The Research On Analysis System Of Vehicle Impact Based On Dynamic Image Sequence

Passive safety of vehicles has become the chief performance when the traffic accidents have become a global problem with the increase of vehicles.Recently there are three ways to study the passive safety:theoretical calculations,testing and accident analysis.Because the latter two take the objective facts as the basis and the results are more practical,nowadays they are focused.Based on dynamic image sequence analysis,the entire process of the collision can be recorded.Meanwhile the qualitative and quantitative analysis of the entire process is irreplaceable in study of passive safety.This paper presents a vehicle collision analysis system method based on dynamic image sequence.There are four main works which list as follow:(1) During the detection to moving target,especially the image captured from dynamic environment(particularly video from the traffic scene) would be affected to a certain extent by various aspects,for instance the changes in weather and light conditions which will result in degradation of image.This paper imports the method of image recovering based on atmospheric physical model to reduce the effect on image quality made by atmosphere and presents the amalgamation method(three frames method) of background subtraction and temporal difference to realize fast detection to moving object.(2) For the realization of camera accurate calibration and later work,it is the premise and key point to identify the feature points from the collection of digital image and pinpoint the precise imaging center the point,as well as feature points on the precise positioning of imaging centers,to achieve accurate camera calibration and follow-up image of the premise and key measurements.In this paper among the recognition of feature point,we improve noise cancellation algorithm which can coarsely calibrate the feature point with the noise recognition,thus effectively decrease the calculating time Of the sub-pixel location method based on the center of mass.The encoding unit is introduced to realize the automatic matching between the calibration points and pixel points among the identification of calibration points.(3) Among traditional camera calibration methods,most need to measure the spot and calibration object and have to control Camera to make strict movement so it needs strict experimental places.In this paper the composite calibration methods can be taken as integration and improvement of above-mentioned methods:fully calibrate the extrinsic and intrinsic parameters based on the coplanar targets step by step;due to the calibration object as coplanar targets,the camera can make without any movement. This lower requirement for spot can be applied to vehicle collision experiments,but also to all types of traffic on-site monitoring.Furthermore calibration of distortion parameters based iterative optimization can effectively improve the accuracy of calibration.Moreover among the identification of calibration points,the encoding unit is introduced to realize the automatic matching between the calibration points and pixel points.Besides that,the algorithm of line detection based on Hough transform and line intersection is taken for the identification of central position,,so as to avoid the measurement error from usual Hough oval detection methods.(4) Using the vehicle impact lab as application environment,the design and realization of image analysis system are completed.The vehicle impact analysis system Based on the dynamic image sequence analysis is applied to the video analysis of vehicle side door impact.The side door strength and stiffness can be reflected according to the data and displacement trajectory graph which can be taken as an important basis for passive safety design of vehicles.Then we not only qualitatively analyze the collision course of the trial,but also to quantitatively analyze,which can provide an important reference to study the safety of vehicle impact.