Spectrum reconstruction using relative-deviation-based kernel regression in temporally and spatially modulated Fourier transform imaging spectrometer.

During the temporally and spatially modulated Fourier transform imaging spectrometer push-broom scanning process, the motion state of the spectrometer platform can vary. Thus, the target interferogram obtained from the image sequence deviates from the ideal interferogram obtained using high platform stability. The recovered target spectrum will not reflect the true target characteristics. We adopted target tracking to acquire the target position in the image sequence via a proposed kernel regression, with a relative deviation method for determining the target intensities, and the recovery of the spectrogram using the nonuniform fast Fourier transform algorithm. We tested our algorithm on simulated and experimentally obtained aerial images and, from comparison with accurate spectrograms, demonstrate the effectiveness of the proposed method.