Research On Image Correction Method Of Accident Scene Based On Inherent Geometric Features Of Road Environment

With the increasing of vehicle registration, road traffic accidents occur frequently,the security of people’s lives and property and the stability of social are under greatthreat. The scene of the accident investigation delays and improper process are leadedby lacking of necessary technology and equipment, which can result in enormousdirect and indirect losses. Therefore, searching for one fast, accurate and efficientapproach of road traffic accident scene processing has great significance to roadtraffic safety. With the continuous development and improvement of computer visiontechnology, photography and digital image processing technology, many newtechnologies and new equipments begin to be applied to the scene of the accidentinvestigation, which aimed at reducing the delay of accident scene investigation,properly processing the scene and drawing the traffic accident scene map.Geometric distortion of images can be caused by photographic equipment,characteristics of imaging system, the viewing position and other factors, thegeometric correction must be done to make the processed image look like shootingfrom the top of the vertical ground, which can show the information more intuitivelyand comprehensively. However, calibration objects are needed by most of currentimage correction methods as calibration reference points, which not only producesmeasurement errors to affect results of investigation, but also increasing thecomplexity of the work because of carrying, displaying and measuring the calibrationobjects. It is not conducive to rapid investigation and processing of the accident scene.The content includes followings:The2D mathematical model of the image geometric correction has been studiedof further. Image space transformation is focused on, and the particularity of accidentscene and the linear model are also seriously considered, to derived2D mathematicalmodel of the image geometric correction, and determine the image pixel gray value.One method named Four-Points-method is used to acquire point coordinate quickly.Defining and classifying the inherent characteristics of the road environment, andthen the dimensional data of the inherent geometric features of the road environmentcan be made sure based on the GB standards. The specific implementation steps andhandling precautions of accident scene image correction method based on the inherentgeometric features of the road environment are put forward. And demonstrate how touse the accident image correction software through the inherent geometric features ofroad environment to realize of image correction process. With close attention to the key problem of image registration how to retain thescene information as much as possible in the corrected image, the accident sceneimage registration algorithm were derived, the image dynamic registration algorithmwas innovatively put forward on the basis of horizontal centerline method, and madesure the feasibility of the image dynamic registration algorithm through severalexperiments.Through field tests and simulating the real road traffic accident scene，the factorsaffecting the accuracy of the image correction based on the inherent geometricfeatures of the road environment can be analyzed, and compared with the traditionalimage correction that need calibration objects to be set up by persons, to prove itsreliability and effectiveness. And at the same time, the accuracy of dynamicregistration algorithm was also verified by experiments to prove it feasibility andeffectiveness.