Abstract
Mobiles robots must navigate successfully in environments in which there are low lying obstacles. This paper presents a method using stereo vision which can detect such low-lying obstacles and build a 2.5 dimensional map. A stereo camera system placed on-top of a mobile robot is calibrated in relation to the robot coordinate reference frame. Calibration is achieved automatically through detection of the ground plane. The stereo system can then produce estimates of variance from the ground plane which are used to form a local map.
