基于改进YOLOv11n的车辆目标检测算法

打开文本图片集

主题词：复杂交通场景 车辆目标检测 YOLOv11n C3k2-CTA C2CGA中图分类号：TP391.41 文献标志码：A DOI：10.19620/j.cnki.1000-3703.20250421

Vehicle Target Detection Algorithm Based on Improved YOLOv11n

Zhou Jianxin，Hao Yingjie，Li Zhongze，Hou Zichuan （North China Universityof Scienceand Technology，College of Electrical Engineering，Tangshan 063210）

【Abstract】To address the degradation of feature representation and detection accuracy caused by occlusions and cluttered backgrounds incomplex trafic scenes，this paper proposeanimproved detector，CCT-YOLObasedonYOLOv11n. First，a C3k2-CTA moduleisdesignedbasedonthefrequency-aware mechanismFCA-CTAto enhancemulti-scalefeature discriminationviafrequencydomainchannelatention.Second，byitegatingagroupingstrategywithspatialatentioCGA， the C2CGA module is proposed，which reduces parameters while improving detectionofoccluded objects.Finall，a TaskAlignedDynamic Head（TADH）isdevised thatemployssharedconvolutions tocutmodel sizeand strengthen fine-detail capture. Experiments on KITTI and BDD10OK show that CCT-YOLO achieves mAP of 89.5% and 52.6%， respectively， with a15.5% reduction in parameters，demonstrating the effectiveness of the proposed approach.

Key words：Complex traffic scenarios，Vehicle object detection， YOLOvl1n， C3k2-CTA，C2CGA

【引用格式】周建新，郝英杰，李忠泽，等.基于改进YOLOv11n的车辆目标检测算法[J].汽车技术，2025（9）：1-9.ZHOUJX，HAOYJ，LIZZ，etal.Vehicle TargetDetectionAlgorithmBasedonImprovedYOLOv11n[J].AutomobileTechnology，2025（9）： 1-9.

1前言

随着自动驾驶技术的快速发展，交通场景的目标检测已成为保障行车安全的核心技术，提升检测算法的实时性、鲁棒性和泛化性，在智能驾驶系统中具有重要的应用价值。（剩余14259字）