基于峰谷形态纸张纹路分割算法.docVIP

基于峰谷形态的纸张纹路分割算法 王富治 黄大贵 (电子科技大学智能机械系统信产部重点实验室, 成都 610054) 摘 要: 针对基于图像纸张计数技术最为关键的纹路提取与分割问题, 本文从峰谷形态角度提出一种新的算法。该方法首先利用测地重建消除产生干扰的次要极值区域, 接着利用区域极值对极值区域和非极值区域进行0, 1标记得到二值条纹后即可实现计数。文章比较了区域极值, 动态阈值以及全局阈值在复杂成像条件下不同的纹路分割性能。实验证明, 极值是比阈值更为优越的明暗条纹判据, 具有更强的抗毛边, 抗油墨, 以及抗光照不均性能, 因而能保证更高的计数准确率。 关键词: 纸张计数；区域极值；图像重建；测地膨胀 中图分类号: TG156　　　文献标识码: A　　　国家标准学科分类代码: 510.4050 Paper texture segmentation algorithm based on peak-valleymorphological feature Wang Fuzhi Huang Dagui (Intelligent Mechanical System Key Laboratory of Information Industry Ministry, University of Electronicand Science Technology of China, Chengdu 610054, China) Abstract: The texture extraction and segmentation is the key to the paper counting technique based on image. A new paper texture segmentation algorithm based on the peak-valley morphological features is presented in this paper. This method firstly uses the geodesic image reconstruction to remove the minor extrema region producing the interference points. Then using the regional extrema to extract the textures by marking the extrema region and non-extrema region with 0 and 1, the paper stripe is obtained, therefore the counting is easily realized. A comprehensive comparison of texture segmentation performance of region maxima, local maxima and global maxima are carried out in this paper. The experiments prove that the extrema is better criterion of the dark and light stripes than threshold because it almost is immune to the print ink and disparity of illumination, as a result, a higher counting accuracy can be obtained. Keywords: paper counting;regional extrema;image reconstruction;geodesic dilation 1 引 言 印刷造纸行业生产过程中, 面临着大量的纸张清点任务, 计数对印刷造纸业具有特殊重要的意义。目前普遍使用的机械式纸张计数机对纸张的厚度材质均有要求, 计数范围小, 速度有限, 容易打坏产品, 计数精度也不高, 给生产与管理都带来很多问题。因此, 研制快速、准确的非接触纸张计数设备具有重要意义。 随着计算机技术以及微电子技术的发展, 图像已成为人类获取信息的重要来源, 运用图像技术对纸张叠层图像进行处理进而计数就成为可能和发展趋势。国内外从成像方式, 计数算法等方面都对此作了不少研究, 其中国外的成果大都以专利形式出 现[1-4]。在国内, 文献[5]对所摄取的图像进行逐行扫描, 对每行像素施以傅里叶变换, 滤波处理后作二值处理提取纹路, 计算量巨大。文献[6]利用DSP构造FIR滤波器代替文献[1]中