Self-calibration And Error Analysis For Rotating Camera

The camera self-calibration occupies very important position in the field of computer vision, which has been well developed in theoretical research. However, there are still several issues to be further discussed. For instance, self-calibration based on rotating camera is theoretically easy to realize, the camera motion for this type of self-calibration is a pure rotation but has often been violated, as a result, the error will affect the calibration result and application. The former research mainly limited to the qualitative research, theory of quantitative research is only for some specific situations, which is difficult to analyze the various factors together; and that, self-calibration is based on image corresponding points, while researches generally assume that corresponding point problem has been resolved and there are little analysis of the influence of the problem of corresponding points on the self-calibration results. Without resolving corresponding point problem will have a strong influence on calibration result. This paper discusses around these problems, and the main work and research result are as follows:1. This paper analyzes the influence of different control points on self-calibration. It is first verified experimentally that the calibrated parameters by using different control points are also different for a given configuration of the camera based on the traditional method; then the relationship between camera self-calibration and choosing of the corresponding points is analyzed theoretically; finally, the influences of choosing different corresponding points on precision of self-calibrated parameters are validated experimentally. Experiment results showed that corresponding points in self-calibration should lie in the space as near as possible to the range of the measured object for improving calibration precision.2. This paper proposes a quantitative analysis of impact factors on rotation-based self-calibration. It uses a lot of experiments on simulation data to analyze the error of the effects of translational misalignment and the noise of image points on self-calibration result by synthetic; moreover, the quantitative analysis of the image number and different levels of noise has been completed, the number of images needed at different noise levels is provided for reliable and stable calibration. These quantitative data can be used to help give a valuable suggestion for real applications.3. This paper presents a new calibration method with varying intrinsic parameters for camera self-calibration based on the infinite homography H∞. This paper discusses the problem of self-calibration with H∞, and presents a linear self-calibration method by strict theoretical derivation.