2024 10th International Conference on Communication and Signal Processing (ICCSP)

An AI based System for Smart Detection of

Leaf Diseases using Machine Learning
Techniques
S. Sharmila Dr. V. Jeyalakshmi
2024 10th International Conference on Communication and Signal Processing (ICCSP) | 979-8-3503-5306-8/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICCSP60870.2024.10543748

College of Engineering, Department of Electronics and Communications

Anna University Engineering, College of Engineering, Anna
Chennai- 600025, India University,Chennai- 600025, India
Email Id: [email protected] Email Id: [email protected]

Abstract-- An Internet of things is a system of MATLAB is a truncation of “matrix laboratory” may be a

connecting various smart devices which are able to multi-paradigm programming dialect and numeric computing
gather and transfer data over the network. The environment concocted by Math works. The MATLAB
expansion of Internet of Things can be used in smart application is built around the MATLAB programming
agriculture to boost the quality and quantity of the crop. dialect common utilization of this program, it makes utilize
Leaf disease prediction is an essential step used in of Advanced Picture Preparing strategies.
agriculture for screening early diagnosis of disease and Computerised picture preparing employments
provide pesticide at early stage, it is an application used computer calculations to form, prepare, communicate and
for indoor plantation or terrace gardening users. By computerised pictures. It changes over signals from a
using AI and IOT technology, we can able to create a picture sensor into computerised pictures, which makes a
digital system for agriculture which can help the users to difference to upgrade clarity, commotion evacuation and
take right decision on correct situation in advance. Here, other highlights of a picture. The input of that framework
we proposed a system which helps to recognize the leaf may be an advanced picture and the framework prepares
disease. In this work, the Mobile camera interface is that picture using efficient algorithms, and gives a picture as
used to cover the effective leaf image. Quality of the a yield. Adobe Photoshop could be a best case. It is broadly
image is improved by Digital Image processing utilised application for handling advanced pictures.
techniques using MATLAB. The temperature and soil A picture has been captured by a camera and has
moisture sensor are placed at various location, which been sent to a computerised framework to perform noise
helps to identify the dryness of leaf and for monitoring expulsion in all measurements, in such a way that the
the soil quality respectively. ARDUINO is a controller quality of the picture will be improved. It gives clear
device which helps to control those sensors. This system pictures. Advanced Picture Handling do picture
can robotically sense leaf disease at early stage as well as improvement to recall the information through pictures. It is
it provides required pesticide to the user through Mobile utilised broadly all over in numerous areas. It decreases the
App. This project is to engage AI and IOT technology in complexity of advanced picture handling. It is basically
agriculture for more accuracy in detection of leaf disease utilised for Picture reformatting (multi-view).Then, it is
at early stage. utilised for Quick picture capacity and recovery.

