Comparison of under-canopy unmanned aerial vehicle, airborne, and ground-based mobile laser scanning for forest field reference measurements
No Thumbnail Available
URL
Journal Title
Journal ISSN
Volume Title
Insinööritieteiden korkeakoulu |
Master's thesis
Authors
Date
2023-12-11
Department
Major/Subject
Geoinformatics
Mcode
Degree programme
Master's Programme in Geoinformatics (GIS)
Language
en
Pages
87+27
Series
Abstract
Knowledge about forests from forest inventories typically rely on field reference measurements of individual trees. The fast collection of data and accuracy of tree attribute estimates obtained is essential. Unmanned aerial vehicle (UAV), airborne, and ground-based mobile laser scanning constitute significantly faster methods of collecting data than manual approaches. Thus, the evaluation of their accuracy in terms of tree attribute estimation is of importance. This thesis aims to compare six different kinematic laser scanning systems using an automatic processing flow on their accuracy in tree detection and tree attribute estimation. The following laser scanning systems were included in the comparison: 1) a high-density airborne laser scanning (ALS) system; 2) two under-canopy UAV systems; 3) three handheld mobile laser scanning systems. The data was collected from eight 32 m × 32 m field plots that represented three forest complexity categories. The obtained point clouds were automatically processed using identical workflows to directly measure the stem curves and estimate tree heights, diameters at breast height (DBH), and stem volumes of the detected trees in the plots. To evaluate the accuracy, the results were compared against a multi-scan terrestrial laser scanning (TLS) reference; the height estimates were compared against the ALS system. The reference data was obtained using the same workflow as the studied data, with the accuracy of the algorithmic workflow evaluated for TLS data. The results find clear variation in the accuracies of the tree attribute estimation that can be attributed to sensor differences. Meanwhile, the stem detection completeness and correctness were high for all under-canopy systems. The sensors that had a narrower vertical field of view were found to produce systematic underestimation of tree heights greater than 15-25 m, and the height underestimation increased with plot difficulty. These sensors also had a higher beam width, and found less accurate DBH, stem curve, and stem volume estimates. The other under-canopy sensors included in the study found results in the easy plots that are close to some of the highest accuracies obtained from previous research. Increased plot difficulties generally reduced the accuracy. The ALS system had accuracies at similar levels to prior studies on ALS, but with a lower completeness rate. Moreover, the results found the algorithm to work well with TLS data, which validated its use as a reference.Description
Supervisor
Vaaja, MattiThesis advisor
Lehtomäki, MattiMuhojoki, Jesse
Keywords
airborne laser scanning, UAV laser scanning, mobile laser scanning, stem curve, tree height, stem volume