AMS-YOLO: Multi-scale Feature Integration for Intelligent Plant Protection Against Maize Pests

leilei Deng^1,2Di Fang¹Aziz Ullah¹qi Hou¹helong Yu^1*

• ¹Jilin Agriculture University, Changchun, China
• ²Changchun University of Science and Technology, Changchun, China

As a major global food crop, maize faces serious threats from pests that significantly affect production output and crop quality. Accurate and efficient pest detection is crucial. However, existing detection methods show inadequate performance when facing challenges including pest morphological diversity, interspecies similarity, and complex field environments. This study introduces AMS-YOLO, an enhanced model based on YOLOv8n, featuring three synergistic modules specifically developed to tackle these issues. First, the SMCA attention approach improves target recognition within complex environmental settings. Second, a MSBlock multi-scale feature fusion module improves adaptability to pests at different growth stages. Third, an AMConv optimized downsampling strategy preserves subtle features necessary for distinguishing similar pest species. On a dataset comprising 13 common maize pests covering relatively complete developmental morphologies, AMS-YOLO demonstrates 90.0% precision, 89.8% recall, 94.2% mAP50, and 73.7% mAP50:95, surpassing the original YOLOv8n by 3.1%, 3.7%, 3.2%, and 4.0%, respectively. Through comprehensive comparative experiments, AMS-YOLO shows superior performance over existing detection methods including SSD, RT-DETR, and various YOLO variants. Deployment tests on Jetson Nano indicate that the model size is only 5.3MB, 15.9% smaller than the original YOLOv8n, with 19.6% fewer parameters and 16% reduced computational requirements, while maintaining low resource utilization, thus satisfying the demands for extended field deployment. The improved model demonstrates the potential for supporting more targeted pest management decisions, which may contribute to more precise pesticide application and resource optimization in field conditions.