Index Terms--Internet of Things, Smart Agriculture, II. EXSISTING SYSTEM

Leaf Disease, Image Processing, Machine Learning
An IOT structure was created for agricultural field
I. INTRODUCTION is utilised to distinguish leaf infection utilising MDFC-
The Internet of Things describes a network of phys RESNET. It combines IOT procedures with deep learning to
ical device with sensors, software, and other technologies us develop IOT framework for crop infection recognizable proof.
ed to connect and exchange information with other devices It consequently recognizes crop infection utilising residual
on the Internet. IoT is an important platform where devices neural network model in deep learning and send diagnostic
can collect and exchange information with each other by results to farmer. The video camera is set up in crop field and
connecting to the decoder can get information than extricate the crop picture
internet. It allows devices to interact like humans and learn from decoded data. Deep learning demonstrate gets crop
fr om each other's experiences. Soil moisture sensors help picture from decoder, compare status with trained model and
estimate soil moisture levels. Temperature sensors are a send result to message center.
simple component that measures heat or cold and converts it The message center gets result and oversees result
into readable data. IOT is essentially the man to man, man to in frame of message line. Show result to computer or smart
machine or machine to machine handle, which makes phone. The Components on future study about can focus on to
human works exceptionally comfortable. IOT consist of Move forward exactness of picture quality, may utilise
various Artificial Intelligent process which process the calibration chart, may utilise different sensors like
human brain can be readable by machines using IOT temperature sensor, climate sensor, soil dampness sensor and
techniques. This paper work represents the early detection of may give treatment proposal for the leaf infection individually
leaf disease for the user of indoor plantation and terrace The downsides of existing framework are measured such as,
gardening. This paper has achieved its goal to detect and • It increments the complexity of design whereas
recognize various types of plants leaf disease using Digital utilizing RESNET.
Image Processing Techniques. It can automatically detect • Skip connection is essentially the personality
leaf disease at initial stage and also it provides required mapping the input from past layer is included
pesticide to the user, on Mobile App.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:19:37 UTC from IEEE Xplore. Restrictions apply.
979-8-3503-5306-8/24/$31.00 ©2024 IEEE 1505
 straightforwardly to the yield of other layer is more technology algorithm for leaf disease analysis and classification.
complex in RESNET process. Image segmentation is one of the important tasks of using
• It does not apply Information Pre-processing genetics to identify leaf diseases. In the future, artificial neural
Methods which will not improve the picture quality networks, Bayesian classifiers, fuzzy logic and hybrid
where though utilizing this methods we are able to algorithms can also be used to improve the recognition of the
alter the picture layout which makes a progress great classification process.
result. The system developed by authors Apeksha Thorat,
Sangeeta Kumari, Nandhakishor [3] provides leaf disease
III. PROPOSED METHOD detection, server-based remote monitoring control,
In this study, we propose a method that facilitates temperature and humidity measurement, etc. It can contain
the detection of leaf diseases using digital imaging many dimensions. Many sensors are placed in different
technology. It used MATLAB software to carry out the locations of the farm and we use the Raspberry pi (RPI)
analysis process, and with the emergence of the Internet of controller to control all these sensors. Leaf diseases can be
Things, it should create a digital farm. Here the mobile detected using cameras connected to the RPI.
camera acts as a connection to capture the image of the leaf. Authors A.m.Ehilazhahi and p.t.v.Bhuvaneshwari
We can improve image quality by using digital image [4] developed a good soil moisture sensor using WSN
processing in MATLAB. This result helps us diagnose leaf technology. It enables the use of humidity monitoring system
diseases. Temperature and humidity sensors are also used to based on wireless sensor networks. An exponentially weighted
detect dryness of the soil and monitor its quality. moving average detection algorithm is used to generate the
conditions. This function can be extended by considering an
These are controlled using a device called an
additional sensor module.
Arduino controller. The system provides users with the Author Sachin D.Khirade [5] proposed a method for
pesticides they need via a mobile application using Wi- diagnosing diseases using machine images. He discussed
Fi. In this system, temperature and humidity values can methods for detecting plant diseases using leaf images. It
be viewed as graphs by the user of the mobile shows some segmentation and feature extraction algorithms for
application. In smart farming, mobile applications plant disease detection. Many strategies for the distribution of
interact with Arduino-based sensor systems to provide plant diseases have been discussed, some methods of removal
users with real-time updates, including disease alerts and and distribution have been discussed, the characteristics of leaf
medication recommendations. This is how the mobile disease have been eliminated and the disease has been
application communicates with the system. Arduino- classified.
Author Nattapol kaewmard and saiyan saiyod [6]
based sensor systems collect data from a variety of
developed a system to isolate plant diseases. Collect and
sensors to monitor the environment, soil health, and manage sensor data using smart phones and wireless
crops. The data is processed by the Arduino networks. It uses TinyOs or Microcontroller (MCU) to send
microcontroller and sent to the mobile application using data and Zigbee ( Xbee) for radio transmission of data signal. It
Wi-Fi and Bluetooth. The advantages of the proposed works with solar cells and batteries. Sending and receiving
method are: voltage as a cable for data coming from the sensor device
• This system facilitates the distribution of fast and communicating on the Atmel AUR microcontroller (Arduino).
good images, and image editing can be done by Author G.Sahitya, Dr.N.Balaji, Dr.C.D. Naidu [7]
digital image processing techniques. represents this system using Arduino as the main component to
• It provides the user with pesticides to prevent create a sensor network. Each node contains set of sensors
related diseases. Using threshold-based image connected to Arduino and Zigbee. The data transmitted here is
segmentation technology, it can be easily segmented measured and sent to the central device, fzigbee. It collects the
by adjusting the threshold and then using the SVM data and stores it in the file assigned to the PAN ID. Routers
classification method to classify the disease on collect data from base stations. Wireless sensor networks can
specific leaves. be divided into structured and unstructured. Distributed sensor
• The SVM algorithm is designed for non-disruptive networks help increase the chances of data collection.
data such as images, text and trees. Once the objects Author Deepti Sehrawat and Nasib Singh Gill [8]
are segmented, the image can be easily classified represent various sensors and sensor-based IoT applications. It
using an SVM classifier. includes sensors embedded in the smart environment, such as
• Leaf Disease Report is very important for users proximity sensors, temperature, humidity, chemicals, location,
planning indoor or patio gardening. This system is motion and pressure sensors. It uses RFID (Radio Frequency
very useful in analyzing the image and delivering Identification Device) to communicate on the network. RFID
the pesticide early. tags integrate RFID tags and RFID readers and read RFID tags
via a wireless interface. Monitor, analyze, Collect and process
IV. RELATED WORK data to detect physical problems.
Authors Vipa Thananant and Thanakorn khunkh
Authors Awf Abdulrahman and Serkan Varol [1] [9] presented two benchmarks and Android applications. The
proposed the use of MATLAB for image segmentation in device measures heart rate and temperature. It displays the
image processing. It involves splitting the blurry image into readings on the LCD screen while instantly sending the
elements and is useful by dividing by color. It provides two readings to the Thingspeak IOT platform via Wi-Fi. The
image segmentation functions: edge and threshold function. In offline system uses an Arduino-UNO temperature sensor is
future research, deep learning concepts can be used in the played on the computer. The kit consists of an Arduino ESP32,
image segmentation process. a heart rate sensor and a body temperature sensor. The data is
Authors Vijai Singh, A.k.Misra [2] used image stored in the Firebase real-time database. Web views are web
segmentation and computational software to generate leaf pages in Android applications. This web page shows the
diseases. It offers robotic analysis and an image segmentation Thingspeak plot created for measuring data sensors.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:19:37 UTC from IEEE Xplore. Restrictions apply.
1506
 V. SYSTEM DESIGN a(x,y)=1 if b(x,y)>TH and
a(x,y)=0 if b(x,y) <= TH
The System Architecture briefly defines about the So segmented image has two pixels i.e. one value is 1 and ot
proposed system work, which helps to make early detection her value is 0. If the threshold TH is constant in processing
of leaf disease for indoor plantation or terrace gardening over the entire image region, it is said to be Global
users. Thresholding. If TH varies over the image region, we say it
is Variable Thresholding. The advantage of the method is
simple calculation and fast processing. Especially when the
contrast between the target and the background is high, the
segmentation effect can be achieved.
a = imread(image file);
b=rgb2gray(a);
th= input (threshold value);
for i=1:r
for j=1:c
if (b(i,j)> th)
out(i,j)=1
else
out(i,j)=0;
While fixing the threshold value as 170 on affected leaf
image, it gives the accurate result on the image. Such as

