Confidence-guided roadside individual tree extraction for ecological benefit estimation

Abstract

Roadside trees are an important component of the urban ecosystem, and extracting their location based on a point cloud obtained through mobile laser scanning (MLS) is essential for the urban ecology, but remains challenging because of heavy occlusions, disturbance due to tree stake systems, and overlaps between furniture on the street and the crowns of trees. To overcome these limitations, this study proposes a confidence-guided method to extract roadside trees. First, pole-like objects are extracted from among the roadside objects, and the confidence with which they can be assumed to be tree trunks is estimated to guide the order of segmentation. Second, the optimized min-cut algorithm is applied to extract the trees by combining the constraints on confidence-based guidance and allometric growth. Third, the morphological parameters and living vegetation volume (LVV) are estimated at the level of individual trees. Finally, the proposed method was verified on four challenging datasets with different point densities, species of tree, planting densities, and occlusions. The results show that it is robust and effective, with a precision of 88.5% and recall of 92%, and is a useful tool for ecological assessment based on the LVV of urban streets.