Signal processing and performance evaluation of a PMD camerafor space docking

Abstract: This thesis evaluates the performance of a Photonic Mixer Device (PMD) cameraconcerning its ability to determine the relative position and orientationbetween a target and the PMD camera in order to provide a basic concept forusing a PMD camera for spacecraft docking. Therefore, the work is dividedinto three main tasks, namely pre-processing, calibration, as well as posesestimation and tracking. Since the frontal image of a docking port normally has a circular or ellipsefeature, the thesis introduces the RANSAC-based ellipse detection in order todetermine the ellipse feature from a 2D image quickly. The relative positionand orientation of the docking port is later calculated from the detectedellipse with the help of distance data. This method is referred to as thefeature-based pose determination. The pose of the docking port can also derive from the 2.5D image, which isreferred to as the shape-based pose determination. The well-knownregistration algorithm, the Iterative Closest Point (ICP) algorithm, canmatch the 2.5D image with the model point data of the docking port. Therelative transformation vector is the result of the registration. Thealgorithm is initialized by the result of the feature-based pose estimation.Because of the ICP-based tracking algorithm, the shape-based posedetermination is independent from the pose initialization by creating theclose loop ICP pose determination. The evaluation between the feature-based and shape-based pose determinationalgorithm shows that the performance of the pose determination from thefeature-based determination is better than from the shape-based concerningthe standard deviation and the computational time. Moreover, the speed of thefeature-based pose estimation is feasible to apply to space docking missions.