Research On Traffic Sign Recognition Algorithm Based On Single Shot Detector

In today's society,traffic accidents are becoming increasingly frequent.These frequent traffic accidents have caused major losses to today's society.Therefore,Advanced Driver Assistance System(ADAS)has emerged at the historic moment.Automatic traffic recognition and detection technology is an important part of it.However,due to partial occlusion,different viewpoints,lighting and weather conditions,there are still many problems in the recognition of traffic signs,and further research is needed.In order to further solve these problems,this paper uses machine learning and other related knowledge to conduct research on traffic sign recognition methods,and selects the SSD(Single Shot multibox Detector)algorithm as the traffic sign target.On this basis,a new traffic sign detection algorithm is proposed,which can improve the detection accuracy,and has high efficiency and real-time performance.The main work and innovations of this article are as follows:(1)The traditional recognition methods of traffic signs are studied,including traffic sign recognition based on visual information and traffic sign recognition based on feature extraction and machine learning.Summarize and analyze the advantages and disadvantages of this method.(2)The algorithms related to neural networks and common target recognition methods are studied and analyzed.Among them,Faster R-CNN and SSD target detection methods are mainly studied,and then they are applied to traffic sign recognition.The advantages of this method are summarized and analyzed.under.(3)By analyzing the difficulties of real-time traffic sign detection,and then analyzing the SSD network and its recognition results,it is found that the low-level traits of the network contain more information about small targets and less information about large targets,while high-level features include With more large target information and less small target information,a traffic sign detection algorithm(MF-SSD)based on feature fusion and enhancement is proposed,which can enhance the effect of effective channels.It achieves the goal of small target detection rate,and has a good detection effect on the recognition of large target traffic signs.