Visual Position Determination for Autonomous Vehicle Navigation

摘要

This report describes a system by which an autonomous land vehicle might improve its estimate of its current position. This system selects visible landmarks from a database of knowledge about its environment and controls a camera's direction and focal length to obtain images of these landmarks. The landmarks are then located in the images using a modified version of the generalized Hough transform and their locations are used to triangulate to obtain the new estimate of vehicle position and position uncertainty.