An Over-Segmentation-Based Uphill Clustering Method for Individual Trees Extraction in Urban Street Areas from MLS Data

In this article, an over-segmentation-based uphill clustering method for individual extraction of urban street trees from mobile laser scanning data is proposed to solve the problem that the existing methods depend heavily on tree trunks and have poor extraction results in complex environments where the tree trunks are blocked by cars and green belts, and the crown touching or interlocking is large. First, supervoxels are generated by over-segmentation, so that the amount of original data is reduced and the boundaries of different objects are effectively preserved. Then, the potential tree crowns and trunks are obtained by extracting typical object structures. Finally, individual trees extraction is realized by extracting independent crowns from the potential crowns via uphill clustering and searching corresponding trunks from the potential trunks. The main contribution of this article is to propose an individual extraction method for street trees based on uphill clustering that does not rely on the extraction of tree trunks, which improves the completeness of extracted results in complex urban environments. The experimental results demonstrate that the proposed method effectively extracted the street trees individually from the test data, with the completeness of 100%, the correctness of 96.4%, and the F -score of 0.98. Moreover, the proposed method also achieves good result for the extraction of greening trees that are heavily blocked in the green belt areas. And the corresponding completeness, correctness, and the F -score are 94.6%, 83.3%, and 0.89, respectively.