Image-Based Wheat Mosaic Virus Detection with Mask-RCNN Model

Abstract

Wheat is one of the most vital crops around the globe. Due to wheat mosaic virus disease, there are a huge amount of yield quality losses. The mosaic virus is transmitted through curl mite summering hosts. Once the mosaic virus is transmitted, the whole wheat plant is damaged which decreases the wheat grain quality. Therefore, the detection of a mosaic virus on the wheat leaf is detected through the Mask-RCNN model. A total of 15,536 wheat images were captured through Canon camera in the Punjab region. Each wheat healthy leaf and mosaic virus in each wheat leaf was labelled through the Visual object tagging tool (VOTT). The individual wheat leaf and mosaic virus labelled data were used as ground truth data. The resnet-50 is used as the backbone in the Mask-RCNN model. The Mask-RCNN model segments each wheat leaf and detects the mosaic virus on each individual leaf. For segmentation of each individual leaf and mosaic virus disease detection, the mask-RCNN model achieves 88.19% and 97.16% detection accuracy properly.