A Traffic Accident Scene Swift Reconnaissance System Of Mobile Vision

【摘要】：The current traffic accident scene investigation method can not meet the demand of the social and transportation,considering the situation,a traffic accident swift reconnaissance system based on single mobile vision is proposed.。Imaged the coded targets distributed in the accident scene by the CCD camera in several locations and orientations,making use of the uniqueness and feature station's geometric constraint of coded target,can realize automatic recognition and match of coded target and precise space scale constraint,then using fast bundle adjustment to calculate the initial external location and orientation of each camera and the 3D coordinate of space feature points,after that,running the process of Post-processing to correct or optimize the parameter.Before running the full-scale bundle adjustment algorithm to get the final 3D coordinate of the space feature points which is needed to been measured,it's better to repeat the process of fast bundle adjustment and Post-processing several times,so the final three-dimensional information of the traffic accident scene will be calculated with high precision.The whole solving process is done automatically,so it is fast and objective.The experiment results demonstrate that the proposed method is feasible and flexible,and the error ratio is less then 0.06%.