Large-dynamic-range Shack-Hartmann refraction measurement based on deep learning

The human eye wavefront aberrator based on the Shack-Hartmann wavefront sensor (SHWFS) has become a common device for detecting eye aberrations in modern ophthalmology clinics. In order to eliminate the problem of spot and subaperture matching in traditional methods, we use deep learning method to directly map Hartmann spot pattern and corresponding Zernike coefficient, so as to expand the dynamic range of measurement. The lightweight network realizes to fully extract high dimensional feature information and achieves high precision measurement of diopter and astigmatism. The experimental results show that the proportion of the network falling into the tolerant error range (±0.25D) in diopter and astigmatism measurement reaches 94.2% and 100%. This method can measure the low order aberrations of human eyes effectively without changing the SHWFS setting, and at the same time ensure the accuracy and dynamic range, which has been verified by the real machine.