Research Of Camera Self-Calibration Method Based On Fundamental Matrix

With the advantages of high precision,high efficiency and non-contact,machine vision measurement technology has been widely used in many fields,such as medical,industrial,aerospace and so on.Camera calibration is an important basis for visual measurement.The accuracy of calibration results and the feasibility of calibration method will directly affect the accuracy and applicability of the visual measurement system.With its advantages of simple operation,strong flexibility and real-time online,the self-calibration method research has become a hot topic in the field of camera calibration in recent years and has been widely used in practice.But for the research of camera self-calibration,there still remains some problems,such as low calibration accuracy,poor robustness,sensitive to noise and initial values.By analyzing and comparing the present situation,we find out the factors that restrict camera self-calibration method,and this paper will be carried out from the following three aspects: the solution of the fundamental matrix,the calibration of camera intrinsic parameters based on Kruppa equation and the calibration of extrinsic parameters.The main contents are as follows:(1)In the traditional binocular vision measurement system,the fundamental matrix can not be solved correctly by the epipolar line distance constraint of the matching points.By introducing the additional constraints added by the third views,the constraint of "point to line" in binocular vision system is strengthened to the constraint of "point to point" in trinocular vision system.The experimental results show that the error of X direction is reduced from more than 10 pixel errors to less than 1 pixel,and the Y direction error is reduced from the original about 2 pixel error to less than 0.5 pixels,and the stability is greatly improved.(2)As the nonlinear optimization of the Kruppa equation is easy to get into local optimal solution during the process of intrinsic parameters calibration,the method of separated calibration is proposed.Firstly,the center point coordinates of the left and right camera are obtained by the method of changing focal length,and then the normalized focal length of the camera is obtained by the simplified Kruppa equation.Take these as the initial value,a further optimization is conducted to get the finally intrinsic parameters.This separated calibration method can effectively avoid the local optimal solution caused by the traditional Kruppa equation,and only six pictures are needed to calibrate the intrinsic parameters of the left and right cameras.(3)In order to get the extrinsic parameters between left and right camera we need to decompose the essential matrix.But there exist multiple combinations of extrinsic parameters and using these parameters we can only get the similar structure of 3D objects.To solve this problem,an extrinsic parameters calibration method based on coordinate transformation is proposed.By establishing the intermediate conversion coordinate system,the solution of the extrinsic parameters between the two cameras is transformed into solve the relationship of two camera coordinates and the intermediate conversion coordinate respectively,thus we can indirectly calibrate the extrinsic parameters between the two cameras.The problem of ambiguity of the extrinsic parameters and the scaling factor of translation vectors by decomposing the essential matrix thus can be solved.