Neural Network Modal Wavefront Reconstruction for Hartmann–Shack Wavefront Sensors

Abstract

Abstract: : Purpose:In order to describe the wavefront patterns in highly aberrated eyes, wavefront reconstruction methods which can classify abnormal shapes have to be applied. The wavefront data are derived from conventional CCD–based Hartmann–Shack sensors as well from large bandwidth custom designed CMOS wavefront sensors. Methods:An artificial neural network was trained with data from highly aberrated eyes or with simulated data, on order to identify abnormalities and measure highly aberrated wavefronts. Adapted architectures and training mechanisms were selected. Results:Simulation studies were performed which established an appropiate architecture of the neural network for modal wavefront reconstruction. The evaluation of real data with the neural network demonstrated the feasibility of the new method. The neural network is capable of controling an active mirror in a closed–loop adaptive optical system. Conclusions:The study established a new method for evaluating an describing optical aberrations in highly aberrated eyes.

Keywords: optical properties • visual search • image processing