Research On Autofocus Algorithm Of Smart Space Camera

The intelligent small space camera can adaptively optimize the parameters,and can effectively assist the satellite to evade other space assets and monitor space targets autonomously,which is the key to improving the ability of space situational awareness.As one of the main core technologies of camera intelligence,autofocus is the main research trend today.At present,on-plane autofocus mainly uses the image-based autofocus method.There is a problem that the rotating satellite platform uses ground target imaging to detect MTF as the image clarity value and cannot achieve real-time on-orbit focusing.Therefore,this article focuses on this issue,the main research content and results are as follows:1)In-depth study of auto-focusing methods and principles,according to the needs of the intelligent development of space cameras,the use of depth-of-focus method as the auto-focusing method,and analysis of the problems and solutions of the image sharpness evaluation algorithm and the positive focus position search algorithm involved in the method.2)Analyze the influence of noise on the evaluation of image clarity and propose a denoising algorithm by mean value of flat area that uses the characteristics of point-like target distribution.It is verified through simulation experiments that the denoised image of the algorithm is similar to the original noise-free image structure with a SSIM degree of up to 0.9056,and in terms of improving the accuracy of MTF detection,the mean square deviation of the MTF curve from the noise-free image is only 0.0124,which is much higher than the traditional Therefore,the denoising algorithm can improve the accuracy and performance of the MTF50 evaluation curve of the defocused image sequence.3)Aiming at the problem of extracting the point-shaped target window for focusing from the deep sky background with many star points,this paper used the star point segement method,adopted the signal-to-noise ratio threshold and energy concentration threshold as the focus window selction basis.According to the basis,and directly using the size of the target connected domain obtained by the segmentation process as the window size calculation parameters,there is no need to calculate additional parameters,which is simple and effective.4)Aiming at the problem that the imaging of ground target cannot achieve realtime focusing,here we use common point-shaped targets combined with MTF detection technology,innovatively apply MTF50 as the basis for image sharpness,and given the method detection thresholds,which are image signal-to-noise ratio threshold 3.7137 and target energy concentration threshold 0.0179 respectively.Compared with the traditional spatial and frequency domain evaluation functions: the sensitivity of the evaluation curve of this method is greater than 0.1200 under different noise standard deviations,and the ratio does not decrease with the increase of the noise standard deviation which average is 18.0906.The average mean square error of the evaluation curve and noise-free images is 0.5168;while the sensitivity and sharpness of the traditional evaluation algorithm decrease with the increase of the noise standard deviation significantly,the average is 0.0918 and 6.2468,and the average mean square error of the curve is 3.7610.5)Aiming at the problems of the traditional positive focus position search algorithm easy to fall into local extremum,low search efficiency,large back-and-forth movement error,and poor versatility,an adaptive jump direction judgment and an overfocus position search algorithm based on the cumulative average growth rate are proposed,And finally use the minimum step three-point method to find the best focus position.Among them,the adaptive direction judgment method moves in small steps and uses the number of judgments as the basis to adaptively jump the local extreme value and the far focal area,improving the search efficiency in the far focal area.The over-focus detection algorithm based on the cumulative average growth rate improves the problem that the traditional variable-step search algorithm needs to formulate corresponding thresholds according to different projects and improves versatility.Experimental results show that the average focusing accuracy of the algorithm is only 14% of the system's geometric half-focus depth,the average focusing step is 16.2857,and the focusing effect is good.6)Finally,the verification software of the auto-focus algorithm was developed,and a real-shot experimental system was established for experiments to verify the rationality and effectiveness of the auto-focus method proposed in this paper.