Accuracy assessment of stem classification obtained from forest point cloud using fsct algorithm

The recent advancements in close-range remote sensing techniques are creating opportunities to explore more about the forest ecosystem more precisely. The Light Detection and Ranging (LiDAR) scanners are a boon to forest applications because they provide detailed and precise information about the forest structure from ground to canopy top level. The point cloud data collected with Terrestrial Laser Scanner (TLS) must be processed to get exact information about the trees, including topographic terrain information in the forests. Tree parameters are essential to calculate the total productivity of the forests and their monitoring. In this research, two forest plots were considered in the Zvolen district within central Slovakia; one forest plot (TLS_Plot1) consists of 49 trees, and the other (TLS_Plot2) consists of 102 trees. A total of nine TLS scans were performed in each of the forest plots. It is also crucial to segregate the unstructured 3D point cloud data into various classes for accurate information extraction of each feature in the forest, such as stem, canopy, terrain, dead wood, etc. Forest Structural Complexity Tool (FSCT) algorithm is one such algorithm that has been recently developed for point cloud classification. So, we are focusing on the classification function of this tool. The classification function is classifying the forest point cloud data into stems, leaves, some branches, wood debris, lying dead wood, etc. We tried to investigate the accuracy of stem classification obtained for the forest point cloud using the FSCT algorithm. But a qualitative assessment must be done to evaluate the accuracy of the point cloud classification obtained. For this reason, stems from both the forest point cloud were manually extracted and compared with the stems point extracted from the FSCT algorithm classification, and the accuracy was evaluated. The FSCT algorithm-based stem classification achieved accuracy for TLS_Plot1 and TLS_Plot2 is 94.80% and 96.28%.