Research On Intelligent Wavefront-Error-Detection Technology For Ground-Based Optical Telescope

There are two main reasons for image degradation of large aperture ground-based optical telescopes: wavefront aberration caused by atmospheric motion and systematic error caused by mechanical motion and alignment.At present,in order to obtain clear and stable images,the large aperture telescope system using adaptive optics usually adopts the way of multistage aberration correction to compensate the error.In the structure of the large aperture ground-based optical telescope,a composite axis tracking detection and correction system is used to detect and correct the low-order errors,and an adaptive optical correction system to detect and correct the high-order errors.In this thesis,the imaging characteristics of composite axis tracking images of optical telescopes and images acquired by Shack-Hartmann sensors are studied,and algorithms are designed to solve the problem of low signal-to-noise ratio in the process of wavefront error detection.(1)When a ground-based optical telescope is observed for a long time,the mechanical structure of the telescope will change slightly due to temperature and gravity and other factors,resulting in the defocus of the imaging terminal image.In order to solve the problems of poor accuracy and low efficiency during manual debugging of the focal length of the secondary mirror,this thesis based on the distribution characteristics of the energy of the target of the telescope fine tracking image,An image sequence autonomous focusing algorithm based on spot center energy characteristics is designed to control the secondary mirror of the telescope system.Combined with the improved single-direction second-order neighborhood difference algorithm,a sharpness evaluation function with better accuracy and monotonicity is obtained.The algorithm satisfies the basic characteristics of the focusing sharpness evaluation function and has better accuracy and noise suppression compared with the traditional algorithm.The sensitivity of the sharpness evaluation function is improved by 2.9708 to 3.356,and the operation time of every 100 frame spot image is shortened by 0.0004 to 0.0025 seconds.(2)Aiming at the problem that the extraction accuracy of traditional barycenter method and its improved method is limited when Shack-Hartmann wavefront sensor conducts centroid detection under strong interference background,this thesis proposes a neural network centroid detection method with higher robustness under the condition of low signal-to-noise ratio.Based on the multi-task and multi-classification neural network method,the algorithm can predict the noise statistics of the sub-aperture,so as to obtain better post-processing effect.Moreover,a noise model that is more suitable for real noise data is designed according to the various noises and stray light situations faced by the actual image in the acquisition process.Under the condition of low signal-to-noise ratio,the centroid estimation error of this algorithm is reduced by 0.053 to 0.324 compared with the traditional method and the algorithm using only the multi-classification neural network.(3)Finally,the two algorithms are designed in parallel through the embedded platform of FPGA+DSP,and a wavefront error detection and processing system is built to test the practicability of the algorithm.Through the comparison test,both the sub-lens focusing algorithm and the centroid detection algorithm of the wavefront sensor in this thesis can obtain the credibility and practicability better than the traditional algorithms in the actual optical acquisition process.