根据散射光成像原理,采用大小两个视场来获取不同精度的暗背景下的亮疵病图像,设计了完整的数字化表面疵病检测系统。该系统采用多区域自适应阈值分割算法对图像进行分割,然后采用基于等价归并标记方法快速提取疵病的特征参数,最后利用BP神经网络对疵病进行分类。实验结果表明该方法既满足实时性需求,又取得了较好的分类检测效果。

Based on the light defect images against the dark background in a scattering imaging system,a digital detection system of surface defects for large aperture optical elements has been presented.In the system,the image is segmented by a multi-area self-adaptive threshold segmentation method,then a pixel labeling method based on replacing arrays is adopted to extract defect features quickly,and at last the defects are classified through back-propagation neural networks.Experiment results show that the system can achieve real-time detection and classification.