Optimization of Disease Detection System for Improved Arecanut

Cultivation by Machine Learning

Karthikeyan M and Vijaychitra S

Kongu Engineering College, Erode, India

Keywords: Areca Nut, Optimization, Deep Learning, Machine Learning.

Abstract: Agricultural growth is crucial for ensuring a consistent food supply for ordinary people. To optimize disease

detection methods and improve crop quality, data from various published research works is collected and

analysed. These efforts aim to protect plants from diseases, enhancing both agricultural productivity and the

country's economic contributions. This review compares machine learning and deep learning techniques for

identifying and categorizing plant diseases. Images of arecanut leaves, trunk, stem and root were used as input

for the research. The study focuses on disease detection systems implemented using various algorithms and

compares their accuracy based on findings from published research. The results indicate that Convolutional

Neural Networks (CNNs) consistently achieve better accuracy than traditional machine learning methods.

Future research can explore advanced deep learning techniques to achieve even higher accuracy in plant

disease detection.

1 INTRODUCTION

Agricultural land expansion has stabilized or even

declined in many regions, despite increased

production. Advancement in farming technology and

efficiency gains also play a significant role, as higher

crop yields now allow the same amount of land to

produce more. Areca nut farming is a significant

agricultural activity in India. Most concentrated in the

southern and northeastern states. Karnataka gave 79%

of India's total arecanut production. The total annual

production of areca nuts in India is approximately

1.37 million tons, grown on around 0.52 million

hectares of land.

Convolutional Neural Networks (CNNs), in

particular, constitute deep learning models that have

gained popularity due to their exceptional

performance in image-based illness categorization.

Research has used architectures including Res Net,

VGG Net, and Mobile Net to accurately diagnose

illnesses like bud rot, yellow leaf spot, and stem

bleeding. The effectiveness of these models has been

further improved by preprocessing methods such

feature extraction, picture augmentation, and

normalization.

Good agricultural practices, effective post-harvest

management, and modern technology would help us

to maintain the quality of areca nut. Fruit rot disease

and other pests can be sustainably controlled using

Integrated Pest Management (IPM). Plant health level

is based on management of farmer practices;

detection of disease has to be done before the

spreading stage of the diseases from one plant to

another. Tomatoes are among the most widely

cultivated crops, with annual production exceeding

180 million metric tons. The export market is about

$10 billion. Tomato cultivation per-hectare yields

range between 50 to 100 tons, and protecting crops

from diseases can ensure consistent productivity and

profitability. Tomato disease detection plays a critical

role in enhancing yield quality.

By gathering vegetation indicators and creating

3D models to measure disease severity, emerging

technologies—such as UAV-based multispectral

imaging—have shown promise for real-time disease

monitoring. These advancements demonstrate how

AI, IoT, and precision agriculture are revolutionizing

the management of arecanut diseases. These tools

give farmers actionable knowledge by facilitating

early detection, categorization, and prediction

insights, which encourage sustainable crop

management and lower financial losses.

Over 125 million metric tons are produced in

India each year, making up about 40% of the

country's total food grain production. The high yield

and export surplus play a crucial role in ensuring

economic stability for millions of farmers while

M, K. and S, V.

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning.

DOI: 10.5220/0013588100004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 2, pages 123-134

ISBN: 978-989-758-763-4

123

strengthening India’s position in global agriculture.

Effective detection not only preserves the quality of

rice but also enhances productivity. To that end, we

have come up with an automated system using

machine learning techniques, a system that will

contribute in country’s agricultural development by

automatically identifying and classifying diseases

from the images of rice leaves. Where efficiency and

accuracy are crucial. Accurately classifying weeds

and crops is a major problem in modern agriculture

since it's necessary for efficient resource allocation

and management. Conventional techniques for

identifying weeds and crops frequently rely on labor-

intensive, time-consuming manual examination

procedures that are prone to inaccuracy.

2 RESEARCH OBJECTIVES

Objectives will help us to sharpen the ideas of

research work. The following objectives are utilized

to study the disease detection on plants for healthy

farming methods:

• To study the early detection of different

diseases in Arecanuts.

• To analyse the methodologies of

Machine learning and Deep learning to

optimize the productivity

2.1 Motivation of study

Since agricultural practices play a major role in

India’s economic considerations, it needs to bring our

country as one which performs well in agricultural

production. Prevention and early detection of possible

diseases will reduce crop death and further to improve

the quality of crop as well as cultivation.

2.2 Role of Artificial Intelligence in

Agriculture

The following are the major roles of Artificial

Intelligence and deep learning technique:

• A form of deep learning technique called

Convolutional Neural Networks (CNNs)

has proven incredibly successful in

automating agricultural activities

including disease detection, crop

classification and yield prediction.

• It has been found that popular designs like

as Mask R-CNN, U-Net, Alex Net,

VGG16, and InceptionV3 perform

exceptionally well in agricultural

applications, achieving exceptional

accuracy and durability even under

challenging field conditions.

2.3 Integration of IoT and Machine

Learning

The following basic ideas about IoT and machine

learning:

• By offering real-time insights, IoT-

enabled data collecting using devices like

Arduino and sensors for temperature,

humidity, and soil conditions has

improved precision farming.

• When combined with sensor data,

Random Forest Classifiers (RFC) and

LightGBM models have demonstrated

efficacy in forecasting crop performance

and illnesses, hence facilitating proactive

decision-making.

2.4 Significance in Arecanut

Cultivation

The following significance are:

• A major crop in many areas, Arecanuts

confront difficulties such fruit rot and

infections like Yellow Leaf Disease

(YLD), which reduce productivity.

Cutting-edge technologies such as CNN

models, IoT-integrated systems, and

UAV-based multispectral imaging offer

encouraging options for early

identification and management.

• High accuracy in arecanut detection and

grading has been demonstrated using

segmentation and classification models

such as Mask R-CNN and hybrid CNN-

SVM systems.

