Global Heading Estimation For Wide Area Augmented Reality Using Road Semantics For Geo-referencing

In this paper, we present a method to estimate global camera head- ing by associating directional information from road segments in the camera view with annotated satellite imagery. The system is based on a multi-sensor fusion framework that relies on GPS, cam- era and an inertial measurement unit (IMU). The backbone of the system is built on a very strong visual-inertial navigation (VIO) pipeline with very low drift rate and the proposed algorithm com- bines relative motion provided by VIO with global cues obtained by an image segmentation module to extract heading information for geo-referenced AR applications over wide area.