场景图像中文本占据的范围一般都较小，图像中存在着大范围的非文本区域。因此，场景图像文本定位作为一个独立步骤越来越受到重视。这包括从最先的CD和杂志封面文本定位到智能交通系统中的车牌定位、视频中的字幕提取，再到限制条件少，复杂背景下的场景文本定位。与此同时文本定位算法的鲁棒性越来越高，适用的范围也越来越广泛。文本定位的方式一般可以分为三种，基于连通域的、基于学习的和两者结合的方式。基于连通域的流程一般是首先提取候选文本区域，然后采用先验信息滤除部分非文本区域，最后根据候选文本字符间的关系构造文本词。基于学习的方式关键在于两个方面：一是不同特征提取方法的使用如纹理、小波、笔画等。二是分类器的使用如支持向量机(Support Vector Machine，SVM)，AdaBoost等。连通域和学习结合的方式一般在提取阶段采用连通域的方式，但是滤除阶段是通过训练样本学习分类器来实现非文本的滤除。-The range of the text in the scene image is generally small, and there exists a wide range of non-text areas in the image. Therefore, the scene image text positioning as an independent step more and more attention. This includes positioning text　the first CD and magazine cover to license plate location in intelligent transportation systems, capturing subtitles in video, and then locating scene text in complex backgrounds with fewer constraints. At the same time, the robustness of the text localization algorithm is more and more high, and the range of application is more and more extensive. Text localization methods can be divided into three types, based on the connectivity domain, based on learning and a combination of the two methods. The process of the connectivity domain is usually to extract the candidate text area first, then to filter some non-text areas by prior information, and finally construct the text words according to the relationship between candidate text characters. Le