一种自适应烟标字符提取方法  

Method of adaptive cigarette mark character extraction

作　　者：殷羽[1] 郑宏[1] 王静[1] 李圣[1] 王震[1] 

机构地区：[1]武汉大学电子信息学院,武汉430072

出　　处：《计算机应用研究》2015年第4期1248-1252,共5页Application Research of Computers

基　　金：高等学校博士学科点专项科研基金资助项目(20110141110044)

摘　　要：针对印刷质量不高、背景复杂的烟标字符难以提取问题,提出了一种自适应的烟标字符提取算法。通过对单通道烟标图像进行背景重构后,根据支持向量机中得到的训练模型自适应选取最优二值化方法;提取标准轮廓定位字符区域;结合标准轮廓,同时根据最优二值化图像投影直方图进行烟标字符提取。经过实验验证,该方法能够自适应地选择最优二值化方法,字符提取精度可达到90%以上,同时原理简单、鲁棒性强并便于硬件实现。This paper proposed a cigarette mark character extraction method which could adaptively solve the problem that cigarette image difficult to extract. By background reconstruction to the single channel images,the method adaptively selected the optimal binarization based on SVM,and extracted the standard contours for the character areas. According to the standard contours and based on the best binary image histogram projection,the method extacted the cigarette characters. After experimental verification,it adaptively selected the best binarization and character extraction accuracy could reach 90%. Simultaneously the method has simple principles and strong robustness and convenient for hardware implementation.

关 键 词：支持向量机 背景重构 N-Otsu法 最优二值化 字符提取 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]