Multi-Object Detection in Complex Traffic Scenarios Based on Improved YOLOv11n

In intricate traffic scenarios, targets can obstruct each other, distant small targets are difficult to detect, and weather and lighting conditions are variable, making it challenging for target detection algorithms to maintain high detection accuracy while ensuring model efficiency. For this reason, this study presents a multi-object detection model, LCTYOLO11, based on YOLOv11n. In order to improve the model's performance without significantly increasing the parameters, this paper designs the Ddws-Conv module to replace the CBS module in the original network. To mitigate the potential feature loss caused by lightweighting, this paper incorporates an EMA attention mechanism layer into the backbone of the original YOLOv11. This improvement bolsters the feature representation capability of the target in intricate traffic scenarios. In addition, in order to improve the model's detection capacity for targets of varying sizes, this article incorporates small target detecting heads into the Head part of the YOLOv11. To verify the validity of the method, this study conducts experiments on the SODA10M dataset released by Huawei. The experimental results indicate that the LCT-YOLO11 model exhibits a substantial enhancement in performance compared to the original YOLOv11, with increases of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4.3 \%, 4.5 \%$</tex>, and 5.8 % for <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$P, R$</tex>, and MAP50, respectively.