基于改进的YOLOv7-tiny的复杂道路环境的目标检测方法OA

To address the problems of low detection accuracy and frequent false/missed detections in small-target identification under complex road scenarios,this paper proposes AT-YOLOv7,an enhanced YOLOv7-tiny-based algorithm for road target de-tection.First,the R-AFPN network is introduced,which adopts a progressive feature-fusion strategy to avoid direct interactions between non-adjacent layers.This design reduces feature discrepancies across non-adjacent layers,alleviates information loss,and enhances multi-scale feature fusion capability.Second,an ELAN-M module is designed to capture richer feature representa-tions and strengthen inter-layer information exchange.Finally,an SPD-Conv module is embedded into the backbone network to improve feature extraction capability for small-scale and low-resolution targets.Finally,an SPD-Conv module is incorporated into the backbone network to improve feature extraction ability for small and low-resolution targets.Experimental results on the KITTI dataset indicate that,compared with the original YOLOv7-tiny,the proposed AT-YOLOv7 achieves significant improve-ments,increasing mAP by 5.1 percentage points,precision by 3.5 percentage points,and recall by 5.8 percentage points.These results verify the superior detection performance of the proposed method and demonstrate its effectiveness for target detection in complex road environments.