A simple oriented search and clustering method for extracting individual forest trees from ALS point clouds

Accurate extraction of individual trees has theoretical and practical significance for improving forest management and productivity levels. Although airborne laser scanning (ALS) technology is frequently utilized for large-scale forest mapping and the detection of individual trees, the challenges of detecting individual trees in multi-layered and deciduous forests remain for most canopy-based methods. This study aimed to develop an efficient individual tree detection method capable of delineating trees with various heights, especially for sub-canopy trees. Inspired by the "from bottom to top" growth characteristics of tree branches, we propose an oriented search and clustering method, which clusters tree points upwards to the local tops, making it more adaptable to the roughly conical growth characteristics of a forest's tree canopy. On the NEWFOR dataset, our approach demonstrated comparable overall performance to that of state-of-the-art methods. In the non- dominant layers of multi-layered forests, our method achieved RMS match values of 30% in the 2-5 m range, 31% in the 5-10 m range, and 55% in the 10-15 m range, demonstrating the best extraction performance. A practical case study was conducted on selected plots of ALS point clouds acquired from the Bavarian Forest National Park (BFNP), with 1704 trees identified. This work provides amore reliable and efficient method for the future forest point cloud segmentation of individual trees; the method better meets the needs of forest inventory and resource management and lays a foundation for good forest management.