Enhancing Crop Disease Classification Accuracy Using Classical Image Processing and Segmentation Techniques

In this study, we present a lightweight, interpretable, and computationally efficient framework for the classification of crop diseases using classical image processing and segmentation techniques. The proposed pipeline incorporates contrast stretching and unsharp masking for image enhancement, followed by Otsu’s thresholding and K-means clustering for precise segmentation of disease-affected regions. GLCM texture descriptors and color histograms are used to extract discriminative features from segmented leaf images. These features are then classified using machine learning models such as Support Vector Machines (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN). The model was tested on both public datasets and UAV-acquired imagery, achieving an overall accuracy of 88%, with F1-scores exceeding 0.90 in several disease classes. The results indicate that classical methods, despite their simplicity, can serve as viable alternatives to deep learning in resource-constrained environments, offering robust performance and real-time potential for deployment on edge devices in agricultural settings.