Camera Calibration And Cross-Camera Physical Panorama Reconstruction In Traffic Scene

The continuous video data generated in the traffic monitoring system lays an important data foundation for traffic data analysis and road information optimization.Camera calibration technology is to establish the conversion relationship between two-dimensional image and three-dimensional space scene based on the monitoring video data.Accurate and stable camera calibration results provide a prerequisite for the acquisition of relevant traffic parameters and practical traffic applications.At present,there are many problems in camera calibration methods in large-field traffic scenes,such as the tedious steps of manual calibration and online manual calibration,the large global calibration error in large scene.At the same time,the coverage for single camera field is limited and the data between cameras is difficult to effectively correlate.In view of the above problems,the specific research work of this paper is as follows:1.Obtain the vanishing point in the horizontal and vertical directions in the road monitoring scene.Based on the results of target detection,the idea of edge detection + slope histogram is used to obtain the local transverse main direction and the method of feature point + track tracking is used to obtain the local longitudinal main direction.Combined with multiple vehicles and multiple frames,the cascaded Hough transform finally is used to obtain the transverse and longitudinal vanishing points.2.Research on two kinds of camera calibration algorithms under single camera.Aiming at the problem that it is difficult to directly use Zhang ’s camera calibration under a fixed camera,a camera calibration method based on a spatial virtual chessboard is proposed.This method builds a spatial virtual chessboard based on the previously extracted horizontal and vertical vanishing points and scene prior information,and according to road cameras The model implements a camera calibration method for a spatial virtual chessboard.Aiming at the problem of large camera calibration error of the double vanishing point under the gimbal camera,an optimization algorithm based on double vanishing calibration was proposed.The algorithm takes the parameter result of VVH calibration method as the initial parameter,builds the nonlinear constraint function about the distance between points and lines according to the key points generated in the scene,and solves the optimal calibration result by iteration.3.Cross-camera model fusion and physical panorama reconstruction.In order to realize the monitoring of the whole road information and the simultaneous analysis of data,this paper,based on the camera calibration,solves the mapping relationship between cameras according to the common feature information between cameras,and integrates the road information of different cameras into the unified world coordinate system,so as to establish the one-to-one mapping relationship between cross camera vision and the unified world coordinate system of the whole Bureau,and puts forward the error problems in the model construction The idea of translation vector is improved,and the road panorama with physical information is reconstructed.Experiments and analysis of the related algorithms in different scenarios were conducted to prove the feasibility and effectiveness of the proposed algorithm.At the same time,the research work in this paper provides a reference for the acquisition of traffic parameters and information association between multiple cameras.