Innovative Surface Feature Extraction for Visualization using LIDAR Intensity and Co-registered Optical Data

Pre-mission rehearsal using 3D, simulated fly throughs derived from remotely sensed data can signicantly enhance warfighter safety and efficiency. Because of its great potential for extracting the location, orientation, identification, and elevation profile of surface features, LIDAR (light ranging and detection) data provide an excellent information base for creating these simulated visualizations. The proposed work will demonstrate the feasibility of using both LIDAR signal return timing and, uniquely, signal intensity to extract surface features. The work will be based on the ELF (Extracting LIDAR Features) algorithms, currently under development by HyPerspectives scientists. The ELF algorithms employ intensity data for scene classification into vegetated and non-vegetated regions, then employ a unique search procedure for feature identification and characterization. The work also will show feasibilty of co-registering high resolution optical spectra to the LIDAR scenes. When used in conjunction with LIDAR, high resolution optical sensors (e.g., hyperspectral) can provide valuable and highly complementary data for surface feature identification and characterization. Phase I proof of concept efforts will employ single return LIDAR, seek to extract three terrain features (buildings, vegetated patches, non-vegetated ground), and require user input for selecting the original input grid search size. In the Phase I Option, we will automate the ELF algorithm search procedures and, like Phase I, benchmark performance for incorporating the extracted data into a visual database. Manual co-registration of optical data will be shown for a single scene in Phase I, then automated in Phase II.