摘要
提出了一种基于流形学习的视频中文文本检测算法。算法重点针对文本图像和非文本图像的特征提取、流形降维、分类器训练等关键部分进行了改进,对人为收集的文本图像样本与非文本图像样本进行特征提取,并使用等距离映射的流形学习算法来完成特征降维,最后使用支持向量机来完成分类器训练,获取文本与非文本检测分类器,完成视频中文文本检测。实验结果表明,算法具有明显优越性,在检测可靠性和准确度上有较大提高,具有一定的实用意义。
Put forward based on the manifold learning video Chinese text detection algorithm is proposed.Algorithm for image and text on the text image feature extraction,manifold dimension reduction,classifier training,and other key parts were improved,and for the collection of text image with the sample text sample image feature extraction,and use mapping manifold learning algorithm is introduced to complete the characteristic dimension reduction,finally using support vector machine classifier to complete the training,get the text with the text detection classifier,complete video Chinese text detection.The experimental results show that the new algorithm has obvious advantages in the test reliability and precision are improved greatly,and has a certain practical significance.
出处
《科技通报》
北大核心
2012年第10期46-48,共3页
Bulletin of Science and Technology
