The Research On Traffic Sign Method On Deep Learning Neutral Network

In recent years,with the increasing maturity of the traffic conditions and the gradual development of the road infrastructure,especially in the more developed cities,road traffic has been dominant and traffic signs are significantly important for the smooth operation of road transport.Furthermore,with the development of intelligent vehicles,more and more researchers pay attention to the automatic recognition of road traffic signs as the one of the basic techniques of intelligent vehicles.In the actual traffic situation,the quality of the collected traffic sign is often poor due to motion blur,light interference,weather conditions and other factors,which makes automatic recognition of traffic signs difficult and puts forward high requests to the accuracy,robustness and real-time of the recognition method.So,how to fast,accurately and efficiently recognize traffic signs has important significance.The recognition of road traffic signs has two basic aspects: the first is to detect traffic signs,including location,extraction and necessary preprocessing;the second is to recognize traffic signs,including feature extraction and classification.The recognition of road traffic signs has two methods,one is the use of "artificial features + machine learning",and the other is based on the deep learning model developed in recent years.This paper has researched on the second method and proposed traffic sign recognition method based on deep neural network and the main researches as follows:(1)Deep learningIn this paper,the theories and models of deep learning are deeply studies and researched which points out that it builds a machine learning model with many hidden layers by simulating the human brain structure to automatically learn more useful features and ultimately improve the classification or prediction accuracy.(2)Traffic sign recognition based on deep convolutional neural networkIn this paper,the basic principles and structure of deep convolutional neural network model are deeply studied,and the method of traffic sign recognition based on deep convolutional neural network is proposed.This method uses deep convolutional neural network model with supervised learning,directly makes the binarization traffic sign images as the input of the network,simulates the human brain's perception of the multi-level structure of visual signals through convolutional layers and sampling layers to automatically extract traffic sign features,and finally uses a fully connected network to recognition the features of traffic signs.Experiment results have shown that this algorithm can automatically extract traffic sign features by using the deep learning ability of convolutional neural network,not only to avoid the extraction of artificial features,and effectively improve the efficiency of traffic sign recognition,with good generalization ability and adaptability range.