中文摘要：

精密表面疵病的自动检测长期以来一直是期待解决的问题。依据显微散射光成像系统对元件表面上的划痕、麻点、气泡等疵病形成暗背景下的亮疵病图像，提出了一个完整的数字化标准的精密表面疵病检测评估体系。首先利用比值匹配原理对冽两方向平移得到的子孔径扫描图像进行拼接。引入数学形态学实现疵病识别。运用膨胀、腐蚀、细化等算法，对二位化图像进行噪声消除，图像分割和特征提取。项目的研究成果已能分辨微米量级的疵病，为建立国内外疵病的数字化标准评价系统提供有力的理论及实验依据。

英文摘要：

The automatic inspection of surface imperfections of precise elements is a durable problem expected to be solved. Based on the defects image are light in dark background in scattering imaging system, a complete digital evaluation system of surface defects, is presented in this paper. Using the xy-scanning system to detect sub-aperture, digital image could be stitched based on isometric serial images. Mathematical morphology is introduced to defect image recognition. How to implement binary image eliminating noise, segmentation and feature extraction by using olilation erosion and thinning algorithm was studied. In this system, the lateral resolution is approximately 1μm. The results provide theoretical and practical evidence to establish the precise surface imperfections evaluation system.