3D position determination using a camera

Abstract

This master's thesis researches pose determination methods with a camera in cooperation with the research group CoSenS of Hasselt University. 3D position determination is a frequently occurring problem in the field of computer vision. In this thesis several methods for determining the 6 degrees of freedom are compared. Especially methods which do not require specialized or expensive hardware are studied. This research has a wide field of application, for example: augmented reality, robotics, healthcare, etc'

In this master's thesis we discuss the necessary procedures for detecting feature points. These procedures are, in order: the calibration and normalization of the camera, the detection of reference points in a known pattern and the calculation of the position of the object. This master's thesis researches classical methods like: DLT, EPnP, OI, NPL, NLL and POSIT and also some newer methods like PRSOI and Moiré patterns. These methods differ in robustness to noise and distortion, limited field of target, resolution, etc'

For example DLT is a method which is very susceptible to noise, but it does not require camera calibration. PRSOI is a one of the more accurate methods in determining the pose, however Moiré patterns can be used to detect the out-of-plane rotations more accurately.