High-Resolution Imaging of Closely Space Objects with High Contrast Ratios

Overcoming this problem of detecting a faint source embedded in the noise of another source, commonly referred to as the problem of identifying closely spaced objects (CSOs), requires both high-resolution and high-contrast imaging. The significant technical barrier that must be overcome is the building of a high-fidelity model of the turbulence in the atmosphere. One step in overcoming this barrier is the use of a large aperture telescope equipped with an adaptive optics (AO) system with a temporal response matched to the Greenwood frequency of the site. However, even with such a system, the limited spatial sampling of the wave front by the wave-front-sensor limits the measurement of the high frequencies in the wave front, thus limiting the effectiveness of the AO correction and thus yielding an image with a resolution lower than that of the diffraction limit of the telescope. This residual blur represents an accumulation of uncorrected faint speckle structure in the PSF which can obscure the presence of any objects or debris near the primary satellite.

Removing this residual blur in the AO-restored imagery requires the use of advanced multi-frame blind deconvolution (MFBD) algorithms. The basis of MFBD is to solve for a single static object scene and a set of PSFs that change in a temporal fashion that is consistent with turbulence induced errors in the wave front. Typically, the number of PSFs that must be estimated for MFBD is on the order of several hundred (1-2 seconds of image data) as a low-earth orbiting (LEO) object changes pose significantly enough on longer time scales that the assumption of a static object scene is violated, thus causing MFBD to produce erroneous non-physical results. For the CSO problem at a geostationary orbit the object will remain static on much longer time scales, and this will allow MFBD to use a much larger volume of data (5-10 minutes), and thus obtain a higher resolution image of the primary satellite over the AO compensated image. Further, we will demonstrate the effects of how a high fidelity PSF model can dramatically improve the background in the image via the proper modeling of the uncorrected speckle structure in the PSF. This will lead to an ability to perform high contrast imaging, which allows the use of MFBD restoration on data where brightness between the satellite and companion will cover a dynamic range on the order of 104.

Scaling the MFBD problem with data volume (from several thousand to several million variables) represents a formidable computational and minimization challenge, that we study using a sequence of high frame rate (1 kHz) images acquired with the SHARK-VIS forerunner at the Large Binocular Telescope (8.4m aperture). We demonstrate the importance of leveraging temporal correlations in the turbulence, that is encoded into Fourier spectra of the imagery by the optical system, to obtain high quality starting guesses for the wave fronts used in the large scale MFBD. We then demonstrate how an improvement in the fidelity of a PSF model from MFBD can improve the use of Recurrence Quantification Analysis (RQA) to statistically discriminate between the signal of the faint satellite companion and the speckle noise in the imagery.