基于图像纹理的纸张计数算法设计  

Paper counting algorithm based on image texture

作　　者：张明阳[1] 陈振跃[1] 王霞[1] 

机构地区：[1]北京理工大学光电学院,北京100081

出　　处：《光学技术》2013年第2期151-156,共6页Optical Technique

摘　　要：纸张等薄片的计数在工业生产中使用得非常普遍,而传统的纸张计数方法大多是通过专门的机械装置甚至人工来完成的,具有成本高、体积大、效率低、准确率低等各种突出问题,迫切需要改进。针对彼此平行放置的纸张,提出了一种沿其放置方向概率统计的算法,在采样位置处统计正交于放置方向的纸张数目,最后选取出现概率最大的一个数据作为统计结果。考虑到纸张之间彼此不平行且出现重叠的情况,提出了一种全局轮廓的统计算法,经过预处理,对纸张边缘轮廓统计计数。实验结果表明,该算法有效地克服了因为遮挡和重叠而引起的统计错误。Counting sheets of paper is a very common work in industrial production. Conventional paper counting usually depends on special mechanical devices or even manual work- which has high cost, big volume, low efficiency and accuracy rate, and needs to be improved urgently. For the paper being paralell to each other, a probability statistics algorithm is put forward according to the paper direction, that is, counting the number of paper being orthogonal of paper direction at the sampling location, and choosing the maximal probility data as the statistical result. While for a more complicated condition where the sheets aren＇t parallel or even maybe folded, a statistical algorithm based on global contour is put forward, that is, counting the paper edge contour after pretreatment. The experiment results show that the proposed algorithm can overcome the errors caused by blocking and overlapping effectively.

关 键 词：纹理分析 纸张计数 数学形态学 概率 轮廓统计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]