Integrating Active Contours and K-Nearest Neighbours for Early Detection of Diseases in Banana Plants
Epsita Medhi *
Department of Information Technology, Gauhati University, Guwahati – 781014, Assam, India.
Nabamita Deb
Department of Information Technology, Gauhati University, Guwahati – 781014, Assam, India.
Banashree Bhuyan
Department of Computer Science and Application, Pandu College, Gauhati University, Guwahati – 781012, Assam, India.
Adity Roy
Department of Computer Science and Application, Pandu College, Gauhati University, Guwahati – 781012, Assam, India.
*Author to whom correspondence should be addressed.
Abstract
Diseases and pests pose a significant threat to agriculture, reducing both production and economic viability. Banana plants are highly susceptible to various pests and diseases, which can severely affect yield and quality if not properly managed. Given the global consumption of bananas, addressing disease issues in these plants is crucial. To tackle this challenge, integrating machine learning techniques can enable early disease detection. Specifically, combining region-based active contours (using Chan-Vese) with K-Nearest Neighbours (KNN) classification offers a comprehensive approach for object detection and classification in images. The hybrid approach enhances segmentation accuracy, adaptability, and generalization across different conditions and data types. Compared to traditional methods, this combination provides more robust, accurate, and efficient segmentation, making it a significant advancement in real-world applications. This technique is also applicable to other agricultural crops and fruits, offering promising results. With an average accuracy of 91.63%, the model demonstrates its effectiveness in detecting and classifying banana diseases and pests.
Keywords: Chan-vese, histogram equalizer, K-Nearest Neighbour (KNN), Principal Component Analysis (PCA)