A Camera Self-calibration Method Based On Hybrid Optimization Algorithm

With the further development of photoelectric technologies and digital sequential image processing and analysis, more and more attention has being paid to the research area of obtaining the 3D information from the 2D plane images of object in space. Moreover, many tasks requiring the motion parameters of the camera and Euclidean information of the scene, such as image measurement, virtual reality and 3D reconstruction and so on, also require the fact that the intrinsic parameters are known. Therefore, the camera self-calibration technique has important theoretical significance and practical value.On the basis of reviewing the camera calibration method at home and abroad, Genetic Algorithm (GA) and Levenberg-Marquardt (LM) algorithm used in camera self- calibration, the relation between sequential images must be set up through three steps as feature point extraction, corresponding matching and estimation of fundamental matrices. Meanwhile, a novel camera self-calibration approach is presented by combining an improved GA with LM algorithm. The main contents and contributions of this thesis are summarized as follows:(1) The properties of the fundamental matrix and important effects on the camera self-calibration are discussed according to linear camera model.(2) The algorithm of Scale Invariant Feature Transform (SIFT) for feature extraction and matching is studied and realized, then false matches are eliminated by using Random Sample Consensus (RANSAC) algorithm, so on the basis the precise feature matching among different images is obtained. The calculation methods of the fundamental matrix are analyzed and can be accurately estimated by using the obtained precise feature matching.(3) A camera self-calibration method combined improved GA and LM algorithm is presented. The method uses the good characteristics of global search ability in GA and improves the poor convergence performance of GA by making use of good local convergence of LM algorithm. The combination of improved GA and LM algorithm is applied to the nonlinear optimization of camera self-calibration. Experimental verification and result analysis is given. (4) A camera self-calibration prototype system is developed based on OpenCV and Visual C++ 6.0. Based on the thought of hybrid programming, the system makes use of the interface technique of OpenCV and Visual C++ 6.0 to realize calling OpenCV function by Visual C++ 6.0, finally Camera self-calibration method is implemented.