Application of UAV-Based Imaging and Deep Learning in Assessment of Rice Blast Resistance

Abstract

Rice blast is regarded as one of the major diseases of rice. Screening rice genotypes with high resistance to rice blast is a key option for ensuring global food security. Unmanned aerial vehicles (UAV)-based imaging, coupled with deep learning, can high-throughput acquire imagery traits about rice blast infection. In this study, we developed a segmented detection model (RiceblastSegMask) for rice blast detection and resistance evaluation. The feasibility of different backbones and target detection models was further investigated. RiceblastSegMask was a two-stage instance segmentation model, which included an image denoising backbone network, feature pyramid and trinomial tree fine-grained feature extraction combination network, and image pixel codec module. The results showed that the model that combining the image-denoising and fine-grained feature extraction based on Swin Transformer and the feature pixels matching feature labels with the trinomial tree recursive algorithm presented the best performance. Overall accuracy for instance segmentation of RiceblastSegMask reached 97.56%. It achieved a satisfying result of 90.29% accuracy for grading unique resistance to rice blast. The results demonstrated that UAV low-altitude remote sensing, coupled with the proposed RiceblastSegMask model, can efficiently calculate the infected area of rice blast, providing a new phenotypic tool to evaluate the resistance to rice blast in rice at a field scale for the breeding program.