Application And Research Of Binocular Stereo Vision In Traffic Accident Survey

Road traffic safety has become the major social issue in the global range. In recent years, itshows rising trend about road traffic accident in China, and the position of traffic safety is stillausterity. The process of handing traffic accidents is mainly divided into three stages: The firstis investigation and evidence-taking of the traffic accident scene, the second is reason-analyzingand process–analyzing of traffic accident, the third is duty-affirming of traffic accidentaccording to the result of analysis. In above three stages, the first stage work completes in trafficaccident scene. The duration time of the first-stage decides the degree of the traffic affecteddirectly. Photogrammetry is a highly effective method for realizing intelligent processing oftraffic accident scene. Binocular stereo vision imitation human by binocular cognitive distanceprinciple, it shoots two images that get from the same scene by two cameras which is fixeddifferent position. This method can obtained the depth information of object and distanceinformation of target by calculating the parallax of spatial point in two images.We cooperate with traffic administration bureau of HeBei. Correlational researchachievement both at home and abroad in handing traffic accidents were studied to put forward acomplete scheme of binocular stereo vision in handing traffic accidents. The paper firstintroduced the principle of digital camera and the model of pinhole imaging, and rangingprinciple of binocular stereo vision is emphatically analyzed. Thus makes it possible tomeasurement of three-dimensional object in handing traffic accidents. The paper summarizesthe usual method of pre-process for image in photogrammetry of handing traffic accidents, andpresents to direction gradient filtering. We use the morphologic edge detection for pro processing of image, and it has obtained a good effect. We find an effective algorithm to contrastenhancement for poor contrast image. It provides fully preparation for later processing of image.Extracting and matching for feature point is emphasis and difficulty in binocular stereo vision.This paper finds an algorithm for feature point extracting and matching which suitable forapplication to photogrammetry in handing traffic accidents, it named SIFT and has theadvantage of anti-constrained scaling and anti-rotation. The experiment show the binocularstereo vision in traffic accident survey system established by the paper theory is sample, rapid,accurate location and easy to complement and repeat measuring. It can reach high measurementprecision and meet demand.