Piston sensing for sparse aperture systems with broadband extended objects via a single convolutional neural network

It is crucial for sparse aperture systems to preserve imaging quality, which can be addressed when fast corrections of pistons within a fraction of a wavelength are available. In this paper, we demonstrate that only a single deep convolutional neural network is sufficient to extract pistons from wide-band extended images once being appropriately trained. To eliminate the object characters, the feature vector is calculated as the input by a pair of focused and defocused images. This method possesses the capability of fine phasing with high sensing accuracy, and a large-scale capture range without the use of combined wavelengths. Simple and fast, the proposed technique might find wide applications in phasing telescope arrays or segmented mirrors.