Detection of Natural Landmarks for Mapping by a Demining Robot

Abstract

To enable automated landmine detection a number of practical problems must be overcome. One of them is the mapping of data from landmine detection sensors using a mobile demining robot in an unstructured outdoor environment. The odometry of the robot used for mapping can be improved by using an additional vision system. It is proposed in this paper to utilize natural landmarks which can be found on the ground for this purpose. New simple algorithms for detection and association of natural landmarks are developed and described. Finally, the experimental results of applying these algorithms on a set of images obtained from a camera mounted on the robot are presented. The results show their robustness for different environmental conditions and robot motions