Road Target Detection Method Based on Multi-Scale Feature Fusion

Aiming at the problem of low detection accuracy and high leakage rate due to variable target detection scales and complex environmental interference in unmanned scenarios, the road target detection algorithm YOLOv8n-APW is proposed based on YOLOv8n. Firstly, the attention scale sequence fusion mechanism ASF is introduced to effectively integrate the feature maps from different levels and scales, to better capture the shapes and sizes of different types of targets, and to improve the detection capability for multiple targets. Second, the P2 detection head is added to the base model to improve the recognition accuracy of small targets and reduce the target miss detection rate; finally, the Wise-IoU loss function is introduced to replace the original loss function to improve the robustness of the model. On the public dataset SODA10M, the mAP0.5 of the improved model is improved by 4.4% compared with YOLOv8n, and the improved method is more suitable for the task requirements of road vehicle detection.