The Research on Traffic Object Detection Algorithm Based on Attention Mechanism and Multi-Scale Optimization

In intelligent driving scenarios, traffic object detection faces challenges such as object occlusion and difficulties in recognizing small-scale targets. To improve detection accuracy and real-time performance in this context, this paper proposes an improved YOLO11-based object detection algorithm. By introducing the CBAM attention mechanism, the model enhances feature selection capability in complex environments; a new smallobject detection layer is added to improve the detection of small targets; and the GhostConv module is employed to reduce computational costs. To verify the effectiveness of the model, comparative experiments with current mainstream deep learning models and ablation studies on individual modules were conducted. The results of the comparison experiments show that the proposed method improves recall rate, mAP50, and mAP50-95 on the KITTI dataset by $8.1 \%, 3.4 \%$, and 3.5 %, respectively, compared to the second-best YOLO11 model. Ablation study results further confirm the effective contribution of each improved module. Overall, the experimental results demonstrate that the proposed algorithm achieves a better balance between accuracy and inference efficiency, providing an efficient and feasible solution for real-time traffic object detection in intelligent driving scenarios.