Practical aspects of visual recognition for indoor mobile positioning

Visual recognition based positioning requires complex image processing algorithms like feature detection, description, grouping and matching, which need considerable processing power from mobile devices. Many algorithms for feature detection have been developed from which some became de-facto solutions for many visual recognition scenarios. Novel techniques have since been published for feature description, grouping and matching. Although the applicability of these techniques for indoor mobile positioning is a subject of recent, ongoing research, only a few papers discuss the peculiarities of mobile visual imaging considering recent technical developments and usability aspects. For a visual recognition based indoor positioning application not only do we need to consider the quality of known algorithms, but we also have to take into account the whole set of boundary conditions, like processing capabilities, memory availability, battery capacity and quality of visual recording of the mobile devices and overall system communication capabilities. In this paper we investigate some of these aspects together through a real visual recognition system and present simulation results to support the usability of feature detection algorithms in mobile environments.