Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Mulberry Leaf Disease Detection

Nagaveni B Nimbal1, Namratha S.2, Arshitha K.3, Chaithra Shetty S.4, Bhuvanashri KB5
Asst. Professor1, Student2,3,4,5
Department of Computer Science
KSSEMKSSEM Bangalore, India

Abstract:- Sericulture is an important domestic essential one in the agriculture field in the beginning of a
industry. In India,it is one of the eco-friendly industries. series of activities to fight the plant disease and reduce their
India is the only country where all five recognized spread as soon as they appear on leaves of the plant.
commercial silks are made, namely mulberry, tropical Technology has grown to such an extent that a machine is
tasar, oak tasar, eri and muga. Sericulture is labor capable enough to predict the leaf disease without human
intensive, providing jobs in India for over 8 million intervention.to predict these diseases we use KNN and
people, and serving Indians as a tremendous source of SVM algorithm. Mulberry silk originates from the
revenue. Silkworm is sericultural foundation. silkworm, Bombyx mori L7 which exclusively benefits
Commercial silk is developed through the production of from the leaves of mulberry plant. These silkworms are
different types of silkworms, of which BOMBYX totally domesticated and reared indoors. Other uses of
MORI, originally from Asia, is the most widely and mulberry leaves are seen in the fields of health and skin
economically used, Mulberry is significant sole care.There is an option for consulting experts, but farmers
nourishment for mulberry silkworm, which exclusively living in remote villages may not be able to afford
benefits from the leaves of mulberry plant. These travelling expenses because he laboratory test results could
silkworms are totally domesticated and reared indoors. take a few days and farmers have to travel many times and
Other uses of mulberry leaves are seen in the fields of government experts visit the farm a few villages, but not
health and skin care. These mulberry plants include a regularly.
high pace of yield disappointment and are over the top
expensive for creation, so should be dealt quite well.Our There is a possibility for counselling specialists, yet
goal is to overcome these problems using a farmer- farmers living in remote towns probably won’t manage cost
friendly system where the result involves cure of the of travelling expenses because the test results from the lab
disease and the fertilizer or pesticide proportion to be could few days and farmer needs to travel ordinarily and in
used are displayed on the user interface. hardly any towns specialists from government division
visits the sector however not on standard premise.
Keywords:- Plant disease, Image preprocessing, disease
classification, EHD, Support vector method, KNN, DT. II. RELATED WORK

I. INTRODUCTION The image of the leaf is taken and then converted into
greyscale image. Using texture feature including contrast,
Sericulture is an agro-based industry. India is second local homogeneity, cluster shades are also used. For image
largest producer of silk. Sericulture has following classification, SVM classifier and minimum distance
highlights: high work potential, gives energy to town criterion is used. The data collection is done by taking
economies, Low Gestation and High Returns, women pictures and then image is preprocessed using image
friendly occupation, Eco-accommodating Activity. India annotation and augmentation. Image analysis is done using
has the one amongst a form qualification of being the main multiple extractors. For experimentation dataset is divided
nation delivering all the five known business silks, into testing, training and validation sets. System can
especially mulberry. Mulberry silk is more produced in classify the diseases using algorithm. the image
India from Karnataka, Andhra Pradesh, Tamil Nadu, segmentation including all existing factors using the HSI
Jammu & Kashmir and West Bengal, whilethe non- color system, here H component is used to segment spots
mulberry silks are produced in the states of Jharkhand, and to reduce illumination. the affected region and leaf
Chhattisgarh, Orissa and north-eastern states.Agriculture is segment area respectively. In the final step, the disease is
a key source of livelihood. Agriculture provides classified using quotient calculation of leaf and lesion area.
employment opportunities for village people on large scale According to this research, this method is fast and
in developing country like India. Most of Indian farmers are calculation of leaf disease severity is accurate, here leaf
adopting manual cultivation due to lagging of technical area is calculated by using threshold segmentation disease
knowledge, when plants are affected by diseases through detection using KNN and SVM algorithm. Feature
leaves that will affect production of agriculture and extraction and statistical features. Different statistical
profitable loss. features energy, sum entropy, covariance, information
measure, entropy contrast. The system can detect the
Leaves are important for fast growing of plant and to disease with accuracy.
increase production of crops. Identifying these diseases in
plants leaves is challenging for farmers. They are integral
part of the plant system they play a major role in
preparation of food in plants. The detection process is

