Abstract: This paper proposes an online real-time image recognition algorithm for pipeline microelectronic dispensing  package location. The color characteristics of the package location circle are analyzed. The R component is extracted to do image gray. Iterative method is used to do threshold segmentation of gray image in order to remove the complex background and obtain binary image. Median filter is used to smooth the binary image to eliminate the noise after segmentation. Sobel operator is used to obtain edge information. The center of circle which is to be identified in the edge information is found out through Hough transform. Then package location is identified. The experimental results show that, machine vision technology can position the microelectronic packaging location quickly and accurately, with a good theoretical and practical value.

Key words: machine vision, microelectronic packaging, image processing, Hough transform

摘要： 提出一种针对流水线上微电子点胶封装位置的在线实时图像识别算法。通过对封装位置所在圆的颜色特征分析，提取R分量进行图像灰度化，采用迭代法对灰度图像进行阈值分割去除复杂背景得到二值图像，利用中值滤波法平滑二值图像消除分割后的噪声，运用Sobel算子提取边缘获得边缘信息，利用Hough变换找出边缘信息中待识别圆的中心位置，完成对封装位置的检测。实验结果表明，利用机器视觉技术可以快速、准确地对微电子封装位置进行定位，具有很好的理论和实际应用价值。

关键词: 机器视觉, 微电子封装, 图像处理, Hough变换