Research On Vehicle Information Recognition Technology

With the progress of the times and the development of the economy,automobiles have become a common way of transportation.In order to develop the intelligent transportation system,manage the vehicle efficiently and reasonably,obtain the real information of the vehicle,and prevent the illegal phenomena such as illegal modification,the information identification technology of license plate,vehicle logo,vehicle color and vehicle type is crucial.This paper mainly studies two aspects of vehicle information recognition technology: license plate recognition and vehicle logo recognition.In the aspect of license plate recognition,this paper studies license plate extraction and license plate character recognition.In the aspect of vehicle logo recognition,this paper mainly improves the classification algorithm of vehicle logo.The part of license plate extraction: A license plate extraction method based on multifeature combination is designed and implemented.Based on the characteristics of domestic license plates,the license plate location method based on color information,edge detection and MSER is designed and implemented.Then,the license plate tilt correction method based on linear fitting is completed.Finally,the SVM license plate classifier is trained to complete the license plate screening based on the relevant characteristics of the license plate candidate areas.Compared with license plate extraction using single feature,the proposed method has better accuracy.The part of license plate character recognition: A method of license plate character recognition based on convolution neural network is designed and implemented.Based on the characteristics of license plate characters in China,an improved character segmentation method based on connected region is designed and implemented.The segmentation task is completed by determining the position of Chinese characters.This method is simple and efficient.A method of license plate character recognition based on convolution neural network is designed and implemented.A 7-layer and 12-layer convolution neural network is built based on LeNet-5 and VGG16 respectively.Compared with the traditional method,the 7-layer convolutional neural network constructed in this paper has higher character recognition accuracy.The part of vehicle logo classification: A vehicle label classification method based on Capsule network is designed and implemented.Applying the Capsule network based on dynamic routing to vehicle logo classification,feature clustering is realized within the network,which greatly reduces the number of weight updates,and at the same time,it can get more structural features which are more conducive to classification.Compared with the traditional method,the proposed vehicle classification method based on Capsule network has higher classification accuracy of vehicle classification.