Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 12 Issue: 04 | Apr 2025

p-ISSN: 2395-0072

www.irjet.net

Automated Crop Disease Detection Using Deep Learning and Image Processing Mr. Pratik H. Gurav1

Dr. Pravin Jangid2

Dept. of Computer Engineering Shree L R Tiwari College of Engineering Mumbai, Maharashtra, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - The tracking and analysis of plant leaf

sickness are critical issues in modern agriculture since the condition of a plant determines the total agricultural yield and productivity. Plant diseases are often linked with various issues, but are typically linked with a failure to identify and evaluate at the right time. In the present work, the authors created a robust methodology for plant leaf disease identification using a combination of deep and machine learning techniques for various crops with a dataset that covers tomatoes, peppers, and potatoes. The model uses Convolutional Neural Networks for classifying images and a multi-species dataset of plants, including tomatoes, peppers, and potatoes. The system has various steps such as image pre-processing, segmentation and feature extraction. Apart from K-Nearest Neighbors (KNN), other was also utilized for the fine-tuning of the classification results. Experimental results indicate high accuracy rates in training (98.45%) and testing (98.15%) for various disease types such as bacterial, fungal, and viral. This model not only identifies diseases with accuracy, but also identifies important features such as the type of disease, the affected region, and the health condition of the plant under consideration. Using CNN allows automatic feature extraction, whereas Random Forest guarantees effective classification leading to better accuracy and precision. This piece of work helps the field of precision agriculture through an effective, low-cost, and scalable solution for real-time monitoring of plant diseases. Key Words: Plant Disease Detection, Deep Learning, CNN, KNN, Image Pre-processing, Feature Extraction, Agricultural Automation, Plant Health Monitoring, Disease Classification, Precision Agriculture.

1.INTRODUCTION Agriculture is the most significant industry of the global economy, supporting food sovereignty and tapping rural economies. Pathogens such as fungi, bacteria, and viruses are extremely devastating to crop health, exacerbating economic, social, and ecological losses. Crop quality and yields largely rely on early diagnosis of these diseases. The employment of fully trained experts for plant visual diagnosis is conventional and not very efficient since it is

Impact Factor value: 8.315

subjective and time-consuming. This requires and more immediacy. Recent developments in deep learning combined with image processing methods have brought about a revolution in the field of plant disease detection. Automatic diagnosis based on neural network models, or in particular Convolutional Neural Networks (CNNs), is applied nowadays with remarkable success since it helps in image classification by automatically identifying different diseases in the leaves and other structures of the plants. CNNs have been proven to be much more precise and quicker compared to conventional methods, which depend on human interpretation of data, since they can extract and learn features independently. Agriculture is the most significant sector of the global economy, supporting food sovereignty and leveraging rural economies. Fungi, bacteria, and viruses are highly devastating to plant health, however, furthering economic, social and ecological losses. The quality and yield of crops rely greatly upon early diagnosing the disease. The age-old practice of applying fully trained experts for visual inspection of plants is inefficient and not very effective since it is slow and subjective. This necessitates and a greater immediacy. Recent developments in deep learning combined with image processing algorithms has brought about a revolution in the world of plant disease detection. Neural network model-based automatic diagnostics, that is, Convolutional Neural Networks (CNNs), are employed today with much success because it facilitates image classification through automatic identification of various diseases on the leaves and in other parts of the plants. CNNs have been found to be more accurate and quicker compared to conventional methods that are based on manual interpretation of data since they can learn features and extract by themselves. The current study attempts to examine the use of machine learning and deep learning methods for diagnosing as well as plant disease classification. Improving the performance and detection accuracy of detection systems is the objective of using CNNs and other machine learning classifiers in the model proposed. We aim to contribute to the development of lowcost and accessible technologies for plant disease management in regions with limited agricultural knowledge through this study. Literature review, methodology, results, and the impact such development can have on agriculture are included in other sections of this paper.

ISO 9001:2008 Certified Journal

Page 325