Research Of Self-Calibration Method Based On Monocular And Binocular Vision Image Sequence

Camera self-calibration is one of the key technologies in3D reconstruction. Itstask is to determine the mapping relation between the projections of2D space pointcoordinates and3D space coordinates, the process of solving the camera parameters iscamera calibration. The precision of the calibration and the robustness of thealgorithm directly affect the accuracy of the3D reconstruction. Therefore, a goodcalibration method is the prerequisite to realize3D reconstruction; also, Cameracalibration technique has a great value.In recent years, some different methods had been presented such as throughseveral views of the same scene to calibrate the camera, but those methods are basedon the same scene and the same camera parameters. In this paper, N pieces of imagesshot by the camera then put the matching points of different frames transformed to thesame coordinate system, the focal lengths of the Image sequences are alwayschanging. As a result, the camera self-calibration with the varying intrinsic parametershas practical significance and great application value.First，tell the basic theories related to calibration, I have done some exploring inthe field of monocular camera self-calibration and binocular camera self-calibration.The main work as follows:(1)Described the research background, significance of the research situation, andtheory value and application prospect of self-calibration.(2)Introduced the basis theoretical of self-calibration.(3) I studied the calibration method of the absolute conic, and then a new CameraSelf-Calibration method is proposed which is based on image sequence, in addition,the intrinsic camera parameters are varying. This method based on Kruppa equations,by two upper triangular matrixes must exist a relational matrix, make the varyingfocal length convert to the relational matrix，thus use the epipolar geometryrelationship of absolute conic to get the intrinsic camera parameters. Experimentalresults show that the method is a high precision, robustness strong self–calibration.(4) With the study the method of layered reconstruction presents a binocularself-calibration method for varying intrinsic parameters. The idea is obtaining aquasi-affine reconstruction to search for the plane at infinity, thus to get the infinitehomography matrix, then to calculate the intrinsic parameters with the constraintproperties of absolute quadric. The method doesn’t require a special scene constraints or camera movement information, after experimental, we analysis each error, theresult show that this method is a high accuracy, strong robustness and practicalbinocular camera self-calibration method.