An efficient image-based indoor positioning approach using ORB and LSH

Indoor localization has recently gained much attention in academia and business. Here, we raise a new approach to deal with this problem, using a variety of indoor images with localization information. Different from previous research, we use Oriented FAST and rotated BRIEF(ORB) instead of SIFT, which saves much time in feature extraction stage. Moreover, we apply Locality Sensitive Hashing(LSH) to do matching work. It reduces time complexity and further improve efficiency. With this novel combination of ORB and LSH, we significantly shortened process time. To test our approach, we use the images in Yorkdale, a shopping mall in Toronto. All the images are labeled with indoor localization information, the coordinate. The result shows that the whole processing time of our approach can be within 0.6 seconds, and all the test images can be matched to the closest picture in the database.