基于Faster R-CNN模型的交通信号灯检测识别研究  

Traffic Signal Detection and Recognition Based on Faster R-CNN Model

作　　者：李靖博 LI Jingbo(Shanxi Intelligent Transportation Research Institute Co.,Ltd.,Taiyuan,Shanxi 030032,China)

机构地区：[1]山西省智慧交通研究院有限公司,山西太原030032

出　　处：《山西交通科技》2024年第3期115-118,共4页Shanxi Science & Technology of Transportation

摘　　要：随着城市交通的不断发展和智能化进程的推进,交通信号灯的准确检测与识别对于交通安全和效率的提升显得尤为重要。在传统识别算法的基础上,研究使用Faster R-CNN算法实现交通信号灯高效准确的检测与识别。通过使用人工标注的数据集,并对数据进行预处理,确保数据质量与准确性;基于PyTorch框架上构建Faster R-CNN模型,并经过训练保证模型收敛;在模型评价方面,使用准确率、召回率等指标对模型性能进行了全面评估。试验结果表明,研究所提出的模型在交通信号灯检测与识别任务中表现出色,预测准确率达到90%以上,对交通管理和智能交通系统的发展具有积极意义。With the continuous development of urban transportation and the advancement of intelligent processes, the accurate detection and recognition of traffic signals become crucial for improving traffic safety and efficiency. Based on the conventional recognition algorithm, this paper proposes to use the Faster R-CNN algorithm to achieve efficient and accurate detection and recognition of traffic signals. The data quality and accuracy are ensured by manually labeled datasets with data preprocessing and labeling. The Faster R-CNN model is established based on the PyTorch framework and is trained to ensure model convergence. In terms of model evaluation, the model performance is comprehensively evaluated using indicators such as accuracy and recall rate. The results show that the proposed model performs well in the traffic signal detection and recognition tasks, with a prediction accuracy of more than 90%, which has positive significance for developing traffic management and intelligent transportation systems.

关 键 词：智能交通系统 交通信号灯 Faster R-CNN 多类别分类 

分 类 号：U491.51[交通运输工程—交通运输规划与管理]