高超声速气动光学效应退化图像复原方法研究

Aero-optical effect is one of the key factors which affect the accuracy and precision of optical imaging-guided hyper-sonic vehicle. In order to solve the shortage of current research on the correction of hyper-sonic aero-optical effect, we present a method to systematically solve the problem of restoring image degraded by hyper-sonic aero-optical effect. Firstly, basing on the optical imaging environment, we build the numerical model of hyper-sonic aero-optical effect, calculate the wave aberration of the scene’s radiation distorted by hyper-sonic aero-optical effect, get the point spread function of hyper-sonic aero-optical effect. Then, we propose a regularization constraint based on block sparse representation to solve the tough trouble cause by spatially varying point spread function, a low-rank regularization constrains based on non-local structure information to make up for the deficiencies of block regularization, a regularization constrains based on generalized auto-regressive heteroscedasticity model to solve the artifact effect in restoration. Finally, the method adaptively optimizes the parameters of regularization item and minimizes the cost function to solve the restoration problem, obtaining accurately restored images. The research is feasible, and can provide high-quality image data supports for hyper-sonic vehicle’s navigation and guidance.

气动光学效应是影响高速飞行器光学成像导航、制导准度和精度的关键因素之一，课题针对高超声速气动光学效应校正和复原研究现状之不足，提出一套系统地解决高超声速气动光学效应退化图像复原问题的理论和方法。方法首先根据高超声速飞行器载光学系统成像环境，建立高超声速气动光学效应数值模型，求解受气动光学效应干扰的场景辐射在出瞳处的波像差，获得高超声速气动光学效应降质模型的空变点扩展函数；然后提出一种基于分块稀疏表示的正则化约束，解决点扩展函数随空间变化带来的复原难题；提出一种基于聚类子块非局部结构的低秩正则化约束，解决分块稀疏表示正则化时将各子块独立复原带来的不足；提出基于广义自回归异方差模型的全图正则化约束，解决“伪像”效应的问题；最后，自适应优化正则化参数的取值和最小化代价函数的求解，实现高超声速气动光学效应退化图像的高精度复原，为高超声速飞行器精确导航、制导提供良好的清晰图像数据支撑。

气动光学效应是影响高速飞行器光学成像导航、制导的关键因素，项目提出了一套系统地解决高超声速气动光学效应退化图像复原问题的理论和方法。首先根据高超声速飞行器载光学系统成像环境，建立了高超声速气动光学效应数值模型，求解得到受气动光学效应干扰的场景辐射在出瞳处的波像差，获得高超声速气动光学效应降质模型的空变点扩展函数；然后提出一种基于分块稀疏表示的正则化约束方法，解决点扩展函数随空间变化带来的复原难题；提出一种基于聚类子块非局部结构的低秩正则化约束，解决分块稀疏表示正则化时将各子块独立复原带来的不足；提出基于广义自回归异方差模型的全图正则化约束，解决“伪像”效应的问题；最后，自适应优化正则化参数的取值和最小化代价函数的求解，实现高超声速气动光学效应退化图像的高精度复原，为高超声速飞行器精确导航、制导提供良好的清晰图像数据支撑。项目发表学术论文4篇，获授权国家发明专利3项，授权软件著作权2项，培养博士研究生1名，培养硕士研究生7名。