Research On Surveying Technology For Traffic Accident Scene Based On Virtual-actual Target Fusion And SUAV Photography

The first step of traffic accident investigation is scene surveying which is significant to the comprehensive and detailed analysis of traffic accident causes.The 2D-DLT method based on single image photogrammetry which is analyzed extensively indicates that much important on-site information,such as vehicle rest positions,tire marks,and debris traces,can be taken from the scene photos.But it has lower survey precision and is not a feasible and practical method because the calibration target cannot cover the whole site region,and it cannot overcome the influence of dip angle of photography on survey precision.Therefore,the 2D-DLT method should be improved to provide new technical methods and theoretical support for traffic accident scene surveying.This study starts with the surveying technology for traffic accident scene,and then obtains a large virtual calibration target through the construction of virtual-actual target fusion model.In order to improve its implementation effect,the actual target is designed,and the detection method of feature point is studied,and the layout analysis is carried out.In order to enhance the applicability of 2D-DLT method,an improved 2D-DLT method based on virtual-actual target fusion is proposed,which can probably overcome the shortcomings of poor precision based on single target.In order to raise the practicability of 2D-DLT method,an improved 2D-DLT method based on SUAV photography is proposed to solve the problem of the low accuracy due to small shooting angle when using ordinary digital camera and 2D-DLT method to survey the traffic accident scene.The main contents of this paper are as follows:The virtual-actual target fusion model is constructed.In order to obtain a large virtual calibration target which is adapted to survey traffic accident scene,the camera imaging process and the lens distortion factors are studied.Through the rotation matrix and the translation vector relative to the camera coordinate system of each actual target in the traffic accident scene,the position of multiple actual targets is unified into the same coordinate system,and then the virtual-actual target fusion model is constructed.The calculation method of the virtual-actual target fusion model's parameter is studied from the calculation of camera intrinsic parameters,the calculation of actual target position and the optimization of the model's parameters.In order to optimize the parameters of the virtual-actual target fusion model,the constraint conditions should included that the euclidean distance of small actual targets is invariant,the position relation between the actual targets satisfies the coplanar relation,and the re-projection error is minimum.Finally,the feasibility and stability of virtual-actual target fusion model construction and parameter calculation are verified by simulation test.Implementation and layout analysis of virtual-actual target fusion is studied.In the construction of virtual-actual target fusion model,actual target is an important foundation.By analyzing the characteristics of the traffic accident scene,the black and white checkerboard of the plane structure is designed.The median image is preprocessed by median filter method,and the characteristic points on actual target are extracted by the SAD method.The comparison of Harris detection method show that SAD method can effectively suppress the appearance of imprecise control points and improve the accuracy of feature point's extraction on actual target.In order to improve the implementation effect of virtual-actual target fusion,its layout is analyzed by indoor experiment.The results show that there are four scattered actual targets in the fixed scene area,which can obtain the maximum coverage area of a virtual target with the minimum number of actual targets.Traffic accident scene surveying based on virtual-actual target fusion is proposed.The virtual-actual target fusion is applied to traffic accident scene surveying,which can overcome the poor precision just using a single target and improve its applicability.By studying the theory of 2D-DLT method,nonlinear optimization method and traditional control point selecting method,the optimal strategy of control points after virtual-actual target fusion is established,and the coordinate data of control points around the area of concern scene are obtained.Then,an improved 2D-DLT method based on virtual-actual target fusion is proposed.The refluence of the region size taken by virtual target after virtual-actual target fusion,the area of actual target when the region taken by virtual target is the same size,and comparison of using a single actual target with using virtual-actual target fusion on the surveying results was investigated by field experiments.The results show that the larger the area occupied by the virtual target,the smaller the relative error of the measured area,the better the result of the photogrammetry.Under the premise of the constant size of the virtual target,the results of photogrammetry are basically the same.The results of surveying when using virtual-actual target fusion are better than those of single actual target.All of this verified the superiority of the improved 2DDLT method based on virtual-actual target fusion.Finally,the feasibility of this improving 2D-DLT method is proved by a specific case study.Traffic accident scene surveying based on SUAV photography is proposed.The ground image can be easily achieved by SUAV photography from the low altitude.Imaging resolution has now met the requirements of the relevant standards.In order to solve the problem of the low accuracy due to small shooting angle when using ordinary digital camera and 2D-DLT method to survey the traffic accident scene,SUAV photography is applied.In order to reduce the problem of solving the illconditioned equation,2D-DLT method based on SUAV photography is proposed,which is firstly to normalize the world coordinates of the object point and the measurement coordinate of the image point,and then to estimate the DLT parameters using singular value decomposition method.The simulation experiments confirmed the validity of the method.Then,the method of estimating the height of photography is studied,which can guide the practical implementation of proposed method.Finally,virtual-actual target fusion is used in 2D-DLT method based on SUAV photography,and the complete method of traffic accident scene surveying based on SUAV photography is constructed.The technique workflow,hardware equipment and test program are designed.The effectiveness of the proposed method is verified by field experiments,and the average error is less than 1%,which was further reduced by using ground photogrammetry with ordinary digital camera.Traffic accident scene surveying based on SUAV photography can not only extend measurement range in a single image photogrammetry and effectively inhibit the shooting angle effect on the 2D-DLT method,but also improve the accident survey accuracy and enhance its practicability.