A Global Image Descriptor Based Navigation System for Indoor Environment

We present a vision based indoor navigation system based on the GIST global image descriptor. The use of lightweight GIST descriptor enables faster computation and low memory requirement. Our system takes care of the key tasks of creating map and localization of user. This work also focuses on finding destination and displaying route from current location to destination. A topological map is created from keyframes that are extracted from a walkthrough video of the indoor environment, using the GIST descriptor. The L2 norm of the GIST descriptor is used for both the keyframe extraction as well as localization.