Fig. 4.1. System Architecture

The input of above system are the leaf image,
temperature and soil moisture’s sensor data, which are send
through WIFI, then the image is processed in MATLAB,
which consist of various processing steps like Image
acquisition, Image processing, Image segmentation and
classification process. The image processed in MATLAB is
compared with database where disease classification images
are already stored and then the result is predicted. And also
the system provides Pesticide for the detected leaf disease.
Simultaneously, the temperature and soil moisture sensors
are deployed at various locations to identify the dryness of
leaf and for monitoring the soil quality respectively. These Fig. 6.1 Image Segmentation using Threshold Algorithm
results will be conveyed to the user of Mobile App.
Image Acquision: The image acquisition is the function of Multi Class - SVM Classification Algorithm
gathering data. Image Processing: This process is for Support vector machine is a supervised machine
converting the image format ie.RGB2Grayscale. Image learning algorithm that can be used in classification and
Segmentation: It is the work of segmenting the image into regression problems. It helps to determine the best possible
various types of classes and this use Threshold Based outcome. This algorithm represents the determination of the
Algorithm for Image segmentation processing. Image boundary between two groups, called hyperplanes, to
Classification: For classification of disease it uses Multi separate them. SVM is also used for object detection and
class SVM classifier. image classification. A multi-class Support vector machine
(SVM) algorithm is an extension of the Binary SVM
VI. WORKING SYSTEM algorithm and is used to solve classification problems over
two groups. SVM is a supervised learning model that
Threshold Based Algorithm analyzes data for classification analysis. When used for leaf
Threshold based algorithm is the simplest disease detection, a multi-class SVM classifier (model) is
algorithm for performing image segmentation process can trained using features obtained from the training process.
be regarded as the process of dividing an image into a Test the trained SVM classifier, test the model, and deploy
foreground from background. This image tool is a type of the trained SVM classifier to leaf disease detection in real
image segmentation that separates objects by converting applications.
grayscale images into binary images. Binary images are • Load a dataset from a source.
images in which pixels have only two values (such as 0 • Split the dataset into "training" and "test" data.
and 1), so by dividing the grayscale image into two regions • Train the SVM classifier in the data.
as thresholds, only one of them needs to store the pixel • Using the classifiers above, predict the labels of the
used. Therefore, pixels in the output image whose intensity test data.
value is greater than the threshold will be considered white • Measure accuracy and see classification.
or 1, while other pixels will be considered black or 0.
In other words, if we have threshold TH, the segmented ima
ge g(x,y) is calculated as

