Research Of Camera Calibration Based On Corrected Template And 3D Reconstruction

As an important part of stereo vision, and a premise and base for computer vision obtaining the three-dimensional(3D) space information, camera calibration has been widely applied in image fields, such as computer vision, photogrammetry and 3D reconstruction. Calibration results have a direct impact on the accuracy of 3D measuring and the effect of 3D reconstruction, while real-time calibration can meet the needs of industrial machine vision of navigation. Therefore, the research of calibration methods have important value in theory and practice.For existing of the existing camera calibration and 3D reconstruction methods have low precision and high complexity, the paper carried on a in-depth study and a large number of experiments, and achieved the following results:1. For traditional calibration methods need calibration object which has high precision, while non-metric calibration or self-calibration methods have low precision and robustness. With an analysis of the relation between various types of lens distortion, a method of camera calibration based on corrected template is proposed. Firstly the method corrects the calibration object, then uses the improved Zhang Zhengyou calibration method to accomplish camera calibration. The experiment shows that the method can achieve higher accuracy of camera calibration, stronger robustness. Compared with the similar methods, the method can gain roughly equivalent calibration accuracy and greatly reduce the complexity of the algorithm.2. Considering the current 3D reconstruction methods based on multi-camera have many constraints for experimental environment and the cost is high, while the methods based on single-camera do not consider the case that the two lines which determine the coordinates of 3D point may locate in the different planes, it will make great errors. The paper presents a new 3D reconstruction method based on single-camera, that is using the geometric properties of the image to correct the distorted image, and then accomplish the matching of corresponding points of image, the next finish solving the camera extrinsic parameters using the location of the experimental images, finally discuss the location of the two lines that are used to obtain the coordinates of 3D point, especially when the two lines locate in the different planes using the midpoint of common perpendicular of the two lines to approximate the point. The experiment shows that the method can greatly reduce the errors compared to the existing 3D reconstruction methods based on single-camera, and it can gain roughly equivalent performance with the methods based on multi-camera.