Land Classification of Wavelet-compressed Full-waveform Lidar Data

Given sufficient data storage capacity, today’s full-waveform LiDAR systems are able to record and store the entire laser pulse echo signal. This provides the possibility of further analyzing the physical characteristics of the reflecting objects. However the size of the captured data is enormous and currently not practical. Thus arises the need for compressing the waveform data. We have developed a methodology to efficiently compress waveform signals using a lossy compression technique based on the discrete wavelet transform. Land classification itself is also a non-trivial task. We have implemented an unsupervised land classification algorithm, requiring only waveform data (no navigation data is needed). For the classification Kohonen’s Self-Organizing Map (SOM) has been used. Finally, the effect of the information loss caused by the lossy compression scheme on the quality of the land classification is studied.