Fast Detection And Recognition Of Traffic Signs Based On Convolutional Neural Network

With the rapid development of my country's economy and society in recent years,the urbanization rate is also getting higher and higher.From the perspective of social structure,more and more rural population will gradually be transformed into urban population,which has brought tremendous development momentum to my country's urban development.At the same time,it also brings various problems,including the traffic congestion caused by the rapid expansion of cities and the increasing road safety problems.In order to solve the current traffic congestion and road safety problems,the intelligence of the transportation system has inevitably received attention.The most important point for intelligent transportation systems,including unmanned driving technology,to enter the practical stage is to be able to informationize various traffic conditions on the road in real time.How to eliminate the interference from environmental factors as much as possible and carry out accurate and efficient traffic sign detection and recognition is a key technical problem that must be solved,which has great research value and broad application prospects.The traditional road traffic sign recognition first uses the characteristics of the image color and shape to detect the area of the traffic sign image,and then uses template matching,SVM classifier or neural network to recognize the traffic sign.In recent years,convolutional neural networks have been greatly developed.Compared with traditional road traffic sign detection and recognition methods,the method based on convolutional neural network can effectively improve the processing speed of traffic sign detection.As an important part of unmanned driving system and automatic driving assistance system,the detection and recognition of traffic signs has a direct impact on overall driving safety.Aiming at the problem that the YOLO network model has low accuracy in identifying small targets,this paper uses the YOLO network model to quickly detect and locate 43 types of traffic signs in the German traffic sign detection benchmark traffic sign dataset.Then,the detected traffic sign areas are roughly classified into four categories:"instruction","warning","ban" and "other".Finally,the cascaded LeNet-5 model is used to quickly and subdivide the detected four types of traffic sign areas.This paper uses YOLO network model as a traffic sign detector,and LeNet-5 model as a fine classifier of traffic signs to detect and recognize traffic signs in street scene images.It improves the accuracy of classifying small target traffic signs in street view images,and realizes rapid detection and recognition of traffic signs in street view images.