayyal, M. A. (2021). Guava Disease Detection Using Deep
Convolutional Neural Networks: A Case Study of Guava Plants. Applied
Sciences, 12(1), 239, pp. 1-19.
Ouyang, C., Li, D., Wang, J., Wang, S., & Han, Y. (2012, October). The research
of strawberry disease identification based on image processing and pattern
recognition. In International Conference on Computer and Computing
Technologies in Agriculture. Springer, Berlin, Heidelberg. pp. 69-77.
Öztürk, Ş., & Akdemir, B. (2018). Application of feature extraction and
classification methods for histopathological image using GLCM, LBP,
LBGLCM, GLRLM and SFTA. Procedia computer science, 132, pp. 40-
46.
84 | Fruits and Vegetables Disease Detection System Based on Indications Using
Machine Learning Approach: A Systematic Review

Pass, G., Zabih, R., & Miller, J. (1997, February). Comparing images using color
coherence vectors. In Proceedings of the fourth ACM international
conference on Multimedia, pp. 65-73.
Prakasa, E. (2016). Texture feature extraction by using local binary pattern.
INKOM Journal, 9(2), pp. 45-48.
Pujari, J. D., Yakkundimath, R., & Byadgi, A. S. (2013). Grading and
classification of anthracnose fungal disease of fruits based on statistical
texture features. International Journal of Advanced Science and
Technology, 52(1), pp. 121-132.
Qin, J., Burks, T. F., Ritenour, M. A., & Bonn, W. G. (2009). Detection of citrus
canker using hyperspectral reflectance imaging with spectral
information divergence. Journal of food engineering, 93(2), pp.183-191.
Singh, U. P., Chouhan, S. S., Jain, S., & Jain, S. (2019). Multilayer convolution
neural network for the classification of mango leaves infected by
anthracnose disease. IEEE Access, 7, pp. 43721-43729.
Ullah, M. R., Dola, N. A., Sattar, A., & Hasnat, A. (2019, November). Plant
Diseases Recognition Using Machine Learning. In 2019 8th
International Conference System Modeling and Advancement in
Research Trends (SMART). IEEE, pp. 67-73.
VerticaPy. (2022). Retrieved 22 June 2022, from
https://www.vertica.com/python/documentation_last/learn/KMeans/
(Last accessed 21 July 2022)
Yang, M., Kumar, P., Bhola, J., & Shabaz, M. (2022). Development of image
recognition software based on artificial intelligence algorithm for the
efficient sorting of apple fruit. International Journal of System
Assurance Engineering and Management, 13(1), pp. 322-330.
Zaki, M. A., Narejo, S., Ahsan, M., Zai, S., Anjum, M. R., & u Din, N. (2021).
Image-based Onion Disease (Purple Blotch) Detection using Deep
Convolutional Neural Network. Image, 12(5), pp. 448-458.