| The distribution and size of branches are the main factors affecting wood quality.It is very important to understand the information of branches comprehensively for improving wood quality.It is a waste of resources to obtain branch information through traditional tree analysis,while terrestrial laser scanning can collect branch data from the forest without damage.In order to verify the feasibility of tree analysis data with terrestrial laser scanning point cloud data,and to comprehensively consider the economic value of Larix plantation,30 trees of different grades of trees in 6 artificial larch forests in Mengjiagang Forest Farm were selected as the research objects.By terrestrial laser scanning to obtain Larix plantation three-dimensional point cloud data,extraction of 30 groups DBH and tree height value and 1266 groups of branch factors information in the form of human-computer interaction,and matching the extracted branches with the measured branches,the extraction accuracy of sample trees and branch factors was analyzed,and the number prediction model of branches was established.In order to reduce the impact of larch needles on point cloud data,point cloud data and tree analysis were collected in November 2019 in this study.The results showed that:(1)The average value of tree height extraction accuracy was 98.15%,and the minimum and maximum values were 91.40%and 99.77%.After removing 4 abnormal samples,the average extraction accuracy of tree height was 99.08%.The average accuracy of DBH extraction was 99.54%,and the minimum and maximum values were 98.67%and 99.90%.The comprehensive analysis showed that DBH and tree height had higher extraction accuracy.(2)The extraction rate of branches was above 98%when the relative height of branches was between 0.1H and 0.6H,84.85%when the height of branches was between 0.6H and 0.7H,and 70.16%when the height of branches was between 0.7H and 0.8H.Among all the branch factors,the extraction accuracy of branch height was the highest,and each part was greater than 96%.The extraction accuracy of branch length and chord length were similar,with average extraction accuracy of 89.97%and 90.58%.The average extraction accuracy of branch angle and branch diameter was 82.50%and 72.52%.The extraction accuracy of branch arc height was the least,and the average extraction accuracy was 70.74%.The extraction rate and the extraction accuracy of branch factors both showed a decreasing trend with the increase of relative branch height.(3)In this paper,the crown was divided into dead crown and living crown,and the living crown was divided into the lower part of the living crown,the middle part of the living crown and the upper part of the living crown from bottom to top.The average extraction rate of the dead crown could reach 100%.The average extraction rates were 99.03%,76.18%and 8.91%for the lower for middle and upper parts of the living crown.The extraction rates for the middle and lower parts of the living crown were higher.The extraction accuracy of branch factor is also different in each part of the tree crown and all of them showed a decreasing trend with the increase of crown height and branch density.In any position of dead and living crowns,the extracted values of branch length,chord length and branch height can replace the measured values.The extracted values of branch diameter and branch angle can replace the measured values in the lower and dead crowns.(4)The tree crown was divided into two parts:above contact height and below contact height.The average extraction rate was 94.12%for the branches below contact height,and 72.40%for the branches above contact height.The extraction rate of the branches below contact height was significantly higher than that of the branches above contact height,and the more adjacent trees,the smaller the extraction rate.In the two parts of the contact height,the extraction accuracy of the branch height has no significant difference,and the extraction accuracy is more than 98%.In addition,the extraction accuracy of all the other branch factors except the branch height was as follows:below the contact height>above the contact height,and the extraction value of each branch factor below the contact height could replace the measured value.(5)The accuracy of branch diameter extraction is low by extracting directly from terrestrial laser scanning point cloud data,a branch diameter prediction model was built based on the information of sample tree and branch factors in the measured tree analytical data.The results showed that when the branch diameter is less than 20mm,the branch diameter prediction model significantly improved the accuracy of base diameter.(6)Based on number as the dependent variable,from the branches of the point cloud data statistics with every factor of the extracted trees and branches data as independent variables to build branch number prediction model,the results showed that the fitting result of mixed effect model was superior to the basic model,will eventually contain a constant term,relative total branch the square of the depth of this two models with stochastic parameters optimal mixed effects models as branches.After the model was adjusted,the coefficient of determination R2 was 0.77,the mean error of the fitting was 1.21 e-14,and the root mean square error was 2.59.In conclusion,the DBH,tree height and branch factor whose relative branch height below 0.8 can be accurately extracted based on terrestrial laser scanning point cloud data,and the extraction rate and extraction accuracy were high.When the branch diameter was less than 20mm,the extraction results can be corrected by establishing a prediction model.On the whole,it was more feasible to replace the branch analytical data with terrestrial laser scanning point cloud data,which was helpful to realize the real-time monitoring of the stand,make scientific and reasonable pruning and management plans in time,improve the wood quality and realize the targeted cultivation of wood. |
PDF Full Text Request