Assessment of the Influence of Survey Design and Processing Choices on the Accuracy of Tree Diameter at Breast Height (DBH) Measurements Using UAV-Based Photogrammetry

Abstract

This work provides a systematic evaluation of how survey design and computer processing choices (such as the software used or the workflow/parameters chosen) influence unmanned aerial vehicle (UAV)-based photogrammetry retrieval of tree diameter at breast height (DBH), an important 3D structural parameter in forest inventory and biomass estimation. The study areas were an agricultural field located in the province of Málaga, Spain, where a small group of olive trees was chosen for the UAV surveys, and an open woodland area in the outskirts of Sofia, the capital of Bulgaria, where a 10 ha area grove, composed mainly of birch trees, was overflown. A DJI Phantom 4 Pro quadcopter UAV was used for the image acquisition. We applied structure from motion (SfM) to generate 3D point clouds of individual trees, using Agisoft and Pix4D software packages. The estimation of DBH in the point clouds was made using a RANSAC-based circle fitting tool from the TreeLS R package. All trees modeled had their DBH tape-measured on the ground for accuracy assessment. In the first study site, we executed many diversely designed flights, to identify which parameters (flying altitude, camera tilt, and processing method) gave us the most accurate DBH estimations; then, the resulting best settings configuration was used to assess the replicability of the method in the forested area in Bulgaria. The best configuration tested (flight altitudes of about 25 m above tree canopies, camera tilt 60°, forward and side overlaps of 90%, Agisoft ultrahigh processing) resulted in root mean square errors (RMSEs; %) of below 5% of the tree diameters in the first site and below 12.5% in the forested area. We demonstrate that, when carefully designed methodologies are used, SfM can measure the DBH of single trees with very good accuracy, and to our knowledge, the results presented here are the best achieved so far using (above-canopy) UAV-based photogrammetry.

Unveiling the Potential of Drone-Borne Optical Imagery in Forest Ecology: A Study on the Recognition and Mapping of Two Evergreen Coniferous Species

“…The choice of flight altitude and final image resolution should be carefully considered and tailored to the specific study and recognition objects. Ideally, researchers should have the opportunity to experimentally determine the optimal flight parameters based on empirical experiences [39,40].…”

A novel post-fire method to estimate individual tree crown scorch height and volume using simple RPAS-derived data

“…While the lack of derived trees can be simply attributed to complete crown consumption, this is still a limitation. The derivation of DAP points that can reliably reconstruct subcanopy trees and individual tree boles could have been improved through the collection of oblique RPAS images (Perroy et al 2017;Ye et al 2019;Moreira et al 2021). Such images could have supplemented the NADIR RPAS images used in this study.…”

Comparison of Low-Cost Commercial Unpiloted Digital Aerial Photogrammetry to Airborne Laser Scanning across Multiple Forest Types in California, USA

“…Two missions were flown over each site, one with images taken at nadir and the other with images taken 30 degrees off nadir. Previous studies recognized greater than 80% overlaps and 15 to 45 degrees of tilted camera as desired options for canopy tree studies [36][37][38]. All missions were flown on days with little to no cloud cover, minimal wind and near local solar noon, with the exception OW which was flown later in the afternoon (approximately 2:00 p.m. local time) due to concerns of heat impacts on UAS batteries.…”

“…Furthermore, standard practices of UAS DAP data acquisition typically include only collecting images at nadir, yet multi-angle DAP has potential in improving characterization of vertical forest structure [33]. By including off-nadir imagery in a DAP dataset, the image sensor has an increased view of the sides of the forest canopy, potentially allowing the photogrammetry algorithm to create a more "complete" model of the whole canopy than with nadir imagery alone [34,[36][37][38]. An example of this can be seen in Figure 1, where the 3D DAP model generated with multi-angle imagery includes more of the lower tree canopy than nadir imagery alone.…”