基于改进YOLOv11的野外中草药目标检测

Object Detection of Wild Chinese Herbal Medicines Based on Improved YOLOv11

摘要: 为解决野外复杂场景下中草药检测存在的小目标漏检、背景干扰大、模型计算冗余等问题，提升检测精度与效率，本文以YOLOv11为基础开展针对性改进研究。方法上，将SPPF模块替换为AIFI模块降低计算复杂度并整合全局信息；采用Dysample模块替代传统上采样，增强遮挡模糊目标的细节恢复能力；新增小目标检测层扩大尺度覆盖，同时嵌入CBAM双注意力模块抑制背景干扰、聚焦关键特征。实验表明，改进模型mAP@0.5为81.2%，较基线YOLOv11n提升2.9个百分点，GPU端FPS为266，参数量仅2.80 M，大幅提升了复杂场景下的检测精度与运行效率，为野外中草药资源的智能化识别、调查与保护提供了可靠技术支撑。

Abstract: To address the issues of small target missed detection, severe background interference, and redundant model computation in Chinese herbal medicine detection under complex field scenarios, and to improve detection accuracy and efficiency, this paper conducts targeted improvement research based on YOLOv11. Methodologically, the SPPF module is replaced with the AIFI module to reduce computational complexity and integrate global information; the Dysample module is adopted to replace the traditional upsampling layer, enhancing the detail recovery capability for occluded and blurred targets; a small object detection layer is added to expand scale coverage, and the CBAM dual attention module is embedded to suppress background interference and focus on key features. Experimental results show that the improved model achieves an mAP@0.5 of 81.2%, which is 2.9 percentage points higher than the baseline YOLOv11n model, with a GPU inference speed of 266 FPS and only 2.80 million parameters. It significantly improves the detection accuracy and operational efficiency in complex scenarios, providing reliable technical support for the intelligent identification, investigation, and protection of wild Chinese herbal medicine resources.