基于主动遥感数据的林木测树因子及用表研建

[Objective]This research employs high-precision laser scanning to acquire non-destructive and efficient single-tree point cloud data,establishes a novel method for measuration tree factors of standing trees,updates the volume of Ginkgo biloba standing trees in the study area,enhances the efficiency and accuracy of forestry surveys,reduces resource waste,and promotes the modernization and sustainable development of forestry management.[Methods]Firstly,the terrestrial laser scanner(TLS)data and UAV-borne laser scanner(ULS)data of the whole stand in the study area were obtained,and the DBH data of Ginkgo biloba plantation were obtained through field investigation,a total of 1192 trees.The tree height and crown width in airborne data were determined.The upper diameter in TLS data was obtained by least square method,and the volume of individual tree was calculated.Then,the one-way volume table is directly compiled and evaluated.The trees were divided into large,medium and small crown lengths according to crown width,and the cutting degree equation of ginkgo biloba was constructed.The wood yield table was compiled with the optimal model,and the model evaluation and index verification were carried out.Then,ULS data processing was used to obtain the characteristic parameters of single tree point cloud,and the volume-tree height model was fitted into three crown levels.The recursive feature elimination method of random forest was used to analyze the correlation between all feature parameters and volume values.The feature parameters with high correlation were selected to enter the random forest model,and the unitary volume-tree height model was inverted to compile the comprehensive single-entry volume table of the tree.And test.[Results](1)The optimal one-dimensional volume formula derived from TLS data is V=0.0002D2+0.0084D-0.089,with R2=0.744.The slope model is d2=0.670D1.595(H-h)2.659/H2.58,with R2=0.867.After testing the two models,it was concluded that there is no systematic bias,and a one-dimensional volume table and outage rate table were compiled.(2)The one-dimensional volume-tree height model and random forest model were respectively used for one-dimensional linear regression analysis of their predicted volume values and actual volume values,and the results showed that the random forest volume model had higher accuracy.(3)The elev_mean and density_metrics[3]feature parameters of the single tree point cloud with the highest correlation with volume were selected and added to the random forest volume model,and the training set data model RMSE=0.162,R2=0.807,RRMSE=0.623,MAE=0.067;the test set data model RMSE=0.142,R2=0.837,RRMSE=0.662,MAE=0.072.(4)A comprehensive single-entry volume table was established based on the random forest model by dividing the crown width level.[Conclusion]The active remote sensing technology of laser scanning provides efficient and non-destructive live tree data for tree measurement research.After compiling the one-way volume table and percentage of outturn table for Ginkgo biloba,it was found that the single crown area is proportional to the trunk diameter.Combining the different crown construction results,the larger the crown area,the higher the total percentage of outturn of Ginkgo wood species,i.e.the higher the relative utilization rate of wood.At the same time,the comprehensive single-entry volume table established based on the onboard data fills the basic data table for the study area,and provides an effective method for establishing forestry data tables based on drone laser data in the future.