Plant Disease Classification Through Image-Based Feature Extraction And Soft Computing Approaches

Abstract

Agricultural production and quality of crop have significant impact by plant diseases. It necessitates the development of accurate and reliable automated disease detection systems. This study proposes a soft computing-based framework for plant leaf disease classification by incorporating image processing, feature extraction, Artificial Neural Networks (ANN) and traditional machine learning techniques. The proposed methodology uses image pre-processing and discriminative texture feature extraction for characterization of disease patterns from leaf images. Several machine learning classifiers like Decision Tree, AdaBoost, Bagged Tree, Naïve Bayes, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) are tested and compared with ANN based classification. The k-fold and holdout validation strategies are embedded to ensure reliable performance assessment and to reduce overfitting. E Experimental results show that models trained without validation reach unrealistically high accuracies, suggesting overfitting. Cross-validation produces more realistic and generalized performance estimates. Comparative analysis shows that the ANN model with 12 neurons trained using the Levenberg–Marquardt learning algorithm achieves the best overall classification performance. This study emphasizes on the need of validation methods for application related to plant disease detection and demonstrating the utility of integrating soft computing methods with image-based feature extraction. The proposed framework is presenting a scalable and adaptable solution for agricultural work-related diagnosis and has potential for future deployment in intelligent systems for crop monitoring.