Hyperspectral Image Processing Method Of Environmental Satellite Based On Wetland Classification And Comparative Study On Its Classification Effect

Tumen River estuary is located in the junction of China,North Korea,Russia,There are many rivers and lakes in the region,and wetland animals and plants species are extremely rich,the research and conservation value of wetland is huge.Location characteristics lead to difficulties in wetland surveys in North Korea and Russia,therefore,remote sensing is the best way to classify wetlands in the area.Hyperspectral remote sensing images have high spectral resolution and have the potential for fine research on ground objects,but there are also problems with low spatial resolution and data quality.Aimed at the demand of wetland classification for Tumen River estuary area and above questions,the artical uses hyperspectral data of domestic HJ-1A satellite to carry out research,development a feature extraction and wetland classify method apply for this region and domestic hyperspectral data.The main research results are as follows:(1)In view of the low spatial resolution of HJ-1A satellite hyperspectral remote sensing image,the spectral fidelity analysis of two fusion methods of principal component(PC)and Gram-Schmidt Pan Sharping was carried out using Landsat-8 full-color image.The results show that Gram-Schmidt Pan Sharping fusion method is better for spectral fidelity of paddy field,water body,woodland and grassland.The principal component(PC)transformation fusion method is better for the spectral fidelity of bare land.Spectrum fidelity is similar.In the 460?516nm band Gram-Schmidt Pan Sharpening fusion image of 'the spectral fidelity is better,516?952nm band principal component(PC)transform fusion image spectral fidelity performance is better.(2)The characteristic spectra of wetland in HJ-1A hyperspectral image before and after fusion were extracted by using feature spectral extraction method based on spectral standard deviation threshold.In the pre-fusion image,the characteristic spectrum of the water body before the envelope removal is more in the visible and near-infrared ranges,and the characteristic spectrum of the paddy field is less,and only the visible light is distributed,and the marsh has no characteristic spectrum.After the envelope is removed,Only the water still exists characteristic spectrum,all distributed in the near infrared range,and the number is also reduced.In the image after fusion,the characteristic spectrum of the water body before the envelope is more,and the characteristic spectrum of the swamp and paddy field is less,all of which are distributed in the visible light range.The paddy field,the water body and the marsh are no longer characteristic spectrum after the envelope removal.(3)The classification experiments of HJ-1A hyperspectral images were analyzed before and after the fusion.Before and after the envelope was removed,the classification results were analyzed before and after the application of all bands or characteristic bands.It is found that the overall classification accuracy and kappa coefficient decrease obviously after the envelope removal process,because the envelope removal enlarges the band noise,reduces the signal-to-noise ratio,and improves the classification effect of paddy field and swamp.The classification effect of the image after the fusion with Landsat-8 panchromatic image and the classification of the characteristic spectral band of the applied image and the kappa coefficient,the improvement of the classification effect of paddy field and swamp.The classification effect of water.body is not obvious in all cases.Based on the HJ-1A hyperspectral image before the pre-fusion envelope,the classification accuracy of the characteristic spectral band classification of wetland and the classification accuracy of paddy field,water body and swamp are the highest.The above research results can be used to further study the improvement of domestic hyperspectral image quality,improve the accuracy of wetland automatic classification of domestic hyperspectral images,and provide remote reference for wetland remote sensing monitoring and domestic hyperspectral image.