Using An Image Of A Line To Calibrate Paracatadioptric Camera

Paracatadioptric cameras are widely used in the field of computer vision because of their simplicity,low cost,large field of view and single viewpoint.Camera calibration is an indispensable step in visual work.In the process of using a line calibration,the accuracy of line image fitting directly affects the accuracy of calibration results.Therefore,there is the important significance of research and the great application value about the algorithm of line image fitting and the Paracatadioptric camera calibration.In this thesis,there is an algorithm about fitting of image of a line and a new calibration algorithm of paracatadioptric camera using an image of a line.The main contents are as follows:On the unit vision sphere,a space line is projected as a large circle,and the mirror contour of paracatadioptric camera corresponds to a large circle that is parallel to the image plane.Determine a rectangle by separately extracting a pair of antipodal points on these two great circles.Two opposite edges of this rectangle intersect at the point of infinity.Take a different image point on the image of the line and the contour of the mirror,respectively.And,then calculate their antipodal image points.By connecting these four image points obtain a set of orthogonal vanishing points.Take multiple sets of image points on a line image to determine multiple sets of orthogonal vanishing points,then compute camera intrinsic parameters.On the basis of the least square,considering gradient distance as the distance metric,combining imaging characteristics of the paracatadioptric camera,using the properties of the image points and antipodal image points,fit an image of a space line.Based on the Matlab2016 a platform,a large number of simulation experiments and real experiments demonstrated the effectiveness and feasibility of the algorithm.In the experiments,noise factors were taken into consideration.Image correction experiments were performed in real experiments and Hough transform was used to detect the correction results.The results showed that the algorithms were effective and feasible.