基于CNN技术和DETR的智能汽车自动驾驶道路智能识别的研究

Research on Intelligent Vehicle A utonomous Driving Road Recognition Based on CNN Technology and DETR

文章针对智能汽车自动驾驶道路智能识别的问题,提出了采用DETR模型进行道路智能识别的方法,并对DETR模型进行了改进。在DETR模型基础上,阐述了在道路智能识别复杂问题处理上的不足,针对存在问题,在DETR加入了Swin Transformer模块,提高了道路目标检测性能,同时采用基于多头自注意力机制,实现了道路多目标的高精度识别,达到了模型优化的目的,通过实验与结果分析,将改进后的DETR网络与其他常见的识别算法进行比较,结果表明改进后的DETR网络在准确率、召回率和平均精度均值优于其他模型。

This article proposes a method of using the DETR model for intelligent road recognition in intelligent vehicle autonomous driving,and improves the DETR model.based on the DETR model,the shortcomings in dealing with complex problems in road intelligent recognition were elaborated.In response to the existing problems,the Swin Transformer module was added to DETR to improve the performance of road object detection.At the same time,a multi head self attention mechanism was adopted to achieve high-precision recognition of road multiple targets,achieving the goal of model optimization,this article compares the improved DETR network with other common recognition algorithms through experiments and result analysis.The results show that the improved DETR network outperforms other models in terms of accuracy,recall,and average accuracy.