Research On Camera Calibration Algorithm

Camera calibration is the estimation of parameters associated with a camera being used for imaging. The parameters of a camera to be calibrated are divided into two classes:intrinsic and extrinsic. The intrinsic parameters describe the camera's imaging geometric characteristics, and the extrinsic parameters represent the camera's orientation and position with respect to the world coordinate system. Existing camera calibration methods can be classified into two main categories:self-calibration and object-based calibration. Self-calibration method is usually chosen for camera calibration in uncontrolled environments because the scene geometry could be unknown. However when no reliable feature correspondences can be established or when the camera is static in relation to the majority of the scene, self-calibration method fails to work. On the other hand, object-based calibration methods are more reliable than self-calibration methods due to the existence of the object with known geometry. However, most object-based calibration methods are unable to work in uncontrolled environments because they require the geometric knowledge on calibration objects.There are large number of object-based algorithm, and the main idea it to find the projection between object and its correspondence. A new calibration algorithm was proposed after analyzing point to point, curve to curve, area to area camera calibration algorithm.In pinhole camera model, a circle whose projection quadratic curve is always a ellipse. Points on the circle and points on the ellipse, the corresponding relationship between them were used to calibrate the camera parameters. Our proposed method was based on the composition of point set graphics area.In this algorithm, the projection of a circle is a ellipse in pinhole camera model. We using the largest oval coincidence degree between the ideal ellipse and the actual ellipse on image as our target to solve the ideal parameters. In the process of calculating ideal camera parameters, we propose two methodsMethod one:An ellipse contains, long axis, short axis, center and rotation angle of the composition, only these five parameters to determine an ellipse. In the case of small distortion, we find the largest oval coincidence objective function to solve camera initial parameters. The objective function is determined by four ellipse parameters. And our approach can quickly find the initial homography.Method two:After obtaining the homography between calibration plane coordinate system and between the image plane, we optimize the results in the next step using our method two to solve the internal pinhole camera. Specific process: Made a number of split lines through the center of the actual ellipse on image, so the actual ellipse and modeled ellipse are divided into lots of small area at the same time. Using the public part of a two-oval area and the maximum ratio of two elliptical target area as our objective function to solve the initial parameters of the camera. In this approach, we use genetic algorithm to find the optimal solution, its initial value is obtained by method one. The Advantage of method two, if there is no basis of method one its hard to finder the optimal parameter, and ellipse divided into more pieces the computation speed will be slower accurate calibration of the initial value, but with the rapid development of computer calculation speed, it will not be a constraint of this' methodThe combination of the two methods were highly accurate initial camera parameters, camera calibration of non-linear process can have more accurate parameters of the camera distortion. Area-based approach than the corresponding method based on the point of camera shake in certain circumstances, the calculated camera parameters smaller. Real experimental results show the algorithm's accuracy and robustness...