在高功率激光系统中，光学元件缺陷的散射、衍射等效应，会造成光能量衰减，影响光学系统的性能；当光功率足够大时，在激光辐射下元件缺陷会发生内爆对元件造成不可逆转的伤害，导致元件报废，甚至会危及整个光学系统。为及时检测和更换存在损伤的光学元件，避免元件内爆对系统造成危害，有必要对系统中的光学元件进行在线监测。然而，在现有的在线损伤检测技术中，多个光学元件的像面会发生重叠，而且在定位损伤元件时，需要采用调焦的方式依次对焦于不同位置处的元件。此外，当光学元件之间间距较近时，上述方式难以准确甄别损伤元件的具体位置。对于光路复杂、光学元件繁多的大型激光系统，这将严重影响在线损伤检测的效率，为解决该问题，本文将单镜头三维成像相机（光场相机）引入到光学元件损伤在线检测与定位技术中。

首先，对于待检光学系统，需要设计光场相机的结构参数，使其景深、深度分辨率、空间分辨率等满足相应光学系统中光学元件损伤在线探测与定位的需求。因此，本文借鉴前人的研究经验，从几何光学成像模型出发，系统总结与推导了光场相机性能参数与结构参数之间的关系，作为设计光场相机的参考指南。另外，为明确微透镜阵列口径较小时光场相机的成像模型，同时为进一步揭示光场相机成像本质，本文从物理光学的角度出发分析并数值仿真了光场相机成像过程。

其次，考虑到光学系统中光路狭长（光路长度一般大约为数米到数十米，通光口径一般大约为几毫米到几百毫米），在采用基于光场相机的在线损伤检测方法时，需要光场相机的景深足够长。目前，在光场相机的设计与研究中，为拓展光场相机景深，多焦距微透镜阵列被引入到光场相机。此外，伴随着光场相机的不断发展，非规则排布微透镜阵列也有可能被引入光场相机，用于优化光场相机性能。多焦距或非规则微透镜阵列对光场相机的标定方法形成了挑战，针对该问题本文提出了基于数学形态学的光场相机自动标定方法。该标定方法无需微透镜阵列参数等先验知识便可实现光场相机的自动化标定，提高了光场相机标定方法的鲁棒性。

最后，为验证基于光场相机的在线损伤检测方法，本文进行了原理验证实验。在实验中，首先搭建了光学系统的模拟光路；然后，根据具体光路设计了光场相机参数，搭建了光场相机实验样机；随后，采集损伤原始图像，实现光场相机复眼图像处理算法，包括损伤特征点提取与匹配、深度测量算法、损伤深度伪彩处理算法等。最终，验证了本论文所提出方法可行性。

本文提出了一种基于光场相机的光学元件损伤在线检测与定位方法，是一种新的在线损伤检测技术。在几何光学模型下系统总结与推导了光场相机性能参数与结构参数的量化关系，用于指导光场相机的参数设计；提出了一种基于数学形态学的光场相机自动化标定方法，提高了光场相机标定的鲁棒性与智能化水平。搭建了光场相机实验样机，实现了光场相机复眼图像处理算法，完成了基于光场相机的光学元件损伤在线检测与定位的原理验证实验，同时也对拓展光场相机的应用领域产生积极的作用。

 

In high power laser, optics damage will result in different degrees of scattering and diffraction, which will lead to performance degradation of optical system. When the power of laser is high enough, the optics will be damaged that will influence the whole optical system. Therefore, optics damage online detection and location is required to reduce the risk to equipment due to an implosion of optics damage, to provide the ability to remove optics before they are damaged beyond repair and to reduce the performance degradation of the laser. However, at present, the images of the optics will be overlapped each other in the online inspection method, and to locate the optics, we must focus the damage in turn. Besides, it's difficult to discriminate which optics the damage occur when two or more optics are in close proximity. For the complex optical system, this will result in performance degradation of efficiency and effectiveness when completing a online inspection. To overcome these problem, in this thesis, the single lens 3D imaging camera (plenoptic camera) is introduced in the optics damage online detection and location method.

Firstly, to different optical systems, the physical parameters of plenoptic camera should be designed properly, so that the field of depth, the depth resolution and the spatial resolution of the plenoptic camera will be applicable to the corresponding optical system. Hence, in this thesis, we summarized and derived the quantitative relationship between the performance parameters and physical parameters of plenoptic camera from the standpoint of geometric optics according to the previous research in the history, which can be a guidebook for plenoptic camera design. What's more, to reveal more truth of plenoptic camera imaging with small diameter microlens, we re-described and simulated the imaging process of the plenoptic camera from the standpoint of physical optics.

Secondly, in general, the optical path is long and narrow in the optical system (the length of the optical path is about several meters to tens of meters and the aperture is about several millimeters to hundreds of millimeters), so the field of depth of plenoptic camera should be large enough when using it in online detection. At present, to extend the field of depth, multi-focus microlens array is applied in plenoptic camera. Besides, with the development of plenoptic camera, irregular arrangement microlens array may also be applied in plenoptic camera. It is a challenge to the available calibration methods. For this problem, an automatic calibration method based on digital morphology is proposed in this thesis. Finally, the plenoptic camera is calibrated without prior knowledge of microlens array parameters, which improves the robust and intelligence of plenoptic camera. 

Thirdly, to verify the on-line inspection method based on plenoptic camera, the principle experiment is done in this thesis. Specifically, we built a simulating optical system firstly, and then design and built the experimental prototype of the plenoptic camera. Subsequently, the raw image processing methods (such as damage extraction and matching algorithm, depth estimation algorithm and pseudo-color depth map) are coded, and the depth map of damage is achieved finally.

All in all, in this thesis, the optics damage on-line detection and location method based on plenoptic camera is proposed. Concretely, the imaging laws of plenoptic camera from the standpoint of physical optics are derived, an automatic calibration method is proposed, and the related algorithms of plenoptic camera application are realized. At the same time, the application scope of plenoptic camera is extended.