Research On Extraction Of Single Tree Parameters And Biomass Estimation Based On LiDAR

Forest is the main body of terrestrial ecosystem,which provides the necessary material foundation for the survival and development human beings.Forest resource investigation is an important way to grasp the quantity,quality and distribution of forest resources,and it is also an important basis for the scientific formulation of forest management planning and the adjustment of ecological structure.The acquisition of single tree parameters is the main work of forest resources investigation,which provides important basic data for forest biomass,carbon storage and economic and ecological value estimation.Forest aboveground biomass(AGB)plays a vital role in the dynamic change of forest resources,climate change and carbon sink.Timely,accurate,efficient and low-cost acquisition of single tree parameters and forest AGB has always been a key issue for forestry practitioners and researchers.The traditional forest resources survey mostly uses the method of sampling survey,which requires the investigators to seize feet of all trees in the sample plot,which costs a lot of time,manpower,material and financial resources,and the investigation cycle is long and the efficiency is low.Light Detection and Ranging(LiDAR)has a strong penetrating force,which can obtain detailed information of the three-dimensional spatial structure of the forest and the terrain under the forest,and then accurately extract single tree parameters and estimate the AGB.In this study,the Beijing area was taken as the main research area,and 11 common tree species in the northern area were taken as the research object.Point cloud data of a total of 12 forest sample plots were obtained by using mobile SLAM LiDAR and terrestrial LiDAR under different forest types(coniferous forest,broad-leaved forest and coniferous-broad forest).Based on the detailed research and analysis of the extraction accuracy of single tree parameter factors and the estimation of AGB,a new method of DBH extraction based on point cloud data fitting polygon cylinder was proposed,and the quantitative structure model algorithm(QSM)was optimized for the AGB estimation.The main research conclusions are as follows:(1)The method of extracting DBH from fitting polygon cylinder proposed in this study uses contact and non-contact features to fit polygon cylinder respectively,and reconstructed the volume of the point cloud sections with the thickness of 40 cm and 20 cm at DBH to estimate DBH.The results showed that the DBH values estimated by using non-contact feature fitting polygon cylinder to reconstruct 40 cm thick point cloud data are closer to the measured reference values and have higher accuracy(R~2=0.989;Bias=-1.01cm,r Bias=-3.78%;RMSE=1.51cm,r RMSE=5.68%).(2)The mobile SLAM LiDAR can accurately extract the position of trees in the sample plots under the canopy with weak GNSS signals,and has high investigation efficiency.Compared with the tree position reference value obtained by using total station,the Bias and RMSE of mobile SLAM point cloud data to extract the tree position were 0.21m and 0.23m respectively.Compared with the actual field investigation,the efficiency of single tree parameter factors acquisition by mobile SLAM LiDAR is higher,which is about 17 times of the actual field investigation.The results showed that the mobile SLAM LiDAR can accurately map the location distribution of forest trees,to a certain extent,it can solve the problem of tree positioning when GNSS signal is weak under the canopy,which can meet the actual needs of forest resource investigation and has the potential for further application in forest resource investigation.(3)The terrestrial LiDAR was used to obtain the high-density point cloud data of forest,and then the least square circle fitting method was used to extract DBH from the point cloud slice of 10cm thickness and the tree height was extracted from the normalized point cloud data.The RMSE and r RMSE of extracting DBH and tree height from terrestrial LiDAR data were 1.15cm and 1.30m,6.08%and 9.67%,respectively,which all met the accuracy requirements of forest resource inventory in China.It was found that the bark roughness had a certain impact on the extraction of DBH from point cloud data.As the bark roughness increased,the extraction accuracy of DBH gradually decreased,but the difference was not significantly obvious.In addition,RMSE of DBH were 1.14cm,1.14cm,1.13cm,1.13cm and 1.13cm for point cloud data of 1cm,2cm,5cm,10cm and 20cm thickness by using the least square circle fitting method,indicating that there was no significant difference in the accuracy of DBH extraction of point cloud data of different thickness within a certain thickness range.It was also found that DBH and tree height extracted from LiDAR point cloud data were generally smaller than the measured reference values,but the extraction accuracy did not increase with the increase of DBH and tree height.(4)The quantitative structure model(QSM)algorithm was optimized based on 100 trees of 10species,then the optimized QSM algorithm was used to directly reconstruct the 3D structure model from the single tree point cloud data,further calculate the model volume,and estimate the single tree AGB combined with the basic wood density value of specific species.The results showed that the tree AGB estimated from the terrestrial LiDAR data had a good consistency with the reference value of the model(RMSE=17.40kg,r RMSE=13.63%,CCC=0.97).Among the ten tree species,the accuracy of AGB estimation of Salix mandshurica and Populus tomentosa was the highest(Salix mandshurica:RMSE=16.23 kg,RMSE=7.74%,CCC=0.99;Populus tomentosa:RMSE=11.62 kg,RMSE=6.52%,CCC=0.96),and Pinus tabulaeformis was the worst(RMSE=24.40 kg,RMSE=26.79%,CCC=0.59),indicating that the non-destructive measurement method could accurately estimate the tree AGB.At the same time,the results of this study will be helpful for forestry managers to formulate relevant management measures and more accurate calculation of forest economic and ecological benefits,and provide reference for relevant researchers who use non-destructive measurement methods to estimate forest AGB.