Efficient Vision Data Processing on a Mobile Device for Indoor Localization

The paper presents an approach to indoor personal localization using a sequence of captured images. The solution is tailored specifically for processing vision data on mobile devices with limited computing power. The presented FastABLE system is based on the ABLE-M place recognition algorithm, but introduces major modifications to the concept of image matching, in order to radically cut down the processing time, and to improve scalability. FastABLE is compared against the original OpenABLE and the popular OpenFABMAP place recognition systems.