| With the development of hyperspectral infrared remote sensing technology,the detection of the atmosphere is becoming more and more precise,and the detection cycle is becoming shorter and shorter,so that the amount of information is increasing.Whether it is on the satellite or under the satellite,the storage and transmission of the detection information is a problem that must be faced in the process of data application.Therefore,in order to achieve the rapid transmission of hyper spectral atmospheric infrared remote image data,and the storage space occupied by small.At the same time to ensure the accuracy of data assimilation and inversion,radiance thinning is a necessary process for hyperspectral atmospheric infrared remote sensing image data.The process can reduce the amount of data,and ensure the minimum loss of information and data of air quality of restoration.Radiance thinning divided into two cases,one is for lossless data compression,and the other is the spectral channel selection.In this paper,two cases of radiation extraction are studied.Firstly,the atmospheric detection and remote sensing data transmission and storage are analyzed in this paper.Typical hyperspectral infrared remote sensing images detected by AIRS detector are used as experimental data,and their spatial correlation and spectral correlation characteristics are analyzed.The feasibility and necessity of lossless compression and spectral channel selection are qualitatively explained.Secondly,considering the greatly spectral correlation of hyperspectral atmospheric infrared remote sensing image,in order to achieve effective compression effect,this paper uses the ICA transform to remove the spectral redundancy,and the ICs components of the image realizes mutual independence in transform domain.After that,the ICs component and the transform coefficients are quantized and unquantized,and the quantized residuals are retained.The quantized data and the quantized residual are predicted,so as to reduce the amount of data.In the coding section,we choose range coder and improve the probabilistic estimation model by using the Stochastic Learning Weak Estimation(SLWE)method to improve the coding effect.At the same time,before encoding,the encoding data is processed positively,so that it is more suitable for interval encoding process.Finally,the typical AIRS experimental data is compressed,and the compression ratio can reach more than 3.35.And compared with some existing classical compression methods of hyperspectral infrared remote sensing images,the compression method studied in this paper has certain advantages in compression ratio.Finally,taking temperature and humidity inversion as an example,based on the typical brightness temperature data and temperature and humidity inversion profiles in AIRS data,the Jacobi matrix of temperature and humidity is obtained by radiative transfer model(RTTOV)analysis.In order to extract the related spectral information and channel from the large amount of AIRS data,the channel selection process based on the principal component influence coefficient is implemented.Focusing on the temperature Jacobi matrix and the humidity Jacobi matrix,so as to avoid the main influence of noise and other physical parameters of the spectral information,and choose the temperature and humidity influence of large band channel.The temperature inversion and humidity inversion are applied to the detection of brightness temperature data by AIRS based on selected channel.The inversion results are compared with the channel inversion results given by the satellite data numerical weather prediction application research group(NWPSAF).It is obvious that the method studied in this paper has less error in temperature profile and humidity profile obtained by temperature and humidity inversion.Furthermore,a channel selection method based on information capacity iteration is given,which is compared with the PC-AIC algorithm in this paper,it is shown that the PC-AIC algorithm can be used for effective channel selection. |
PDF Full Text Request