Research On Extraction Of Individual Tree Parameters And Estimation Of Trunk Biomass Based On Hand-held LIDAR

Hand-Held Mobile Laser Scanning(HMLS),as a new type of ground mobile laser scanning technology,can quickly obtain forest stand 3D structure information with certain accuracy,and has the characteristics of high efficiency and low data acquisition cost.In this study,a hand-held mobile LIDAR ZEB-REVO was used to obtain stand point cloud data,and the extraction accuracy of individual tree structure parameters from HMLS stand point cloud was calculated.The optimal DBH calculation method of HMLS point cloud under different pretreatment conditions was evaluated.And finally the nearest neighbor geometric feature values calculated by different KNN methods and the random forest classifier are used to achieve the classification of HMLS stand point clouds and the refined estimation of individual tree trunk biomass.The results of the study are as follows:(1)Accuracy evaluation of individual tree structure parameters extraction based on HMLS.In terms of point cloud quality,the individual tree trunk of HMLS was scanned completely,and there were many point clouds missing in the canopy part.From the perspective of the estimation accuracy of individual tree structure parameters,the estimation accuracy of HMLS DBH is high,with R2 0.92 and RMSE 1.60cm,which meet the accuracy requirements of DBH acquisition.However,the estimation error of tree height and canopy projection area is large,and the R2 is below 0.5,so there is still a lot of room for improvement.(2)Comparison of HMLS point cloud DBH estimation algorithms based on different preprocessing conditions.The optimal point cloud thickness of the 2D DBH estimation algorithm is 5cm,and the optimal thickness of the 3D cylinder fitting algorithm is 15cm.The relative accuracy of each algorithm can reach 90%.The increase of tree inclination Angle and the decrease of point cloud integrity will reduce the accuracy of DBH estimation.Under different inclination degrees,the RMSE difference of 2D algorithm is about 0.5cm,while that of 3D algorithm is about 0.18cm.The relative accuracy difference of DBH fitting of fast convex hull algorithm is 7.29%under different point cloud integrity,and the relative accuracy difference of other three algorithms is about 3%.(3)Classification of stand point clouds and biomass calculation based on HMLS.KNN algorithm based on adaptive distance calculation of eight neighbor geometry characteristics compared with other KNN method showed better classification accuracy,using the random forest classifier classification model of stand point cloud the overall classification accuracy of 93.09%,in addition to the canopy point cloud in part due to geometric feature similarity is not easy to distinguish,the rest of the basic realized the accurate classification;The Delaunay triangulation model based on PowerCrack algorithm can truly reconstruct the 3D structure of individual tree trunk,and the biomass of individual tree trunks calculated on this basis is in good agreement with that calculated by empirical formula(R2=0.88,rRMSE=0.19).In conclusion,HMLS can complete small-scale digital measurement and estimation of 3D structure of stand.HMLS data acquisition efficiency is higher and data processing is more automatic.The accuracy of DBH estimation and biomass extraction of individual tree trunk based on HMLS can meet the precision requirements of forest resource survey.The results of this study expand the application scope of ground moving 3D scanning technology in forest resource survey,and provide an effective reference for efficient digital extraction of forest structure parameters and point cloud classification.