IJISRT23APR1763 www.ijisrt.com 1725

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. MULBERRY DISEASES

A. Powdery Mildew
Powdery Mildew Disease of mulberry is caused by
fungal Pathogen, Phyllactiniacorylea. The major symptom
of this disease is an appearance of white powdery patches
on the lower surface of the leaves. When the disease is
severe, the white powdery patches turn in to brownish-
black; the leaves become yellow, coarse and lose their
nutritive value. The severely infected leaves will be become
powder brittle even after gently crumple by hand.

Fig. 3: Leaf rust

D. Leaf curl
Curly leaves of mulberry revealed the presence of
Taeniothrips, attacking mulberry leaves. The attack of this
taeniothrips to mulberry leaves is known from India and Sri
Lanka. They injure epidermal tissue and affected leaves
show early maturity, depletion of moisture, reduction in
crude protein silkworm rearing. Taeniothrips affected
leaves generally show streaks in the early stage of attack
whereas blotches are observed at the advance stage and
ultimately become yellowish-brown on maturity.

Fig. 1: Leaf spot

B. Leaf Spot Disease

Leaf Spot Disease of mulberry is caused by a
fungus, Cercosporamoricola.

At the beginning of the disease there will be small

light brown irregular spots appear on mulberry leaf surface.
Later, these spots enlarge and join together leaving with
characteristic ‘shot hole’ and yellow patches around the
brownish spot and wither off.

Fig 4: Leaf curl

IV. LITERATURE REVIEW

Detection of plant leaf diseases using image

segmentation and soft computing techniques. Information
Processing in Agriculture.

A generic algorithm for image segmentation is

presented. Soft computing techniques are used for image
segmentation. Here, the images are converted into
greyscale, and texture featuring methods such as local
homogeneity, cluster shades are also used. SVM classifier
Fig. 2: Leaf spot and minimum distance criterion are used in image
classification.
C. Leaf rust disease
Mulberry leaf rust disease is caused by Peridiospora A. Plant Disease Detection in Image Processing Using
mori fungus.At the rust disease initial stage, the symptoms MATLAB
is a circular pin head sized light brown spots and becomes Plant leaf disease detection is done using KNN and
darkish brown spots as disease advances. The leaf loss is up SVM algorithms. The different statistical features energy,
to 5 %. Rust severity significantly increased with increasing sum entropy, covariance, information measure, entropy
shoot age, irrespective of pruning time. contrast is being considere.

IJISRT23APR1763 www.ijisrt.com 1726

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Feature Extraction
In this feature extraction the shape, colour& texture is
identified. The aim of this phase is to extract features such
as colour and shape. Two shape features such as area and
perimeter are extracted This includes two phases: image
resize and image filtering.

C. Classification
The final stage of the work is selection of suitable
classification algorithm for classification of leaf disease to
the category they belong.Support Vector Machine (SVM) is
a supervised machine learning algorithm that can be used
for classification. SVM is popularly used in texture and
classification. After extracting colour and texture features
of leaves the classification.

