The collapse of buildings often result in human victims becoming trapped within rubble. This environment is dangerous for emergency first responders tasked with locating and extricating victims. Recent work in scene mapping using photometric colour and metric depth (RGB-D) data suggest the possibility of automatically identifying potential access holes into rubble interior. This capability would improve search operation by directing limited resources to be concentrated on areas where access holes might exist.
This thesis presents an approach to automatically identify access holes in rubble. The investigation begins by defining access holes in terms of their functional utility, that allow for their algorithmic identification. From this definition, a set of hole-related features extracted from RGB-D imagery are proposed for detection. Experiments were conducted using data collected over a real-world disaster training facility. Empirical evaluation indicates the efficacy of the proposed system for successfully identifying potential access holes in disaster rubble scenes.