3 LITERATURE REVIEW

For the KR Sharathkumar. et al., (SharathKumar, and,

Mohan, 2020) suggested a disease prediction

detection system to track soil moisture (%), wind flow

(m/s), rainfall (cm), temperature (°C), and humidity

(%) in order to forecast the emergence of various

illnesses. Environmental variables caused certain

illnesses to manifest as expected. The illness impact

was predicted using machine learning techniques

such as Support Vector Machine Regression (SVMR)

and Random Forest Classifier (RFC). Environmental

data is gathered using a soil humidity sensor and a

DHT-11. Certain illnesses (Mahali, fruit rot,

INCOFT 2025 - International Conference on Futuristic Technology

124

Koleroga) can only be predicted by research that is

restricted to certain places.

S. Anupama Kumar. et al. (Kumar, Anupama, et

al., 2024) used Convolutional Neural Network

techniques, concentrated on identifying the three

main illnesses that harm areca trees: Yellow Leaf

Spot, Stem Bleeding, and Mahali Disease (Koleroga).

Various CNN algorithms, including MobiNet,

ResNet, and VGG, are used to identify these

disorders. Three separate metrics—the total

parameters, trainable parameters, and nontrainable

parameters—are used to compare the models based

on three network architectures and the plant disease

that is tracked. It was shown that accuracy rose as the

number of non-trainable parameters decreased. The

accuracy of the ResNet architecture is 79%, MobiNet

is 86%, and VGGNet is 96% based on data gathered

over 50 epochs. VGGNet outperformed the other two

networks in terms of accuracy and overall

performance.

Namra Mahveen. et al., (Mahveen, Namra, et al. ,

2023) proposed a system which was based on

Convolutional Neural Networks (CNN) for detecting

diseases in arecanut crops, specifically targeting

leaves, trunk, and fruit. CNN, a potent deep learning

algorithm designed specifically for image analysis,

analyzes input photos by giving different visual

aspects learnable weights and biases. The CNN

algorithm is able to differentiate between healthy and

unhealthy arecanut plants by identifying patterns in

these photos. Datasets from published literature that

included a wide range of photos showing both healthy

and ill plant samples were used in the CNN model's

construction and assessment. In addition to accurately

identifying diseases, the system sought to help

farmers maintain healthy crops and increase output by

offering helpful advice and recommendations. Pre-

processing, feature extraction, model training, and

classification were all phases of the system's

operation. The accuracy percentage attained by the

model was almost 88%.

Ms. Shwetha Kamath. et al., (Kamath, Shwetha,

et al. , 2023)concentrated on machine learning

methods to analyze different soil factors and apply the

ID3 algorithm to identify illnesses early. The

suggested technique detected the presence of the

fungus that causes YLD in soil samples taken from

arecanut crops. In order to predict crop health, this

study used soil characteristics, such as

micronutrients, in addition to pH, nitrogen (N),

phosphorus (P), and potassium (K) levels. The ID3

technique was used to build a decision tree model

based on user-provided inputs. The results were

binary, meaning that the existence of a disease was

either "Yes" or "No."

Tejaswi R. et al., (Tejaswi, Mysoremath, et al. ,

2024) used the ResNet model to identify a number of

illnesses that impact arecanut crops, including yellow

leaf disease, fruit rot, and foot rot. RGB photos were

processed using sophisticated methods such as global

average pooling. Learnable weights are assigned by

the model to differentiate between healthy and

unhealthy plants. With a 97.5% accuracy rate on a

self-gathered dataset, this system demonstrated its

ability to help farmers detect diseases early and

minimize crop losses. The study also demonstrates

how well ResNet handles vanishing gradient

problems, laying the groundwork for more extensive

precision agricultural applications.

Dr.P.Sreenivasulu. et al., (Sreenivasulu, Lakshmi,

et al. , 2024) applied a CNN-based techniques to

approach arecanut diseases and achieved an accuracy

of 94.8% across multiple conditions like fruit rot and

yellow leaf disease. It involved robust preprocessing

methods, such as normalization and augmentation,

and employed advanced architectures like ResNet.

The model demonstrated the feasibility of leveraging

image-based deep learning techniques to assist

farmers with accurate diagnosis and timely

intervention, marking a significant step toward

precision agriculture.

Rajashree Krishna. Et al., (Krishna, Rajashree, et

al. , 2022) focused weather parameters such as

rainfall, humidity, and temperature to predict fruit rot

disease in areca nut. A unique dataset combining

weather data and field surveys was created and

validated using machine learning models like

Random Forest Regression, achieving the best

performance with a mean absolute error of 0.9. This

approach emphasized proactive disease management

and supports farmers in mitigating losses through

early prediction.

Dhanuja K C. et al., (Dhanuja, and, Kumar, 2020)

focused on the application of machine vision for

detecting and classifying diseases in areca nuts,

utilized advanced image processing techniques such

as Gabor wavelets and GLCM. It has emphasized the

role of precision agriculture in improving crop

management and demonstrated the effectiveness of

texture-based grading systems. By employed dataset

of 700 images spanning various disease classes, the

research achieved a classification accuracy exceeding

91%, showcasing the efficiency of convolutional

models and neural network classifiers. The study

further suggested that future work could focus on

optimizing automated systems for real-time

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning

125

applications to improve scalability and usability in

agricultural operations.

Kuo-Yi Huang. (Huang, Kuo-Yi, et al. , 2012)

applied machine vision techniques to classify areca

nuts. He used geometric and colour features along

with a Detection Line (DL) algorithm for defect

identification. It has employed a Back-Propagation

Neural Network (BPNN) to sort areca nuts into

quality categories with an accuracy of 90.9%.

Combined image smoothing, feature extraction, and

neural network classification, the study demonstrated

an effective and automated solution to assess nut

quality. This innovative method reduced manual

effort and enhances consistency in quality evaluation,

making it a significant contribution to agricultural

technology.

Ajit Hegde. et al., (Hegde, Shetty, et al. , 2023)

focused on convolutional neural network (CNN)-

based system for detecting and classifying various

diseases affecting areca palms. They classified

diseases such as fruit rot, bud rot, and yellow leaf

disease, achieved accuracy of 94.8%. It emphasized

the practical benefits of deploying such models in

agricultural settings in reduced manual inspections

and improved crop management strategies. It also

highlighted the potential of AI in advancing

sustainable and data-driven crop management

practices.

