Map Construction for Mosaic-Based Vehicle Position Estimation

Abstract

Mosaic-based vehicle position estimation is a way of tracking the motion of a vehicle using real-time imagery. A vehicle equipped with a downward-looking camera, for example, can track its motion by matching camera images with a high resolution "map" image of the floor. This paper describes our current technique for constructing such a map. The challenges of the map building process include collecting the raw map images, aligning the raw images in a common global reference frame, insuring smooth transitions between map images at map intersections, and reducing the total amount of memory required to store the map. To date, four maps have been constructed and used by four different vehicles equipped with a mosaic-based vehicle position estimation system. Results have shown this technique to be a quick, cost effective way for providing autonomous vehicles with position estimation.