Wavefront phase difference detection method based on long short-term memory deep network

Abstract

The invention discloses a wavefront phase difference detection method based on a long short-term memory deep network. The wavefront phase difference detection method comprises the steps of inputting characteristic parameters of a certain actual optical system and generating a training data set; generating a focus plane PSF image i1 (x, y) and an out-of-focus PSF image i2 (x, y) from the training data set according to the Fourier optics principle; extracting feature vectors of the focal plane PSF image i1 (x, y) and the out-of-focus PSF image i2 (x, y) as input data; extracting a feature vector of a PSF image sequence collected from an actual optical system; inputting the PSF image sequence into the trained convolutional neural network model according to a time sequence t to determine a wavefront distortion phase so as to obtain a series of distortion phase parameters for training; and generating input data and output data for LSTM deep network training, initializing network parameters, and repeatedly training the LSTM deep network until a loss function of the LSTM deep network is converged.

Description

technical field [0001] The invention relates to the field of network model training, in particular to a wavefront phase difference detection method based on a long short-term memory deep network. Background technique [0002] The emergence of adaptive optics technology enables large-aperture telescopes to greatly overcome the random disturbance of the wavefront caused by the atmosphere, and wavefront phase difference detection technology is one of its key technologies. Phase diversity (PD) uses two cameras on the focal plane and defocus plane of the optical system to collect images of wavefront distortion, and uses digital image processing methods to calculate the wavefront in real time under the action of randomly expanding targets phase difference to reconstruct a high-definition target image. [0003] From the perspective of reference objects, there are three types of detectors with wavefront phase detection capabilities: (1) active lighting facilities inside the measuri...

AI Technical Summary

Problems solved by technology

However, the CNN-based PD technology also has a contradiction between accuracy and calculation amount, and it only recovers the image deviation caused by the wavefront phase distortion under special conditions.

The influence of the time-varying nature of seeing and atmospheric environment parameters on the resulting dynamic wavefront estimation is not considered, resulting in inconsistent results with the instantaneous disturbance of the atmosphere within the observation range of the telescope

Images