Vision based system for camera tracking in eye-in-hand configuration

This paper presents an effective approach to estimate the fixed internal and varying external parameters of the camera for real time experiments using 2D-3D point correspondences. Images are acquired at each time step, a pose estimation algorithm is then employed to determine the camera pose w.r.t the object. A simple homogenous transformation is derived between the camera and end-effector to determine the position of the manipulator end-effector, as camera is mounted on the tool in eye-in-hand configuration. The paper focuses on determining the pose accurately and to look upon those issues that we encounter in real time. The major contribution of this paper is in two folds: camera pose parameters are easily and accurately recovered from 2D to 3D point correspondence; second is that experiments using real images are conducted, which presents good results.