Research On Multi-Camera Calibration Based On Sphere

With the development of computer application technology,computer vision technology is widely used in the industrial testing,medical image processing,artificial intelligence,and other fields.In the field of computer vision,camera calibration is the essential steps of recovering the real 3-D target geometry information from 2-D images and also is the fundamental of computer vision applications.At present,planar calibration target is widely used to calibrate camera parameters.As the price of the CCD camera coming down,multi-camera systems are more and more widely applied in the practical engineering.Among them,the inward-looking multi-camera systems: 1)The efficiency is higher than single camera;2)Multiple cameras shooting moving target simultaneously,it can realize 3-D reconstruction of moving target;3)In the 3-D reconstruction of big scene,it tend to apply multiple cameras from different perspectives.However,multi-camera systems are faced with problems: the plane 2-D target can't be seen by different perspective camera simultaneously.Therefore,it is necessary to find the non-planar camera calibration method which is suitable for multi-camera systems.For multi-camera systems,in this paper a new camera calibration method based on a sphere is proposed.For the first time,we put forward building a space plane with three centers of spheres and two new methods to get the relative distance of centers of spheres.So,we make the camera calibration based on a sphere and the camera calibration based on a space plane consistent.Firstly,this paper introduces some basic concept of projective geometry and camera calibration and also analyzes the principle of the classical camera calibration with the 2-D plane.Secondly,this paper proposes a new camera calibration method based on the projective geometry feature of a sphere.It extracts the projection of a sphere on the image plane by the subpixel edge detection algorithm.Then,obtain the position of centers of spheres in 3D space by explaining the relation between spheres and their project conics.With the centers of spheres and the midpoints of any two centers of spheres,the homographic matrix between the plane in the space and image plane are calculated.The projection of circle points is calculated by multiplying the homographic matrix with the circle points on the space plane.So,internal camera parameters are obtained with Cholesky decomposition.According to the known internal camera parameters,the distance between the center of a sphere and the camera optical center is calculated.Thus,obtain the external parameters of camera.Finally,this paper study the positioning precision with using sphere as a maker.In this paper,we analyze the accuracy and stability of the method presented in this paper according to the simulation experiment and real experiment.The simulation experiment and real experiment results show that the method presented in this paper can achieve the same result with the classic calibration method based on a plane.And the real positioning experiment results show that the positioning precision can reach to 0.02 mm by using table tennis ball as a marker.