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:19:37 UTC from IEEE Xplore. Restrictions apply.
1507
 of the soil, deviations from optimum conditions can be
GroupTrain={'1' '1' '1' '1' '1' '1' '2' '2' '2' '2' '3' '3' '3' '4' '4' '4' detected, indicating the presence of dangerous diseases.
'4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '4' '5' '5' '5' '5' '5'
'5' '6' '6' '6' '6' '6' '6' '7' '7' '7' '7' '7'}

TestSet=QF;
c = cell(1,7);
Y=GroupTrain;
classes = unique(Y);
rng(1);
for j=1:numel(classes)
indx = strcmp(Y',classes(j));
svmStruct=svmtrain(DF,indx,'Kernel_func
tion','linear','BoxConstraint',1);
end
for j=1:numel(classes)
score= svmclassify(svmStruct,xGrid);
zscore(:,j) = score(:,1);
end
C. Analysis of Diseases on Various Leaves
VII. OUTPUT
The system provides users with fungicides and insecticides
A. Image Classification with Pesticide
to prevent diseases. Using threshold-based image
segmentation technology, the input image is easily
Classification of images is performed on stored
segmented by setting the threshold and then SVM
training data. It is based on the multi-class SVM classifier in
classification is used to classify the disease on specific
MAT LAB. It is a SVM application designed to solve
leaves. Once the objects are segmented, the image can be
classification problems over two groups.
easily classified using an SVM classifier. A foliar disease
report is then generated with soil, temperature, and moisture
results, as well as the name of the leaf disease and the
pesticide name. The report will be displayed on the Farmers
Mobile app.

B. Soil Moisture and Temperature Analysis

A soil moisture sensor is used to measure the moisture
content in the soil. It provides valuable information about
the soil's water content, which helps prevent overwatering
or underwatering plants, leading to more efficient water
usage and healthier plants. The temperature and humidity
sensor (DHT11) is used to monitor the temperature of the
place where indoor plantation or a terrace garden is done.
Then controller transmits the digital value of temperature
for easy analysis of temperature and humidity.

