Agricultural Pest and Disease Detection Method for UAV Based on Improved YOLOv11n

To address the problems of small targets, significant background interference, and high missed detection rates in low-altitude Unmanned Aerial Vehicles (UAV)-based agricultural pest and disease detection, this study proposes a detection algorithm based on improved YOLOv11n of You Only Look Once (YOLO). The algorithm enhances detection accuracy and efficiency through multi-module optimization. The proposed algorithm introduces Asymmetric Padding Bottle Neck (APBN) into the backbone network, leveraging the asymmetric padding and grouped convolution mechanism of pinwheel-shaped convolution to enhance feature extraction capability. It designs the embedding of Multi-Scale Channel Attention Module (MSCAM) into the Convolutional block with Parallel Spatial Attention (C2PSA) module, utilizing multi-scale convolution kernels to fuse features from different levels, thereby improving detection capability for multi-scale pests and diseases. In the neck network, it designs Content-Guided Simple Attention (CGSA), which integrates spatial, channel, and pixel-level attention to strengthen target feature representation. Experimental results show that compared with other algorithms, the proposed APBN-MSCAM-CGSA-YOLO (AMC-YOLO) algorithm achieves significant improvements in Precision (P), Recall (R), and Mean Average Precision (MAP), while maintaining lightweight advantages in parameters and floating point operations (FLOPs). It provides an efficient technical solution for agricultural pest and disease control.