Early Identification of Coconut Leaf Diseases using Inception V3 for Precision Agriculture

Leaf diseases detection of coconut tree are providing the maximum yield and sustainability of agriculture. This paper presents a novel framework for early detection of coconut leaf diseases by utilizing deep learning-based approach that incorporates image processing and classification techniques to develop an automated system for precision agriculture. The method starts by converting the RGB images into HSV format so that the color difference improves and the complex backgrounds can be eliminated. Binary images are then created based on hue and saturation values, since it contains useful information to separate healthy from diseased areas. Capsule Networks are utilized for feature extraction since they can maintain spatial hierarchies and capture intricate patterns in image data. The features that are extracted later are classified with the Inception V3 model which is known to be quite a powerful model for image recognition due to its efficient architecture. The proposed approach evaluated on Mendeley Data coconut leaf disease dataset. It is performing awesomely with an overall classification accuracy of 98.59%. The Utilization of HSV-based preprocessing, then using CapsNet as a feature extractor, then adding Inception V3 as a classifier significantly improves detection. The study offers a fast and accurate method for early detection of diseases in coconut cultivation, aiding in early intervention and greater crop management in precision agriculture.