Anilkumar M G. et al., (Anilkumar,

Karibasaveshwara, et al. , 2021) utilized

convolutional neural networks (CNNs) for the early

detection of diseases affecting arecanut, such as

Mahali, stem bleeding, and yellow leaf spot. A dataset

of 620 images, split into 80% for training and 20% for

testing, formed the basis of the model and achieved

an accuracy of 88.46%. Pre-processed steps like

image resizing and augmentation, enhanced the

feature extraction and improved classification

reliability. The research highlights the practical utility

of AI in agricultural disease management, offered

both detection and remedy suggestions. This

approach presented a significant step toward

promoting smart farming and sustainable crop

management. The findings emphasized the role of

machine learning in minimized yield loss.

Shuhan Lei. et al., (Lei, Shuhan, et al. , 2021)

applied UAV-based multispectral imaging and

machine learning techniques to monitor yellow leaf

disease in arecanut. Five vegetation indices (e.g.,

NDVI, LCI) were extracted from high-resolution

UAV images, enabled the quantification of disease

severity. Machine learning models, including BPNN

and SVM, achieved the highest classification

accuracies of 86.57% and 86.30%, respectively. The

research further correlated the severity of the disease

with the living vegetation volume (LVV) using UAV-

constructed 3D models. This quantitative approach

advances traditional qualitative disease assessments,

offering precision in disease monitoring.

Shabari Shedthi Billadi. et al., (Billadi, Shedthi, et

al. , 2023) applied machine learning techniques to

find unique solution for good and defective arecanuts

based on their color, texture and density value.

Machine vision method is used to grade the arecanut.

Various segmentation algorithms are available, and in

this work, simple k-means clustering is applied to

distinguish the arecanut from the background in the

image. Classifiers like logistic regression, k-NN,

naive Bayes classifiers, support vector machine

(SVM), and ANN are used to identify the healthy

arecanut from the spoiled one. The results indicate

that among the classifiers tested, ANN demonstrates

superior performance. Incorporating the newly

derived density feature, the arecanut grading system

achieves an overall accuracy of 98.8%. Experimental

results demonstrated that the machine vision system,

enhanced with density features, delivers a high

success rate. Density features play a crucial role in

arecanut classification.

Pallavi P. et al., proposed a new system using

Convolutional Neural Networks and detected the

diseases of arecanut in leaves and its trunk. The

proposed method consists of three phases training

phase, testing phase and GUI phase. Developed

system recognized the features and anticipated crop

diseases using data mining and data science

techniques. Algorithm detected arecanut disease and

provided solutions for the detected disease. The

overall accuracy of the system is estimated to be

81.35%.

Suresha M. et al., (Suresha, Danti, et al. 2014)

proposed work on diseased and unhealthy arecanut by

implementing texture features of Local Binary

Pattern (LBP), Haar Wavelets, GLCM and Gabor.

Statistical method of correlation determined the query

sample and histogram of training set. KNN classifier

have been used for classification. The process was

carried out in two stages, in first stage LBP is

generated, hence it achieved 92% of success rate.

Later in second stage Haar Wavelets, GLCM and

Gabor resulted high success rate of 100% and high

degree of discrimination has been found. This work

can be extended to identify disease of fruits, flowers

and seeds etc.

K. Rajashree. Et al., (Rajashree, Prema, et al. ,

2022) applied Artificial intelligence method based on

deep learning model to detect fruit rot disease. Sliding

window concept were compared with Vanilla LSTM,

INCOFT 2025 - International Conference on Futuristic Technology

126

stacked LSTM, and bidirectional LSTM models.

Models are measured by validation loss and training

loss with the help of Mean Square Error (MSE).

There are 3,152 samples in the dataset, 80% of

samples are used for training, and 20% are used for

testing purposes. From 630 testing samples, only 100

are predicted. LSTM model predicted disease in crops

such as rice and cotton. Based on RH, sunshine,

temperature, and disease score data, with 67.4%

accuracy. Vanilla GRU is the best model compared

with all other models, with a low error rate of

1.3(MSV) and less processing time due to its fewer

gates.

Jiawei Guo. Et al., (Guo, Jin, et al. , 2022)focused

on remote sensing-based detection of Areca Yellow

Leaf Disease (AYLD) using Planet Scope satellite

imagery. Employed 13 spectral features that

optimized through Correlation Analysis and

independent t-tests to develop disease classification

models. Random Forest algorithm achieved the

highest accuracy 88.24%, Backpropagation Neural

Network achieved 85.29% and AdaBoost achieved

67.65%. RF demonstrated superior error

minimization and robustness for classifying healthy

and diseased plants. It was emphasized the

significance of feature optimization in improving

model accuracy precise disease monitoring in

agricultural and forestry applications.

Krishna A N, et al., (Anitha, Dhanesha, et al. ,

2022)] investigated deep learning-based methods for

segmenting arecanut bunches from field images,

implementing Mask R-CNN and U-Net architectures.

Both models eliminated pre-processing and handle

challenges like illumination, occlusion, and complex

backgrounds. A dataset of 1017 images (ripe and

unripe) was used, split into training and testing sets.

Mask R-CNN, pre-trained on COCO and fine-tuned,

achieved superior accuracy with IoU of 67.88% and

F1-score of 79.36%, outperforming U-Net accuracy

with IoU 57.04%, F1 70.78%. These results

demonstrated Mask R-CNN’s is effective for precise

segmentation, enabling automated yield estimation

and harvesting, that offering a robust solution for

agricultural automation.

Smita Nair, et al., (Anitha, Dhanesha, et al. , 2022)

introduced a real-time PCR method for detecting the

phytoplasma responsible for Arecanut Yellow Leaf

Disease (YLD). The methodology involved designing

primers targeting the 16S rRNA gene, followed by

validation using spindle leaf tissue samples. Among

four primer sets, QPF2/QPR2 exhibited consistent

amplification, achieving accurate detection with a

unique melting peak at 82.3 ± 0.5°C. This approach

outperformed conventional PCR by reducing cross-

contamination and enhancing sensitivity. The

method’s rapid and precise detection capabilities

support early disease diagnosis, enabling better

management of YLD in arecanut palms.

