Geo-Mapping and Visual Stitching to Support Landmine Detection Using a Low-Cost UAV

摘要

This paper describes the development of an aerial system applied for the terrain mapping and geo-detection of explosive landmine-like objects. In practice in Colombia, a large percentage of the anti-personnel mines that still remain across the country are hand-crafted and partially exposed on the terrain's surface so that they can be triggered. This scenario facilitates the use of a vision-based approach for the detection of these artifacts. Our goal is to integrate computer vision algorithms within a low-cost UAV robot suited for the Colombian scenario. The aerial system enables: (i) terrain mapping based on a visual stitching method to generate a mosaic image of the covered terrain, and (ii) the visual detection of landmine-like objects in real-time. Despite the hardware drawbacks and the camera limitations of the used UAV, we demonstrate that our low-cost platform could be used as a complementary tool for demining missions in Colombia. After briefly reviewing the state of the art regarding the use of robots for mine clearance, we present a field report that confirms the feasibility of our aerial-based system featuring in approximate scenarios.