ABSTRACT In line with the forestry principles adopted by the countries, it is of great importance for sustainable development to protect the trees, which play a key role in preventing problems such as climate change and global warming, which have been on the agenda in recent years, and to segment their current status with remotely sensed data and to process individually into forest inventory information. In this paper, a novel approach based on the two‐stage Improved Local Maximum method is proposed for individual tree crown segmentation from point clouds. Additionally, four different tree metrics (tree height, crown width, diameter at breast height, crown base height) were estimated as a result of the segmentation for individual trees. The workflow of the proposed approach consists of detecting individual trees using the existing Marker‐Controlled Watershed Segmentation (MCWS), followed by an alternative solution with the Improved Local Maximum method supported by Bézier Curves in two different planes for the improvement of erroneous tree clustering in adjacent trees. Existing and improved allometric equations are used for metric estimation of the resulting tree clusters. Aerial LiDAR (UAV‐LiDAR) and photogrammetric point clouds of stone pine ( Pinus pinea ) woodlands in Akdeniz University and spruce ( Picea orientalis ) woodlands in Maçka (Trabzon) were used as datasets to test the proposed method. According to the object‐based accuracy analysis, the average segmentation accuracy is 79% and the spatial error at the tree centers is ±0.785 m. In addition, the proposed method provides ease of work and speed compared to large‐scale ground‐based tree segmentation and metric extraction methods.
Bahadir et al. (Tue,) studied this question.