Research On Road Traffic Sign Recognition Based On Depth Learning

With the rapid development of China's modernization,a lot of gratifying achievements have been made in the construction of transportation infrastructure.However,the traffic safety problems are still grim.According to reliable estimates,the annual economic losses caused by traffic safety in China are as high as those of the previous year Billions of dollars,but also brought a tragedy of a large number of casualties,these tragedies for the victims and their families,whether psychological or physical harm will have a very important impact,which will be directly or indirectly Therefore,it is obvious that developing intelligent transportation systems to standardize traffic safety and guiding drivers to drive safely is an urgent problem to be solved.Although many intelligent sign recognition systems have been born,But the ability to accurately identify traffic signs is still limited.Therefore,how to improve the traffic sign recognition ability is an important issue.Deep learning as a key component of machine vision,plays an important role especially in image processing.The Convolutional Neural Network?CNN?has always played a very important role.At present,In many fields,it has achieved good results.It has achieved remarkable success in image processing and target detection and recognition tasks.Therefore,this paper applies it to the task of traffic sign recognition and makes some improvement and optimization of convolution neural network It is more adaptable to the identification of traffic signs.Experiments show that the OVGG?referred to as optimized VGG,referred to as OVGG?model in this paper has a better performance in traffic sign recognition tasks.The main contents of this paper are as follows:?1?Using convolutional neural network to recognize traffic signs.Traditional methods for feature extraction have many shortcomings.For example,the extracted information is not comprehensive,the features are damaged or the computational complexity is increased due to too much manual intervention.The most important point is that the traditional feature extraction method takes a long time,Can not meet the real-time needs of traffic sign recognition.To solve this problem,this paper presents an optimized VGG model for road traffic sign recognition.Experiments show that compared with the traditional traffic sign recognition method,the traffic sign recognition based on improved convolutional neural network VGG model proposed in this paper has higher recognition rate and lower time-consuming.?2?The mainstream convolutional neural network VGG model such as VGG₁6,although it has a good recognition effect,has its own shortcomings,such as the complexity of the model calculation and the computer resource occupancy,and as the model level deepens,training The parameters are also more and more complex,time-consuming and more and more,which for real-time requirements there is a big gap.In view of the fact that the depth of VGG₁6 is too long,which leads to the increase of training time,too many parameters and slow convergence,this paper constructs a 9-layer optimized convolutional network OVGG model.At the same time,the methods of data enhancement and migration learning are used to speed up the convergence of the network model.The experimental results show that the OVGG model in this paper has higher recognition effect and lower time consumption.