目的 为解决卷烟厂包装机烟盒涂胶检测中，单一的可见光相机或红外相机对于涂胶的有无、位置、面积、均匀性等难以同时检测的问题，文中采用PCA图像融合算法作图像预处理，用于FX-2型包装机视觉检测系统。方法 检测系统需要在涂胶检测处安装红外相机，同时捕获可见光图像和红外图像，然后将可见光图像与红外图像作主成分分析，替换主成分分量后进行图像融合，最后将融合后的图像输出至后端处理系统。结果 实验证明，融合后的图像纹理细节丰富，同时包含了边缘信息与温度信息，对比度高，可检测性强。结论 PCA图像融合算法在涂胶检测的前端处理中非常有效，融合后图像经过后端处理，可以快速检测出包装纸上涂胶的有无、位置、面积以及均匀性，嵌入FX-2型包装机视觉检测系统后，可实时检测出涂胶不合格的包装纸。

The work aims to pre-process the images with PCA image fusion algorithm for the visual detection system of FX-2 packaging machine, in order to solve the problem that it is difficult for a single visible or infrared camera to sim-ultaneously detect the presence, location, area and uniformity of the coating during the glue coating detection of cigarette packers in cigarette factory. The detection system needed to install an infrared camera at the glue coating detection site, and capture visible and infrared images at the same time. Then, a principal components analysis was conducted on the visible and infrared images. The images were fused after the principal components were replaced. Finally, the fused images were outputted to the back-end processing system. The experimental results showed that, the fused image contained rich texture details, edge information and temperature information, and had high contrast and detectability. PCA image fusion algorithm is very effective in the front-end processing of glue coating detection. After the fused image is processed in the back-end, it can quickly detect the presence, location, area and uniformity of glue coating on packaging paper. After the visual inspection system of FX-2 packaging machine is embedded, the unqualified packaging paper can be detected in real time.