Research And Application On The Algorithms Of Camera Calibration On Computer Stereo Vision

One of the basic tasks of computer vision is computing geometric information of the object in the 3D space from the 2D image information captured by camera, then reconstruct and recognize the object. The interrelation, between the 3D geometric location of the point in the surface of space object and the corresponding point of the 2D image, is decided by the geometric model of camera imaging, and these geometric model parameters is camera parameters. These parameters can get by experiment and computing in major conditions, this process is called camera calibration. Generally speaking, the technique of the camera calibration can be divided into two classes: the traditional camera calibration method and self-calibration method.The thesis applies the calibration method to the ARToolKit based on video through researching the camera calibration methods of the computer vision. ARToolKit is a secondary package of the AR system, explored by the Japanese Hiroshima City College and University of Washington. Through researching the process of the camera calibration on the ARToolKit, it can find that the process is using two special pictures , then computing the parameters in real-time through hand moving the pictures, and these interactive modes are hard realized in video. So in the research of the Augmented Reality based on the video data, it can not realize the camera calibration in the ARToolKit, though the default camera parameters can not match the request. In computer vision, the ways of computing the camera calibration parameters can be divided into two conditions: one is computing the integrated matrix of the camera model directly, the other is computing the camera intrinsic and exterior parameters separately. The virtue of the front one is computing is relatively easy; the behind one is more bothered, but increases flexibility in applying parameters. The parameters which need in the ARToolKit is the intrinsic parameters of the camera, so the thesis uses Tsai two-stage method which can do the camera calibration for the video included calibration object, calculates the camera intrinsic and exterior parameters separately. The process of the ARToolKit has its own system, the calibration parameters computed need embed the data flow of ARToolKit, so it need pass the intrinsic parameters, which can get from the calibration results of the Tsai two-stage method, to ARToolKit. At last, it can achieve the combination of graphics and image, using the technology of recognition tracking and image synthesis to the marks in ARToolKit, and also extends to the application faced to video for ARToolKit.