Current indoor positioning systems typically base upon active radio signals and require a costly setup of appropriate infrastructure. In this dissertation, a vision based positioning system is developed. Position and orientation is determined by matching images, taken by a smart phone, against a database of previously recorded reference views. This is achieved in fractions of a second via content based image retrieval techniques. Contributions of this thesis include a system architecture for the efficient integration of prior knowledge, approaches to facilitate localization in poorly textured environments, and a technique to allow for a meter level accurate positioning by introducing virtual reference views.
«
Current indoor positioning systems typically base upon active radio signals and require a costly setup of appropriate infrastructure. In this dissertation, a vision based positioning system is developed. Position and orientation is determined by matching images, taken by a smart phone, against a database of previously recorded reference views. This is achieved in fractions of a second via content based image retrieval techniques. Contributions of this thesis include a system architecture for the...
»