Multi-Scale Feature Alignment Aggregation Method for Enhanced Object Detection in UAV Images

Object detection in Unmanned Aerial Vehicles (UAV) images faces numerous challenges, including multi-scale changes, complex scenes, and distant small targets. Existing multiscale feature fusion methods integrate feature information across different levels to address the issues arising from multi-scale variations. However, the integration often lacks interactive information exchange due to the use of concatenation operations, leading to suboptimal results. In response to these challenges, we have enhanced the general-purpose object detection algorithm YOLOv8 by introducing a novel multi-scale feature alignment aggregation method. The Feature Alignment Aggregation Module (FAAM) in the neck pyramid network addresses misalignment and degradation from fusing multi-resolution features, thereby improving target localization across various scales. Additionally, we incorporate the Convolutional Block Attention Module (CBAM) attention mechanism to enhance feature fusion. Furthermore, our Bidirectional Path Aggregation Network (BPAN) and additional small target detection head enhance multiscale feature fusion capabilities. Experimental results on the UAVDT dataset demonstrate that our method outperforms other YOLO series methods with an average precision mAP50 of 34%, indicating superior detection efficacy in scenes with multi-scale variations.