基于机器视觉的卷烟小盒商标纸表面缺陷在线检测技术

  目的  为检测卷烟小盒商标纸表面多种质量缺陷，提高缺陷检测准确率和检测速度。  方法  以标准图像的定位点通过偏移和相似度量实现快速定位配准，并改进图像差分算法进行实验。  结果  （1）待测图像与标准图像存在一定的偏移，其中最大偏移量为3.6 mm，最大偏移角度为2.1°，最快配准99张图像只需2.484 s。（2）传统差分算法检测图像速度为18.24 s，改进算法检测最快速度为15.62 s，改进后的平均准确率提高了15.23%。  结论  卷烟小盒商标纸表面缺陷在线检测技术速度快、准确率高，减轻了人工检测商标纸的工作量。 

On-line detection technology of label paper surface defects of small cigarette packs based on machine vision

  Objective  This study aims to detect a variety of quality defects on the surface of the label paper of small cigarette packs, and improve the accuracy and speed of defect detection.  Methods  The positioning points of standard images are used to realize rapid positioning and registration through offset and similarity measures and the image difference algorithm is improved.  Results  (1) There is a certain offset between the image to be tested and the standard image, where the maximum offset is 3.6 mm and the maximum offset angle is 2.1°. It only takes 2.484 s to realize registration of 99 images. (2) The detection of the traditional difference algorithm can be finished within 18.24 s, and that of the improved algorithm can be finished within 15.62 s. The average accuracy of the improved algorithm is increased by 15.23%.  Conclusion  The on-line detection technology for the surface defects of the label paper of small cigarette packs is fast and accurate, which reduces the workload of manual label paper detection.