| Gradually improve and rising vehicle ownership in the national road network makes the traffic safety issue has become a down to ordinary people, the country attaches great importance to the issue. In the accident, the accident caused due to vehicle overrun about 70%of the total accidents. Many scholars dedicated to overrun detection, but the focus will be overrun detection on overweight detection, ignoring the contour of the vehicle overrun issue.At present, the overrun detection method of vehicle contour is very limited, or even stay in manual measurement levels in China. In order to meet the needs of road safety and intelligent detection,Against the current problems and other issues of traffic safety caused by vehicle overrun, designed the vehicle contour overrun detection system based on machine vision,detect the target vehicle, over-long, over-wide, over-high problem.Article relatively comprehensive expounded current China’s vehicle over-long, over-wide,over-high detection methods. Analyzed the advantages and disadvantages of each method and put forward a detection system of vehicle overrun based on machine vision. Use CCD cameras get the video of vehicle in the speed limited lane, after processing and measurement,display the results to the owner at the toll station.The video processing is divided into contour segmentation of target vehicle, extraction and positioning, measuring of the target vehicle. In this paper, image preprocessing filter out interference, then use three video frame difference and Canny operator processing and get the preliminary extraction image of the vehicle, and after use the vote filtering based on pixel template, gets the final extracted image. According to the invariance of the lane in the image, preliminary extraction of the vehicle image position,then use the smallest rectangle method accurately locate the position of the vehicle. Extracting the vehicle contour length information under pixel planes in the smallest ROI, by modeling for the camera calibration of the measurement system calculation and use the calibration algorithm based on virtual calibration object,converted the vehicle contour length of the pixel to the actual contour information in world coordinate system. Finally, compare the measured results with the national legal requirements and get the information whether the vehicle is overrun.Through the vehicle field shooting video and OPEN-CV image library, process the acquired video, verification the feasibility of algorithm of this paper. Finally, built hardware platform and use VC++ software design PC interface, achieve the full algorithm that has beenachieved. The proposed system verify the feasibility that the vehicle contour overrun problems based on machine vision inspection, opened up a new prospect. |