Traffic sign recognition based on YOLOX in extreme weather

With the rise of machine learning and deep learning, the accuracy and speed of traffic sign recognition in the direction of computer vision have been continuously improved in recent years, but most of the literature and research are based on clear unobstructed This paper studies the problem of traffic sign recognition that becomes blurred in rainy and snowy weather, fog and haze weather, and the speed of the vehicle is too fast. The newly proposed YOLOX model is used to train and test the data set. The final experimental results show that the accuracy of using the YOLOX model to identify pictures after deblurring is improved, which is better than other models, but the types of pictures are not rich enough. The similarity with the real traffic scene needs to be improved.