Advances In Algorithms For Image Based Wavefront Sensing

Author : Scott W.. Paine
Publisher :
ISBN 13 :
Total Pages : 130 pages
Book Rating : 4.:/5 (112 download)

DOWNLOAD NOW!

Download or read book Expanding the Capture Range of Image-based Wavefront Sensing Problems written by Scott W.. Paine and published by . This book was released on 2019 with total page 130 pages. Available in PDF, EPUB and Kindle. Book excerpt: "Image-based wavefront sensing using phase retrieval is a method to determine wavefront error in a system using images taken with that system. It has been used to determine the error in the Hubble Space Telescope and develop the necessary prescription to correct the aberrations in this telescope. This method is being used by the James Webb Space Telescope (JWST) in order to characterize wavefront error and perform correction both in ground-based testing and on orbit. This thesis considers methods by which to improve the efficacy of phase retrieval and expand its usage for image-based wavefront sensing. Phase retrieval algorithms operate by using point-spread functions (PSFs), which are images of unresolved stars. A numerical model to simulate the aberrated system is developed, and then nonlinear optimization (NLO) adjusts the parameters of this system to minimize an error metric which quantifies the difference between measured PSFs and the simulated PSFs from the numerical model. In our case, the parameters we adjust are related to the wavefront of the system, and we use NLO to recover this wavefront aberration based on the imaged PSFs. This method is reliable when the starting guess for the aberrations is close to the true solution. If this is not the case, then the NLO can fall into a local minimum of the error metric function and fail to recover the true wavefront aberration. When the starting guess does not converge to a good solution, it is considered to be outside of the capture range of the phase retrieval algorithm. This capture range problem is present with piston errors in segmented telescopes, where the error metric value has multiple minima. By examining the error metric function, we show that by imaging with broadband light, we developed a grid search to move out of these local minima and recover the real wavefront error in a system with piston errors on the order of several waves. In order to fall within the capture range for a certain phase retrieval problem, a good starting guess is necessary. Classically, one can use many random starting guesses for parameters in an attempt to land within the capture range of the true solution. Alternatively, we show that one can train a convolutional neural network (CNN) on simulated PSFs to predict Zernike coefficients, which can then be used as starting guesses for phase retrieval. We show that this trained neural network outperforms random starting guesses in simulation for large amounts of wavefront error on a single PSF image. Another telescope, on which phase retrieval will be used, is the Wide-Field Infrared Survey Telescope (WFIRST). Scientists intend to use this telescope to measure the ellipticity of galaxies as a way to measure dark energy. To perform this tasks, the ellipticity of PSFs must be known extremely well. It therefore becomes important to consider the ect of polarization on the simulated PSFs, and determine if it is necessary to simulate polarization aberrations to accurately measure PSF ellipticity. We show that it is possible to add polarization terms separately from the scalar wavefront terms, and examine the effect of these additional polarization terms on the error in the recovered ellipticity of the system PSFs. The detector for WFIRST is not located on the optical axis, and therefore is not located in a plane normal to the exit pupil of the system. This will warp the underlying grid of the wavefront aberration, requiring resampling. If the grid is expressed using Zernike polynomials, then no interpolation becomes necessary, since these polynomials are continuously defined. We developed a method by which the coefficients of the Zernike polynomials can be changed, and compare it to applying the Zernike polynomials to a warped grid."--Pages xiii-xv.