【作者】 官燕燕；

【导师】 刘昕；

【作者基本信息】 西安理工大学 ， 制浆造纸工程， 2007， 硕士

【摘要】 凹印正向高速化、自动化方向发展,而目前的印刷质量检测技术还不能满足这种发展需求;尽管国内外的科技工作者对这方面进行了大量的研究,但是检测效果还不能达到满意的要求,经常出现漏检、误检现象。为了进一步提高检测效率,降低损耗,节约成本,实现印刷质量的过程控制,本文提出了一种凹版印刷质量检测的方法。应用数字图像处理技术和统计学原理,建立了基于统计阈值法的印刷质量检测系统模型,对凹印产品进行检测;以峰值信噪比作为评价指标,选择去噪声处理算法;根据匹配正确率和时间,确定图像定位算法;依据检测需要,设计缺陷检测流程图;采用方差分析法和直方图分析法处理实验数据。通过比较五种去噪声模板的峰值信噪比,确定了用高斯滤波去除图像噪声;分别计算了三种图像定位算法的匹配正确率和时间,选取正确率高、时间短的二次综合匹配法进行图像定位;分析图像灰度值波动情况,建立了标准图像模板并确定各个像素点的灰度统计阈值;根据设计的检测流程图编写了凹版印刷质量检测软件,并对统计阈值法进行验证。结果表明:高斯滤波去除凹印图像噪声,峰值信噪比为28.89 dB,能够减少图像信息丢失;二次综合匹配法兼有相关系数法与误差法的优点,在确保匹配正确率满足要求的条件下,把时间缩短至误差法的1/5,适用于凹印图像的定位;统计阈值法能有效解决传统算法的边缘误检问题,在相同的检测精度下,检测时间比传统算法减少一半。更多还原

【Abstract】 High speed and automation are the developmental trend of gravure technology, but the inspection technology of printing quality can’t satisfy the demand at present. Although many experts in interior and exterior have done plentiful researches, there are still lots of defects which couldn’t be checked. This paper put forward a new method inspecting gravure printing quality for improving the inspection efficiency, reducing the wastage, saving the cost and achieving the progress control.The model of gravure quality inspection system based on statistical threshold method was established with the technology of digital image processing, and defects of gravure image could be detected. The peak out signal-to-noise was used to evaluate the effect of the noise process algorithms. According to the match right probability and time, the image registration algorithm is selected. The workflow of defect inspection was designed on the base of inspection demand. The variance analysis and histogram analysis were used to process experimental data.Comparing with the five algorithms, the Gauss Filter was chosen to wipe off noise. The match right probability and time of the three kinds of image registration algorithm were calculated, and the algorithm of twice integrate match with high match right probability and short time was selected. By analyzing image gray value fluctuation, the standard model image was built and the gray threshold of each pixel was gained. According to the workflow of defect inspection, the gravure quality inspection software was compiled to test and verify the statistical threshold method.It has indicated: the peak out signal-to-noise of Gauss Filter was 28.89 dB, and the image information loss would be reduced by using Gauss Filter to wipe off noise in gravure image. The algorithm of twice integrate match having the virtue of correlation coefficient method and the error method was suit for the gravure image registration, and the match time was one fifth of error method under the condition of insuring the match right probability. The algorithm of printing defect inspection based on the statistical threshold method could resolve the problem of outline false inspection, and the work time was the half of the conventional algorithm with the same precision.更多还原