Motion Imaging Error Modeling And High-precision Correction For Pushbroom Hyperspectral Imaging On Satellite Platform

In the imaging process of satellite-based pushbroom hyperspectral imager,nonideal attitude variation and motion errors of satellite platform(defined as motion error of platform) will cause pixel offset on the detector,i.e., motion imaging,leading to degradation of hyperspectral image quality(defined as motion error of image). Exploring the degradation mechanism accurately,compensating and correcting the degraded images is one of the most concerned issues needed to be solved. We carry out research in image degradation and spectra aliasing caused by the motion error of satellite platform. Reasons of degradation of hyperspectral image, compensation and correction methods are be explored to improve the imaging precision of satellite-borne hyper-spectral imager.First,optical structure and convolution factor of each modules of typical dispersive imaging spectrometer is analyzed according to the hyperspectral imaging mechanism. Then established the static and dynamic model of dispersive imaging spectrometer. Relationship between satellite attitudes and pixel offset and pixel mixing process is studied. After that a pixel mixing model of spacial imaging is presented. By the model,mean mixing ratios of each pixel are easy to calculate by a universal coefficient matrix,as long as the satellite attitude parameters of each moment are known. Based on spectral imaging principle of remote imager with slit,we propose an in orbit spectral imaging degradation model. The synthecical effect of different attitudes of vibration are considered in the model. The common expression of mean mixing ration under simple and complex vibration is derived separately,then the simulation degraded spectral image data cube is achieved.A simulation experimental platform with multiple degrees of freedom is designed and set up. According to controlling of motion model, simulation analysis and experimental study on the effects of satellite vibration frequency and amplitude on hyperspectral image are carried out. Results show that simulation is in good agreement with reality. Mean mixing ratio can reflect the degradation extent of hyperspectral data directly. Satellite vibrations bring about spatial quality deteriorate of hyperspectral image,and lead to the aliasing of spectrum comes from different ground object. The degradation extent of hyperspectral data is determined mainly by vibration amplitude. The influence of frequency, expecially high ones, is limited and declined sharply with frequency. Accuracy of the degradation model and reliability of the simulated data are also verified by above studies.Afterwards,research on the high-precision compensation and correction of degraded hyperspectral images is conducted. Various image restoration algorithms are comparative analyzed. On the basis of that,we propose a hyperspectral image correction algorithm based on pushbroom mechanism. Adopting the established degradation model,the correction algorithm calculates point spread function(PSF)by measured parameters of satellite attitudes. Considering the problems of common rely on accuracy prior and sensitivity to sub-pixel calculation involed conventional deconvolution algorithms,and the space-variability of PSF brought by pushbroom mechanism, dimension riasing(resolution expansion), slicing technique and gradual splicing are emploied. Then the problems are significantly improved. Next,hyperspectral data is corrected combining the image deblurring algorithm based on maximum posterior probility(MAP). Degradation and correction experimental results show that spacial and spectral data quality are both improved by our correction algorithm. It can be widely used in the correction of degraded hyperspectral images resulting from different ground object and different attitudes of satellite vibration.At last,blind restoration research on degraded hyperspectral images is discussed against the situation without prior of imaging parameters and satellite attitudes in practice. Considering the limitation of existing blind restoration algorithms and band selection methods as used to correct hyperspectral images,we present a adaptive band selection method based on hyperspectral image characteristics. Bicubic interpolation is applied to estimate the PSF more precisely. Finally,a hyperspectral image blind restoration algorithm based on adaptive band selection is proposed. Restoration experimental results show that the proposed method can compansate and correct the image motion errors introduced by various unknown satellite motion. Image sharpness are improved,and spectral distortion is also reduced.