Calibration Paracatadioptric Camera Using Sphere Images Based On The Polar Correspondence

Computer vision is a comprehensive disciplinary in which camera calibration plays an important role.Camera calibration has been used in many ways like three-dimensional reconstruction and measurement and helps to provide a wider angle of view than ordinary camera does.As a calibration item,without self-shield,its projection under a central catadioptric camera is in the form of conic.In this thesis,my focus is to use the image of sphere to calibrate paracatadioptric camera.The projection of sphere is a circle on the unit viewing sphere,Camera optical center and the sphere images create an oblique cone,when the cone intersects with unit sphere,a circle comes into being.Based on geometric relations in space,we get circle center and its projection consequently,which two have a polarizing correspondence relation,that is,polar line of the center against small circle is a line at infinity.Because the circle center is center,it shares affine property.According to that,we conclude that there is polarizing correspondence between image of center and conic,through which,we get vanishing line on image plane.When vanishing line intersects with the projection of sphere,we get the images of circular points.My thesis writes to calibrate camera by using three images mentioned above.At last,I start a simulation and real experiment by MATLAB to justify my algorithm is workable,practical and effective.