Image Information Processing Technology Of Fast Investigation For Road Traffic Accident Scene

In recent 20 years, traffic accident disputes have increased yearly, which had a serious influence on human society, resulting in great loss of life and property, meanwhile, causing a series of legal issues. As a result, it is required to improve the ability of analyzing traffic accident investigation. The lack of the necessary technology and equipment, leading to the delay and mishandled investigation of traffic accidents, particularly the major traffic accidents, inducing a chain reaction of accident and propagation effects of economic loss (domino effect of the accidental loss), what is more, causing enormous indirect economic loss. Therefore, the systematic study and the establishment of traffic accident scene image processing technology system for the fast and effective investigation are necessary. This dissertation provides theoretical references and algorithm basis for the fast, accurate and effective investigation of the traffic accident scene, at the same time, it is of great significance in solving the lack of practical technical problems when investigate traffic accidents.Mapping of the traffic accident scene investigation drawing is the main part of the formation of traffic survey delay. Seeking for a fast and effective investigation technology which is not highly dependent on the operators has become an urgent requirement for the relevant research field. With the existing image processing techniques, we can establish the fast and effective investigation technology system which is based on the image information processing techniques. According to the feature of the traffic accident scene investigation drawing, road traffic accident scene image processing technology system which is based on the geometry correction of Bird's-eye-photo for traffic accident scene was proposed. The system is the expansion of the image mosaics technology of photos which have been geometry corrected; it is also a supplement to the photogrammetry technology based on the 3D reconstruction method. The system provides basic data for the generation of automatic mapping system files so that it can realize fast and effective investigation.This dissertation is based on the Technology of Fast Investigation and Disposal for Major Traffic Accidents, which is one of projects on the 863 State Plan.(2009AA11Z201). According to problems of fast investigation for traffic accident scene, combined with relevant image processing techniques, with the accurate survey and reasonable disposal measures, for the aim of the fast and effective investigation of traffic accidents, an in-depth and extensive research of the road traffic accident scene image processing technology has been done. The detailed contents are as follows:(1) Theory of traffic accident scene image processing technology was analyzed deeply, a fast and effective investigation of traffic accident scene image processing system was established, the two-dimensional approach used in Bird's-eye-photo and the three-dimensional theory used in 3D photogrammetry technology were analyzed in detail. On the basis of the two-dimensional approach, two kinds of gray value mapping method were analyzed. It analyzed the image matching and image fusion about the method of image mosaics systematically. The basic principles of geometric transformations, as well as several basic two-dimensional geometric transformations were analyzed deeply. It analyzed the gray value mapping method using interpolation algorithm. On the basis of the three-dimensional approach, with the relations among the three-dimensional theory used in the world coordinates system, camera coordinates system and image coordinates system, the dissertation analyzed the camera's imaging process and the linear camera model. In order to improve the measurement accuracy of 3D photogrammetry technology, considering the nonlinear distortion of camera lens, the nonlinear camera model was analyzed. Finally, analysis of the theories of 3D reconstruction of point, 3D reconstruction of line and 3D reconstruction of quadratic curve was given.(2) Geometry correction of Bird's-eye-photo for traffic accident scene based on the two-dimensional approach was studied. First of all, on the basis of the two-dimensional coordinate transformation mathematical model, reference point coordinates of the image acquisition methods were analyzed. It used the fast acquisition methods to acquire the actual coordinates of the reference point. Using semi-automatic and automatic ways to obtain the input image coordinates of the reference point. It used the weighted average method to make the image graying. Furthermore, it used the Otsu method to acquire gray image binary segmentation, then corroded the image to extract the reference point coordinates automatically. After that, methods of image registration, proportion determination and color recovery have been done to realize image processing. Finally, the field experiment verified the correction algorithm and proved that method discussed above was practical, in addition, calibration accuracy was analyzed.(3) Image mosaics technology of photos which have been geometry corrected was mainly to show a complete scene of a traffic accident in the process of acquiring information. First of all, it analyzed image matching approaches of image mosaics technology. On the basis of MAD, NCC, and SSDA which are measures of similarity algorithm used commonly, regional matching method was analyzed. By Moravec operator, Harris operator and Forstner operator, corner detection algorithm was studied. On the basis of character point matching method, according to the feature of photos which have been geometry corrected., the characteristic line matching and registration technology based on region growing method have been analyzed. Furthermore, for the image fusion technology of photos which have been stitched, the principle of determining pixel value was confirmed, using the gaussian template to smoothing the mosaics image. Finally, experiments proved the feasibility and effectiveness of the image mosaics technology.(4) When correcting and mosaiking the Bird's-eye-photo for traffic accident scene, information processing approach is limited if points are at different planar surface with a certain height. In addition, some related dimensions of the correction and mosaic of Bird's-eye-photo for traffic accident scene need spatial measurement to obtain. Therefore, photogrammetry technology based on the three-dimensional theory which is determined to acquire additional measurements needs to be studied. The use of Shen Jun operator was carried out to acquire step edge detection, and infinite impulse response (IIR) filter was used to achieve ridge line edge detection. Settings technology of the index of quantity point was studied, then analyzed the identification and sequencing of the index of quantity point based on cluster analysis.Through the least square method and the index of quantity point optimization method, analysis of the linear calibration techniques, and the non-linear calibration methods considering radial distortion have been done. Experiments proved that the least square method and the non-linear calibration method were feasible and practical.