| As an important part of terrestrial ecosystems,forests have the characteristics of large area,wide distribution,and complex composition.Traditional forestry investigations are time-consuming and labor-intensive,and they are destructive to forests and not suitable for large-scale investigations.With the application of remote sensing technology in forestry,the limitations of traditional forest survey methods,such as large consumption and limited survey scope,have been made up,making large-scale,low-cost forest resource surveys possible and meeting the needs of survey accuracy.As the first professional three-dimensional mapping satellite in China,ZY-3 is equipped with a panchromatic camera and a multispectral camera with multi-angle observations.It can acquire stereoscopic images and spectral information of ground features,and provide new research for the extraction of forest parameters.In this study,the management area of the Forestry Bureau of Wangqing,Jilin Province was used as the research area,and the digital surface model of the research area was established using the ZY-3 triple linear-array stereo image.The DEM data was used to be the terrain under forest to establish the research area CHM.Multispectral data was used to extracts variables such as the spectral value,vegetation index,and texture characteristics,and we uses these variable factors to build a model for estimating canopy height and forest aboveground biomass.Due to the small amount of measured data,space-borne lidar ICESat-GLAS data with high accuracy were used to calculate the tree height in each spot of the study area.The ELM is used to establish the relationship model between GLAS estimated forest parameters and ZY-3 CHM and spectral information.Than apply ELM to estimate the regional distribution of canopy height and forest aboveground biomass.The main research contents are as follows:(1)Based on the three-line-array stereo image of ZY-3,the image was performed absolute orientation,relative orientation,and regional network adjustment to generate three types of DSM:front-back view DSM,front-down view DSM,and rear-view DSM,and an adaptive method is used to generate a three-line-array DSM.Accuracy verification with DEM,root mean square error is 3.66m.The three-line-array DSM error is the smallest.(2)The correlation between texture statistics and forest parameters under different window sizes is analyzed,and 11×11 pixel is selected as the best window for calculating textures.Modeling factors such as CHM,texture,and vegetation index are extracted.CHM has a high correlation with tree height and biomass.Among the spectral characteristics,the green band has the highest correlation with tree height and biomass.(3)Estimation of biomass and tree height based on resource No.3 CHM combined with multispectral data.Establish a tree height estimation model based on CHM data and a tree height estimation model based on the combination of CHM and multispectral variables.The RMSE of only the CHM data model is 1.99-2.21m.After adding the vegetation index and texture features to the model The RMSE of the model is 1.47m-1.61m,and the accuracy of the model is significantly improved.Three sets of biomass estimation models were established based on the combination of multispectral variables,CHM,and two types of data.The model combining CHM and multispectral data had the highest accuracy,with RMSE of 13.3 t·hm-2-15.25t-hm-2.(4)Use GLAS data to estimate the forest tree height of a large number of spots in the study area.Establish a relationship model between the GLAS tree height and ZY-3 CHM and multispectral data variables based on the ELM algorithm to estimate the continuously distributed canopy height in the study area,map the regional distribution of tree height and biomass,and analyze the results. |
PDF Full Text Request