Fruit Disease Detection Using Colour, Texture Analysis and CNN

Abstract

With the rising demands in modern agriculture, optimizing fruit growth and improving yield has become increasingly important. Traditionally, farmers depend on manual inspection to monitor the various growth stages and identify diseases, a process that is not only labor-intensive but also prone to inaccuracies, often requiring expert input. To overcome these limitations, this work introduces a smart farming solution designed to enhance crop monitoring while minimizing human involvement. The proposed system focuses on identifying and classifying visible fruit diseases by utilizing image processing and deep learning methods. Unlike conventional approaches that rely predominantly on text-based information—often hindered by language constraints—this method emphasizes visual analysis, offering a more user-friendly and universally accessible solution. The system incorporates the OpenCV library for initial image preprocessing and leverages Convolutional Neural Networks (CNNs) for feature extraction and classification. CNNs effectively capture critical visual features such as color, texture, and shape, enabling precise disease identification. Two image datasets are used: one for training with labeled samples and another for testing with new, unseen images. This dual-dataset strategy enhances both the accuracy and real-time applicability of the system, making it a practical tool for smart agriculture.