| As emerging Li DAR scanning technologies,UAV-Li DAR technology and hand-held mobile laser scanning(HMLS)technology have the advantages of high data acquisition efficiency and low cost.However,when these two technologies are applied to complex forest surveys,it is difficult to collect complete tree information due to platform limitations.So it has become an inevitable trend of information processing to integrate data from two platforms to obtain complete tree information.Therefore,this study proposes an unlabeled fusion method of UAV-Li DAR and HMLS point cloud data,and performs data compression,individual tree segmentation,structural parameter extraction,model construction,and biomass and carbon stock estimation on the fused data.The results are as follows:(1)Propose an accurate fusion method for UAV-Li DAR and HMLS point cloud data.The fusion method is based on Delaunay triangulation and iterative nearest point(ICP)algorithm,and optimized by simulated annealing algorithm.In the sample area,the coordinates of projection points of tree position of the two platform point clouds after registration were compared,and the average coordinate offset distances of projection points of Pinus sylvestris plot and Quercus mongolica plot were 0.19 m and 0.25 m,respectively.According to the offset error calculated by the marker set during data collection,the mean square error of the fusion results of Pinus sylvestris plot and Quercus mongolica plot were 0.0512 and 0.0802,respectively.The fusion accuracy of cloud data of Pinus sylvestris plot was higher than that of Quercus mongolica plot.(2)Propose data compression and individual tree segmentation method based on fusion data.The point cloud data compression method is based on slice technology and angle-chord deviation method,and the individual tree segmentation algorithm is based on overlapping method and mean-shift algorithm.The result shows that the compression time is shorter and the compression effect is better when the slice thickness,chord deviation threshold and angle deviation threshold are properly selected.The compression rates of point cloud data of Quercus mongolica plot and Pinus sylvestris plot were 86.84% and 89.47%,respectively.The effects of data compression on DBH,height and crown width of trees before and after compression were negligible.The result shows that the recall rate,accuracy rate and F-score were 0.76,0.86 and0.81 for Quercus mongolica plot,and 0.86,0.84 and 0.85 for Pinus sylvestris plot,respectively.(3)Extract tree structural parameters,build models and estimate tree biomass and carbon stock based on individual tree data.Three tree structure parameters,DBH,tree height and crown width,were estimated based on individual tree data.The regression analysis between the estimated values and the measured values showed a good correlation between the estimated values and the measured values.The quantitative structure model(QSM)algorithm was used to reconstruct the three-dimensional model of individual trees and estimate the biomass and carbon stock of individual trees according to the volume of branches.The regression analysis between the reference value of biomass calculated based on allometric growth equation and the estimated value showed that the estimated value of biomass was in good agreement with the reference value and the error was small.In conclusion,the integration of UAV-Li DAR and HMLS point cloud data is conducive to the complete extraction of forest information.In this study,the point cloud data obtained from the air and the ground are combined to realize the complementarity of remote sensing data from different platforms and promote the wider application of Li DAR data in forest resource survey and other fields. |
PDF Full Text Request