用于叠层纸张图像分析计数的快速直线增长算法  

A FAST STRAIGHT-LINE GROWTH ALGORITHM FOR ANALYSING AND COUNTING STACKED-PAPER IMAGES

作　　者：杨硕[1] 彭双[1] 肖昌炎[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410082

出　　处：《计算机应用与软件》2015年第9期188-191,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61172160);湖南省自然科学常德联合基金项目(12JJ9019)

摘　　要：叠层纸张数量检测是包装印刷生产过程中重要的工艺环节,其计数精度直接影响企业经济效益。基于机器视觉成像的计数方法具有非接触、无物理损伤、实时测量等优点,在高端印刷产品质量控制中得到了广泛的应用。针对叠层纸张图像端面条纹检测难题,通过改进基于水平线方向引导增长的直线段提取算法,可以提高条纹识别的准确度和完整性,从而实现叠层数量的高精度测量。主要工作包括:引入单向梯度算子,消除同一条纹多重响应的问题;组合梯度幅值和水平线方向信息,并改善线支持区域增长;采用连接元分析,弥补条纹局部断裂异常。最后,通过大量不同类型印刷纸张的实际测量实验验证了算法性能。结果表明:长期测量误差小于0.75‰,完全可以满足工厂应用要求。Quantity measurement of stacked-sheet is an essential technical step in packaging and printing production process, and its counting accuracy has a direct impact on the economic efficiency of related companies. With its merits of noncontact, non-destructivity and real-time measurement, the counting method using machine vision imaging has been widely applied to quality control in high-end printing products. Aiming at the end face fringe detection problem in stacked-sheet images, it is able to improve the precision and completeness of fringe identification by improving the line-segment extraction algorithm which is based on guiding the growth along horizontal line direction, so as to realise the high accuracy measurement of stack quantity. Our work mainly consists of three parts： 1 ） Introducing the unidirectional gradient operator to eliminate the multiple responses on a single fringe; 2 ） Combining the gradient magnitude and horizontal line direction, and improving the growth of line supported regions; 3） Using connected element analysis to make up the anomaly of fringe＇s partial rupture. At last, the performance of our algorithm is verified in experiments of actual measurements using a large number of various printed-papers. Results show that the long-term measurement error is less than 0.75‰ and it is sufficient to meet the requirement of factory applications.

关 键 词：机器视觉.纸张计数 直线分割 LSD 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]