Utilizing Convolutional Neural Networks and KMeans Clustering for Efficient Plant Leaf Disease Detection

The agricultural sector is pivotal to many economies, but its productivity is frequently hampered by various plant diseases, especially those affecting leaves. Timely and precise detection of these diseases is vital for preserving field health. While image segmentation, clustering, and deep learning are prevalent techniques, our study presents a novel hybrid method for image classification. This strategy integrates the k-means clustering algorithm with a Convolutional Neural Network (CNN). Initially, k-means pinpoints the diseased portions of a leaf, which is followed by the CNN determining the exact disease. For validation, the PlantVillage dataset was employed, which includes various crops facing diverse disease challenges. The determination of the optimal k-value was assessed using the Silhouette coefficient, Elbow method, and Kneedle Algorithm. The Kneedle Algorithm emerged as the most consistent in selecting the optimal k-value. Images were then segmented using k-means clustering based on the k value derived from the Kneedle Algorithm. A CNN was subsequently trained to classify the type of disease from these segmented leaf images, achieving high accuracy and highlighting the effectiveness of the proposed model.