多方向自然场景文本检测  

Multi-orientation natural scene text detection

作　　者：何思楠 郭永金 张利[1] He Sinan;Guo Yongjin;Zhang Li(Dept.of Electronic Engineering,Tsinghua University,Beijing 100084,China;Systems Engineering Research Institute,Beijing 100094,China)

机构地区：[1]清华大学电子工程系,北京100084 [2]中国船舶工业系统工程研究院,北京100094

出　　处：《计算机应用研究》2018年第7期2193-2196,共4页Application Research of Computers

基　　金：国家自然科学基金重点项目(61132007)

摘　　要：针对自然场景图像背景复杂和文本方向不确定的问题,提出一种多方向自然场景文本检测的方法。该方法利用颜色增强的最大稳定极值区域(C-MSER)方法对图像中的字符候选区域进行提取,并利用启发式规则和LIBSVM分类器对非字符区域进行消除;然后设计位置颜色模型将被误滤除的字符找回,并利用字符区域中心进行拟合估计文本行倾斜角度;最后通过一个CNN分类器得到精确的结果。该算法在两个标准数据集上(ICDAR2011和ICDAR2013)测试,得到F-score分别为0.81和0.82,证明了该方法的有效性。The background of text regions in natural scene images was complex,and the orientation of the text-line was uncertain,to address this problem,this paper proposed a multi-orientation natural scene text detection method. First the method used the color-enhanced maximally stable extremal region（ C-MSER） to create character candidates for natural scene images,and used the heuristic rules and the LIBSVM classifier to eliminate non-character regions; then it designed a position-color model to retrieve the false negative candidates,and used the centers of the character regions for linear fitting to get the inclination angle of text-lines; finally,the method trained a CNN classifier to obtain the accurate results. The method gets the F-score of 0. 81 and 0. 82 on two standard benchmark datasets（ including ICDAR2011 and ICDAR2013）,which demonstrates the effectiveness of the proposed multi-orientation natural scene text detection method.

关 键 词：自然场景文本检测 颜色增强的最大稳定极值区域 特征提取 多方向估计 分类器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]