Extraction Of Underforest Topography And Individual Tree Parameters Of Southern Plantation By Oblique Photogrammetry

The accurate remote sensing extraction of plantation resources information is of great significance for the scientific administration and management of plantation resources.It is difficult to obtain large-scale and high-precision plantation terrain,individual tree height,crown and biomass parameters by traditional measurement methods.In recent years,the oblique photograunetry technology has overcome the shortcomings of traditional measurement techniques and has become the new method of acquiring three-dimensional geographic information.In order to explore the specific methods and application methods of extracting artificial forest structure information by using UAV,we selected GaoFeng Forest Farm which located in Guangxi province of china as the research area,and 15 sample plots including Eucalyptus,Chinese fir and native broad-leaved trees were set up and ground investigation was conducted.Firstly,we acquired the UAV and Lidar data for the study area.And then the SFM-PMVS method was used to restore the 3D point cloud of the survey area acquired by the UAV oblique image.The cloth simulation filter,the progressive TIN encryption filter and the progressive morphology filter algorithm were used to extract the ground points in different canopy artificial forests,and different inter:polation methods were adopted to generate a complete DTM.Lidar DTM were used to verify data accuracy,and it was also possible to normalize the entire range of UAV oblique photography point clouds and DSM.The individual tree maxima was extracted by a local maximum method combining different smoothed windows and fixed windows,and the CHM value at the vertex of the tree was extracted as the tree height.Based on the accurate extraction of the apex of the single tree,the marker-controlled watershed segmentation algorithm and the seeded region growing algorithm were used to segment the single-wood canopy.Power function model fitting by extracted individual tree height and its corresponding above-ground biomass can be applied to estimate regional biomass and map.All individual tree parameter extraction results were evaluated by measured values.The results showed that:(1)There is a negative correlation between canopy density and extraction accuracy of undergrowth terrain obtained by oblique photography.In the canopy closure of 0.5 of plantation stands,we used cloth simulation filtering combined with TIN interpolation,the RMSE of the oblique photography DTM is 0.70 m;In the canopy closure 0.6,0.7 plantation forests,we used progressive TIN encryption filtering combined with TIN interpolation,and the RMSE of the oblique photography DTM was 0.99 m and 1.18 m,respectively;For the plantations stand with a canopy density of 0.7 or more it is impossible to extract the undergrowth terrain by oblique photogrammetry technology.(2)Using the local maximum method,the overall recognition rate of individual tree in the southern plantation stands was 79%,the accuracy rate was 89%,and the F measure was 84%.The best window of different stand can obtain the best single tree recognition effect.For tree height extraction,the RMSE of individual tree height of mature,middle-aged and young Eucalyptus forests were 0.63 m,0.30 m and 0.43 m,with relative accuracy of 98.2%,97.2%and 98.1%,respectively.RMSE was 0.65 m and relative precision was 96.5%.The average RMSE of height extraction from native broad-leaved trees was 0.37 m,and the relative accuracy was 89.0%.(3)For crown area extraction,the study showed that the seed point region growing algorithm is more accurate for Eucalyptus and native broad-leaved tree species,while the marker-controlled watershed segmentation algorithm is more suitable for Chinese fir stand.The crown area RMSE of mature,middle-aged and young Eucalyptus plantations were 4.17m2,2.87 m2 and 4.95 m2,with relative accuracy of 58.2%,72.6%and 65.6%,respectively.Chinese fir uses marker-controlled watershed segmentation algorithn to extract crown area RMSE of 3.39 m2 with relative accuracy of 76.0%.(4)For biomass estimation,RMSE for mature,middle-aged and young Eucalyptus forests was 25.3 kg,5.2 kg and 6.3 kg respectively,with determinant coefficients of 0.55,0.85 and 0.84,respectively.The estimated individual biomass RMSE of Chinese fir was 0.33 kg and the individual tree biomass model R2 was 0.82.The position,height,crown area and biomass parameters of individual trees in Southern plantation can be extracted effectively by inclined photogrammetry combined with lidar DTM.