Niranjan Murthy Chandrashekarappa. et al.,

(Chandrashekarappa, Murthy, et al. , 2022) focused

on designing and developing an efficient monitoring

mechanism also called as efficient data sensing and

monitoring (EDSM). Proposed model minimized the

energy, reduced the false alarm rate, and enhanced the

detection accuracy. It follows four steps, first step

included the sensing device condition, second step

will update the data strategy, third step follows data

validation, and the fourth step include sensed data

optimization. Proposed model outperformed the

existing model with significant optimization and

achieves 100% detection rate and failed to get even

single false alarm rate. EDSM are proven to be

efficient and solves various problem of existing

model.

Vikram Kumar1. et al., (Kumar, Vikram, et al. ,

2020) approached work uses two different data like

one from researched data and another from real time

collected from IoT (Internet of Things). Arduino Uno

board and the sensors DHT-11 and soil humidity

sensors are used to collect real time data. Work is

based on machine learning algorithm to process the

data. RFC (Random Forest Classifier) algorithm is

used for classification of data. 300 data trees are used

for prediction. Based on the MIN-SCORE output the

disease is predicted. The predicated score from the

RFC was most accurate. Compared to other

algorithms RFC gave high performance.

Pushparani M.K. et al., (Pushparani, Kumar, et al.,

2019) developed a system for efficient grading

system based on computer vision technology.

Arecanuts classified into different grades by using

MATLAB Toolbox. Arecanuts differ from one

another based on colour, size, and texture. colour

feature use HSV Histogram colour transform, texture

recognition uses Gabor filter and Gabor wavelet

transforms. Support Vector Machine is used as a

classifier. The application of main GUI was built

using the MATLABs own UI building tool called

GUIDE. processing based arecanut grading is a novel

approach to sort arecanuts, implementation in real

time give more accuracy in high performance for

farmers.

Dr. Kalai Selvi T. et al., (Revathi, Arivuselvam, et

al. , 2024 ) focused in comprehensive approach that

combine recommendations of crop to detect diseases.

System proposed based on leverages a hybridized

technique utilizing the LightGBM (LGBM) model.

The model focuses on developing robust LightGBM-

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning

127

based crop recommendation system and fine-tuning

deep-learning models for accurate disease detection.

Ensemble. Innovative developed system

revolutionized agricultural technology by accessing a

user-centric web application interface for farmers and

stakeholders. Model's accuracy in validation phase

shows 94.8% and test phase shows 94.2%. ResNet-50

based crop recommendation model predicts

promising results.

Mr. Ajith Kumar Shetty. Et al., (Shetty, Pandu, et

al. , 2024) developed a model to train the ResNet90

for improving automated weed and detecting the

crop. Goal of this work is to improve automated

classification systems' precision and to support more

productive. Cutting-edge methods improved

agricultural operations by differentiating images of

weeds and crops used by both deep learning and

conventional machine learning algorithms. 13

different input images fed to represent different

species or crop. Logistic Regression showed

relatively good accuracy, but not as high as the RF

higher accuracy in image classification tasks.

Sandeepa Prabhu. et al. (Prabhu, Sandeepa, et al.

, 2024) concentrated on drone technology by

exchanging data and gathering data in real time. The

fungus Phytophthora areeae causes koleroga, or fruit

rot, which affects the majority of areca nuts. In this

system, disease detection is done using drones. A

camera placed on a drone takes pictures. To process

the photos and distinguish between the healthy and

afflicted ones, several techniques are employed, such

as machine learning or computer vision algorithms.

They identify patterns suggestive of illness and

pinpoint areas of concern. An average detection rate

of more than 90% was obtained. With the use of

sophisticated imaging technologies and effective

image processing algorithms, drones can now

diagnose diseases with high accuracy.

Dong Xu. et al., (Xu, Dong, et al. , 2023) aimed to

monitor the severity of Areca Yellow Leaf Disease

(YLD) using UAV-based multispectral and thermal

infrared imagery, addressing the urgent need for

effective disease management in areca cultivation.

The methodology involved collecting imagery from

UAVs, utilizing the ReliefF algorithm for feature

selection, and employing machine learning models

like Random Forest to predict disease severity. With

an R2 of 0.955 and a substantial correlation (0.753)

between canopy temperature and disease severity, the

results show great prediction accuracy. This study

demonstrated the potential for wider applications and

offered a scalable, accurate method for agricultural

disease surveillance.

Balipa. et al., (Balipa, Mamatha, et al. , 2022) used

CNN and SVM, two machine learning techniques, to

create a system for identifying different arecanut

disorders. To enhance training, the dataset—which

included photos from farms in Shivamogga,

Karnataka—was produced and enhanced. GLCM and

GLDM matrices were used to transform the pictures

into texture-based features. Comparing CNN and

SVM classifier performance has been part of the

technique. With an accuracy of 90% as opposed to

75% for SVM, CNN fared better than SVM,

demonstrating its applicability for intricate image-

based illness identification. This study showed how

deep learning may be used to automate disease

monitoring in arecanut farming, resulting in better

crop management and early identification.

S.B. Mallikarjuna et al., (Mallikarjuna,

Shivakumara, et al. , 2021) aimed to enhance the

classification accuracy of arecanut images affected by

various diseases, including rot, split, and rot-split.

The authors propose a novel approach that combines

multi-gradient images with deep convolutional neural

networks (CNNs). By applying multiple Sobel masks

to the input images, fine details are accentuated,

providing vital clues for disease classification. These

enhanced images are then processed using a CNN

model, specifically AlexNet, to extract distinguishing

features. The proposed method demonstrated superior

performance on a four-class dataset, achieving higher

classification rates, recall, precision, and F-measures

compared to existing techniques. This research

offered a promising direction for automated detection

and classification of multiple arecanut diseases,

potentially aiding in timely disease management and

control.

Rashidul Hasan Hridoy. et al., (Hridoy, Hasan, et

al. , 2022) focused on the detection of betel plant

diseases using deep learning to support healthy

production and mitigate economic losses. The authors

utilized a dataset of 10,662 leaf images and applied

transfer learning with various architectures, including

AlexNet, VGG16, ResNet50, Inception V3, and

EfficientNet B5. Efficient Net B5 achieved the

highest test accuracy of 98.84% and the lowest

misclassification rate, outperforming other models

like AlexNet and VGG16. The results validated the

approach as a reliable solution for precise disease

diagnosis, enhancing agricultural management and

productivity in the betel leaf industry.

