For adaptive-optics systems to compensate for atmospheric turbulence e
ffects, the wave-front perturbation must he measured with a wave front
sensor (WFS), and key parameters of the atmosphere and the adaptive-o
ptics system must be known. Two parameters of particular interest incl
ude the Fried coherence length r(o) and the WFS slope measurement erro
r. Statistics-based optimal techniques, such as the minimum variance p
hase reconstructor, have been developed to improve the imaging perform
ance of adaptive-optics systems. However, these statistics-based model
s rely on knowledge of the current state of the key parameters. Neural
networks provide nonlinear solutions to adaptive-optics problems whil
e offering the possibility of adapting to changing seeing conditions.
We address the use of neural networks for three tasks: (1) to reduce t
he WFS slope measurement error, (2) to estimate the Fried coherence le
ngth r(o), and (3) to estimate the variance of the WFS slope measureme
nt error. All of these tasks are accomplished by using only the noisy
WFS measurements as input. Where appropriate, we compare our method wi
th classical statistics-based methods to determine if neural networks
offer true benefits in performance. Although a statistics-based method
is found to perform better than a neural network in reducing WFS slop
e measurement error neural networks perform better in estimating the v
ariance of the WFS slope measurement error, and both methods perform w
ell in estimating r(o). (C) 1995 Optical Society of America