iFind: Image-Based Indoor Navigation System

Abstract

In this time of life, people always rely upon navigation systems to help them get around. From GPS devices embedded in a car to GPS embedded in mobile devices, it provides accurate navigation. This project is motivated by three current scenarios happening at three different locations i.e., the difficulty in finding a classroom in the faculty, the difficulty in finding a specific room in the hospital, and difficulties in finding shops in the shopping mall. GPS networks are bad when indoors because the signal cannot penetrate the thick structure of the building. Therefore, finding a specific location in a building is a difficult task. iFind is a navigation system that utilizes image recognition techniques and path-finding algorithms. The techniques used for localization are ASIFT and Flann-Based Matcher. The combination of the two algorithms produces a very accurate image matching system and the accuracy of this system can be tested while the image was taken at 3 m or less and have enough light source. A navigation module using the A-Star algorithm is added to complete the system. Experiment for both localization and navigation has been made and the chosen algorithm shows a lot of promise. The system runs on a server which is then used to take input images and targets. The output of the system is a map that can be used by the user to navigate to their wanted shops. This project contributes to the GPS community towards making and developing the best method for indoor navigation systems.