Tree point clouds can be used to estimate tree structure information in a nondestructive manner, which is very useful for studying forest ecosystems. Modeling and analysis of tree structures are critical in the investigation of tree topologies and biomass. One popular tree modeling method is a priori hypothesis modeling, which uses tree branches as cylinders and point clouds as input to model the branches by the topological structure of the tree skeleton. Tree quantitative structure modeling (TreeQSM) is an a priori hypothesis modeling method that enhances the regularity of tree models to rapidly obtain tree structures and is currently a mainstream tree modeling method mostly used for tree model reconstruction, biomass estimation, and other aspects. This study analyzed the accuracy estimations of tree height, diameter at breast height (DBH), and volume using TreeQSM based on multi-precision and multi-site scans of two artificial model trees and five real apricot trees.
To achieve a more effective evaluation of the TreeQSM method, experiments were conducted using both artificial model and real apricot trees. The model trees were constructed from smooth logs without bark or from rough logs with original bark. The real trees were in the leaf-off stage. The DBH, height, branch length, and diameter of each tree were manually measured, and the tree volume was calculated. A RIEGL VZ-400 scanner was used to collect multiscan tree point clouds. Single- and multi-scan tree point clouds with different scanning parameters were used to construct tree structure models using the TreeQSM algorithm. The tree models were then used to evaluate tree parameters such as tree height, DBH, and volume. The estimated and actual values were compared, and the absolute and relative errors were calculated and analyzed.
The estimation accuracies of the tree heights and DBHs of smooth trees reach 98.51% and 100%, with average accuracies of 96.19% and 98.06%, respectively, whereas those of bark trees reach 100%, with average accuracies of 98.67% and 92.90%, respectively. Overall, the TreeQSM method is shown to be relatively accurate in estimation. In terms of volume estimation, the estimation accuracy of the smooth, bark, and apricot trees reach 99.88%, 96.86%, and 82.88%, respectively, and the average accuracies are 94.37%, 89.62%, and 71.32%, respectively. Volume estimation is underestimated, but its accuracy can be improved by fusing multi-station data. First, results show that the configuration and selection of the scanning angular resolution and number of scans have distinct effects on the modeling accuracy of the TreeQSM method. The selection of appropriate parameter configurations when using the TreeQSM method enables accurate estimation of tree height, DBH, and other indicators. Second, more multiview scans with excessive overlap may not improve the accuracy of estimation results and may even cause noise superposition, resulting in reduced estimation accuracy. Third, the TreeQSM algorithm has a certain degree of randomness in classifying and retrieving data, and multiple processing results for the same data are not unique. Averaging multiple results can reduce the error of single estimation.
We use seven artificial and real trees to evaluate and analyze the performance of the TreeQSM method. Using TreeQSM, we analyze a relatively suitable scanning angular resolution and number of scans to estimate tree height, DBH, and volume. Results show that this method can obtain a better tree model by using appropriate scan parameters such as angular resolution and number of scans. Although the TreeQSM method performs well in estimating tree height and DBH, the method should be improved to estimate more accurately the complex topological structures of tree branches and trunks. This study demonstrates that the TreeQSM method still has significant biases in terms of modeling and volume estimation for complex branch structures. In future research, we will continue to study related topics such as tree trunk recognition errors and underestimated branch diameters in volume estimation.
