DEEP-LEARNING-BASED IDENTIFICATION OF CORN PESTS AND DISEASES: RECOGNITION AND FAST ANALYSIS

Pests and diseases seriously affect the quality and yield of maize. Therefore, it is important to carry out disease diagnosis and identification for timely diagnosis and treatment of maize pests and diseases and to improve maize production quality and eco-nomic efficiency. In this study, an improved Resnet50-based maize pest identification model was proposed to efficiently and ac-curately identify maize pests and diseases. Based on convolution and pooling operations for shallow-edge feature extraction and data compression, further effective channels (environment–cognition–action) were introduced into the residual network module to solve the problem of network degradation, establish connections between channels, and extract deep key features. Finally, ex-perimental validation was performed to achieve 96.02% recognition accuracy. This study recognized maize leaf blight, Helmin-thosporium maydis, gray leaf spot, rust disease, stem borer, and corn armyworm, which can provide useful guidance for the intel-ligent control of maize pests and diseases.