Research On Calibration Of Long-focus And Wide-baseline Stereo Vision System

The binocular stereo vision is based on the trigonometry principle.Similar to the eyes of human beings,stereo vision systems acquire different pictures from two different views,then,threedimensional surface information can be obtained according to the calibrated parameters and matched points.High-precision camera calibration is the precondition of accurate measurement on the binocular stereo vision systems.Meanwhile,valid processing and application of point data is also the guarantee of the spread of the binocular stereo vision systems.This paper is based on the analysis of the state-of-the-art study in the world.In the paper,to satisfy the demand of the project,a one-dimensional calibration method based on infinite homography matrix is proposed for long-focus and wide-baseline stereo vision systems.Moreover,the extraction of three-dimensional hair line is used as example to detailed discuss the processing and application of point cloud data.The main work and innovation is as followed:1.The characteristics of the long-focus and wide-baseline binocular stereo vision systems is analyzed base on the mathematical model of binocular stereo vision.Then,we proposed the definition of long-focus and wide-baseline binocular stereo vision systems and build the mathematical model for this systems.All the system parameters to calibrate are confirmed.2.As for long-focus and wide-baseline stereo vision system has the characteristics of far distance and small field of view,a one-dimensional calibration method based on infinite homography matrix is proposed.The method uses a one-dimensional calibration target,which can effectively avoid the problems that the large-scale calibration target is difficult to manufacture,difficult to move and cover itself.In addition,with infinite homography matrix,the optimal solution of the system parameters can be obtained by combining linear and nonlinear optimization.Through simulation with C++ and 3ds Max and practical experiments,the feasibility and effectiveness of the method are verified.3.The processing and application of 3D point cloud are studied.Taking the extraction of the hairline in the human head color 3D point cloud as an example,a method of directly extracting the three-dimensional hairline is proposed.The hairline extracted by the method is compared with the hairline extracted based on the two-dimensional image,in which the accuracy of the extracted result is confirmed.The 3D hairlines are directly extracted from multiple color point cloud,which verifies the validity and stability of the proposed method.