The paper describes an approach to tree species classification based on features that are derived by a waveform decomposition of full waveform LIDAR data. Firstly, 3D points and their attributes are extracted from the waveforms, which yields a much larger number of points compared to the conventional first and last pulse techniques. This is caused by the detailed signal analysis and the possibility to detect multiple pulse reflections. Also, constraints are embedded into the mathematical model of the decomposition to avoid erroneous 3D points caused by the system electronics. Secondly, special tree saliencies are proposed, which are computed from the extracted 3D points. Subsequently, an unsupervised tree species classification is carried out using these saliencies. The classification, which groups the data into two clusters (deciduous, coniferous), leads to an overall accuracy of 80 % in a leaf-on situation. Finally, the results are shortly discussed.     «

The paper describes an approach to tree species classification based on features that are derived by a waveform decomposition of full waveform LIDAR data. Firstly, 3D points and their attributes are extracted from the waveforms, which yields a much larger number of points compared to the conventional first and last pulse techniques. This is caused by the detailed signal analysis and the possibility to detect multiple pulse reflections. Also, constraints are embedded into the mathematical model o...     »