Forest parameters is an important investigation content in forest resource monitoring.How to measure the tree trunk factor and trunk volume accurately is an urgent problem which needs to be solved by traditional survey methods.The ground-based radar technology can figure out the map.However,there are still several problems need to be solved,including but not limited.First,how to process the large number of point cloud data which detected by ground-based radar.Second,it is difficult to simulate the three-dimensional trunk surface based on point cloud data,therefore,the precision of measured trunk volume is affected.This paper will explain the survey result of three-dimensional laser scanning on the Huangfengqiao State-owned Forest Farm in Hunan Province,the forest parameters are extracted,and the trunk volume will be estimated.The main conclusions are as follow:(1)Present exact extraction algorithm of forest parameter measurements which based on ground-based LiDAR point cloud data.In this research,the ground-based LiDAR point cloud data are intercepted horizontally each 50 centimeters that along the vertical direction of height.In addition,the intercepted point cloud is simplified by the quadrant angle of 15 degrees,which can drop the data size as 1/200 of before.The isoline model can be constructed by simplified point cloud data.According to convex hull algorithm and circumference measurement principle,the tree measurement factors such as DBH,diameter and crown width at different heights of tree trunk are extracted.The overall accuracy of extracted results exceeds 90%.(2)In this research,the optimal taper equation is established by several variables of 62 Chinese fir samples.Combined with simple taper equation model and variable parameter taper equation model,the optimal taper equation can be obtained by comparing and analyzing the accuracy evaluation factors.The determination coefficient R2of the fitting equation is 0.974,and the root mean square error(RMSE)is 0.887.Based on the fitting equation,the trunk diameter of the missing part of TLS data can be calculated and analyze,which also can validate the missing point cloud data.(3)An effective method to repair the missing point cloud of tree trunk surface model is proposed,and the Delaunay triangular network model of tree trunk three-dimensional surface is constructed.It is found that the taper equation and quadrant angle simplification method combined with diameter fitting circle principle can be used to fit the trunk point cloud corresponding to the part covered by the crown.The fitting accuracy of point cloud missing data is basically consistent with the measured data.The distribution of residual errors is random and the fitting accuracy of parameters is high.The determination coefficient is 0.992 and the RMSE is 0.513.Based on fitting point cloud data and high part point cloud data under trunk branch,the Delaunay triangular network model of Chinese fir trunk can be constructed and the volume is extracted.Compare with measured data,the accuracy of volume extraction is excellent.The determination coefficient is 0.987,and the RMSE is 0.042.(4)Based on extraction of forest parameters that detected by ground radar,the volume model of Chinese fir is constructed.According to the one-dimensional volume model and two-dimensional volume model,the correlation analysis and stepwise regression between the modeling factors and volume has been made,which improved the two-dimensional volume model.Feather more,the relationship model between different volume segments and total volume;the volume model of Chinese fir based on simultaneous equations are constructed.The evaluation index of each volume model achieved certain accuracy,and the decision has been made.The coefficients are almost above 0.90.In conclusion,the result about volume model of 31 Chinese fir on Ground LiDAR method is better than the traditional volume model.The most accurately volume model is the power function relationship model between the volume of different segments and the volume of the whole volume,which is constructed by the volume of trunk below 2/3 height(V2/3H)and the volume of whole volume(Ⅴ)respectively.The determination coefficient is 0.999,and the RMSE is 0.011.In addition,it is found that crown width(CW),volume of different segments(Vx)and height under branches(UBH)can effectively improve the accuracy of volume model due to simultaneous equations,but these factors cannot be constructed volume model separately,but must be combined with tree height or DBH. |