Landmark tracking and camera calibration

Pose estimation is an important operation for many robotic tasks. In this thesis, we present a new algorithm of pose estimation based on the volume measurement of tetrahedra composed of feature-point triplets extracted from a quadrangular shaped target and the lens-center of the vision system. The contribution of this method is the fast recovery of the vectors joining the effective focal-point and each of the target points using an all-geometric close-form unique solution. This solution can be obtained even if the effective focal-length of the camera is not given with the condition that the optical-axis of the camera is not normal to the target plane. Taking advantage of all the geometric information inherent in the target and its image, each of these vectors is recovered in six different ways. Later, using the optimization technique, this redundancy is exploited in order to minimize the effect of random errors in the recovery of image coordinates of target points. Knowing the relative position of the target with respect to a fixed coordinate system, the exterior orientation parameters of the camera are recovered in the form of a matrix transformation relating a fixed coordinate system to the camera coordinate system. The decomposition of this matrix into a translation and three rotations about the major axes provides the final pose of the camera.

Simulations are conducted to provide for the analysis of the sensitivity of this algorithm. Finally the improvement of the accuracy of the solution achieved with the optimization process is presented using simulated and real data.