基于机器视觉的印刷标签检测系统的改进

Improvement of machine vision based defect detection system for printed labels

针对图像采集的非同步性和传送平台存在的抖动等因素造成采集图像质量降低的问题,提出了一种简单可行的高可靠性机器视觉印刷标签检测系统。通过比较几种边缘检测算法,采用Canny这种相对高可靠性的边缘检测算法制作边缘掩膜,通过在模板图像上加盖边缘掩膜,在差影比较后对差影图像进行形态学去噪来消除轮廓伪影和人眼难以识别的微小缺陷。该方法运用在印刷标签质量检测系统中,有效地降低了印刷标签误检率,并且符合人眼识别特性。

To deal with the problem of decreased quality of the captured image in machine vision based printed label defect detection system, which is caused by non-synchronization of image acquisition and by transmitting platform shaking, a simple and feasible label printing detection system based on machine vision is proposed. Through the comparison of several edge detection algorithms, a relative high reliability of the edge detection algorithm is used to make edge mask, and then the edge mask is covered on the template image. After image subtracting, morphological filtering is performed, and the shadow of contour and tiny defects, which is difficult to identify, is dispelled. This method has been applied to the label printing detection system, and can effectively lower the false detection rate and meet the human visual characteristics.