R. Sujatha. et al., (Sujatha, Radhakrishnan, et al. ,

2021) evaluated the efficiency of Deep Learning (DL)

and Machine Learning (ML) models in identifying

diseases in plant leaves. The study compared ML

methods like Random Forest and SVM with DL

INCOFT 2025 - International Conference on Futuristic Technology

128

architectures such as VGG-16, VGG-19, and

Inception-V3. Results indicate DL models

significantly outperform ML counterparts, achieving

higher classification accuracy, with VGG-16 yielding

the highest accuracy (89.5%). They have

demonstrated the superior capability of DL for plant

disease detection and classification.

Sharada Mohanty. et al., (Mohanty, Sharada, et al.

, 2016) explored the application of deep learning for

diagnosing plant diseases through leaf images. The

study aimed to utilize advancements in computer

vision for smartphone-assisted disease diagnosis,

addressing global food security challenges. Using a

dataset of 54,306 controlled images of healthy and

diseased leaves, the authors trained a deep

Convolutional Neural Network (CNN) to classify 14

crop species and 26 disease types, achieving a high

accuracy of 99.35% on a test set. However, testing

with diverse online images revealed lower accuracy

(31.4%), highlighting the need for varied training

data. The findings suggested a scalable method for

global agricultural disease management through

smartphone applications but emphasize the necessity

of diverse datasets for better generalization.

Xin Yang and Tingwei Guo (Yang, Xin, et al. ,

2017) examined how machine learning (ML) is

utilized to detect, classify, and understand plant

diseases, contributing to advancements in precision

agriculture. It emphasizes the use of various ML

algorithms, such as Support Vector Machines (SVM),

Artificial Neural Networks (ANN), and Random

Forests (RF), to diagnose diseases like

Huanglongbing in citrus and powdery mildew in

tomatoes. These methods have achieved

classification accuracies as high as 92.8% in some

cases. The research highlights challenges, such as the

need for robust preprocessing and diverse datasets to

improve accuracy, while also discussing potential

applications in disease prediction and quantification.

The study concludes that integrating ML with

technologies like aerial imaging and sensors can

significantly enhance agricultural practices and plant

disease management.

Surampalli Ashok. et al., (Ashok. et al. 2020)

aimed to develop an efficient system for identifying

diseases in tomato plant leaves through image

processing and deep learning methodologies. The

authors employed image segmentation and clustering

techniques to preprocess leaf images, followed by the

application of Convolutional Neural Networks

(CNNs) for disease classification. The proposed CNN

model achieved a classification accuracy of 98.12%,

outperforming other models such as Alex Net and

Artificial Neural Networks (ANN), which attained

accuracies of 95.75% and 92.94%, respectively. This

research demonstrated the potential of deep learning

approaches in accurately detecting and classifying

tomato leaf diseases, thereby contributing to

improved agricultural practices and crop

management.

Arsenovic Marko. et al., (Arsenovic, Marko, et al.

, 2019) aimed to make deep learning techniques for

plant disease diagnosis more useful. Key issues in the

subject are noted by the authors, including the need

for improved generalization across a variety of

scenarios, the scarcity of datasets, and limitations

with real-world environmental circumstances (such

as complicated backdrops and changing lighting). A

unique two-stage CNN architecture that is intended

for excellent accuracy in practical applications was

put forth by the authors. Tests show that the model

can manage situations involving many diseases and

get an accuracy of 93.67% in difficult settings

Amrita S. Tulshan. et al., the work focused on the

existing classification techniques like leaf disease

detection to detect the diseased plant from images.

Two classifiers of algorithms are used to compare the

results, existing one is Linear SVM Classifier and

proposed algorithm is KNN Classifier. Following

steps carried out like image preprocessing, image

segmentation, feature extraction. Existing one used

150 images for dataset, had accuracy of 97.6% and

proposed algorithm used 75 images as dataset, had

accuracy of 98.56%.

Halil Durmus. et al., applied Deep learning

techniques to detect the disease on the tomato plant.

Work is about to train a robot to detect the disease on

real time based on different algorithm training data

fed to robot. AlexNet and SqueezeNet are two

different architectures used for the system. Tomato

leaf input Data set was sent from the plant village.

Training and testing are done using mobile

supercomputer Nvidia Jetson Tx1 has 256 CUDA

cores, quad core ARM processor,4GB RAM, 16 GB

eMMC. PlantVillage data set contains 54.309

labelled images for 14 different crops, AlexNet

performed with accuracy 0.9722. AlexNet performed

slightly better than SqueezeNet. But SqueezeNet is

good candidate for the mobile deep learning

classification due to its lightweight and low

computational needs.

Kawcher Ahmed. et al., the goal of the system is

early detection of diseased plant by automated

method instead of manual method that takes

prolonged time for detecting the disease. Machine

learning techniques were implemented for this

research work. The most common rice plant diseases

are leaf smut, bacterial leaf blight and brown spot

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning

129

diseases are detected in this work. Affected white leaf

images are used as input. Supervised classification

algorithms were applied to detect three diseases of

rice. Logistic Regression, K-Nearest Neighbour,

Decision Tree, Naive Bayes Classifier are classifiers

used. Among them decision tree performed the best

with 97.9167%accuracy on test data.

Murk Chohan. et al., developed plant disease

detector based on deep learning. Using neural

network plant disease detection model is developed.

Initially augmentation is used for the dataset to

maximize the sample size, then convolution and

pooling layers are used. CNN architectures like

simple CNN, VGG, and InceptionV3 are trained by

Jupyter notebook and Keras API of Tensor flow.

Images from 38 different classes and 20% of each

class randomly chosen for testing. Model gave us

more than 95% accuracy. Testing dataset gave more

than 98% accuracy. It concludes that from accuracy

CNN is highly suitable for automatic detection and

diagnosis of plants. This system could be integrated

into mini-drones to live detection of diseases from

plants in cultivated areas.

Vishal Dineshkumar Soni. et al.,work

investigated on the data mining to predict the disease.

