| Due to the limitation of atmospheric Anisoplanatism,the traditional Adaptive Optical(AO)can only correct the influence of atmospheric turbulence within the isoplanatic patch(angular scale is about 10'')to reach the imaging resolution close to the diffraction limit of telescope,and cannot meet the requirements of high resolution observation of solar activity region(angular scale is usually 1'~2').Multi-Conjugate Adaptive Optics(MCAO)detects three-dimensional wavefront of atmospheric turbulence to obtain wavefront aberrations caused by atmospheric turbulence at different altitudes.The deformable mirror located at the conjugate position of the strong turbulent layer is controlled,and the wavefront distortion is corrected by layers to achieve high resolution imaging with large field of view.The solar MCAO system detects wavefront aberrations on multiple lines of sight simultaneously by using large field-of-view correlating Shack-Hartmann sensors.Compared with the traditional AO system,the calculation of wavefront slope increases dramatically.In addition,the solar AO system works in daytime and has poor seeing conditions.Compared with the night-time AO system,the sampling frequency and system bandwidth of the sensor are higher,and the requirements of calculating wavefront slope in real time is higher.Therefore,the wavefront slope measurement of solar MCAO system should not only ensure the accuracy of slope calculation,but also meet the real-time requirements.Based on the multi-core CPU hardware platform,aiming at the characteristics of large-field-of-view correlating Shack-Hartmann sensor image,the calculation method of wavefront slope is accelerated and optimized on the premise of ensuring accuracy.The main contents and innovations of this paper are as follows:For large field-of-view correlating Shack-Hartmann sensor of solar target,correlation algorithm is used to calculate wavefront slope.The commonly used correlation algorithms are normalized cross-correlation algorithm in spatial domain and phase correlation algorithm in frequency domain.In this paper,two algorithms are optimized and analyzed from two aspects of computation and parallelism.In order to optimize the computational complexity of the normalized cross-correlation algorithm,the computational complexity of the algorithm itself is simplified through the boundary partial correlation,fast Fourier transform and the improved integral graph method proposed in this paper,starting from reducing the computational complexity of the search area,molecules and denominators.In this paper,the inversion operation of fast Fourier transform and the butterfly operation structure are optimized to optimize the computational complexity of phase correlation algorithm.In parallel degree optimization,two algorithms are optimized from threading level,instruction level and data level.For the normalized cross-correlation algorithm,a fine-grained parallel algorithm is proposed in this paper.The image convolution is expanded into matrix vector multiplication to facilitate thread-level parallelism,and the appropriate allocation method is used to balance the load among the branches.Parallel optimization of phase correlation algorithm mainly considers data-level parallelism.In this paper,the algorithms before and after optimization are simulated and analyzed,and their running time,mean,maximum,standard deviation and accuracy are compared.Based on the 4-core CPU,the slope of 2000 large-field-of-view correlating ShackHartmann sensor images is measured.The simulation results show that the average time for the optimized normalized cross-correlation algorithm to calculate the slope of a frame of images is 11.514 milliseconds,which is about 7 times faster than the normalized cross-correlation algorithm without any optimization,and about 1.4 times faster than the phase-correlation algorithm based on fftw3 library.In addition,compared with other slope measurement algorithms,the optimized normalized cross correlation algorithm has less standard deviation,stronger robustness and lower frame loss risk.The simulation results show that the optimized normalized cross correlation algorithm performs better in real-time and accuracy.Otherwise,in the simulation experiment of fixed offset sliding window,the improved normalized cross correlation algorithm can get the correct measurement results.Therefore,the optimized normalized cross-correlation algorithm is transplanted into the indoor solar MCAO experimental system,and the correctness of the algorithm is verified by the sinusoidal perturbation experiment of the extended source target and achieved a measurement frame rate greater than 900 Hz on the indoor test platform based on Intel i7-6950 X.The experimental results show that the optimized normalized cross correlation algorithm can realize the real-time measurement of the wavefront slope of the system.The content of this paper provides technical support for wavefront slope measurement of solar MCAO system and similar application scenarios.It comes up with new ideas and has both research significance and engineering value. |
PDF Full Text Request