大口径精密光学元件表面疵病快速检测方法  

Fast method for large aperture optical elements surface defects detection

作　　者：杨程[1] 卢蓉[1] 范勇[1] 陈念年[1] 柴立群[2] 

机构地区：[1]西南科技大学计算机科学与技术学院,四川绵阳621010 [2]中国工程物理研究院激光聚变研究中心,四川绵阳621900

出　　处：《计算机工程与设计》2012年第6期2476-2480,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(10776028);四川省教育厅重点基金项目(2006C074)

摘　　要：为确保对大口径精密光学元件表面的疵病进行高效和准确检测,针对以往疵病检测系统不能处理单像素宽目标、提取疵病特征参数速度过慢等问题,提出了一种基于顶点链码与离散格林相结合的快速几何特征参数提取算法。算法把针对目标的曲面积分变为曲线积分,提取速度是以往方法的3倍左右;使多种几何特征参数的计算统一在该算法框架下,且能处理单像素宽的非闭合疵病;并在确保其泛化能力的基础上,运用机器学习的方式使疵病的分类识别率达到了90%以上。The effective defects detection of the optical element surface becomes the key guarantee to the quality of the optical system.Aimed at the shortages of the defects detection system,which can＇t handle the target with a single-pixel wide and the speed of the defects geometric feature parameters extraction is too slow,a rapid feature extraction algorithm based on the combination of vertex chain code and discrete green is designed.The method turns the surface integral into line integral and improve the extraction speed up to about three times of the traditional methods,and the calculation of varied parameters is unified under the framework of this algorithm,and the non-closed single-pixel wide defects is handled.Moreover,under its better generalization ability,the recognition rates of defects classification using machine learning methods reachs more than 90%.

关 键 词：疵病检测 顶点链码 离散格林 几何特征参数 机器学习 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]