Voxel-based analysis attenuation corrected waveforms
Project title
Volumetric forest structure reconstruction from full-waveform airborne laser scanner data
Funding
German Research Foundation (DFG)
Motivation
Full-waveform laser scanning has become a very popular tool for an efficient acquisition of forest inventory related data. Recently, operational full-waveform laser scanner systems have become commercially available, delivering a time-resolved digitization of each laser scanner pulse. In addition to the digital terrain model and the digital crown height model obtained from conventional laser scanner data, these waveforms contain very valuable information about the vertical vegetation structure and density.
AimS and taskS
The chief aim of the project is to develop methods for generating volumetric forest stand representations of high geometric and radiometric quality from full-waveform airborne laser scanner data. We expect that these volumetric representations will facilitate the determination of parameters such as crown density, open stem area and the density of understorey vegetation. Applying 3D segmentation and single tree delineation techniques to multi-temporal data, biomass and biomass changes can be determined at high accuracy.
The generation of a voxel representation requires a series of geometric and radiometric transformations of the raw full-waveform laser scanner data. To achieve a better radiometric correction, models are developed which approximate the interaction between a laser pulse and the traversed vegetation. After a preprocessing of the voxel space containing gap closing, redundancy handling and data filtering, tasks like voxel space segmentation, individual tree detection and change detection analysis of multi-temporal data can be fulfilled.
The following figure shows two different voxelspaces of a subplot with three spruces which were generated inserting the entire waveform information. The voxel attributes were obtained from the raw waveform sample amplitudes (left) and from the attenuation-corrected amplitudes (right). After correction the enhancement of the attributes of the lower voxel is obvious.
Additionally, the following video shows the corrected voxelspace overlaid with the corresponding extracted point cloud and tree positions as vertical lines which were obtained automatically from the voxelspace.
Visualisation of attenuation-corrected voxelspace overlaid with extracted point cloud and tree positions which were automatically extracted from the voxel space
Related publications
acknowledgment
Thanks to Milan Geoservice GmbH for data acquisition and support.
Contact
- Prof. Dr. habil. Hans-Gerd Maas (project management)
- Dipl.-Ing. Katja Richter, Dipl.-Ing. Nadine Stelling (project work)