Research On Traffic Sign Recognition Algorithm Based On Deep Learning

With the rapid development of the social economy and the improvement of people's living standards,the output of cars is increasing year by year,road traffic congestion and traffic accidents have occurred frequently,and intelligent transportation system have emerged.As an important part of intelligent transportation system,traffic sign recognition has been widely concerned by scholars at home and abroad.However,the actual road environment is variable and complex,which brings challenges to the detection and recognition of traffic signs.The traffic sign recognition with high accuracy and good real-time performance needs to be solved urgently.The research on the detection and recognition of traffic signs has theoretical value and practical significance.The main work of this paper is as follows:1)In view of the low accuracy of small target traffic sign detection,the algorithm is improved based on the existing Tiny-YOLOv3 traffic sign detection algorithm.The level of the Tiny-YOLOv3 backbone network is deepened to extract higher-level semantic information,a feature pyramid layer is added to obtain different scale feature map to predict the location and category of the image.The experimental results show that the improved Tiny-YOLOv3 network model has high detection accuracy,good generalization ability,and good robust performance.The proposed algorithm is examined on the LISA dataset,the mAP(mean Average Precision)reached 93.91%,which is 16% higher than the Tiny-YOLOv3 network.2)Aiming at the problems of low accuracy and poor real-time performance of traffic sign recognition,an improved convolution neural network algorithm for traffic sign recognition is proposed.Based on the LeNet-5 network,the convolution kernels of all convolution layers are replaced by 3×3 convolution kernels.The Inception module group is introduced to increase the width and depth of the network.During training,the batch normalization method is used to process the input batch samples.The proposed algorithm has achieved 98.59% recognition accuracy on the German Traffic Sign Recognition Benchmark(GTSRB),and the recognition time of each image is about 0.158 ms.The experimental results show that the algorithm has high accuracy and good real-time performance.