On demand vison-based indoor localization

摘要

Many indoor localization systems are based on earth-magnetic or radio field strength fingerprinting which is considering its low computing power requirement is very well-suited method for mobile phone applications. However, certain parts of navigation areas might not cover with any radio signal which makes the localization very difficult and imprecise. In this case the mobile phone camera can be used as a temporary localization sensor. Since the mobile phone camera is usually not stereo or depth sensing camera and since the image processing capabilities of the mobile phone is limited, a specific algorithm needs to be developed. The algorithm is expected to work without using any special markers, using only existing features of the surrounding environment, recognizing them, using a database to lookup their position and using a Perspective-n-Point algorithm to determine the user's relative position. The markers can be any specific feature of the building or any logo, sign or anything visually specific to the given location. This article summarizes the development, tuning and testing of a mobile phone application using marker-less vision-based localization algorithm..