Research On Camera Calibration Method Based On Two Vertical Lines

Stereo vision is an important part of the field of the computer vision.Through stereo vision technology,The position information of the target object in the actual three-dimensional space can be obtained through the two-dimensional image,so it has wide range in the various fields of robot positioning,3D reconstruction and automatic driving.These visual problems must be determined by camera calibration to determine the projection relationship between the image and the actual three-dimensional space.The accuracy of the calibration results will directly affect the accuracy of the visual system's final calculation of the spatial depth information.This paper is based on this background.From the theoretical research of camera imaging and parameter calibration,aiming at the improvement of current popular methods,this paper proposes an estimation method of camera internal reference initial value based on two vertical lines.The main work of this paper is as follows:(1)The current application is more commonly used by Zhang Zhengyou in 2000 to use a checkerboard for camera calibration.The key innovation is to combine the linear imaging model with the nonlinear imaging model.The internal and external parameters and distortion coefficients of the camera can be calibrated at one time,and the checkerboard calibration accuracy used is low and easy to obtain.This method simplifies the process of calibration of the entire camera while obtaining more accurate calibration results.This article provides an in-depth analysis of this calibration method.However,it is also found that this method requires manual input of the corner arrangement and square size information of the checkerboard calibration plate,which increases the complexity of the calibration process and introduces unnecessary errors in the estimation of the homography matrix.These two points are insufficient.This paper proposes a camera parameter calibration method based on two vertical lines.This method does not require the size of the input calibration plate,nor does it have to solve the parameters by means of homography matrix estimation.(2)The principle analysis and theoretical derivation of the principle of the proposed method are carried out,and the simulation and experiment are carried out in the computer and real environment,which proves the effectiveness of the proposed method..The re-projection error is used to compare the calibration accuracy of the method with Zhang Zhengyou.Using the same calibration image,the reprojection error calculated by the parameter result obtained by Zhang Zhengyou calibration method is 0.484322 pixels.The reprojection error obtained by the method is 0.470307 pixels..It proves that the accuracy of the calibration method in this paper has indeed been improved.Finally,from the two aspects of the influence of the location of calibration board and the effect of the number of calibration images,the practical performance analysis of the method is carried out.The main contribution of this paper is that compared with Zhang Zhengyou's camera calibration method,the proposed method simplifies the calibration process,including no need to measure the physical size of the checkerboard and the order of the input checkerboard corners,and the calibration accuracy has also been improved.