Forest Parameters Extraction Based On BRDF Corrected Hyperspectral Remote Sensing Data

As one of the most important ecosystems in terrestrial ecosystems,forest ecosystem play a vital role in the global climate and environment.How to accurately understand and understand forest ecosystems from all aspects and scales has become one of the global concerns.In recent decades,the research on forest ecosystems has gradually shifted from small-scale artificial ground surveys to remote sensing and ground surveys.Dynamic monitoring and quantitative analysis of various physical and chemical characteristics of large-scale forest vegetation from spatial and temporal scales.The spectral resolution of hyperspectral remote sensing data can reach nanometer level,and it can fully extract the narrow-band spectral information reflecting the structure and physiological condition of forest canopy.It has great potential in extracting forest parameters by remote sensing.However,due to the anisotropic scattering characteristics of the forest canopy,the hyperspectral data obtained by remote sensing is affected by other imaging environments and solar-observation geometry,and the obtained forest canopy spectral reflectance exhibits a Bidirectional Reflectance Distribution Function(BRDF)phenomenon.The varying spectral reflectance will directly affect the subsequent remote sensing extraction of forest parameters and other quantitative research accuracy.At the same time,the hyperspectral remote sensing data with different imaging mechanisms and observation platforms have their own data characteristics,and the factors affecting their spectral reflectance are not the same.Therefore,how to effectively improve the characteristics of hyperspectral remote sensing data of different imaging mechanisms and observation platforms,and improve the BRDF correction method to accurately estimate the forest canopy spectral information,become an important premise for estimating and extracting forest parameters by remote sensing.Based on the above background,this paper mainly carries out the following aspects:(1)Time-stratified BRDF modeling method is adopted for long-term non-imaging hyperspectral remote sensing data of forest canopy with different forest ages.This method can weaken the sky condition and canopy physiological factors in data acquisition.The effect of spectral information can correct the BRDF effect of the data(DF49:R~2=0.73,RMSE=0.0065;HDF88:R~2=0.79,RMSE=0.0048)effectively.By combining the flux data and Li DAR point cloud data,it was found that the complex vertical structure of canopy would lead to higher light use efficiency(R~2=0.87,RMSE=0.0026).This conclusion is helpful to scale up the canopy light use efficiency research to satellite level.(2)Kernel dictionary is constructed,and the suitable combinations of volume scattering kernel and geometric optical kernel for different types/structures of ground objects in high resolution airborne hyperspectral images are analyzed and screened.It is found that for high spatial resolution airborne hyperspectral images,the information contained in the pixels is usually the local branches and leaves of a single tree.The geometric shape and internal structure of the pixels belong to the dense distribution,which makes the pixels belong to the expression of thick nuclei in both geometric scattering and volume scattering(Ross-Thick and Li-Dense combination).The BRDF correction image based on this combination of kernels is used for forest Leaf Area Index(LAI)modeling and inversion(R~2=0.93,RMSE=0.29),and its accuracy is much better than that of the image without BRDF correction(R~2=0.23,RMSE=2.02).(3)Considering the BRDF effect caused by wide-field-of-view observation and topographic factors,the kernel function expression suitable for slope pixels is analyzed and improved from the mechanism,and an algorithm framework(RT-BRDF)for BRDF correction of airborne hyperspectral images on undulating terrain is proposed.The functional richness index extracted from RT-BRDF corrected airborne hyperspectral images has a positive correlation with biodiversity index(R~2=0.60),which can reflect the spatial distribution of biodiversity at remote sensing level.In general,hyperspectral remote sensing data have abundant spectral information and have certain advantages and potential in reflecting forest canopy structure and physiological parameters.However,influenced by BRDF effect,the use value of spectral information is limited,especially in complex forested areas.The semi-empirical linear kernel-driven BRDF model used in this study can effectively act on hyperspectral remote sensing data of multiple scales and imaging forms.The model demonstrates its parameter expansion and reliable correction ability in the correction of airborne hyperspectral remote sensing data on flat and rugged terrain.At the same time,when extracting forest parameters from hyperspectral remote sensing data and quantitative inversion of remote sensing data,it is necessary to carry out BRDF correction processing,so as to reduce spectral variability,increase the consistency and comparability of data,and improve the accuracy of extracting forest parameters.