3DFin

3DFin detects stems in a terrestrial lidar scan and fits cylinders to them, returning diameter at breast height, tree height and taper. It uses no learned weights, no training data and no stochastic inference.

The same scan therefore produces the same numbers on every run, and the method can be described in full in a publication. Where a neural network’s output would invite scrutiny, 3DFin provides a more readily defensible result, and a baseline against the deep learning segmenters.