From collection of huge diseased data, the hidden

patterns information found in medical field based on

data mining. Reliable prediction models are built

using the data mining techniques. In this method,

classifier learning is carried out in one step before the

genetic algorithm. Selection is based on genetic

algorithms combined with the SVM classification.

Method has been tested with Various other classifier

of different types on the datasets for breast cancer,

Lung cancer and Diabetes.

Zamani. et al., explored the combined use of

machine learning algorithms and image processing

techniques for the detection of plant leaf diseases.

This study demonstrated the efficacy of a hybrid

strategy in which machine learning models

effectively manage the classification problem while

image processing improves the feature extraction

stage. This study's integration of machine learning

models like Support Vector Machines (SVM) and

Random Forest for illness classification with image

processing techniques including feature extraction

and picture enhancement is one of its important

features. This method's benefit over simple image

classification techniques is its capacity to increase

illness detection accuracy by preprocessing pictures

to emphasize significant characteristics prior to

classification. The final findings demonstrate that

integrating machine learning models with image

processing produces high illness detection accuracy,

with Random Forest and other machine learning

models offering the greatest classification

performance.

3.1 Observation from the Literature

Review

Agriculture is the foundation of any nation, for past

several years. Focusing precision farming helps us to

maintain high yield quality. Machine learning, Iot and

deep learning technologies are used to detect and

predict the disease at early stage. Plant is susceptible

to disease by comparing the data values, which we got

using Iot sensors like DHT-11 [1]. The performance

of VGG algorithm has high accuracy when compared

with Resnet and Alexnet. But data storing space

required is high for VGG. Even using soil samples

detection of yellow leaf disease can be done at early

stage. Machine learning techniques detect fruit rot by

weather parameter data collected. In this method

Random Forest regression (RFR) gave minimum

error in prediction. Image processing detects the spot

Table 1: Pros and cons between two methods

Methods Advanta

es Disadvanta

Machine

Learning

ID3 algorithm

can handle both

categorical and

continuous data

and its powerful

tool for decision

making [4].

Had low scalability,

when training large set

of data.

Less

computational

power and Fast

response for less

data sets.

Predicted low accuracy

for complex task.

Deep

learning

Higher accuracy

than machine

learning

Techniques.

When deeper networks

are considered, there is

a degradation problem

where accuracy

initially rises with

increased depth but

then reaches saturation

decreases with further

th increases [5].

Used in real

time detection

system.

Segmenting specific

region of plant based

on supervising learning

required expertise in

that domain.

and classify the quality of areca nut with CCD camera

accurately and efficiently but is unable to inspect for

covered blades [9]. CNN techniques evaluate

probability score and provide solutions with

responses to the end user that is displayed. Table 1

INCOFT 2025 - International Conference on Futuristic Technology

130

explains the advantages and disadvantages of both

machines learning and deep learning methods.

4 METHODOLOGY

Deep learning and machine learning methods are used

to classify, identify, and predict diseases in different

types of plants. Data were collected from the

published research papers to study and compare the

multiple techniques used to detect the disease.

Algorithms like ResNet, MobileNet, VGG, image

resizing, Support Vector Disease (SVD), clustering,

and Efficient Net B5 are based on deep learning

methodology used in this paper to detect the diseases.

Artificial Neural Network (ANN), LightGBM

(LGBM), Support Vector Machine (SVM), Random

Forest (RF), Decision Tree, Back Propagation Neural

Network (BPNN), and AdaBoost are the various

algorithms based on machine learning implemented

here to detect the diseases. Table 2 shows the

different methodologies used to detect the disease and

visualized the infected plant.

5 PROPOSED METHOD

Figure 1: Flow chart for Disease detection process

Fig 1 shows the review work based on the data

collected from various works, based on the study the

work can be categorized and processed based on the

proposed algorithm.

Table 3: Process for detection

Stage Description Techniques

Involve

Types of

Diseases in

Plants

Identification of biotic

(e.g., pests, fungi,

bacteria) and abiotic

factors (e.g., drought,

nutrient deficiency).

Categorized as

Biotic or Abiotic.

Steps for

Disease

Detection

Acquisition,

Preprocessing,

Segmentation, and

Classification of

Images to detect plant

diseases.

Each step is

detailed below.

Acquisition

of Images

Collecting images of

plants or leaves

affected by disease

using digital devices

like cameras o

drones.

Multispectral, Iot-

Enabled cameras

Preprocessi

ng of

Images

Enhancing image

quality by removing

noise, resizing, and

standardizing the

dataset for better

analysis.

Histogram

Equalization,

Filtering, Resizing,

Normalization.

Segmentati

on of

Images

Dividing images into

regions to isolate

diseased areas for

furthe

anal

sis.

Thresholding,

Genetic Algorithm

(GA), K-means

Clusterin

Classificati

on of

Diseases

Using machine

learning and deep

learning models to

classify plant diseases

based on the processed

images.

- Support Vector

Machines (SVM)

- Decision Trees

- K-Nearest

Neighbou

(KNN)

- Artificial Neural

Networks (ANN)

- Convolutional

Neural Networks

(CNN)

Disease

Categories

Applying the above

methods to specific

crops and diseases,

such as:

- Wheat Crop

Disease Detection

- Tomato Crop

Disease Detection

- Areca Nut Crop

Disease Detection

Results

Evaluating the

accuracy, efficiency,

and applicability of

models for identifying

diseases.

Output results for

specific crops

based on the

method used.

Stages of process, first one is to identify the types

of diseases in plants, then processing them for disease

detection in plants, then comparison with existing

method for detection for better accuracy and finally

the Results. Types of plants have two categories one

is Biotic and another one is Abiotic. Disease detection

has three stages like, segmentation of images, pre-

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning

131

processing of acquired images and classification of

diseases. Segmentation process implements R-CNN,

K-means clustering, U-Net. Disease can be classified

based on K-nearest neighbour (KNN). Support vector

machine (SVM), Genetic algorithm (GA), Decision

Tree, Convolution neural network (CNN), Artificial

neural network (ANN) are applied in this review

paper for better results. Table 3 explains step by step

for disease detection in plants with a small

description.

6 RESULTS AND DISCUSSION

If From above the study, from various published

research works based on machine learning and deep

learning. The accuracy of disease detection system

depends upon the data collected from the field;

algorithms used for detection. For small set of data

machine learning predictions are more accurate than

deep learning. For complex applications, deep

learning is more accurate.

