中文摘要：

由于作者在前一段工作中提出的基于工艺匹配的显微图像边缘提取算法（MPTM—MIED）无法适应微小构件的实时自动检测，本文利用BP神经网络技术重新设计实现了MPTM-MIED，并提出了一种新的自动提取显微图像边缘的方法（AMIED）。为了验证该方法的有效性，利用AMIED对4种工艺实现的微小型结构件显微图像的边缘进行了提取，并对线切割工艺零件的尺寸进行了测量。边缘提取的分析结果表明：AMIED提取出的显微图像边缘与MPTM—MIED提取出的基本一致；与常用的边缘检测算法相比，AMIED提取出的显微图像的边缘线形连接程度较好。测量尺寸的分析结果表明：MPTM—MIED和AMIED测量的尺寸基本相同，比Canny法测量得到的结果更接近万能工具显微镜测得的尺寸。由于在测量过程中不再需要手工选取边缘过渡区域，提出的方法极大地提高了检测速度，可用于实时自动测量微小型零件的尺寸。

英文摘要：

As Micro Processing Technology Matching based Micro Image Edge Detection method （MPTM-MIED） developed by previous research can not automatically detect the micro image edges of micro accessories in real time, this paper designs the MPTM-MIED based on BP neural network a- gain. Then, it proposes a novel Automated Micro Image Edge Detection method （AMIED） to extract edges of micro images automatically. To verify the feasibility of the proposed method, the edges of micro images from micro accessories fabricated by four different methods are extracted by AMIED and the sizes of micro accessories by line cutting are measured. Obtained results show that the detected ed- ges by AMIED and MPTM-MIED are almost the same, and the AMIED has better edge-connectivity as compared with some common detection algorithms. Furthermore, the analysis results indicate that the measured sizes by AMIED are almost equal to those measured by MPTM-MIED and they are more close to those measured by the universal tool measuring microscope as compared with that of the Can- ny algorithm. Because the method has no more need of selecting edge transition region, it improves the detection speed and can measure the sizes of micro accessories in real time.