CCSS: Small: Neural Mapping for Wavefront Estimation: Uniqueness, Geometry, Applications

About This Grant

Advanced long-range imaging systems are critical for defense missions involving tracking moving targets and identifying adversarial objects, and astronomical imaging applications for ground-to-sky observations. Today’s long-range imaging systems are often equipped with adaptive optics to compensate for the distortions caused by the atmosphere. At the core of adaptive optics is the wavefront estimation problem where one needs to recover the phase information from the measurements. However, existing hardware solutions are limited by the resolution and hence they cannot retrieve high-order aberrations, whereas computational techniques generally require multiple measurements to ensure uniqueness. These, in turn, put many restrictions on the problem where it is generally difficult to estimate wavefronts involving moving targets. This research aims to advance wavefront estimation techniques with the goal to ultimately allow wavefront estimation from a single image. Achieving this technical goal will transform today’s long-range imaging systems, hence supporting new defense and space applications. The educational activities of the project stress on workforce development by training scientists and engineers for critical missions in national security, space exploration, and scientific imaging. The technical approach taken by this project is to develop an optics-algorithm co-design framework through neural mappings. By simultaneously seeking the optimal geometry of the aperture and recovering the phase using a new set of neural representations, the project positions itself as a potential solution for ultrafast wavefront estimation. The research consists of three thrusts: (1) theoretical foundations which study the optimality, symmetry breaking, and neural representations; (2) dynamic imaging conditions which involve the development of new models, recoverability analysis, and the decomposition of motion trajectory and phase aberrations; (3) computational and optical applications involving optical neural networks and phase-based image deconvolution algorithms. On the education front, the project supports new course development in camera physics, optics, and machine learning. A summer outreach program on machine learning and computational imaging is planned to serve students in grades 9-12. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.