Research On Calibration Technique For Omnidirectional Vision System

Panoramic vision systems have an imaging range which is greater than hemispherical view, and can observe 360°scene simultaneously. It widely used in robot navigation,3D reconstruction, surveillance, virtual reality and image-based metrology and many other fields, which has become one of the central issues in recent years. Since the camera calibration is an indispensable step to recover 3D information from 2D images and is necessary for effective image distortion correction, the thesis is focused on omnidirectional camera calibrationCatadioptric camera systems usually consist a perspective camera and a surface of a revolution mirror, and ideally have a single view point constraint. But because of the reasons such as production crafts or man-made errors, it causes lot of trouble to the panoramic image unfolding and other image processing. According to the situation, the paper proposes a general catadioptric camera model that consists not only traditional camera intrinsic and extrinsic parameters but also the parameters about mirror posture. Based on this kind of camera model we propose a catadioptric camera calibration method, which includes conventional camera calibration and mirror posture estimation. On the other hand, our method can estimate the full degrees of freedom of mirror posture and is free from the volatility of nonlinear optimization. During the process of calibration we uses the mirror boundary on conic fitting, then can estimate up to four mirror postures. Because an estimate of the mirror posture is not unique, we also propose a selection method for finding the best one.Under the statement above the main work can be concluded as followed:First, it has been introduced that the corner detection and image contour extraction method. Through feature images processing and Acquisition such as conic fitting, the calibration can extract the features of the image and obtain 2D information quickly, accurately and efficiently; Secondly the paper has introduced the general content about conventional camera calibration, discribed linear and nonlinear camera model, and also the traditional calibration methods, and then focused on the calibration of Zhang's method based on planar template. At last, according to the theory of panoramic camera calibration the experiment has been done. And we verify the calibration results by putting calibration into panoramic image adjustment, analyze the accuracy. The experimental results show that this panoramic calibration algorithm can calibrate all the structural parameters of the panoramic system effectively and accurately.