Research On The Self-calibration Method Based On Binocular Image Sequences Of Moving Object

To obtain three-dimensional information from two-dimensional image the cameracalibration is one of the key technologies in the field of computer vision, and be widelyused in the field of three-dimensional reconstruction, navigation, visual surveillance.So its theoretical research has become one of the hot research fields of computer vision.camera intrinsic parameters are solved by the characteristics of the image itself in thecamera self-calibration, compared to conventional Initiative calibration, the method ison-line and can be applied in three-dimensional reconstruction, and therefore it has awidely foreground.The method of self-calibration has been widely carried out, but basically based onthe camera self-calibration of monocular vision. In practical application, the accuracyof self-calibration based on binocular vision is better than monocular vision. Soself-calibration based on binocular vision is more practical significance. The paperresearches the theoretical basis of a self-calibration, some methods of self-calibrationof binocular vision have explored. The main content of the work are as follows:(1). Describes the background and significance of self-calibration, and analysesthe research status at domestic and international on the value of self-calibration theoryand application prospects, focuses on the theoretical foundation for self-calibration ofthe camera model and multi-view geometry.(2). The field of self-calibration for the current study did not consider thecamera’s nonlinear parameters, therefore, the precision of calibration is not accurate,camera self-calibration method based on binocular vision is proposed. The method firstthe matching point coordinates to be normalized calculate basis matrix, then solvingKruppa equation to obtain the camera parameters, then using the same name point oflight intersect at a point relative orientation and obtaining two-camera relative positionand attitude that is outside the parameters of the camera, re-introduce cameranon-linear distortion of regional adjustment to improve the accuracy ofcalibration.Finally, using the distance of two cameras as a known to determine theamount and size of objects, using object-side model side three points above theabsolute orientation of public to restore the true three-dimensional coordinates of thescene. Through the experiment, the feasibility of the method has been verified.(3). This paper presents a self-calibration method of varying internal camera parameters based on binocular stereo vision; In this method, using a binocular visionsystem to obtain a focal length of the image, and get the foundation matrix by using thematch points, then build the double linear equations about focal square via utilizing thekruppa equation. Secondly, make relative orientation to get the all external parametersget through letting the baseline length equal the distance between the two cameras.Finally, make the area network adjustment with importing the camera’s nonlineardistortion parameters using the above initial value and the characteristic of camera’srelatively fixed position and posture to get the all precise camera parameters. Theexperimental results show that the back projection of the method can be controlled inabout5pixels, and this method has certain ant noise trait.(4). Finally, Summarize the full text and look forward to the future research.