Construction Of Tree Volume Model Based On UAV Aerial Survey Data

Tree volume is the key index to evaluate the economic benefits of trees.It is an important task to get tree volume by non-destructive and efficient way.In order to reduce the workload of traditional ground survey,based on the aerial view of UAV remote sensing,it can accurately and efficiently get the information of forest canopy density,tree crown,tree height,and carry out the research and construction of forestry data table.In this paper,three five poplar plantations of different ages in the yellow sea seashore National Forest Park of Dongtai City,Jiangsu Province are taken as the research object.By using the methods of linear regression and nonlinear regression estimation,the correlation between DBH,tree height and crown diameter is studied by tree height level,and the optimal DBH estimation model suitable for each tree height level is established;The relationship between volume(V)and height(H),crown diameter(c),DBH(d)was studied in the paper,and the optimal volume estimation model for each tree height grade was established.The results are as follows:1)The accuracy of canopy factor and tree height based on UAV remote sensing is verified.Based on the Digital Orthophoto Image,the canopy density was extracted by the method of manually sketching the outer contour of the tree crown.After linear fitting with the measured value,the decision coefficient(R~2)of the two was 0.9669,and the root mean square error(RMSE)was 0.0374.The object-oriented method was used to extract the crown of the forest,and the results of the linear fitting were as follows:when the canopy density is less than 0.7,R~2 of the fitting equation is 0.9829,RMSE is 0.2810;when the canopy density is greater than or equal to0.7,R~2 of the fitting equation is 0.7379,RMSE is 1.0156.The results show that the object-oriented extraction method is feasible for the extraction of forest crown in the middle and low canopy density stands,but when the canopy density is large,the extraction accuracy of the crown is significantly reduced.Based on the canopy height model,the tree height was extracted.The R~2 of the two models was above 0.9303 and RMSE was below 0.9352.2)The optimal D-C model,D-H model and D-C,H model are constructed at the tree level.It was found that the correlation between DBH and crown diameter decreased and the correlation between DBH and height increased with the increase of tree height.At the same time,the binary DBH regression model established by crown and tree height is introduced,R~2 is significantly higher than that of the univariate model,which effectively improves the accuracy of DBH estimation;with the increase of tree height level,R~2 of the binary regression model decreases.Combined with previous research results,it is possible that with the increase of forest age,site conditions,competition among trees,climate conditions and other factors have an impact on forest factors,and the correlation between forest factors is gradually weakened.3)According to different tree height grades,based on tree height,crown diameter and DBH,the optimal one-way volume model,two-way volume model and three-way volume model are established respectively.Comparing the optimal V-D,V-H and V-C models in different tree height classes,we can see that:within each tree height class,the correlation between volume and DBH is the highest,R~2 is between 0.804 and 0.851;within each tree height range,the optimal models are all in the form of linear or power function,which shows that the correlation among DBH,crown,tree height and wood volume can be fitted by linear equation or power function.Comparing the optimal V-D and H models in different tree height classes,we can see that:in each tree height class,the optimal binary volume equation is V=a+bD~2+cD~2H+dH;in each tree height range,the R~2 of the optimal binary volume model is above 0.898,which is significantly higher than that of the unitary volume model.Comparing the best V-C and H models in different tree height classes,we can see that:when the tree height is below 20 m,R~2 of V-C and H models is slightly higher than that of V-D and H models;when the tree height range is greater than 20 m,R~2 of V-C and H models is significantly lower than that of V-D and H models.Comparing the optimal V-D,H and C models in different tree height classes,we can see that the form of ternary volume model is V=a+bD~c+dH~e+fC~g;in the range of tree height,the optimal volume estimation model is ternary volume model,R~2 is above 0.908.4)The optimal single volume model,two volume model and three volume model are verified by the test data.The variance analysis and multiple comparison of the volume estimated value and the volume measured value under different models show that the difference between the volume estimated value and the measured value based on the three-dimensional volume estimated model is not significant in each tree height grade,which shows that the three-dimensional volume estimated model can be used to carry out the tree volume with high accuracy in different tree height grades of the three five poplar estimate.