摘要
提出了一种基于Adaboost的视频文本定位的新方法。首先我们提取视频图像中的连通域,经过对视频文本区域分析,提取了视频文本的5类特征,然后利用这5类特征经过分类与回归决策树构造了Adaboost强分类器,最后将候选文本区域送入强分类器,得到正确的文本区域。实验结果表明本方法不仅对视频帧图像中字体、大小和颜色多变的文本具有很好的定位效果而且还实现了视频文本定位要求的快速性和准确性的特点。
This paper proposes a new method of video text location based on Adaboost. First extracting theconnected domains from the video image, we get five classes features of text after analyzing the text areas. Then weconstruct a strong classifier of Adaboost with CART (Classification And Regression Tree) using the five classesfeatures. Finally, we send the candidate text regions into the strong classifier to get the correct text areas. Theexperimental results show that not only can this method achieve a good effect on the text location in the video imagesincluding the text of various fonts , sizes and colors but also realize rapidity and precision that the video textlocation requires.
出处
《哈尔滨理工大学学报》
CAS
北大核心
2017年第1期103-108,共6页
Journal of Harbin University of Science and Technology
基金
黑龙江省教育厅科学技术研究项目(12541119)
关键词
文本定位
文本识别
连通域
强分类器
分类与回归决策树
text location
text recognition
connected domain
strong classifier
decision or classification tree