V. THE PROPOSED METHOD

A device is being built that automatically identifies

leaf disease using image recognition. This in turn helps the
farmers in the identifying the diseases at an early stage and Fig. 5: Design of Application
provide useful information to control it. This system does
few image processing techniques like image acquisition, A. Disease identification is done using following steps:
image pre-processing, feature extraction and classification.  Image Acquisition.
Modern agricultural practices assure great development of  Image Pre-processing
cultivation.We have other smart agriculture technology  Feature Extraction
models. The Project aims at classifying the diseases in plant  Image Classification
leaf at the earlier stage prevent the plants from the diseases.
The model is based on Supervised Learning Algorithms to  Image Acquisition:
detect the disease which is more accurate for classification Image acquisition can be definedas the retrieval of an
purposes. SVM, KNN, DT techniques are used in image from some source,typically a hardware-based source,
classifying the plant leaves into healthy or diseased, if it is a so that it can be passed through whatever processes need to
diseased plant leaf these algorithms will give the name of occur later. Performing image retrieval in image processing
the particular disease and control measures. At first an input is the first step of the workflow sequence, because
image is selected from the folder which under goes image processing without a image is not achievable. The captured
acquisition, pre-processing where it undergoes rescaling, picture is totally unprocessed and is the outcome of
converting into grey scale and removal of noise. Then the whatever hardware was used to produce it, which can be
training images are subjected to feature extraction. The very useful for providing a stable reference on which to
texture features extracted are used for classification. The operate on certain areas. The images obtained using the
classification results detect and categorized based on the digital camera is subjected to further pre-processing.
disease of leaf. The output is obtained at last which
 Image Pre-processing:
indicates the disease name along with its control measures
The data collected are typically noisy and come from
for the better management of the crop.For feature extraction
different sources. In order to feed this collected data into
we have used EHD, HOG and GLCM. For classification we
the machine learning model, it should be in standard format
have used SVM, KNN and decision tree, where we have
or noise free. More often, preprocessing is used to perform
applied cross validation for SVM.After the image under
steps that minimize complexity and improve the accuracy
goes Acquisition, Pre-processing, Feature Extraction and
of the algorithm implemented. Pre-processing is meant to
Classification the result will be obtained in the user
enhance image data by removing unnecessary artifacts or
interface. This contains cure of the disease and the fertilizer
enhance other image features required for further
or pesticide proportion to be used and also future preventive
processing. To increase the image quality, the photographs
measures which helps the farmer to manage the disease.
collected from the camera are subjected to preprocessing.

 Resize: Some images caught by camera and fed to our AI

algorithm fluctuate in size therefore we should establish a
base size for all images fed into our AI algorithms.
 Grayscale: To store a single colour pixel of an RGB
colour image we will need 8*3 = 24 bits (8 bit for each
colour component), but when we convert an RGB image
to grayscale image, just 8 bit is required to store a single
pixel of the image. So, we will require 33 % less memory
to store grayscale image than to store an RGB image. It is

IJISRT23APR1763 www.ijisrt.com 1727

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
also easier to recognize features of an image when we
manage with a single layered image.
 Noise removal: Noise is a random variation of image
Intensity and visible as a part of grains in the image.
Here, we use median filtering to remove the noise.
Median filtering is the nonlinear process which is useful
in reducing impulsive. The median filter is broadly
utilized in computerized image processing just because it
preserves edge properties unlike Gaussian blur technique.

 Feature Extraction:
Feature extraction is the name for methods that select
and /or combine variables into features, effectively
reducing the amount of data that must be processed, while
still precisely and completely describing the original data
set.

 EHD: The edge histogram descriptor (EHD) is one of the Table 1: Semantics of local edge bins
widely used methods for shape detection. It basically
represents the relative frequency of occurrence of each  Image Classification:
local area of 5 types of edges, called a sub-image or Image classification refers to image labelling in one of
image block. The sub image is characterized by non- several predefined classes. Classification of images refers
overlapping blocks dividing the image space into 4x4. to the task of extracting classes of information from a
So, the partition of image certainly makes 16 equivalents multiband raster image. The resulting image classification
estimated regardless of the size of the original image. To raster can be used to create themed maps. There are two
define the features of the image block, histogram is forms of classification, based on the relationship between
generated for edge distribution for every image block. the analyst and the machine during classification:
supervised and non-supervised.Classification of images
focuses on the identification of images in one of several
classification techniques.

 SVM: Here we introduce multiclass Classification

model with one-against-all method where there would be
a binary SVM to distinguish members of that class from
members of that group for each class. It builds templates
for k SVM where k is the number of groups. For each of
the examples in the ith sequence, the ith SVM is equipped
with positive labels, and all other examples with
negative labels remain. Thus, given l training data
(x1,y1),….,(xl,yl), where xi Ꞓ Rn , i=1,….l, and yi
Ꞓ {1,….,k}, the xi class, the ith SVM solves the following
problem:

