Abstract: In order to solve the machine vision detection problems of O-type metal seal under limited vision condition in real production, this paper proposes a precise subpixel dimension detection method. Using an improved edge detection algorithm based on fractal interpolation and wavelet transform to detect subpixel image edge; using an improved algorithm based on region growing of gradient and histogram of Euclidean distance to realize dimension detection and roundness verification. The experimental results indicate that compared with traditional method, the calculation speed of the new method proposed in this paper is faster, accuracy of edge detection is higher and dimension detection accuracy can achieve micrometer level.

Key words: machine vision, subpixel, edge detection, O-type dimension detection

摘要： 针对在现实生产中，硬件受限的O型金属密封件机器视觉检测的问题，提出了在亚像素图像条件下的精确尺寸检测方法。利用分形插值和小波变换的改进边缘检测算法，实现亚像素图像的边缘检测；并用结合梯度的区域生长和距离直方图的改进算法，实现O型的尺寸检测和圆度校验。实验结果表明，提出的方法与传统方法相比计算速度快、边缘检测精度高、尺寸测量精度可达微米级。

关键词: 机器视觉, 亚像素, 边缘检测, O型尺寸检测