超声速前视共形光学系统图像复原方法研究

	超声速前视共形光学系统图像复原方法研究
	朱瑞飞
学位类型	博士
导师	贾宏光
	2014-07
学位授予单位	中国科学院大学
学位专业	机械制造及其自动化
摘要	导弹作为目前高科技战争的主要武器，正不断地向着高速和精确的方向发展。超声速前视共形光学系统能够有效地改善导弹的空气动力学性能，提高导弹的飞行速度，因而受到国内外研究机构的广泛关注。一方面，由于共形整流罩表面子视场的不对称特性，使得系统随观察视场的变化产生动态像差；另一方面，超声速飞行器在大气层内高速飞行时，整流罩与流场之间会形成严重的挤压和摩擦，使目标图像产生偏移、模糊等气动光学效应。这两方面都严重影响着系统探测、识别和跟踪的能力，成为制约超声速前视共形光学系统应用的关键。针对上述问题，论文从以下四个方面逐层开展研究工作：论文首先分析了超声速前视共形光学系统的图像退化特性，说明了系统退化图像产生的根本原因，分别给出了共形光学系统和超声速湍流流场的图像退化模型，利用类高斯函数模型模拟了不同程度的湍流的点扩散函数，为超声速前视共形光学系统退化图像的复原与校正提供了理论基础。然后，提出了基于邻近算子分裂的自适应正则化图像复原算法。利用邻近算子分裂法将全变分正则化模型分解成两个简单的问题求解，简化了模型求解难度。为进一步提高算法的运算效率，重点研究了正则化参数的取方法，所提出的方法不仅可以自适应地选择合适的正则化参数，而且当正则化参数收敛时，峰值信噪比达到最大值，图像获得最佳复原效果。在以上研究的基础上，对超声速前视共形光学系统退化图像进行了复原与校正。首先利用光线追迹的方法建立了瞄视误差与框架角的关系，对系统瞄视误差进行了校正，设计了瞄视误差测量实验，实验表明校正后系统瞄视误差小于一个像元尺寸（30μm）。然后针对湍流随机性的特点，建立了随机点扩散函数的图像退化模型，提出了利用连续多帧湍流退化图像复原图像的方法。最后对共形光学系统退化图像和湍流退化图像进行了复原实验，实验结果表明：(1)图像复原方法可以用来校正共形光学系统产生的动态像差，为共形光学系统的像差校正提供了一种新的思路；(2)对于湍流退化图像，所提出的算法对噪声有明显的抑制作用，复原效果优于单帧图像的全变分算法，退化图像的帧数一般取小于10帧。最后，为了使复原后的图像对比度更加清晰，提高系统对红外目标识别的能力，首先提出了基于参数化对数图像处理模型的平台直方图均衡增强算法，利用图像评价函数EMEE分析了模型参数的选取方法。然后设计了硬件实验平台，对复原后的红外图像进行了增强实验。实验结果表明：算法能够在基本不丢失图像细节的情况下，增强图像对比度。
其他摘要	As the main weapon of the high-tech war, missile is developing constantly towards high speed and precision. Because supersonic forward-looking conformal optical system can effectively improve the aerodynamic performance and increase the flight speed of the missile, it has become the focus at home and abroad. On the one hand, the dynamic aberrations with the change of field of view are introduced, due to the asymmetry in the respective sub-field of view of the conformal optical system. On the other hand, when the supersonic aircraft flies at a high speed within the atmosphere, it can produce severe squeeze and friction between the dome and flow which may cause aero-optical effects such as displacement and blurring of the target image. Both of the above two aspects seriously affect the ability of detection, identification and tracking of the system, and become a key that restrict the development of supersonic forward-looking conformal optical system. To solve above problems, this paper will be divided to four parts: First, the characteristics of degraded image of the supersonic forward-looking conformal optical system are analyzed. The reason of image degradation is illustrated. The image degradation models of the conformal optical system and supersonic turbulent flow field are given respectively. The point spread function of turbulence is simulated by the similar Gaussian function model. This part of work provides a theoretical basis for image restoration and correction of the supersonic forward-looking conformal optical system. Then, the adaptive regularization method for image restoration based on proximity operator splitting is proposed. To simplify the difficulty of solving the problem, the problem is decomposed into two sub-problems using the theory of proximity operator splitting method. In order to further improve the computation efficiency of the algorithm, the selection method of the regularization parameter is studied. The proposed method can adaptively choose the appropriate regularization parameter. When the regularization parameter is convergent, the peak signal-to-noise ratio of the restoration image reaches maximum and the restored image reaches best restoration effect. On the basis of above research, the degraded image of the supersonic forward-looking conformal optical system is restored and corrected. First of all, the relationship between the boresight error and the gimbal angle is established using the ray tracing method, thus the boresight error of the system is corrected; the measurement experiment of boresight error is designed and the experimental results show that the boresight error is less than the size of one pixel (30μm) after correction. Then, in view of the stochastic characteristics of the turbulence, the image degradation model of the stochastic point spread function is established and the restoration algorithm using multi-frame turbulence degraded image is proposed. Finally, the image restoration experiment is carried out for the degraded image of the conformal optical system and the degraded image of the turbulence. Experimental results show that: the image restoration method can be used to correct the dynamic aberration of the optical system, which provides a new train of thought for the aberration correction of the conformal optical system; the proposed algorithm has obvious inhibitory effect to the noise and the restoration effect is better than the total variation algorithm of single frame image. The frames of the degraded image are usually less than 10. Finally, in order to make the contrast of the restored image more legible, improve the ability of the target identification of the infrared system, the plateau histogram equalization enhancement algorithm based on Parameterized Logarithmic Framework is first proposed, and the selection of the model parameters is studied in detail using the image evaluation function EMEE. Then, the hardware platform of image enhancement experiment is designed, and the restored image has been enhanced. Experimental results show that the algorithm can enhance the image contrast without loss of image details.
语种	中文
文献类型	学位论文
条目标识符	http://ir.ciomp.ac.cn/handle/181722/41514
专题	中科院长春光机所知识产出
推荐引用方式 GB/T 7714	朱瑞飞. 超声速前视共形光学系统图像复原方法研究[D]. 中国科学院大学,2014.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
朱瑞飞.pdf（5552KB）			开放获取	CC BY-NC-SA	请求全文