Study On GPU Accleratson Of Wavefront Sensorless Adaptive Optics Correction

The wavefront sensorless adaptive optics system has good correction capability.The wavefront correction technique based on the stochastic parallel gradient descent(SPGD)algorithm has been applied to the wavefront distortion correction experiment maturely.At the same time,wavefront sensorless adaptive optics system has the advantages of low cost,strong environmental adaptability and convenient constraction.However,the correction speed of the wavefront sensorless adaptive optics system is too slow to meet the requirements of real-time distortion correction.Therefore,it is mainly investigated to improve the convergence speed of SPGD algorithm by Graphic processing unit(GPU)parallel computing to raise the dynamic phase compensation ability of the correction system in this thesis.The main works of the thesis are as follows:The structure of the wavefront sensorless adaptive optics system,wavefront control algorithm,and beam quality evaluation index are introduced.The Zernike polynomial expansion and power spectrum inversion methods are used to simulate the turbulent phase screens with different intensities,and the optical simulation of the beam transmission in the atmosphere is carried out.The SPGD algorithm is simulated to correct the wavefront distortion,and the factors affecting the convergence speed of the algorithm are analyzed.2.The parallel characteristics of the SPGD algorithm are analyzed,and a parallel operation scheme of SPGD algorithm is designed,meanwhile,the change of computational complexity of parallel SPGD algorithm is discussed.The GPU parallel computing is used to accelerate the processing of the deformed mirror voltage update,image filtering and spot centroid calculation in the wavefront correction process,which improves the correction speed of correction system.The comparison of correction simulation is carried out to verify the correction capability and speed advantage of the GPU-accelerated correction system.3.The performance of the GPU-accelerated wavefront correction system was verified by laboratory experiments and external field coherent light experiments.In the indoor correction experiment,the time acceleration ratio can reach 2.5,and the strelh ratio(SR)can come up to 0.8 or more.In the field coherent light correction experiment,the time acceleration ratio achieved 8.6,and the spot energy was more concentrated.Moreover,the beam quality was improved,and the SR reached 0.8 or more.The results show that the SPGD algorithm parallel computing scheme designed in this thesis has lower computational complexity and can be applied to the wavefront sensorless adaptive optics correction system.After GPU accelerated,the correction speed of the wavefront correction system is effectively improved,that is a good correction effect.