Application Of Neural Network Model In Frequency Conversion Shack-Hartmann Wavefront Sensing Technology

As the beam aberration measurement method,wavefront sensing technology has been widely used in astronomical observation,the laser beam clean-up systems and optical detection.Nonlinear optical frequency conversion technology is an effective method to expand the laser wavelength.Combining this method with wavefront sensing technology,the detection range of high-performance wavefront sensors in visible and near-infrared wavelengths can be extended to the direction of infrared wavelengths,which has important application value for improving the wavefront detection performance of infrared laser.However,the phase mapping relationship fitting in the Second harmonic generation(SHG)and the accurate detection of the second harmonics(SH)wavefront seriously limit the development of frequency conversion wavefront sensing technology based on the wave-front mapping relationsDue to the progress of computer technology in recent years,this paper applies the neural network model to frequency conversion wave-front detection for the first time.Taking the optical frequency conversion Shark-hartmann wavefront sensing technology as the research object,we comprehensively used the analysis results of the theoretical model and the strong fitting performance of the neural network model to conduct in-depth research on the complex wave-front mapping relations in SHG and the accurate detection of SH wavefront.The main contents are as follows:Firstly,in order to fitting the complex wave-front mapping relations in the SHG,the theoretical model of SHG with wave-front distortion was derived and the accuracy was verified by simulation and experiments.On the basis of the theoretical model,the wave-front mapping relation between FW and SH are analyzed.For the first time,the linear mapping relation between the FW and the SH on Zernike coefficient is obtained under the negligible walk off effect,and the numerical calculation process of SH wave-front is simplified based on this linear relationSecondly,a SHG wave-front prediction algorithm based on feedforward neural network model is proposed to overcome the slow calculation speed of the numerical model.We take the wave-front Zernike coefficients of the FW and the SH as the target output and input values of the network model respectively,and optimize the network parameters.The numerical results show that the proposed algorithm can accurately fit the phase mapping relationship between the FW and the SH under different degrees of walking-off effect.The prediction time is less than 0.1 second,which is three orders of magnitude smaller than the traditional numerical calculation method.The proposed algorithm provides the possibility for real-time detection of frequency conversion wavefront.In addition,a neural network Shark-hartmann wavefront reconstruction algorithm is proposed for the SH wave-front detection.This algorithm uses the neural network model to fit the nonlinear influence of high order aberrations in the sparse sub-apertures.Based on the numerical model,the basic structure and optimization process are given,and the wavefront reconstruction accuracy of the proposed algorithm is compared with the traditional model method under different sub-aperture sampling numbers.Simulation and experimental results show that the neural network restoration algorithm achieves accurate restoration of the first 65th order Zernike mode aberrations under the number of 6×6 sub-apertures.Compared with the mode method,the error is reduced by nearly 80%.The proposed algorithm breaks through the limitation of the number of sub-apertures on the spatial resolution,and finally realizes the accurate restoration of the SH wave-front with sparse subapertureThis paper has carried out theoretical analysis,numerical simulation and related experimental research around the optical frequency conversion Shark-hartmann wavefront sensing technology.Based on the neural network model,two key problems in frequency conversion wave-front detection are solved:the problem of fitting the wave-front mapping relations between the FW and the SH,and the problem of accurate detection of the SH wavefront.It lays a solid foundation for the development of optical frequency conversion Shark-hartmann wavefront sensing technology.