Research On Image Degradation Mechanism And Image Restoration Algorithm For Aero-optical Effect

When aircraft flying at high speed,the images or videos obtained from the observation window are usually blur or quivering because of the uneven density and reflective index of atmosphere,which is called the aero optical effect(AOE).AOE greatly increases the difficulty of subsequent tasks such as target observation and recognition.In this paper,reconfigurable data dimension reduction algorithms are applied to analyze the image degradation mechanism caused by AOE,and the image degradation model under certain conditions is predicted.In the end,the influence of degradation model prediction on image restoration is studied.The main works of this paper are as follows:1)Aiming at the image degradation caused by AOE,a simplified physical model is proposed to describe the light propagation and imaging process according to the objective reality.The two-dimensional hypersonic shear layer was selected as the research object,and the convective Mach number(Mc number)was selected as the single variable.The optical path difference(OPD)was used to characterize the optical wave distortion.The simplification of the model is useful for analyzing the relationship between OPD and Mc number.For the two prediction algorithms,the averages of the similarity R between the reconstruction result and the truth are 83.2% and 89.6% respectively,which are both at a high level and indicate that the prediction results is credible.At the same time,the reconstruction accuracy based on VAE is better than CDL.2)For the simplified AOE model,couple dictionary learning(CDL)and variational auto-encoder(VAE)are used separately for reconfigurable dimensionality reduction of OPD.The OPD is mapped from the original high dimensional space to a low dimensional space.In low-dimensional space,the regularity of OPD with different time and different Mc numbers is more intuitively discovered.Then the prediction of OPD in low-dimensional space is realized.Finally,corresponding reconstruction algorithms are used to map the prediction results to the original high dimensional space,which means the prediction of OPD is realized.3)According to the OPD prediction result,point spread function(PSF)of image degradation can be inferred.Different image restoration algorithms are applied to compare image restoration results with and without OPD prediction.By comparing several evaluation indicators,when the OPD prediction result is known,the image restoration result is better than the prediction result when the OPD is unknown,which reflects the significance of the work of OPD prediction.