Research on small target detection method of UAV in complex environment: based on multiscale feature fusion and attention mechanism

This paper proposes a small target detection method for UAV based on multi-scale feature fusion and attention mechanism, aiming to solve the problem of small target scale and strong background interference in complex environment. Based on the YOLO framework, this method integrates feature maps of different resolutions by constructing a multi-scale feature fusion module to improve the feature expression ability of small targets. At the same time, it introduces an optimized CBAM attention mechanism to focus on key areas and suppress background noise. Experimental results show that the proposed method achieves 0.813 in mean average precision ([email protected]), which is 8.9% higher than the baseline YOLOv5s. The small target detection accuracy is significantly improved while maintaining the real-time performance of 128 FPS. Ablation experiments verify the independent contributions of multi-scale feature fusion and attention mechanism, and prove their effectiveness in enhancing adaptability to complex environments. This method provides an efficient solution for the target detection task of UAV in complex urban scenes and has high application value.