Table 4 shows the accuracy of detected diseases

based on the algorithms. By implementing higher

accuracy methods avoids the error of detection that

protects the farmer from early disease control.

Disease control plays major role in farming to

improve quality productivity. Future research work

based on deep learning methods to disease detection

in various plants gives better results. For an effective

approach, extensive research needs to be studied and

applied for disease detection.

Table 4: Accuracy of detected diseases

Type of

diseases

Techniques Algorithm Accuracy

Koleroga CNN Res Net 97.5%

MobiNet 86%

VGG 96%

Image

resizing

88.46%

Yellow

Leaf

Disease

CNN Res Net 97.5%

SVD 86.30%

Planet

Scope

Satellite

imaginar

RF 88.24%

BPNN 85.29%

Ada Boost 67.65%

Based on

Colour,

Texture

and

Density

Machine

learning

ANN 98.8%

Classificat

ion of

Deep

learning

Machine

vision

90.9%

Areca

Diseases

Crop

Prediction

Machine

learning

LGBM 94.20

Different

Diseases

of areca

Machine

learning

Image

Processing

90%

Betel leaf Deep

learning

Efficient

Net B5

98.8%

7 CONCLUSION

Farmers need to be conscious in early detection of

diseases that prevent them from spreading disease to

other plants even these leads to production of low-

quality yield. From above the study of research work

paper published applied various algorithms to detect

the diseases. Different diseases from various plants

like areca nut, tomato, wheat plant, citrus gives

variations in accuracy based on the method used.

REFERENCES

SharathKumar, K. R., and K. Mohan. "Arecanut Crop

Disease Prediction using IoT and Machine Learning."

Journal of Science & Technology (JST) 5, no. 3

(2020): 160-165.

Kumar, S. Anupama, Andhe Dharani, M. B. Deepak, and

Aishwarya K. Kamble. "An Automataed Deep

Learning Model to Classify Diseases in Areacanut

Plant."

Mahveen, Namra, Umaira Shahneen, S. M. Hasanuddin,

Ayesha Fatima, Saheba Aijaz, and S. M. Qutubuddin.

"Enhancing Areca nut Plant Wellness: Innovative

Disease Detection using Deep Learning Algorithms."

Kamath, Shwetha, Prakhyath Devadiga, Prathiksha G K,

Prathiksha S, and Shrinidhi Shervegar. "Machine

Learning Based Early Detection of Yellow Leaf

Disease in Arecanut Plant Using Soil Samples."

International Journal of Novel Research and

Development 8, no. 4 (April 2023): e546-e551. ISSN:

2456-4184.

Tejaswi R, Rajesh M Mysoremath, Pradhan D Prabhu,

Honey Jain, and Mahitha G. "Detection and Prediction

of Diseases in Arecanut Plants." International Research

Journal of Engineering and Technology 11, no. 5 (May

2024): 517-522.

Sreenivasulu, P., K. Pavithra Lakshmi, Sk. Chandini, A.

Vimala, and J. Sai Pallavi. “Detection of Diseases in

Arecanut Using Convolutional Neural Network.”

International Journal of Research in Engineering, IT

and Social Sciences 14, no. 6 (June 2024): 564–570.

Accessed November 27, 2024

Krishna, Rajashree, K. V. Prema, and Rajath Gaonkar.

"Areca nut disease dataset creation and validation using

INCOFT 2025 - International Conference on Futuristic Technology

132

machine learning techniques based on weather

parameters." Engineered Science 19 (2022): 205-214.

Dhanuja, K. C., and Mohan Kumar HP. "Areca nut disease

detection using image processing technology."

International Journal of Engineering Research 9, no. 8

(2020): 823-826.

Huang, Kuo-Yi. "Detection and classification of areca nuts

with machine vision." Computers & Mathematics with

Applications 64, no. 5 (2012): 739-746.

Hegde, Ajit, Vijaya Shetty Sadanand, Chinmay Ganapati

Hegde, Krishnamurthy Manjunath Naik, and Kanaad

Deepak Shastri. "Identification and categorization of

diseases in arecanut: a machine learning approach."

Indonesian Journal of Electrical Engineering and

Computer Science 31, no. 3 (2023): 1803-1810.

Anilkumar, M. G., T. G. Karibasaveshwara, H. K. Pavan,

and D. Sainath Urankar. "Detection of diseases in

arecanut using convolutional neural networks."

International Research Journal of Engineering and

Technology (IRJET) 8, no. 5 (2021): 4282-4286.

Lei, Shuhan, Jianbiao Luo, Xiaojun Tao, and Zixuan Qiu.

"Remote sensing detecting of yellow leaf disease of

Arecanut based on UAV multisource sensors." Remote

Sensing 13, no. 22 (2021): 4562.

Billadi, Shabari Shedthi, Madappa Siddappa, Surendra

Shetty, and Vidyasagar Shetty. "Classification of

arecanut using machine learning techniques."

International Journal of Electrical & Computer

Engineering (2088-8708) 13, no. 2 (2023).

Jenitta, J., and L. Swetha Rani. "Arecanut Disease

Classification using CNN." (2023).

Suresha, M., Ajit Danti, and S. K. Narasimhamurthy.

"Classification of Diseased Arecanut based on

Texture." International Journal of Computer

Applications 975: 8887.

Rajashree, K., K. V. Prema, G. Rajath, and S. Angad.

"Prediction of fruit rot disease incidence in Arecanut

based on weather parameters." (2022).

Guo, J., Y. Jin, H. Ye, W. Huang, J. Zhao, B. Cui, F. Liu,

and J. Deng. Recognition of Areca Leaf Yellow Disease

Based on PlanetScope Satellite Imagery. Agronomy

2022, 12, 14. 2021.

Anitha, A. C., R. Dhanesha, Shrinivasa Naika CL, A. N.

Krishna, Parinith S. Kumar, and Parikshith P. Sharma.

"Arecanut bunch segmentation using deep learning

techniques." International Journal of Circuits, Systems

and Signal Processing 16 (2022): 1064-1073.

Anitha, A. C., R. Dhanesha, Shrinivasa Naika CL, A. N.

