基于多向投影新方法的捆钞印章个数统计  

Statistics of the Number of Seals Based on Multi-directional Projection

作　　者：宋昀岑[1] 陈镇龙[1] 罗颖[1] 刘霖[1] 叶玉堂[1] 陈伟[1,2] 刘少荘 

机构地区：[1]电子科技大学现代光电测控及仪器实验室(摩米实验室),四川成都610054 [2]成都术有科技有限公司,四川成都610054

出　　处：《计量学报》2015年第3期242-246,共5页Acta Metrologica Sinica

基　　金：国家自然科学基金(61205004);粤港关键领域重点突破项目(东科2012205106,20091683);总装预研基金(9140A01020507DZ0215)

摘　　要：提出了一种快速自动统计印章个数的多向投影新方法。该方法基于图像处理理论，将采集到的捆钞图像从RGB空间转为LUV空间，然后通过聚类分析算法，得到捆钞的二值化图像，再利用多向投影的思路，依次在垂直和水平方向投影变换，完成印章个数的统计。理论分析及实验结果表明，该方法可以有效去除背景噪声的影响，不依赖于印章印迹的好坏，提高了印章个数自动统计的效率和准确率。A multidirectional projection method which can count the number of seals on bundled banknotes automatically and quickly is introduced. The fundamental principle of this method is that, based on the theory of image processing, converting bundled banknotes image from RGB space to LUV space in the first, processing it by using cluster analysis algorithm, then a binary image of the tied note can be gained. Using thought of multidireetional projection, projection transformation in the vertical direction and the horizontal direction then be applied to in turn. Finally the number of the seals can be counted. Theoretical analysis and the experimental results show that, this method can remove the impact of background noise effectively, do not depend on the quality of seals sigil, and improve the efficiency and accuracy of the automatical number statistics of the seals.

关 键 词：计量学 图像处理 印章个数 聚类分析 多向投影 机器视觉 

分 类 号：TB96[机械工程—光学工程]