Fig. 6: Five Types of Edges in EHD

The complete semantics for the EHD of 80 histogram

bins are listed in Table 1. Each histogram bin meaning The penalty variable is where the xi training data is
should of course be standardized and quantified. For the converted to a multidimensional space by function ϕ and
normalization process, the amount of edge occurrences function C. Minimizing (1/2)(w)T wi implies we want to
inside the sub-image for each bin is separated by the total optimize 2/||wi||, the difference between two data groups. It
number of image-blocks. The image-block can be a simple is not possible to segregate the data linearly, there is a
device to collect details about the face. This implies that for penalty term C Σlj=1 εij that can reduce the number of
increasing picture object, it decides if there is at least one training errors. The core idea underneath Algorithm is to
edge and which edge is prevalent. When there is an edge, find a stability between the terms (1/2)(w)T wiand the
the predominant type of edge between all the 5 edge training errors.There are functions for decision k after
segments is determined as well. The histogram size of the solving the above equation (1).
corresponding edge bin would then increase by one. Then
the picture-block does not include an edge for the
monotonous area inside the picture.

IJISRT23APR1763 www.ijisrt.com 1728

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
through traversing the tree. Test is applied at the root
node to decide which child node the record will come up
next. This process is repeated until the record reaches at
leaf node. Each leaf node has a unique path from root to
We say that x falls within the class with the greatest root. The path is a rule used to sort the record.
value of the decision function.
B. Disease Management
After the image under goes acquisition, pre-processing,
feature extraction and classification the result will be
obtained in the user interface. This contains cure of the
 KNN: The k nearest neighbor algorithm (k-NN) is a non- disease and the fertilizer or pesticide proportion to be used
parametric approach that is used for classification and and also future preventive measures which helps the farmer
regression. The last performance in the KNN group is a to manage the disease.
time association. The item is listed by plurality vote of its
neighbors. Consider the object that is close to the k value, VI. OUTPUT
where k represents a positive integer. We have
implemented Euclidean, Minkowski, Chebychev, The implementation of our project is done in
Cityblock, Cosine distances formulas. MATLAB. We have basically concentrated on 6 types of
mulberry leaf diseases. All these data sets were captured by
us through mobile camera. We have used 600 images where
420 are used for training set and 180 are used as testing set.
At first the image is selected from the folder and after
processing happens in MATLAB the output is obtained
which indicates the disease of the leaf along with its control
measures for the better management of the crop.

Fig. 7: Working of KNN

The execution of a KNN classifier is essentially

dictated by K's decision and also the connected division Fig. 8: Nitrogen Deficiency
metric. The gage is determined by the affectability of
region measure K assessment, on the grounds that the Figure 8 represents Nitrogen Deficiency. Nitrogen
sweep of the surrounding locale is defined by the division deficiency leaf shows Gradual and sluggish plant
of the nearest neighbor to the query from the Kth and broad development with lower cell growth / vigoration. The
K yields distinctive restricted class likelihood. Through young green leaves display chlorosis, the stem is thin and
introducing the deviations from various groups, an yellowish green and the root development stunted.
exhaustive estimate of K essentially renders the gage over
smoothing and the scheme implementation debases.The
estimation of k-NN is perhaps the simplest of all
measurements in machine learning.
 Decision Tree: The decision tree algorithm is the
supervised learning process. It's being used to solve both
regression and classification issues. It uses tree
representation to deal with the problem. One class
represents the leaf node. Initially we consider the whole
training set to be the root. Recursive distributions are
made on the basis of attribute values. Information Gain
used in each level for the identification of the root node
attribute. To accurately define gaining information, we
need to define a measure commonly used in entropy, the Fig. 9: Cercospora Leaf Spot
information theory. The classification of test data is done

IJISRT23APR1763 www.ijisrt.com 1729

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Figure 9 represents Cercospora Leaf Spot.Based on
the weather and variety, it decreases leaf yields by around
10 – 20 percent. Symptoms are in beginning phases, they
occur as little light brown, small dots mostly on stems. As
when the disease worsens, the spots are formed to enlarge,
coalesce, and shot-holes. Severely infected leaves grow
yellowish and prematurely drop down.

Fig. 12: Powdery Mildew

Figure 12 represents Powdery mildew.

