Research On Camera Calibration Method For 3D Reconstruction

Camera calibration is not only the premise of computer vision to obtain the spatial information of object,but also is the key step of 3D reconstruction,and the calibration results have a significant influence on the accuracy of 3D reconstruction.In this paper,the camera calibration was focused,and two important steps in the calibration process that are the angle point detection and the solution of camera calibration parameters were studied.In order to solve the existing disadvantages that the method of angle point detection cannot automatically detect the angle corner point of the calibration template images and the low calibration accuracy,the main contents include the following three parts:(1)The paper proposed an angle detection algorithm based on black and white checkerboard images.The calibration template images was preliminarily extracted the angle points using the Classical Harris angle detection algorithm.Then,in order to determine the inner angle points of calibration,the improved black and white detection operator performed the outer and false angle points of the calibration template images that had been preliminary extracted angle points.Then,the extraction accuracy of the inner angle points were improved to sub-pixel level by the quadratic polynomial approximation.The slope relationship of the angle points was utilized to solve the ordering of the inner angle points and thus achieve the automation of detecting and sorting the inner angle points,which did not require traditional manual intervention and improve the accuracy and efficiency of detecting inner angle points.(2)In order to reduce the influence of camera lens distortion factors during calibration process,a camera-imaging model was initially established under ideal conditions.Then,the single factor matrix calculation and the solution of internal and external parameters under the constraint conditions were calculated successively.Considering the disadvantages of the Zhang Zhengyou calibration algorithm,the lens distortion model was established,and the multi-stage radial distortion and tangential distortion coefficients were calculated.Finally,the accurate camera parameters were concluded by optimizing these parameters with the LM algorithm.(3)As the calibration experiment results showed,the calibration methods proposed by this paper overcame the influences of complex calibration environments and non-uniform illumination on the extraction of angle points.The angle points detection method possessed high accuracy,excellent robustness,and good anti-noise performance.A series of experiments were performed to analyze the factors affecting the accuracy of the calibration results.The calibration curves were plotted against the influence of various factors based on these experimental data.Analyzing the results of these curves,the influence mechanism of various factors on calibration accuracy were also determined.