Research On Forest Stand Information Extraction From UAV-based Point Cloud

The purpose of the forest resource survey is to find out the current distribution and growth pattern of forest resources in China,and to provide a basis for accurate management of forest resources and sustainable development decision-making,or forest survey.At present,forest surveys are still completed through field surveys,which are time-consuming,laborious,long-cycle and slow to update.With the development of drone photogrammetry technology,by carrying different remote sensing equipment can obtain high-definition air band images,laser point clouds,forest canopy multi-spectral data,etc.,through the simple processing of professional software can extract the required forest survey data,this digital forestry survey has a wide range of advantages over traditional manual surveys,high efficiency and more objective.In this paper,based on drone photogrammetry technology and using a high-precision RKT drone equipped with a high-definition digital camera,we produce orthophoto images(DOM),digital surface models(DSM)and dense point cloud data using the Xishan ten thousand mu ecological park and Taiyuan Forest Park in the core forest area around Taiyuan City as the study area,and use the point cloud data application to extract the required forest stand information from the sample survey and model inversion to achieve the following objectives and conclusions.(1)Based on the visual extraction of the DOM and the field measurements,the main tree species in the study area were Sophora japonica and Pinus armandi,and the total number of trees was 2859,of which the maximum canopy area was 25.682 m~2,the minimum value was 0.729 m~2,and the average value was 9.121 m~2.The visual extraction of the single-tree positioning and canopy area unfolded error analysis between the visual extraction and the sample site,with medium errors of 0.282 and 0.203,indicates that the single-tree information extracted visually has high accuracy.(2)Single-tree segmentation of the study area based on DSM and dense point cloud,and extraction of single-tree canopy area,tree height,single-tree positioning,etc.After statistical analysis,the extraction rate of single tree split was 83.45%,the maximum and minimum values of canopy area were 25.752 m~2,0.549 m~2,average area was 9.254 m~2,total deviation was 662.085 m~2,relative deviation was 4.9%,root mean square error was 2.591m~2,relative root mean square error was 3.7%,R~2 was 0.87363;the maximum value of tree height was 5.53 m,minimum value was 0.84 m,average value was 3.24 m,absolute error was 1.5 m,minimum value was 0.12 cm,relative error was 29.5%,minimum value was0.26%,average relative error was 4.16%,root mean square error was 0.72 m,relative root mean square error was 6.9%,R~2 was 0.70969,indicating that the extracted stand information from point cloud split met the accuracy requirement.(3)Through SPSS software for point cloud segmentation of tree height and canopy area as independent variables,the observed experimental area for diameter at breast height as the dependent variable a yuan,dual diameter at breast height prediction of inversion model,calculation and statistical experimental area maximum were 12.538 cm,15.214 cm diameter at breast height,minimum 3.467 cm,5.082 cm,respectively,the average of 6.649 cm,10.293cm respectively.Above is utilized to extract all the information volume in north China,biomass and carbon model to calculate the average DBH of xishan ten thousand mu ecological garden broad-leaved forest is 10.293 cm,the average tree height of 4.13 m,stand density is 693 trees per hectare,crown density is 54.1%,the volume of 51.55 m~3,biomass is46.94174 Kg/m~3,carbon is 39.66373 Mg,carbon density is 33.79376 Mg/hm~2;The average DBH of taiyuan forest park coniferous forest was 6.649 cm,the average tree height was2.517 m,the stand density was 801 trees per hectare,the canopy density was 29.3%,the stock volume was 13.8 m~3,the biomass was 20.26004 Kg/m~3,the carbon storage was52.17782 Mg,and the carbon density was 21.10924 Mg/hm~2.In summary,the forest stand information extracted through the drone image point cloud can meet the accuracy of forest survey while reducing the intensity of field survey work,which helps to realize the integration of forest resource survey with the internal and external industry.