大动态干扰下基于光学图像的自主导航技术

针对高超声速飞行器在平流层内飞行过程中自主天文导航时，气动光学效应及运动模糊等大动态干扰严重影响观星和高精度导航的问题，提出一种基于正则化思想的高超星图半盲复原改进算法。该算法首先借助已有的高超星图模糊核提取算法获得先验模糊核，为复原提供便利条件。接着结合天文图像灰度及梯度的稀疏先验分布特性，设计了一种针对高超星图的正则化盲复原模型，并结合先验模糊核所提供的信息，构建了改进的半盲复原方法，能在先验模糊核精度较低或缺失时转化为盲复原方法，提高了算法的鲁棒性。将本算法与传统星图复原算法、其他最新正则化复原算法进行星图复原与导航效果比较，数值仿真结果显示本算法的复原效果更佳，峰值信噪比（PSNR）达到36.97，并且质心提取成功率达到99.23%，能够明显改善星图识别的成功率，从而大幅提升高超声速飞行器在平流层中的大动态干扰下的自主导航能力。

Autonomous navigation technology based on optical image under large dynamic disturbance

During autonomous navigation in the stratosphere, the hypersonic vehicle will suffer from large dynamic disturbances, such as aero-optical effect and motion blur, which will have great influence on star observation and high-precision navigation. To solve the problem, an improved semi-blind restoration algorithm is proposed based on regularization. Firstly, the priori kernel is extracted by using the existing kernel extraction algorithm for hypersonic star image, so as to provide convenient conditions for image restoration. Then, based on the sparse prior distribution characteristics of gray and gradient of the star image, a regularized blind restoration model for the hypersonic star image is developed. According to the information provided by the prior kernel, an improved semi-blind restoration method is proposed, which can be converted into the blind restoration method when the prior kernel is of low accuracy or is missing, so as to improve the robustness of the algorithm. Numerical simulation results show that compared with the traditional star image restoration algorithm and other latest regularization restoration algorithms, the proposed algorithm has better restoration effect, with PSNR reaching 36.97 and the centroid extraction success rate reaching 99.23%. The success rate of star image recognition can be significantly improved, thus greatly improving the autonomous navigation ability of hypersonic vehicles with large dynamic disturbances in the stratosphere.