Normalized cuts for single tree isolation from LiDAR

Abstract

This study demonstrates a developing method for the isolation of individual trees in a canopy from 3-dimensional laser scanning data. This method is capable of delineating single trees from a storied canopy, and is effective for individual tree isolation. This research compared the Normalized Cut (NCuts) method of single tree segmentation to a more commonly applied Hierarchical Watershed Transform (HWT) within an urban setting. For this study two field sites located within the urban build-up of San Francisco, CA were selected. The results of this work demonstrated an implementation of the Normalized Cut method coded using freeware libraries, and showed significant differences between the two methods, especially in storied canopies. This study also demonstrated the effectiveness and feasibility of using freeware coding tools for research, thereby enabling a wider range of research possibilities without the expense of large software packages.