Model Identification Method Based On Sparse Coding

Urban traffic control system,one of the key branches of the Intelligent Transportation System.Urban Transport for the development of intelligent traffic regulation a central role,in the transport sector as well as pattern recognition,are playing a pivotal role.Along with social and economic development,and enhance the level of public life,motor vehicle showing an increasing trend year by year,the growth of private owned vehicle,it is leaps and bounds,so,they induce a series of traffic problems.An increase in traffic violations and accidents,to the city traffic and personal safety and security.Therefore,traffic management,higher demands,because the reality can not be on the road and crossing all the cities within the 24-hour all-round "human surveillance," if want to solve this problem,we need to rely on scientific and technological means,to take "machine vision and image processing" technology,the creation of intelligent traffic monitoring system,the implementation of the models from the vehicle detection and recognition.In this paper,models classification based on the vehicle image a study by a multi-view vehicle classification and sparse coding,to design an effective vehicle recognition method,the camera captured video of the vehicle within the image,the implementation of positioning and vehicle classification,specific as follows:1)Boundary localization.On the computer,analyzing and processing the collected images,obtaining contour information of the vehicle,namely the Department of length,width,higher data of the vehicle.2)The HOG features and SVM,classifier constructed multi-angle,vehicle classification treatment Perspective image estimation.3)To determine the viewing angle of the vehicle image data through different perspectives,using sparse coding extract effective models feature;under each perspective separately sparse coding.4)The sparse coding and the results of the Perspective estimated each perspective,the merger,the final design of the vehicle characteristics,using SVM for vehicle classification,judged models.The results show that the correct classification rate sparse coding method in traditional word packet method,adopt a multi-perspective means of identification,can improve recognition accuracy by combining multi-class classifier means of implementation model image classification division,but also to effectively enhance the classification accuracy.