光学元件损伤的检测和修复:传统与深度学习(特邀)

Inspection and Repair of Optical Damage in Tradition and Deep Learning(Invited)

光学元件的损伤在高功率激光系统的终端光学组件中较为普遍且对激光系统的正常运行有重大影响。为提高元件的使用寿命和保证激光光路正常运行,首先要做的是检测和判断出损伤出现的位置、大小、类型。在线检测中终端光学损伤检测装置是一种重要的方法,它能够直接、实时地对元件的损伤情况进行成像并分析,另外还有一种间接获取损伤图像的方式,即用衍射环检测损伤,通过相关公式求出损伤点的大小和位置。针对更小的损伤的检测,深度学习这一工具能够处理大量数据,是目前研究该问题不可或缺的一类方法,它能够减少人工,并提高效率和准确率。修复损伤的主要方式是快速熔融缓解,即二氧化碳激光熔融损伤区,该方法是目前最常见、最有效的修复方式。对损伤问题处理的前提和关键在于精确定位更小的损伤点并分类不同类型的损伤,以便确定后续修复步骤。损伤的检测和修复是光学循环回路策略的重要部分,传统方法有一定的局限性。近些年,受到深度学习在图像处理和目标识别领域的优势的影响,未来会有越来越多深度学习的方法能够被用在与损伤检测相关的研究上。这对高功率激光系统长期稳定运行和正常发展有重要意义和作用。

This paper mainly introduces the inspection and repair methods of optical damage in high power laser system.Because the optical element damage is common in the Final Optical Assembly(FOA)of high power laser system and has great influence on the normal operation of laser system,it is necessary to inspect real time and repair in time,so as to achieve the purpose of recycling optical elements.In online inspection,Final Optical Damage Inspection(FODI)is an important method,which can image and analyze the damage of optical components in real time.In addition,there is an indirect way to obtain damage images,which is to detect the damage by diffraction ring.The size and location of the damage point can be calculated by the relevant formula.For the detection of smaller damage,the tool of deep learning,which can process a large amount of data,is an indispensable method for studying this problem at present.The on-line detection device proposed by them has been a very effective means of detection.With the development of deep learning in image and data processing,convolutional neural networks and decision trees are used to identify and judge the location and size of damage points,so that we can quickly find the damage points.Accurate detection and identification is a premise for the protection and recovery of optical components,then,damage repair needs effective technical means to repair the component and bring it back to the original quality standard as much as possible.The main method of repairing damage is Rapid Ablation Mitigation(RAM),which is the most common and effective method of repairing damage.The premise and key to the damage site treatment is to accurately locate smaller damage points and classify different types of damage so as to determine the subsequent repair steps.Of course,different application scenarios require different technical means.Finally,the structure and flow of optical element damage detection and repair are introduced by the optics recycle loop strategy.This process is very helpful to realize multiple utilization of optical components,save cost and improve utilization rate.Damage detection and repair is an important part of optics recycle loop strategy.Influenced by deep learning in the field of image processing,it is believed that more and more methods of deep learning can be used in researches related to damage detection and repair.In a word,optical component damage detection has developed towards the direction of online detection to improve resolution,and deep learning to help improve classification accuracy and accurate positioning.Damage inspection and repair is an important and indispensable part of the optics recycle loop strategy.