A Comprehensive Study on Machine Learning and Deep Learning Models for paddy diseases and weeds detection

Abstract

Agriculture plays a crucial role in economic development by providing food, raw materials, and employment. With the global population increasing and limited agriculture land, enhancing food production is essential. Precision farming, which utilizes advanced technologies like sensors, GPS, and automated systems, aims to improve crop productivity and reduce resource usage. This research focuses on paddy (Oryza sativa), a staple food for many Asian countries, examining its structure, prevalent diseases, and common weeds. Key diseases such as Tungro and Bacterial Leaf Blight, and weeds like Barnyardgrass and Purple Nutsedge, significantly impact yields. Traditional manual inspection methods for disease and weed detection are labor-intensive and error-prone. Implementing advanced monitoring and early detection strategies is vital for effective crop management. By integrating real-time data and precision farming techniques, farmers can optimize their operations, reduce costs, and ensure sustainable agricultural practices, ultimately contributing to global food security. In this paper, various related works on paddy disease detection and weed detection in paddy field is studied.

Deep Learning (DL), Machine Learning (ML), and Image Processing (IP) are revolutionizing agricultural practices, especially in paddy disease and weed detection. IP has long utilized remote sensing to capture and analyze high-resolution crop images. Traditional ML methods, like k-Nearest Neighbors (k-NN) and Support Vector Machines (SVM), rely on manually engineered features for classification but can struggle with large datasets. DL, particularly Convolutional Neural Networks (CNNs), offers automated feature learning and end-to-end processing, enhancing accuracy and scalability in image analysis. The integration of these technologies improves disease and weed management in paddy cultivation.