A comparison of photogrammetric software for deriving structure-from-motion 3D point clouds and estimating tree heights

Authors: Volga Lipwoni, Michael S. Watt, Robin J.L. Hartley, Ellen Mae C. Leonardo, Justin Morgenroth
Publication: New Zealand Journal of Forestry, Volume N.Z.J.For. 2021, Issue N.Z.J.For. 66(4) 2022, pp 18–26, Jan 2022
Publisher: New Zealand Institute of Forestry

---

Abstract: The use of structure-from-motion (SfM) photogrammetry from unmanned aerial vehicles (UAVs) is becoming an increasingly popular means of characterising key forestry biophysical variables such as tree height. Despite the wide array of software that is available to process 3D point clouds from SfM, little research has investigated how the precision of predictions vary between software. This study compared the accuracy of tree height estimates for a young Pinus radiata trial (height range 1.4 - 6.1 m) obtained from 10 different software packages, which were used to derive canopy height models (CHMs) from UAV-acquired SfM point clouds. To ensure a fair comparison, the default parameters for each software were used without any data tuning. Predictions of tree height ranged widely in terms of both precision (R² range: 0.61 - 0.86) and bias (mean bias error (MBE) range: 0.28 - 3.37 m). Height predictions with the highest precision and lowest bias were made using 3DF Zephyr (R² = 0.86; MBE = 0.58 m), Pix4DMapper (R² = 0.78; MBE = 0.28 m) and Maps Made Easy (R² = 0.85; MBE = 0.85 m). The availability of numerous software options provides choice to the user and this study helps to identify the best software for estimating tree heights from SfM-derived point clouds.

---