Abstract: Aimed at the problem of edge detection high precision of micro parts in microscopic field, a real time detection system for micro parts is designed, the system overall design is given and real time image transmission and processing are completed. A sub-pixel edge detection algorithm for micro parts is proposed. Non-orthogonal quadratic B-spline wavelet transform algorithm is used in the pixel level edge detection. The moment invariants of Zernike moment is used for the edge detection of the pixel level in order to get the precise positioning of sub-pixel level edge, algorithm principle is given and the edge detection results of pixel level and sub pixel level are analyzed. Results show that parts size can reach 0.01~10 mm, the detection precision can reach 0.01%~0.1%. The system can accurately identify the edge of the micro parts, improve the detection accuracy to the subpixel level, and can meet the needs of micro parts detection in microscopic view.

Key words: micro parts, edge detection, sub-pixel edge, wavelet transform, Zernike moment

摘要： 针对显微视场下微小型零件边缘检测精度要求高的问题，设计了一套微小型零件实时检测系统，给出了系统总体设计，完成了图像实时传输和处理；提出一种微小型零件亚像素级边缘检测算法：采用非正交二次B样条小波变换得到微小型零件的像素级边缘，利用Zernike矩算法的矩不变性对像素级边缘进行亚像素级精确定位，给出了算法原理，分析了像素级和亚像素级的边缘检测结果。实验数据表明：该系统检测零件尺寸可以达到0.01~10 mm，检测精度可以达到0.01%~0.1%，可准确识别出微小型零件的边缘，将检测精度提高到亚像素级，能够满足显微视场下微小型零件检测需要。

关键词: 微小型零件, 边缘检测, 亚像素级边缘, 小波变换, Zernike矩