Research On 3D Reconstruction Technology Of Accident Scene Based On Computer Vision

With the rapid development of China's economy and the continuous improvement of transportation facilities,the number of motor vehicles in China has also been rising,but the number of traffic accidents has remained high,leading to major casualties and property losses.In the handling of traffic accidents,China still uses manual measurement methods,which have long measurement time and low efficiency,and may also cause secondary accidents.Therefore,in order to ensure road traffic safety,this paper researches and develops an accident scene 3D reconstruction system and visual display platform based on 3D reconstruction technology of accident scene.This system can restore static traffic accident scene to 3D model and archive it to the accident scene visualization platform.The on-site measurement time is saved for the traffic police and road officials to extract the information to assist them in further analysis of the accident,and to review the dissenting part at any time.At the same time,through the distribution map and heat map in the visualization platform,the accident occurrence point is visually seen,and the relevant management personnel provide support during the black spot management of the accident.The research in this paper is mainly carried out from the following aspects:(1)Research on 3D sparse reconstruction of accident scene based on motion recovery structure method.By analyzing and comparing the advantages and disadvantages of several 3D reconstruction algorithms and their applicable scenarios,this paper selects the Motion Recovery Structure Method(SFM)as the method of 3D sparse reconstruction.Firstly,the principle of feature point detection and matching of traffic accident images is studied.Then,the coordinates of feature points in three-dimensional space are obtained by using the triangle point method.Then,the incremental reconstruction method is used to reconstruct the three-dimensional sparse point cloud,and finally the bundle adjustment is performed.Get accurate 3D sparse point clouds.The results show that the results of 3D sparse reconstruction can only roughly describe the contour of the scene,the surface details are rough,and there are holes,which can not be directly used for the measurement of 3D reconstruction of the accident site.It is necessary to further optimize the generated 3D sparse point cloud.(2)Three-dimensional dense reconstruction and texture reconstruction of the accident site.According to the 3D sparse reconstruction of the accident scene,the point cloud is too sparse.Based on the 3D sparse reconstruction,the generated 3D sparsepoint cloud is further densely reconstructed by PMVS method,and then the dense reconstruction result is meshed.Finally,The scene surface texture mapping is completed.The results show that the surface texture reconstruction results in rich surface texture,which can better reproduce the 3D scene of the accident scene.(3)Design and development of 3D reconstruction system for accident scenes.Aiming at the 3D reconstruction algorithm studied in this paper,the desktop program of the accident scene 3D reconstruction system was designed and developed.At the same time,a visualization platform based on Web page form was developed for the result display.(4)Experimental case verification.Combined with the accident scene 3D reconstruction system developed in this paper,a high-efficiency scene shooting scheme for traffic accidents is proposed through experimental tests.At the same time,the experiment demonstrates how to use the system to quickly measure the accident site.The results show that the reconstruction result of the system can reach 97.98%,and the time of the accident site survey can be greatly reduced,so as to restore the road traffic operation as soon as possible and guarantee the accident scene.The personal safety of the survey personnel.