The Research Of Resolution Algorithm For Camera External Location And Orientation Of Traffic Accident Vision 3D Swift Recovery

With the rapid development of our country’s economy and rapid increase of cars, the traffic accidents are obviously increasing. All these can cause huge personnel casualty,property loss and serious traffic jams. So how to rapidly impersonally deal with traffic accidents and lead transportation becomes the focus of related researchers of each country.With the development of computer vision, photography is widely used for traffic accident scene recovery. In the process of recovery based on vision measurement, it concerns several steps such as camera parameters resolution, feature points recognition, feature points matching, 3D reconstruction and so on. And camera external orientation resolution is the most important thing we need to do because its accuracy determines the precision of the whole recovery system directly.There have been mainly two traffic accident vision swift recovery systems researched and adopted currently. One is 2D recovery system, and it is possible that the recovery can be accomplished in the even and straight road, but much error exists in the ramp or crook; The other is 3D recovery system, because the 3D feature points collected from accident scene are no-coded targets, the corresponding points do not match automatically and must be matched manually. Therefore the workload increases and recovery error also augments due to wrong matches. Meanwhile much recovery error exists because the estimation length among targets is regarded as scale basis. For accident vision swift recovery, many pieces of foreign recovery software such as Photomodeler, Pc-rect, Shapecapture and Iwitness adopt special camera external orientation resolution methods, but the technique is not published.For the requirement for external orientation resolving precision of CCD camera in the traffic accident vision 3D swift recovery, this paper adopts resolution method based on coded target. The coded targets distributed in the accident scene are imaged by the CCD camera in several locations and orientations. Utilizing the space distance constraint among coded target feature locations, the absolute location of initial camera base station is determined based on co-angular constraint pose estimate algorithm, and the relative location of camera stations can be got through coplanar condition epipolar constraint resolution method and uniqueness correspondent relation among coded targets. The precise relative location resolution of camera stations relative to word coordinate system can be resolved through space coordinate transformation. Because the space, coding and scale information of coded target and full-scale bundle adjustment algorithm are considered, this method can get high resolving precision and can be used in the accident vision 3D swift recovery. The camera external orientation results of different resolving methods demonstrate that the proposed method is feasible and flexible.