Camera rotation recovery from vanishing point matching
Date
1996
Authors
Leung, John Chung Hon
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Recovering camera rotations is an important operation of any stereo vision system and photograrnmetric work. This thesis examines the method of recovering camera rotations using matched vanishing points between two images. Due to imperfection of automated vanishing point detection algorithms, natural line intersections can be mistaken as vanishing points. These false vanishing points can cause severe difficulties in automated vanishing point matching algorithms. This thesis examines the use of an invariant property exhibited by a group of true vanishing points to perform the matching. In this thesis, relaxation labelling is used to match vanishing points. The method performed successfully recovering camera rotations from matched vanishing points. The method was also useful in discriminating true and false vanishing points resulting from the imperfection of automated vanishing point detection techniques.