Parameters Extraction Research Of Quercus Glauca Individual Tree By Terrestrial Laser Scanning

The terrestrial laser scanner(TLS)can observe object quickly and accurately and save the superficial information as point cloud,which have unparalleled advantages on short time series and small-scale feature observation compared with other remote sensing means.The Laser beams from TLS can penetrate gaps in the forest canopy so the point cloud can be used for forest parameter extraction.It has become a current research hotspot to use TLS for qualitative and quantitative analysis of forest resources.Most of research on TLS forestry resource survey are on the urban woodland or plantation areas.The methods of TLS data acquisition and processing in natural forest environments need to be researched further.Taking the natural forest based on Lutou Experimental Forestry Farm of Central South University of Forestry and Technology as the study area,and the data was collected from Quercus glauca Thunb.by FARO FOCUS 3D X330 terrestrial laser scanner.Firstly,the individua tree point cloud data of Quercus glauca was isolated from the original scan data.Then using the improved classification and regression tree(CART)model to get the trunk point cloud data from it.Finally,trunk skeleton constructed by hierarchical clustering algorithm from trunk point cloud data was used to improve the traditional individual tree parameter extraction algorithm.Analysing the results of parameter extraction and the main findings of the study were as follows:(1)A standardized approach to the acquisition and automated processing of TLS point cloud data from natural forest is established.This study presents a TLS point cloud data registration mechanism based on spherical targets for natural environments,specifically focusing on natural forest of Quercus.Through extensive experiments,scanner layout plans corresponding to different types of sample plots were summarized,and the proposed mechanism was used to supplement the existing methods of TLS point cloud data acquisition and automated processing.(2)An improved model for leaf and wood classification is proposed.In order to achieve high-precision point cloud classification,the study constructed 8 feature descriptors based on the neighbourhood features of the point cloud,improved the CART model by gradually introducing variables,and eliminated the model overfitting problem by adjusting the model structure.The improved CART model is better in accuracy and stability,which has a simpler structure without overfitting.(3)An improved method for extracting diameter at breast height(DBH)based on the trunk skeleton is proposed.The study used a hierarchical clustering method to separate skeleton points from the trunk point cloud,and constructed the topological relationships between skeleton points by using hierarchical shortest path algorithm based on graph theoretical knowledge and ecological theory.Calibrating the point cloud slice of breast height by calculating the trunk tilt angle of the tree at breast height.The DBH were extracted by convex hull algorithm and least squares circle fitting algorithm.The improved least squares circle fitting algorithm got the best result on DBH extraction with a coefficient of determination R2 of 0.927,Root Mean Squared Error(RMSE)of 1.951 cm and Relative Root Mean Squared Error(RRMSE)of 10.891%,which can provide the reference for the extraction of diameter at breast height parameters in forest resource surveys by TLS.(4)There are significant differences between height and trunk length extracted by skeleton method from individual tree point cloud in natural forest environments.The study explored the feasibility of extract tree height parameters with trunk point cloud skeleton.The tree height parameters were extracted by calculating the height difference,straight-line distance and skeleton path length from the tip point to the base of skeletons.Compared with the traditional method,the results show that the accuracy of tree height extraction by the trunk skeleton is lower than traditional method.Most of the results are low,the results from trunk skeleton height difference have the largest difference to the measured value,which average deviation is-1.767 m,and the relative average deviation is about 16%.In the three methods by the trunk skeleton,the extracted results by skeleton path distance are closest to the measured value,but they are overall high and their residuals are significantly different to the traditional method.In natural forest environment,the accuracy of parameters extraction on tree height by using TLS point cloud data can’t meet the requirements of forestry survey.The paper presents a systematic study on the applicability of tree parameter extraction algorithm with TLS data in natural forest environment,and improves the algorithm by incorporating the stem shape characteristics of the Quercus glauca.The results of the improved algorithm can meet the needs of forestry investigation.The result of the study can provide methodological and technical reference for the application of TLS in forest resources survey and monitoring.