Fusion of wifi and vision based on smart devices for indoor localization

Abstract

Indoor localization is an important problem with a wide range of applications such as indoor navigation, robot mapping, especially augmented reality(AR). One of most important tasks in AR technology is to estimate the target objects' position information in real environment. The existed AR systems mostly utilize specialized marker to locate, some AR systems track real 3D object in real environment but need to get the the position information of index points in environment in advance. The above methods are not efficiency and limit the application of AR system, so that solving indoor localization problem has significant meaning for the development of AR technology. The development of computer vision (CV) techniques and the ubiquity of intelligent devices with cameras provides the foundation for offering accurate localization services. However, pure CV-based solutions usually involve hundreds of photos and pre-calibration to construct an densely sampled 3D model, which is a labor-intensive overhead for practical deployment. And a large amount of computation cost is difficult to satisfy the requirement for efficiency in mobile device. In this paper, we present iStart, a lightweight, easy deployed, image-based indoor localization system, which can be run on smart phone and VR/AR devices like HTC Vive, Google Glasses and so on. With core techniques rooted in data hierarchy scheme of WiFi fingerprints and photos, iStart also acquires user localization with a single photo of surroundings with high accuracy and short delay. Extensive experiments in various environments show that 90 percentile location deviations are less than 1 m, and 60 percentile location deviations are less than 0.5 m.