Signal Processing Techniques for Feature Extraction and Classification using Small-Footprint Full-Waveform Airborne LIDAR

Full-waveform digitizer hardware integration within discrete return LIDAR systems provides an enhanced capability to resolve vertical structure within the laser line of sight and potentially classify specific surfaces or objects. However, the subject of waveform signal processing as it applies to surface classification is fairly underdeveloped. This research includes the examination of LIDAR waveform pulse characteristics for known targets and vegetation types. Using these data, a number of signal processing techniques were investigated as precursors to classification engines without prior knowledge of surface slope, or obscuration density. Relevant waveform features were extracted using both Gaussian decomposition method and raw waveform features revealing surface classification distinctions. Preliminary results from this data set indicate that the total integrated waveform energy provides an efficient and rapid methodology for discrimination of vegetation from built surfaces. These results indicate that metrics derived from the full waveforms can be utilized to characterize and classify (to a limited degree) an environment without prior knowledge.