目的 提出一套应用于实际工业干扰环境下的印刷品缺陷检测算法，解决在工业环境中由于受灰尘干扰和传送装置振动干扰等因素影响，一些传统的印刷品缺陷检测算法在检测速度和精度等方面已经无法胜任的问题。方法 算法主要包括图像预处理、模板制作、模板匹配、灰尘处理、缺陷提取等5部分，其中图像预处理主要包括图像去噪、图像旋转、图像归边、图像剪裁和图像平移等环节。结果 该算法在实际工业流水线上的运行结果表明，若检测400 mm×400 mm的印刷品，检测速度最快可小于 1 s，检测精度可达到0.1 mm。结论 该算法能够对传统预处理添加图像归边处理，以及对印刷品帧差结果图进行形态学处理，克服了工业环境下灰尘和传送装置振动等的干扰。

The work aims to propose a set of printing defects detection algorithms applied in the actually disturbing industrial environment, so as to solve the problem that some traditional printing defects detection algorithms are incompetent in detecting speed and accuracy, etc. due to such influencing factors as interferences of dust and vibration of transmitter. The algorithms were mainly composed of image preprocessing, template fabrication, template matching, dust treatment and defects extraction. Wherein, image preprocessing mainly included image denoising, image rotation, image edge normalizing, image cutting, image translation and other links. The running results of such algorithms in the actual industrial assembly line showed that, if the 400 mm × 400 mm printing matters were detected, the fastest detection speed could be less than 1 s and the detection accuracy could reach 0.1 mm. In conclusion, the proposed algorithms can add image edge normalizing to traditional preprocessing, carry out morphological processing on the frame difference result diagram of printing matters, and overcome the interferences caused by dust and vibration of transmitter, etc. in the industrial environment.