Temperature and soil moisture sensors are often chosen

for smart agriculture because they both play an important
role in monitoring environmental and soil health, which
is related to crop growth and diseases. Temperature
sensors are often placed at different heights in the crop
canopy to detect temperature variations and changes.
They can be placed in the shade and exposed to
microclimate conditions. Monitoring soil moisture can
help you improve your watering schedule and ensure
plants receive adequate water without being overwatered
or flooded. Soil moisture sensors are placed at
different root zone depths to detect changes in soil
moisture. Sensors can be installed at different locations
in the field to account for soil heterogeneity and enable
monitoring.
Authorized Byusemonitoring
licensed limited to: VITthe temperature
University- and moisture
Chennai Campus. Downloaded on November 07,2024 at 07:19:37 UTC from IEEE Xplore. Restrictions apply.
1508
 VIII. CONCLUSION using Convolution Neural Network” 2019-2020 Published
In conclusion, the scope of this work says that earlier by Elsevier.
detection of leaf disease can be performed and the same will be [12]. AwfAbdulrahman and Serkanvarol, ‘A Review of
conveyed with the respective pesticides to the user through Image Segmentation Using MATLAB Environment’-
mobile app with more accuracy. In addition to this, this model [June,2020].
can also able to convey about the temperature, soil and [13] DeeptiSehrawat and Nasib Singh Gill.’ Smart sensors:
humidity level of the specific area. It is also useful for indoor Analysis of different types of IOT sensor’-[2018]
plantation and terrace gardening users. This paper has achieved [14].ApekshaThorat, SangeetaKumari, Nandhakishor,’An
its goal to detect and recognize various types of plants leaf IOT Based Smart Solution for Leaf Disease Detection’-
disease using Digital Image Processing Techniques and with [2017].
the classification algorithm. Creating digital farming systems [15].A.m.Ehilazhahi and p.t.v.Bhuvaneshwari.’IOT Enabled
with new IoT technology and artificial intelligence should help plant Soil moisture Monitoring using Wireless sensor
users or farmers make decisions on a case-by-case basis. Networks’-[2017].
The future enhancement of this work may be [16].Vijai Singh, A.k.Misra.‘Detection of Plant Leaf
extended to large amount of various crops with the help of Diseases using Image Segmentation and Soft computing
Raspberry for controlling and storing data, which helps to Techniques’.-[2016].
upgrade this system. With more accuracy of this proposed [17].Sachin D.Khirade.’Plant Disease Detection using
model the user can easily detect the disease at early stage and Image Processing’-[2016].
provide pesticide respectively. [18].Nattapolkaewmardandsaiyansaiyod.’Sensor data
collection and control on vegetable crop using smart phone
REFERENCES and wireless sensor Network for smart farm’-[2016].
[19].G.Sahitya,Dr.N.Balaji,Dr.C.D.Naidu.’WirelesssensorN
etwork for Smart Agriculture’-[2016].
[1] Haseeb Nazki, Sook Yoon, Alvaro Fuentes, Dong Sun [20].YummingDu,JingTian, LinaGai .’Digital Image
Park (2020) “Unsupervised image translation using Processing Teaching Auxiliary System Based On
adversarial networks for improved plant disease MATLAB’-[2016]
recognition” Published by Elsevier B.V.
[2] Nilay Ganatra, Atul Patel (2020) “Performance Analysis
of Fine-Tuned Convolutional Neural Network Models for
Plant Disease Classification” Published by
InternationalJournal of Control and Automation Vol. 13,
No. 3,pp. 293- 305
[3] Battala Lakshmi Thushara,T. Mahaboob
Rasool(2020) “Analysis of Plant Diseases Using
Expectation Maximization Detection with BP-ANN
Classification” Volume XIII, Issue VIII, AUGUST
2020
[4] Chaitanya and Yasudha (2020) “Image Based Plant
Disease Detection Using Convolution Neural Networks
Algorithm”, Volume 5, Issue 5, May – 2020, Published by
International Journal of Innovative Science and Research
Technology ISSN No:-2456-2165
[5]Shanwen Zhang, Subing Zhang, Chuanlei Zhang,
Xianfeng Wang, Yun Shi (2019) “Cucumber leaf disease
identification with global pooling dilated convolutional
neural network” 2019 Published by Elsevier B.V.
[6] Uday Pratap Singh, Siddharth Singh Chouhan, Sukirty
Jain, And Sanjeev Jain (2019) “Multilayer Convolution
Neural Network for the Classification of Mango Leaves
Infected by Anthracnose Disease” 2019
[7] Vijai Singh (2019) “Sunflower leaf diseases detection
using image segmentation based on particle swarm
optimization” 2019 Published by Elsevier.
[8]Sumita Mishra, Rishabh Sachan, Diksha Rajpal (2019)
“Deep Convolutional Neural Network based Detection
System for Real-time Corn Plant Disease Recognition” 2020
Published by Elsevier B.V.
[9]Utpal Barman, RidipDevChoudhury, DigantoSahu.’
Comparison of convolution neural networks for smartphone
image based real time classification of citrus leaf disease’-
[2020]
[10]Parul Sharma, Yash Paul Singh Berwal , Wiqas Ghai
(2019) “Performance analysis of deep learning CNN models
for disease detection in plants using image segmentation”
open access 2019 Published by Elsevier B.V.
[11]Mohit Agarwal, Abhishek Singh, Siddhartha Arjaria,
Amit Sinha, Suneet Gupta (2019) “Leaf Disease Detection
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on November 07,2024 at 07:19:37 UTC from IEEE Xplore. Restrictions apply.
1509