Research On Traffic Accident Reconstruction Based On Multi-mode Images

Fast,accurate,comprehensive,and secure fixation of evidence on the scene of traffic accidents is a necessary guarantee for ensuring the truth of the accident and the fair,just,scientific and objective handling of traffic accidents.The current domestic and foreign fixation of evidence on the scene of traffic accidents have remained at the stage of two-dimensional line drawings and two-dimensional on-site photographs due to limitations of equipment,methods,and vision,which lead to the loss of a large amount of evidence and the misjudgment cases happened.In order to solve this problem,in 2012 the research team of the Third Military Medical University took the lead in using multi-rotor UAVs in the rapid collection and analysis of evidence on the scene of traffic accidents,and obtained evidence on the scene of the accident through a series of overhead sequence images on the road site.It has been widely used throughout the country.However,the traffic accident is a three-dimensional closed-loop system composed of four elements: human-vehicle-road-environment.The traditional evidence collection method only can be cut in two dimensions,resulting in the loss of a large amount of useful information.Foreign countries used three-dimensional laser scanners to record on-site evidence ever.However,the application of this technology is limited in practical,due to its high price,complicated operation,and complex analysis process.Therefore,it has become a hot topic for current accident appraisers to search for a analytical method of three-dimensional evidence fixed that is economical,practical,easy to operate,and easy to analyze.This project mainly relies on camera measurement and three-dimensional reconstruction techniques to reconstruct traffic accident scenes.The reconstruction includes accident scene pavement,accident deformed vehicles,pedestrian corpses and other information,and the using of model point cloud three-dimensional coordinate information and tilt photography volume the principle of measurement to extract information such as accident site pavement and vehicle deformation etc,and analyze its application in the accident speed,the cause of the accident and compare with traditional methods and error analysis,which mainly solves the following problems:1.Due to the lack of GPS information,the normal camera can not obtain the real size of the model.This paper builds a model coordinate system by manually adding control points on the basis of existing image calibration principles,and carries outself-calibration modeling of accident vehicles,the results indicated that the relative error of the model after calibration is 0.81%.2.In view of the difficulty in measuring road slope angle and curve arc length by using a tape measure in real situations,the model uses three-dimensional coordinate points of the model for distance measurement,including the conversion between the WGS84 coordinate system and the national coordinate system,and the inverse solution of the height difference of the plane distance.Calculation,calculation of arc length radius,format conversion between 3D processing software and 2D industrial software,experimental results show that the 3D reconstruction can obtain information such as road surface curve,slope angle,etc.The relative error of measurement can be controlled below 1.9%.This is of great significance for the identification of traffic accidents.3.For the problems of irregular deformation measurement of vehicles,traditional two-dimensional methods with large measurement errors and low efficiency,this paper makes use of the principle of tilt photography for earthwork volume measurement to calculate the volume of accident vehicles before and after deformation,and verify the irregular volume through experiments.The relative error of measurement is 2.9%;Calculate the pre-collision velocity of a real case using the deformation volume measured in this article.Compared with the traditional deformation measurement method,the measurement accuracy is improved by 5.8%.The experimental results show that the multi-mode image based 3D reconstruction and information analysis provide a solution for the rapid collection of road traffic accident scene data,and establish a method for the acquisition and analysis of three-dimensional evidence of traffic accident scenes that can be used for actual combat.This method has strong practical significance and application value.