Target Detection Method for Cotton Leaf Diseases and Pests Based on YOLOv11

This paper proposes a target detection method for cotton leaf pests based on YOLOv11. By enhancing the YOLOv11 framework, three key modules are introduced, integrating multi-head efficient convolution (FMHEC) to achieve multi-scale feature extraction, and hypergraph-based cross-level and cross-location representation network (HyperC2Net) to achieve cross-level and cross-location feature fusion, and Shared Enhanced Convolution Head (SECH) to reduce parameter redundancy by sharing enhanced convolution heads. Experimental results show that the improved YOLOv11 achieves an accuracy of 86.69% on the cotton leaf pest dataset and an average accuracy of 79.29% under the 50% cross-merge ratio threshold (mAP50), while maintaining high computing efficiency and solving the limitations of feature fusion and computational redundancy, thereby improving the real-time monitoring capabilities of agricultural pests and diseases.