Light Detection and Ranging (LIDAR) Surface Feature Extraction Tool

The proposed effort will demonstrate a new approach to fully automated extraction of texture mapped 3D polygonal descriptions of manmade objects from multiple viewpoint LIDAR data and co-registered visible imagery. We will apply robust algorithms that have shown excellent results in use with LIDAR imagery for automated targeting applications. Expert system techniques based on typical building layout will be used to better separate building features from natural clutter. Once extracted, building descriptions will be constrained to fit common-sense properties of human building design, allowing the polygonal descriptions to be significantly simplified while retaining high standards of accuracy. Once fully proven in this domain, our algorithmic techniques will be extended to show the feasibility of future goals. Our Phase I and Option effort will demonstrate that databases produced can be run in real-time without exceeding commercially available graphics and processing hardware limitations, and will lay the groundwork for a Phase II demonstration of an algorithm suite supporting the automated extraction of intermixed natural (i.e., terrain, tree) objects and manmade (building, road, etc) objects using airborne or spaceborne LIDAR data and co-registered visible imagery.