Comparison of estimation algorithms for individual tree diameter at breast height based on hand-held mobile laser scanning

Abstract

ABSTRACT Hand-held mobile laser scanning (HMLS) can quickly and effectively obtain tree point cloud data, which has great potential to conduct forest resource survey at the plot or stand scale. To improve the estimation precision of Diameter at Breast Height (DBH) for individual trees based on HMLS data, four algorithms were compared and the ideal conditions (slice thickness, inclination, point cloud integrity) of use for each algorithm were evaluated. First, the original point cloud data were denoised, sliced, layered, and clustered based on the normalized point cloud data. Then, the tree DBH was estimated by the four algorithms under different conditions. Results showed that: (1) The optimal point cloud thickness for estimating tree DBH was 5 cm for the two-dimensional algorithms and 15 cm for the three-dimensional algorithm. (2) The root mean square error (RMSE) varied by approximately 0.5 and 0.18 cm with the change of inclination for two-dimensional algorithms and three-dimensional algorithm, respectively. (3) A reduction in point cloud integrity can affect the accuracy of all four algorithms, while the fast convex hull algorithm was the most impacted. The two-dimensional algorithms were computationally more efficient than the three-dimensional one. However, the former was less robust than the latter.