Holmgren, Johan
- Department of Forest Resource Management, Swedish University of Agricultural Sciences
Conference paper2017Peer reviewedOpen access
Holmgren, J.; Tulldahl, H.M.; Nordlöf, J.; Nyström, M.; Olofsson, K.; Rydell, J.; Willen, E.
A system was developed for automatic estimations of tree positions and stem diameters. The sensor trajectory was first estimated using a positioning system that consists of a low precision inertial measurement unit supported by image matching with data from a stereo-camera. The initial estimation of the sensor trajectory was then calibrated by adjustments of the sensor pose using the laser scanner data. Special features suitable for forest environments were used to solve the correspondence and matching problems. Tree stem diameters were estimated for stem sections using laser data from individual scanner rotations and were then used for calibration of the sensor pose. A segmentation algorithm was used to associate stem sections to individual tree stems. The stem diameter estimates of all stem sections associated to the same tree stem were then combined for estimation of stem diameter at breast height (DBH). The system was validated on four 20 m radius circular plots and manual measured trees were automatically linked to trees detected in laser data. The DBH could be estimated with a RMSE of 19 mm (6%) and a bias of 8 mm (3%). The calibrated sensor trajectory and the combined use of circle fits from individual scanner rotations made it possible to obtain reliable DBH estimates also with a low precision positioning system.
Forest inventory; Mobile laser scanning; Remote sensing; SLAM; Tree detection; Tree map
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
2017, volume: 42, number: 3-W3, pages: 59-63
Publisher: International Society for Photogrammetry and Remote Sensing
Frontiers in Spectral imaging and 3D Technologies for Geospatial Solutions, 25–27 October 2017, Jyväskylä, Finland
Remote Sensing
Forest Science
https://res.slu.se/id/publ/130287