Image-Based Localization for Indoor Environment Using Mobile Phone

Abstract

ABSTRACT:Real-time indoor localization based on supporting infrastructures like wireless devices and QR codes are usually costly and labor intensive to implement. In this study, we explored a cheap alternative approach based on images for indoor localization. A user can localize him/herself by just shooting a photo of the surrounding indoor environment using the mobile phone. No any other equipment is required. This is achieved by employing image-matching and searching techniques with a dataset of pre-captured indoor images. In the beginning, a database of structured images of the indoor environment is constructed by using image matching and the bundle adjustment algorithm. Then each image's relative pose (its position and orientation) is estimated and the semantic locations of images are tagged. A user's location can then be determined by comparing a photo taken by the mobile phone to the database. This is done by combining quick image searching, matching and the relative orientation. This study also try to explore image acquisition plans and the processing capacity of off-the-shell mobile phones. During the whole pipeline, a collection of indoor images with both rich and poor textures are examined. Several feature detectors are used and compared. Pre-processing of complex indoor photo is also implemented on the mobile phone. The preliminary experimental results prove the feasibility of this method. In the future, we are trying to raise the efficiency of matching between indoor images and explore the fast 4G wireless communication to ensure the speed and accuracy of the localization based on a client-server framework.

Image and WLAN Bimodal Integration for Indoor User Localization

“…Localization based on stereo-imaging has also been studied as stereo images can provide depth insights for 3D reconstruction [29], [32]. An indoor localization algorithm based on an efficient database search using robust matching algorithms is presented in [33].…”

Image-tag-based indoor localization using end-to-end learning

“…Image-based indoor localization [1][2][3][4][5][6][7][8][9][10] finds many use scenarios such as drone and robot navigation. In such methods, images of the environment where localization takes place are pre-captured, structured, and stored in a database, using methods such as image matching and bundle adjustment algorithms.…”

“…Each image's relative pose such as its position and orientation is estimated and the semantic locations of the images are tagged. The camera's position (or the position of a user) is localized by matching the camera's captured image to one of the images in the database, for example, by combining quick image searching, feature matching (such as Scale Invariant Feature Transform (SIFT) features), [4][5][6] and relative orientation. Image-based methods still suffer from issues such as high computational requirements, blind spot issues, and being not robust.…”