Fast Mapping Techniques And Applications Used In Traffic Accident Scene Based On DLT Theory

With the rapid economic development and continuous improvement of transport facilities, vehicles have become an increasingly popular means of transport in our country. At the same time, a large number of automobile accidents occurred, resulting in serious casualties and property losses. In the processing of traffic accidents, on-site staff is required to complete the information collection and mapping quickly and accurately, so as to provide evidence for accident analysis and judgment for the responsibility of the accident. Therefore, the accident scene mapping plays a very important role in the processing of traffic accidents. But in our country, traditional methods, such as tape measuring and manual mapping are still mainly used in the survey of traffic accident scene. As we know, traditional methods are low-efficient, time-consuming and low-precision, and it is not good for the recovery of the traffic, which means it may lead to secondary accidents. Particularly, with our country's highways and elevated highways' rapid development, in order to keep the roads open, higher requirements are put forward for the processing time of the accident scene.In recent years, a large number of experiments and research are carried out at home and abroad. New technologies and methods are used in the analysis of accident scene, such as Total Station, GPS, Laser measuring device and so on. Although these new methods and new technology have played a positive role in improving the efficiency and automation of the accident scene investigation, however, due to high price of equipments, inconvenience to carry and technical expertise needed, they are difficult to be widely applied. In response to these issues, based on the close-range photogrammetry theory, this paper developed fast calibration and mapping technology of independent intellectual property for traffic accident scene, which can shorten the time of the accident scene mapping and improve the quality and accuracy of the mapping. Shanghai Traffic Police Corps have carried out experiments based on the theory of this paper and achieved good results.Based on the close-range photogrammetry research and direct linear transformation theory, by using common non-metric digital camera to take photos of traffic accident scene, this paper proposes four kinds of calibration methods: Control Point Calibration, Linear Equation Calibration, Points and Axis Paralleled Lines Calibration and Automatic Calibration. Control Point Calibration is the basic method of calibration. By using more than four calibration points, it can complete the two-dimensional reconstruction of the accident scene. Based on DLT theory, another method of calibration is derived in this paper, which can get the solution of DLT transformation matrix by coefficient of linear equations. Points and Axis Paralleled Lines Calibration can finish the image calibration and mapping by using more than two points and more than four lines that are paralleled to axes. Experiments show that this method has a more practical application value. Automatic calibration, which is more efficient and faster, can realize automatic recognition and calibration if environmental condition is allowed. Each of these calibration methods can individually complete calibration task of the accident scene, and establish the relationship between pixel coordinate and real-world coordinate, and then to complete the two-dimensional reconstruction of the scene.Meanwhile, this paper has studied shooting method and the calibration points setting methods in accident scene survey. And by a series of experiments, the accuracy and reliability of the methods presented in this paper is validated. With the method of nonlinear optimization method, as well as geometric information in the accident scene, such as distance, angle, parallels or verticals, coefficients of DLT'transformation matrix will be optimized for higher accuracy. Experimental validation and discussion are carried out for the mathematical model. In addition, based on DLT theory, this paper proposes a method to solve the velocity of the vehicle in the surveillance video, which is successfully applied to the analysis of several traffic accidents.Finally, based on the methods proposed before, the paper analyzed several real traffic accidents to get the geometric information needed for mapping and then built the 2D model. Meanwhile, the calibration methods used in traffic accident scene proposed by this paper is used in the experiment for validation, and good results have been achieved.