| The crown profile not only directly reflect the size of trees,but is also the basis for studying the dynamic changes of crown and spatial characteristics of trees.Previous studies on the outline of tree crowns have mainly focused on the single-wood level,using the analysis data of the branches of single trees after felling to simulate crown profile,which is difficult to obtain data,so the amount of data obtained is small and less representative.Terrestrial Laser Scanning(TLS),a ground-based 3D laser scanner,is capable of acquiring complete forest structure information from the ground to crown inside the forest,making it possible to reconstruct the complete crown structure of individual trees.The advent of ground-based LIDAR provides a new approach to the study of crowns structure within forests in detail.In this paper,the point cloud data of 6 artificial Pinus koraiensis plantation and 30 artificial Korean pine different grades of analytic timber in Mengjiagang Forest Farm in Jiamusi City were used as data sources,from which the easily measurable factors(tree height,diameter at breast height,the high to live crown base,crown wide)were extracted,the crown radius at different crown heights was automatically extracted by using the point cloud layered projection method.Then the accuracy of extracting forest parameters and crown radius at different heights using point cloud data was analyzed using every arbor dimension data of six artificial Pinus koraiensis population and branch analysis data of 30 artificial Pinus koraiensis analysis trees as reference data.Finally,based on the crown radius extracted using the sample point cloud data,a nonlinear mixed model technique was used to establish a nonlinear mixed model of the maximum crown outline shape of a single level tree crown based on the sample plots effect and the sample tree effect,and simulate crown profile of different trees.Results show that:(1)Under the sample conditions,the overall extraction accuracy of the easy-to-measure factors was high.The average extraction accuracy of diameter at breast height(DBH)was96.80%,tree height(H)was 96.20%,the height to live crown base(HCB)was 87.00%,and crown width(CW)was 86.56%.Overall,the accuracy of extracting single tree factors using ground-based Li DAR point cloud data meets the needs of forestry survey,which proves that ground-based Li DAR can be used to assist in forest stand survey.(2)In the point cloud projection stratification method,0.6m is the best layered spacing,and the overall extraction accuracy of crown radius is 89.23%using the measured branch factors of30 artificial Pinus koraiensis analysis trees as reference.The extraction accuracy of radius in different relative crown depth ranges differed,and the best relative crown depth range was 0.15-1,and the extraction accuracy of radius in this range was stable at about 90%;the less effective range was 0-0.15,and the extraction accuracy in this range was stable at about 80%,while the extraction accuracy in the range of 0-0.05 was the worst at 70%.(3)The 3-parameter Weibull function was the optimal base model.By reparametrizing the optimal base model,the three stand ease of measurement factors of diameter at breast height(DBH),height-to-diameter ratio(HD)and crown width(CW)were finally introduced into the model.The applicability of the model was enhanced after the reparameterization,and the goodness of fit was significantly improved,with R_a~2of the model increasing from 0.7637 to0.8146 and RMSE decreasing from 0.4412m to 0.3902m.(4)The mixed model fits significantly better than the base model,and the mixed model based on the sample tree effect has a better fit superiority.By comparing different variance structures and autocorrelation matrices,it is found that the generalized positive definite matrix and ARMA(1,1)possess the optimal fit,so the generalized positive definite matrix and ARMA(1,1)are finally used as the assumed form of the optimal matrix of the mixed model.After adding the random effects based on sample tree,the R_a~2of the model improves from 0.7707to 0.8738 and the RMSE decreases from 0.3873m to 0.3086m. |
PDF Full Text Request