基于纹理分析的张数检测算法  

Number detection of layered structures based on texture analysis

作　　者：刘东昌[1] 王欣刚[1] 李锐[1] 

机构地区：[1]中国科学院自动化研究所,北京100190

出　　处：《中国图象图形学报》2011年第3期413-419,共7页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60873160);国家高技术研究发展计划(863)项目(2009AA01Z320)

摘　　要：针对印刷行业中产品张数检测问题,提出一种基于纹理分析的印刷品张数检测算法。该方法通过对印刷品的侧面图像进行分析得到张数值。主要处理过程分为4步:首先对输入图像进行预处理,即构造一个Gabor滤波器对图像纹理进行增强;其次把预处理后图像沿某一方向分段投影转化为一系列的1维投影信号;再次利用经验模式分解(EMD),从含有复杂噪声的信号中抽取周期性的张数估计值;最后统计上述估计值并用求众数的方法确定最终的张数结果。接下来设计了两个实验来测试该算法的稳定性和精度,并证明了算法的有效性。In order to detect the sheet number of layered structures in print industry, a texture analysis based method is proposed. This method detects pages of print by analyzing side image of the target. The main process has four steps： At first, a Gabor filter is designed to enhance the texture in input image. Then, a serial of one-dimensional signals are extracted from the enhanced image using projection method. Third, empirical mode decomposition （EMD） is used to analyze these signals, which contain complex noise, to obtain the corresponding estimate of sheet number. Finally, the result is determined by the mode of all estimates. We designed two experiments to test stability and accuracy of the algorithm and proved its effectiveness.

关 键 词：纹理分析 张数检测 经验模式分解 GABOR滤波器 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]