In order to accurately detect real-time changes in traffic signals, and traffic signals belong to small targets with high recognition difficulty, this paper proposes to use YOLOV5L to solve the problem of traffic signal recognition. This system can achieve recognition of traffic signals under target images and real-time monitoring. This system uses the Pytorch framework, which can provide high-precision real-time monitoring and recognition functions. Finally, experimental operations were conducted on a self-made traffic signal dataset, and the experimental results were obtained mAP@.50 Finally, it stabilized at 78.6%. The YOLOV5L model can basically meet the detection and recognition of traffic signals.