Research On Ultraspectral Data Compression Based On Key Information Of Spatial/Spectral/Time Prediction

In recent years,with the advancement of detection technology,the ultraspectral atmospheric infrared data obtained by the ultraspectral detector observing the earth's atmosphere generally contains thousands of spectral channels,and thus has extremely high spectral resolution.In addition,as the detection frequency becomes higher and higher,the time interval between the acquired images becomes smaller and smaller.The huge amount of data generated every day places a great burden on the transmission and storage of information.Therefore,the study of fast and efficient compression algorithm is very necessary.High-resolution ultraspectral atmospheric infrared data has extremely high information redundancy between the sp ectral channels and the time dimension,especially in the spectral dimension,it is not necessary to apply all spectral channels to atmospheric inversion and assimilation calculations,and It will increase the calculation burden and cause ill-posed results.Therefore,before performing atmospheric assimilation or inversion calculations on ultraspectral data,it is necessary to perform radiation thinning on all ultraspectral data to obtain data that can be directly used for assimilation inversion calculations.The key data,which is called key information in this article,is transmitted preferentially in compression.Based on this,this paper designs a lossless compression and decompression algorithm based on the key information space-time prediction that takes into account both the assimilation inversion calculation and the compression effect.First,starting from the imaging characteristics and physical characteristics of ultraspectral data,the data processing flow from the acquisition to the assimilation i nversion application of the data is introduced,and the various data products obtained in this flow are introduced.Based on this,IASI as the experimental data,the L1 C product quantitatively calculates the spatial,spectral,and temporal correlation leve l information of ultraspectral data,which qualitatively illustrates the compressibility of ultraspectral data.Secondly,since only some key information can be used to perform atmospheric parameter inversion and numerical assimilation,a key information extraction scheme that takes into account the compression effect and assimilation inversion application is proposed.The scheme includes two parts: channel selection and spatial sparse.Extract the key information in the spatial dimension,and on this basis,use the method of reconstruction and prediction of the key data to remove the data redundancy of the spatial and spectral dimensions.Interpolate and reconstruct the data in the spatial dimension.The residual and key data are input into the interval encoder to obtain a compressed code stream.The decompressing end can use the opposite processing method to the compression end to restore the image losslessly.In the experiment,the IASI L1 C data was losslessly compressed,and the final average compression ratio reached 2.48.Finally,considering that the ultraspectral data obtained from the METOP series of satellite observations of the polar regions of the earth has great time correlation,in order to improve the compression efficiency,a time-dimensional key information prediction model based on online learning prediction is designed.Then,the key data of the reference time slice and the predicted key data of the current time slice are separately subjected to spatial spectrum reconstruction prediction to obtain the prediction residual.After further decorrelating the residual,the final residual and the key data of the reference time slice are separately divided.The encoding process obtains a compressed code stream.The decompressing end can use the opposite processing method to the compression end to restore the image losslessly.In the experiment,the six data blocks of the IASI L1 C were losslessly compressed and compared with the existing compression schemes for IASI data.The results show that the compr ession ratio designed by this method can be almost the same as the highest compres sion ratio currently available.Up to 2.52.The compression method in this paper focuses on protecting the key subset of ultraspectral data,and reduces the amount of data while taking into account the subsequent inversion and assimilation application process of ultraspectral data.