Krishna, Parinith S. Kumar, and Parikshith P. Sharma.

"Arecanut bunch segmentation using deep learning

techniques." International Journal of Circuits, Systems

and Signal Processing 16 (2022): 1064-1073.

Chandrashekarappa, Niranjan Murthy, SanjayPande

Mysore Bhagwan, and Kotreshi Shivabasappa Nagur.

"Efficient data sensing and monitoring model for areca

nut precision farming with wireless sensor network."

Indonesian Journal of Electrical Engineering and

Computer Science 25, no. 3 (2022): 1549-1562.

Kumar, Vikram, Sudhanshu Ranjan, and K. Mohan. "Crop

Disease Prediction Using IoT and Machine Learning."

International Journal of Research in Engineering,

Science and Management 3 (2020).

Pushparani, M. K., D. Vinod Kumar, and Abdulla Gubbi.

"Arecanut grade analysis using image processing

techniques." International Journal of Engineering

Research Technology (IJERT) (2019): 2278-0181.

Revathi, Mrs KP, and B. Arivuselvam. "Machine Learning

Approaches for Crop Prediction and Disease Detection

System with Personalized Recommendations."

Mr Ajith Kumar Shetty Artificial, Ms Smita Pandu Naik

Bindya, Mr Deepak S. Shettigar, Mr Shaikh Mohd

Arqam Altaf, and Ms Smita Pandu Naik. "Weed and

Crop Image Classification Using Deep Learning

Techniques." IJRAR-International Journal of Research

and Analytical Reviews (IJRAR) 11, no. 4 (2024): 41-

45.

Prabhu, Sandeepa, Ashwini Gond, Kaushalya Naik,

Keerthana MM, and Manya MM. "ARECARE:

DRONE POWERED DISEASE DETECTION AND

MANAGEMENT USING IMAGE PROCESSING."

IJRAR-International Journal of Research and

Analytical Reviews (IJRAR) 11, no. 3 (2024): 41-44.

Xu, Dong, Yuwei Lu, Heng Liang, Zhen Lu, Lejun Yu, and

Qian Liu. "Areca yellow leaf disease severity

monitoring using UAV-based multispectral and thermal

infrared imagery." Remote Sensing 15, no. 12 (2023):

3114.

Balipa, Mamatha, Pallavi Shetty, Arhath Kumar, and B. R.

Puneeth. "Arecanut Disease Detection Using CNN and

SVM Algorithms." In 2022 International Conference

on Artificial Intelligence and Data Engineering (AIDE),

pp. 01-04. IEEE, 2022.

Mallikarjuna, S. B., Palaiahnakote Shivakumara, Vijeta

Khare, Vinay Kumar, M. Basavanna, Umapada Pal, and

B. Poornima. "CNN based method for multi-type

diseased arecanut image classification." Malaysian

Journal of Computer Science 34, no. 3 (2021): 255-265.

Hridoy, Rashidul Hasan, Md Tarek Habib, Md Sadekur

Rahman, and Mohammad Shorif Uddin. "Deep neural

networks-based recognition of betel plant diseases by

leaf image classification." In Evolutionary Computing

and Mobile Sustainable Networks: Proceedings of

ICECMSN 2021, pp. 227-241. Singapore: Springer

Singapore, 2022.

Sujatha, Radhakrishnan, Jyotir Moy Chatterjee, N. Z.

Jhanjhi, and Sarfraz Nawaz Brohi. "Performance of

deep learning vs machine learning in plant leaf disease

detection." Microprocessors and Microsystems 80

(2021): 103615.

Mohanty, Sharada P., David P. Hughes, and Marcel

Salathé. "Using deep learning for image-based plant

disease detection." Frontiers in plant science 7 (2016):

1419.

Yang, Xin, and Tingwei Guo. "Machine learning in plant

disease research." March 31 (2017): 1.

Tomato Leaf Disease Detection Using Deep Learning

Techniques

Arsenovic, Marko, Mirjana Karanovic, Srdjan Sladojevic,

Andras Anderla, and Darko Stefanovic. "Solving

current limitations of deep learning based approaches

Optimization of Disease Detection System for Improved Arecanut Cultivation by Machine Learning

133

for plant disease detection." Symmetry 11, no. 7 (2019):

939.

Asif, Muhammad, Tabarka Rajab, Samreen Hussain,

Munaf Rashid, Sarwar Wasi, Areeb Ahmed, and

Kehkashan Kanwal. "Performance Evaluation of Deep

Learning Algorithm Using High‐End Media Processing

Board in Real‐Time Environment." Journal of Sensors

2022, no. 1 (2022): 6335118.

Srivastava, Prateek, Shailendra Pratap Singh, Ankit

Vishnoi, and Varun Sapra. "Classification of Corn Leaf

Diseases using Various Pre-trained Deep Learning

Networks and Performance Comparison." In 2022

International Conference on Advances in Computing,

Communication and Materials (ICACCM), pp. 1-5.

IEEE, 2022.

Singh, Aadarsh Kumar, Akhil Rao, Pratik Chattopadhyay,

Rahul Maurya, and Lokesh Singh. "Effective plant

disease diagnosis using Vision Transformer trained

with leafy-generative adversarial network-generated

images." Expert Systems with Applications (2024):

124387.

Guo, J., Y. Jin, H. Ye, W. Huang, J. Zhao, B. Cui, F. Liu,

and J. Deng. Recognition of Areca Leaf Yellow Disease

Based on PlanetScope Satellite Imagery. Agronomy

2022, 12, 14. 2021.

Srivastava, Prateek, Shailendra Pratap Singh, Ankit

Vishnoi, and Varun Sapra. "Classification of Corn Leaf

Diseases using Various Pre-trained Deep Learning

Networks and Performance Comparison." In 2022

International Conference on Advances in Computing,

Communication and Materials (ICACCM), pp. 1-5.

IEEE, 2022.

Karthik, Challa, and Nagul Ulaganathan. "Application for

Plant’s Leaf Disease Detection using Deep Learning

Techniques." International Research Journal of

Engineering and Technology (IRJET) 7, no. 8 (2020):

1-7.

INCOFT 2025 - International Conference on Futuristic Technology

134