Research On Vehicle Recognition Algorithm Based On Video Image

Nowadays,with the increasing number of automobiles,traffic accidents,traffic congestion and so on are becoming more and more serious,which is difficult to solve by relying solely on labor management.Therefore,intelligent transportation system has received extensive attention.In recent years,it promotes the development of intelligent transportation system with the rapid development of computer vision technology and pattern recognition technology.Intelligent transportation system is the development trend of traffic;and vehicle recognition technology is an important part of intelligent transportation system.Vehicle recognition technology has been introduced into real life applications,such as traffic management,public car parks,freeway toll stations and so on.However,due to a wide variety of vehicles,the mixed traffic scene,the realization of technical complexity and so on,vehicle recognition technology has not yet achieved good application in practice;therefore,vehicle recognition technology has great development space and research significance now.This paper mainly studies the vehicle recognition algorithm based on video images,which is based on the relevant research results at home and abroad.In this paper,vehicle recognition system is designed,which is based on SVM.The video images which are collected by the video camera device are inputted to the vehicle recognition system;then the system uses the extracted features to classify the vehicles in the video.Based on the study and research of the existing vehicle recognition algorithm,the vehicle recognition system is divided into three parts: moving vehicle detection,vehicle feature extraction and vehicle recognition.(1)Moving vehicle detection.After the video image preprocessing,the foreground image of the moving vehicle is obtained by combining the background difference method with the efficient background modeling method which is based on sparse representation and outlier iterative removal.Due to video images are inevitably polluted by noise in the process of acquisition and transmission,this paper uses the adaptive weighted median filtering to solve this problem,which ensures the accuracy and integrity of the vehicle extraction;and the shaded portions generated when the vehicle is moving is removed by the color feature information of the vehicle image.(2)Vehicle feature extraction.According to the requirements of vehicle classification,we focus on the geometric features,the histogram of gradient and the local binary pattern.In this paper,the method of feature fusion is used to identify the vehicle type,which can solve the problem that a single feature is easy to be affected by illumination,weather,shadow and so on.The experimental results show that the HOG-LBP fusion feature can improve the accuracy of vehicle recognition.(3)Vehicle recognition.A vehicle classification system is designed based on SVM(Support Vector Machine)classifier.The experimental results show that the system is very good to achieve the car,truck and bus three models of identification and classification.