Analysis And Classification Of Several Typical Vegetation Hyperspectral Characteristics In Xishan Forest Park, Kunming

The identification and classification of vegetation is the basis and basis for the management of vegetation resources.How to efficiently and accurately identify vegetation types is an urgent problem to be solved by forestry remote sensing.Different vegetations have extremely subtle differences in spectral characteristics.Using hyperspectral remote sensing to detect and analyze these differences between vegetations can provide prior knowledge for vegetation remote sensing classification and the possibility of distinguishing vegetation with similar spectral characteristics.On the other hand,it can provide theoretical basis and technical support for monitoring vegetation growth and regional application of vegetation.In this study,the study area is located in Kunming Xishan Forest Park.The SOC710 VP ground object spectrometer is used to collect 11 typical vegetation ground hyperspectral images in central Yunnan,and the UAV equipped with S185 format airborne high-speed imaging spectrometer to collect 9 typical vegetation hyperspectral images.The reflectance of typical vegetation is extracted from the measured hyperspectral images,and the ground hyperspectral reflectance curve and airborne hyperspectral reflectance curve of the typical vegetation are respectively subjected to first-order differentiation,reciprocal logarithm of the spectrum,and continuum removal transformation.Fully understand and analyze vegetation reflectance spectrum and three transformation curves,use vegetation index,trilateral parameters and peak and valley characteristics to quantify the spectral characteristics of vegetation,and extract vegetation based on spectral characteristics analysis and Mahalanobis distance statistical analysis methods to distinguish band.By unmixing the mixed pixels of GF-5hyperspectral images,extracting pure end elements,based on the characteristic bands extracted from ground vegetation and airborne vegetation,the vegetation classification and recognition in the study area is realized based on the spectral angle mapping method.Discuss the classification accuracy of the classified results.The main findings are as follows:(1)Based on vegetation spectral characteristics and transformation analysis,it is found that although vegetation has similar spectral characteristic curves,the spectra of different types of vegetation have certain differences in specific bands.The spectral differences between different vegetations can be magnified through related mathematical transformations.Different individuals of the same type of vegetation The performance of the spectral characteristic curve on the ground object spectrometer image and the airborne hyperspectral image is not exactly the same.(2)Aiming at the extraction of characteristic bands of vegetation spectral data measured by ground object spectrometer and airborne hyperspectrometer,the research shows that the range of spectral characteristic bands increased by the reciprocal logarithm and continuum removal transformation is the largest,which is the most ideal spectral data processing method.(3)According to the research on the spectral feature band classification of 11 planting quilts measured by the ground feature spectrometer,the research shows that the overall classification accuracy and Kappa coefficient of the original spectral feature band are the lowest,and the classification effect is average.The classification results of the three transformed spectral characteristic bands are highly consistent with the field vegetation,and the classification effect is good.(4)Aiming at the classification of the spectral characteristic bands of the vegetation measured by the airborne hyperspectrometer,research shows that the overall classification accuracy and Kappa coefficient of the first-order differential characteristic band are the smallest,and the classification effect is average.The original spectrum,reciprocal logarithm,and continuum removal transformation feature band classification is highly consistent with field vegetation and the effect is good.To sum up,this article is oriented to the need for refined monitoring of forest vegetation in Kunming Xishan with hyperspectral remote sensing.Based on multiplatform hyperspectral data of the sky and ground,the spectral angle mapping method is used to achieve accurate identification of different vegetation types.It is applied to forestry resource management to provide certain ideas and basis.