Unmanned Scanning
Moritz Bruggisser et al.: Influence of ULS acquisition characteristics on tree stem parameter estimation 10.08.2020
The paper present an approach for automatically detecting the positions of tree trunks, for determining their corresponding diameter at breast height (DBH), and for assessing the shape of tree trunks from 3D point clouds derived from unmanned aerial vehicle borne laser scanning (ULS). The experiments are carried out with point clouds from both a RIEGL miniVUX-1DL and from a RIEGL VUX-1UAV. The results reveal that the autonomous stem detection recognizes 91.0% and 77.6% of the stems, respectively, and that the DBH can be modeled with biases of 2.86 cm and 0.95 cm for 80.6% and 61.2% of the trees, when compared to field measurements. The authors further demonstrate that, compared to terrestrial laser scanning (TLS) data, the stem diameters along the tree can be estimated with biases below 3.4 cm and 1.4 cm for the two systems, respectively, up to a tree height of 12 m for stems with a DBH above 20 cm. Their experiments further reveal the accuracy of diameter estimations to be mainly dominated by the tree’s diameter with better accuracies for larger stems, while the completeness, with which a stem is covered by points, has little influence, as long as half of the stem circumference is captured. The absolute point count on the stem does not impact the estimation accuracy of all stem parameters, but is critical to the completeness with which a scene can be reconstructed. Conversely, the authors demonstrate the precision of the laser scanner to be a key factor for the accuracy of the stem diameter estimations, as in their experiments, they found the accuracies of the estimations from the VUX-1UAV to be higher than the ones from the miniVUX-1DL. The findings of their study assist to evaluate the potential of ULS for forest monitoring and management and allow for conclusions regarding the required point cloud qualities and, thus, the mission planning of ULS acquisitions, in order to deliver data products, which fulfill the requirements for an operational application in forest inventories.
The full article was published on ScienceDirect can be found here.