Predicting Banana Leaf Diseases: Feature Extraction with BL-FEOT and Enhanced Classification using the BAT-KNN Hybrid Algorithm

Banana cultivation, fundamental to many rural economies, confronts persistent threats from many foliar diseases. Rapid and precise disease identification is critical for effective management and containment. This research introduces a pioneering method for detecting and classifying diseases on banana leaves, specifically targeting Cordana, Pestalotiopsis, and Sigatoka. Our process coined the Banana Leaf Feature Extraction and Optimization Technique (BL-FEOT), is a systematic approach encompassing manual background elimination, green color removal to emphasize the disease manifestations, and contour detection to mark out the compromised zones. Distinctive features are extracted for each disease type, culminating in a comprehensive dataset tailored for disease classification. Incorporating an extensive suite of feature extraction techniques, our methodology ensures the maximal retrieval of pivotal information from the infected sites. We leverage the Bat Optimization Technique to fine-tune the extracted features, especially within the RGB color space. This streamlines the feature dimensions and zeroes in on the most relevant components, amplifying the efficiency of the ensuing classification phase. The seminal contribution lies in integrating the Bat Optimization Technique with the K-Nearest Neighbors (KNN) algorithm, resulting in the novel hybrid algorithm, BAT+KNN. This algorithm is applied to the refined feature set for classification purposes. To substantiate the efficacy of BL-FEOT, its performance metrics are juxtaposed against prevailing algorithms using the same feature dataset. A thorough evaluation, encapsulating metrics such as precision, recall, accuracy, F1 score, ROC curve, and error rate, is presented. The experimental results, derived from available datasets, underscore the superior capabilities of our hybrid BAT+KNN algorithm in banana leaf disease identification. This research asserts that the BL-FEOT, powered by the BAT+KNN hybrid algorithm, offers a ground-breaking avenue for the automated and precise detection of banana leaf diseases. Its potential integration into real-time monitoring systems could revolutionize early disease detection and intervention in banana plantations.