Abstract: The complexity and variability of traffic scenarios challenge existing human-vehicle target detection algorithms, especially when dealing with occlusion, illumination changes and multi-scale targets, existing algorithms tend to have insufficient accuracy and low computational efficiency. To solve the above problems, an improved detection model, CDC-DETR (CPPA-DWRC-CGNET-DETR), was developed based on the RT-DETR (Real-Time DEtection TRansformers) architecture. Firstly, a Context Pre-activated Pooled Attention (CPPA) module was designed to enhance long-range dependencies and optimize feature extraction. Secondly, a Dilation-wise Residual Connection (DWRC) module was introduced to improve multi-scale feature representation. Thirdly, a lightweight Context Guided Block (CG Block) was proposed to fuse local, surrounding, and global information and reduce computational cost. Finally, these modules were integrated to construct a high-accuracy and efficient real-time human-vehicle detection model suitable for complex traffic scenarios. The experimental results show that, compared to RT-DETR, CDC-DETR improves the mean Average Precision (mAP) when the Intersection over Union (IoU) is 0.5 by 6.1%, increases recall by 4.35% on the BDD100K dataset, while decreasing the number of floating-point operations by 11.2%, significantly enhancing computational efficiency and providing an effective solution for deployment on edge devices.

Key words: assisted driving, human-vehicle detection, Transformer, intelligent traffic perception, multi-scale feature fusion 
 

摘要： 交通场景的复杂性和多变性对现有的人车目标检测算法提出了挑战，尤其在处理遮挡、光照变化和多尺度目标时，现有算法通常精度不足且计算效率较低。为解决上述问题，在RT-DETR(Real-Time DEtection TRansformers)模型的基础上，提出一种改进型检测模型CDC-DETR(CPPA-DWRC-CGNET-DETR)。首先，设计了上下文预激活池化注意力（CPPA）模块，以增强远距离依赖，优化特征提取；其次，引入膨胀残差连接（DWRC）模块，提升多尺度特征表达能力；再次，提出轻量化的上下文引导模块（CG Block），融合局部、周边和全局信息，降低计算成本；最后，结合上述模块，构建了一个适用于复杂交通场景的高精度、高效率实时人车检测模型。实验结果表明，CDC-DETR与RT-DETR相比，在数据集BDD100K上，当交并比(IoU)为0.5时，检测平均精度均值（mAP）提高了6.1%，召回率提升了4.35%，浮点运算量减少了11.2%，显著提高了计算效率，为边缘设备部署提供了高效的解决方案。

关键词: 辅助驾驶, 人车检测, Transformer, 智能交通感知, 多尺度特征融合