ImPos: An Image-Based Indoor Positioning System

Recent years have witnessed growing interests from both academia and industry in developing effective Indoor Positioning Systems (IPSes). An IPS allows users to learn their locations and navigate in large unfamiliar indoor venues such as shopping malls, hospitals, and airports, where GPS signals are often absent or unreliable. Among different types of IPSes, images-based IPSes estimate a user’s location from one or more pictures the user took of nearby landmarks, which explore the deep penetration of smartphones into people’s everyday life and do not require any costly infrastructure upgrade. While several image-based IPSes have been proposed in the literature, most of them suffer from large processing delay due to computationally intensive 3D reconstructing or low positioning accuracy caused by inaccurate angle estimation. In this paper, we introduce the design and evaluation of ImPos, a novel image-based IPS that achieves high positioning accuracy by improved angle estimation and fully utilizing all recognized landmarks. Detailed experiment studies confirm the significant advantages of ImPos over prior image-based solutions.