Classifying Compressed Lidar Waveform Data

Full waveform recording is becoming increasingly affordable and, consequently, available in today's state-of-the-art LiDAR systems. Therefore, there is no practical limitation to the complexity of pulse detection and other methods that can be applied in post-processing mode. Analyzing the entire return signal, the full waveform can provide additional geometrical and physical information about the reflecting surfaces. Currently, most LiDAR applications are based on utilizing only the geometry of the point cloud, where both the precision and density of these points primarily depend on the peak detection method used in real-time during data acquisition. Analyzing the properties of the full waveform in post-processing, additional information can be obtained that can provide a better geometrical description of the surface (point cloud) and object classification information that can be used, for example, for land cover classification. Since the storage requirements for waveform are quite significant for modern LiDAR systems because of the high-pulse rate, compression is an obvious choice to reduce storage and data transmission requirements. Though the original waveform can be fully reconstructed from the compressed format and used for waveform analysis in post-processing mode, an interesting question is whether the compressed format can be directly exploited for classification and/or peak detection. The objective of this paper is to investigate the feasibility of using the compressed domain waveform data for land cover classification and peak detection.