Multi-sensor indoor pedestrian navigation system with vision aiding

Abstract

In many indoor security and rescue missions, it is essential for the deployed personnel to have accurate positioning information available. If the scene is not explored yet, a map has to be created during the mission and exchanged between different teams, so that subsequent team members benefit from the knowledge about previously visited locations. In this paper, the multi-sensor navigation system called IndoorGuide is presented which accomplishes these tasks. It comprises two lightweight micro-electro-mechanical inertial measurement units for tracking the motion of the user. A monocular camera is integrated in the navigation system which provides attitude and position aiding. The attitude information is extracted by tracking vanishing points in the images of the camera. Position aiding is realized by mapping and recognizing visited locations. For the data fusion, a Kalman Filter and a robust optimization technique based on graphs are compared. Results from experimental test runs are presented and discussed.