Research On Camera Calibration And Related Technique

Human apperceives and understands the outer world mainly by vision. Now with the development of computer technology and with the continual increase of the demanding of people on digital information, both computer vision and binocular vision have been developing greatly. As an important part of Binocular Vision, and a premise and base for computer vision obtaining the spatial information of 3D object, camera calibration has gained great attention recently and the research in this field is boosting up increasingly.In this dissertation, we research the theory, method and technique on the camera calibration and the related field, and gain the parameters about the camera to construct the relation between the 3D object and the 2D image, and provide the reliable data and establish the helpful base for the next research on computer vision and binocular vision. We mainly work on camera calibration and its associate fields: the image corner detection. The main results obtained can be summarized as following:1. Based on the defect of Susan corner detection algorithm, an easy and quick method of selecting the threshold automatically is proposed in this paper, and the improved algorithm can avoid the problem of detecting corners in error because of setting the threshold too low or to high in the Susan algorithm, and enhance the reliability and availability of Susan algorithm, and realize the automatization of selecting the threshold.2. A linear approach for camera parameter calibration is proposed, and it only requires a coplanar target without camera's motion. The method is simply and accurate highly. All key parameters of the camera are linearlyderived through step-by-step decomposition algorithm because we introduce a new distortion model. This not only overcomes the instability and iteration of nonlinear problems, but also implements the calibration of partial intrinsic-parameters which the other linear methods fails to do. 3. Validate the linear approach for camera parameter calibration from two aspects by design the experiment scientifically: on one hand we apply it to the Zhang's data and compare its result with the result from Zhang's calibration; On the other hand, we carry a real experiment and apply our algorithm to it. And the results are all good and consistent. In addition, the result in our experiment can proved the algorithm's validity in the corner detection improved by us. In the finally, we analyze and discuss the results and data in the experiment and arrive at some helpful conclusion.