Combining Fisheye Camera with Odometer for Autonomous Parking

Simultaneous Localization and Mapping (SLAM) is one of the key technologies for autonomous driving. This paper focuses on the autonomous parking problem. A fisheye camera with a very large field of view is combined with the odometer inside the car to provide the localization information in an underground garage. The odometer provides an initial estimation of pose increment, and then the odometer and camera measurements are jointly optimized by graph optimization. The proposed strategy is evaluated on an autonomous driving platform, and high accuracy is achieved for the trajectory estimation with real scale.