Vine Disease Detection UAV Multi Spectral Image using Segnet and Mobilenet Method

Deep Learning-based vine disease detection has garnered significant attention from the community, particularly with the utilization of UAV multispectral images for grapevine disease detection. However, identifying vine diseases in various crop and horticultural conditions remains a complex challenge, especially under mobile and edge computing conditions. The vine disease detection process makes use of the PlantVillage dataset, which includes unlabelled data. Data normalization is performed and UAVs are involved for data capture, while the SegNet architecture is utilized for segmentation. This enables the separation of healthy and unhealthy vines for detection. Subsequently, classification is performed using MobileNetV2, with layers split to detect diseases all combined with UAV spectral images and larger image sizes, greater than 32 × 32, resulting in better performance. The proposed method achieves high performance, with an accuracy achieve 99.50%, precision at 99.42%, recall at 99.39%, and mean average precision (MAP) at 99.20%. These metrics are compared to existing methods such as Deep Convolutional Neural Network (DCNN) and Inception V2.