Research On Registration Of Raw Image Sequence Of LASIS

Large Aperture Static Interference Imaging Spectrometer (LASIS) is a type ofpushbroom Fourier transform imaging spectrometer. The distribution of spectrum of thetarget can be reconstructed from its interferogram. The accuracy of distribution isdetermined by the accuracy of the interferogram of the target. The interferogram isacquired from the raw images of LASIS. A whole interferogram of the target can beacquired by a whole field of view pushbroom. Because of pushbroom flat attitude jitter,the wrong raw images was acquired. So the final data cube was distorted in both spacedomain and frequency domain.There are two methods for solving above problems. The first method is to find apushbroom flat with high attitude stability. The second method is to correct the rawimages sequences base on the flat attitude.For airborne remote sensing, no suitable pushbroom flat can be found for LASIS. Thesecond method yet can not do the job. So the only way is to employ both of methods.In this article, analyzed the difference of registration on LASIS image and theregistration on ordinary remote sensing image, and analyzed the impact on LASISimaging caused by change of flight attitude; Basic methods of image registration areresearched, and the appropriation of Fourier-mellin algorithm is researched; A nomailizedcorrelation coefficient algorithm for performing image-to-image registration has beenderived. The images sequence error are restricted to translations. The computersimulation presented here show that the error of new algorithm is less than 0.1 pixel,while translation is less than 3 pixels.