Phyllactiniacorylea karst tends to cause powdery mildew.
This has been the most infectious illness and it spreads
rapidly. The fungus corresponds to the Erysiphaceae group,
Fig. 10: Leaf Curl category Ascomycetes Erysiphales order.
Figure 10 represents Leaf Curl. Curly Mulberry leaves VII. RESULT AND ANALYSIS
showed the existence of Taeniothrips, assaulting leaves of
mulberry. The assault on mulberry leaves by these The disease classification performance analysis of the
taeniothrips is reported from India and Sri Lanka. They KNN classifier on leaf dataset. The leaf disease detection
injure epidermal tissue and significantly impacted leaves and classification result has been simulated on MATLAB.
show early maturity, humidity depletion, decrease in the The segmentation output of the KNN classifier has been
rearing of crude protein silkworms. compared with the manually segmented result of the leaf
image.

 Accuracy: Accuracy indicates the number of correctly

predicted data points out of all the data points.

 Accuracy = (TP+TN) / (TP+FP+FN+TN)

 Specificity: Specificity is the ratio between how much is

correctly classified as negative to how much is the actual
negative.

 Specificity = TN / (TN+FP)

 Sensitivity: Sensitivity is a measure of how well a

machine learning model can detect positive instances.
Fig 11: Leaf Rust  Sensitivity = TP/(TP+FN)
Figure 11 represents Leaf rust. Ceroteliumfici triggers
herb rust infection. Under the genus Irnperfect fungi, the
pathogen suborder Uredinales and family Uredinaceae. It is
a common disease which occurs during the winter season
(November-February). Developed leaves seem to be more
susceptible to this illness.

Table 2: Result Comparison

IJISRT23APR1763 www.ijisrt.com 1730

Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
REFERENCES

[1.] 24th International Conference on Information

Technology (IT) Zabljak, 18 – 22 February 2020
Image-Based Plant Disease Detection: A
Comparison of Deep Learning and Classical
Machine Learning Algorithms DraškoRadovanović
1, Slobodan Đukanović 1, 2, Member, IEEE.
[2.] International Conference on Communication and
Signal Processing, April 4-6, 2019, India Machine
Learning for Plant Leaf Disease Detection and
Classification – A Review L. Sherly Puspha
Annabel, Member, IEEE, T. Annapoorani and P.
Deepalakshmi.
[3.] 2018 International Conference on Computer,
Control, Informatics and its Applications, Machine
Learning-based for Automatic Detection of Corn-
Plant Diseases Using Image Processing
EHD+KNN EHD+SVM EHD+DT BudiariantoSuryoKusumo, Ana Heryana, Oka
Graph 2: Result comparison Mahendra, and Hilman F. Pardede, IEEE
[4.] 2019 International Conference 7-9 February, 2019, A
From the above result it is observed that EHD with
Color and Texture Based Approach for the Detection
KNN gives 97.50% accuracy, 0.9778 specificity. It is
and Classification of Plant Leaf Disease Using KNN
observed that for this combination of algorithm we have
Classifier Eftekhar Hossain1, Md. Farhad Hossain2
obtained better precision which indicates that the ratio of
and Mohammad Anisur Rahaman3, IEEE
the correctly positively predicted data by our program to all
[5.] Proceedings of the Second International Conference
positive data is accurate.
on Intelligent Computing and Control Systems
EHD with SVM gives 75.56% accuracy, 0.5769 (ICICCS 2018), Diseases Detection of Cotton Leaf
specificity. It is observed that for this combination of Spot using Image Processing and SVM Classifier
algorithm we have obtained better precision which indicates Namrata R.
that the ratio of the correctly positively predicted data by
our program to all positive data is accurate.

EHD with DT gives 78.50% accuracy, 0.9640

specificity. It is observed that for this combination of
algorithm we have obtained better specificity which
indicates the true negative rate which means that the
proportion of actual negatives, which got predicted as the
negative.
VIII. CONCLUSIONS

In this paper, we briefly explained the Mulberry leaf

disease detection. We were able to detect the disease using
SVM and KNN algorithm. After the dataset is required and
given as input, preprocessing takes place and will detect the
mulberry leaf disease and their solution automatically once
the disease is identified.Our experimental results indicate
that EHD with KNN gives the best result with 97.50% of
accuracy, 0.9778 of specificity.

ACKNOWLEDGEMENT

We are thankful to the resourceful guidance, timely

assistance and graceful gesture of our guide Mrs. Nagaveni
B Nimbal, Assistant Professor Department of Computer
Science Engineering who has helped us in every aspect of
our project work.

IJISRT23APR1763 www.ijisrt.com 1731