| Vehicle violation identification systems based on video integrate thecomputer vision technology into video surveillance, which determines whether avehicle is illegal or not through the analysis and processing of video se quencesand then keep a record of it. This thesis uses the image processing techniques toidentify vehicle violation behaviors, such as the covering of traffic-line, notdriving along the road signs, drivers not wearing a seatbelt and using mobilephones while driving, and so on, from vehicle traffic surveillance video images.In this paper, the main research work are as follows：（1）In order to overcome the influence of vehicle occlusion while extractsthe road line, this paper first uses the method of mixed Gauss model to extractthe background of the scene, and then on this basis, a kind of road routeextraction algorithm based on edge detection and bipolar fusion is studied.（2）Vehicle route violation includes traffic-line covering and not drivingalong the sign direction. Based on Canny detection algorithm and morphologicalimage processing method, and combining the detection of target centroid,histogram, texture characteristics, this paper uses the local feature matchingmethod to achieve the vehicle moving target tracking.（3）A projection histogram based on image edge detection algorithm isproposed to locate the windows and the driver, then uses the linear filteringtemplate for the driver’s figure for filtering, and then analyzes the linear imageobtained from Hough transform according to different cases, which realizes therecognition of driver’s seatbelt wearing.（4）Establish the Gauss model of color to find out maximum likelihooddegree of skin color， which contributes to the human skin color blocksegmentation. Then according to the characteristics, filters color blocksincluding facial skin and extracts its features, the resulting features as the inputsof BP artificial neural network for identification, which can recognize whetherdrivers use cell phone while driving. |
PDF Full Text Request