【作者】 章慎锋；

【导师】 杨淑莹；

【作者基本信息】 天津理工大学 ， 计算机应用技术， 2006， 硕士

【摘要】 汉字识别的研究工作,一直被看作是十分困难的模式识别问题,并被视为字符识别的最终目的。汉字的识别最早可以追溯到上世纪60年代。我国对汉字识别研究工作始于70年代,从80年代开始,汉字OCR的研究获得了许多重大成果。本文对印刷体文字识别的整个过程作了研究,综合应用了视频流的播放和捕获、图像处理、模式识别等各种技术,并给出了一个印刷体文字识别的设计方案——基于USB口汉字识别系统。本文用DirectShow捕获USB口摄像头的图像,得到文字图像。然后对图像进行预处理,预处理包括图像的灰度变换、图像二值化、噪声消除、汉字的行字切分,文字大小规范化,细化处理,彩色图像灰度化处理等图像处理技术。接下来需要提取汉字特征,汉字的特征包括结构特征和统计特征,本文提取了笔划穿透数目特征、粗网格特征、粗外围特征、内孔个数等结构特征,统计特征提取Gabor小波变换特征,为了尽可能反映某个汉字,我们提取了大量的特征,但是太多的特征给以后的识别判断带来了麻烦,因此我们对提取的特征进行了主成分分析,对特征进行了降维。在研究了特征提取与选择方法之后,本文研究了字库的建立及汉字的识别方法。采用本文中所研究的方法我们实现了基于USB口汉字识别系统,该系统能对普通USB口摄像头所拍摄的文字图像进行在线识别,实验结果表明本文研究的技术切实可行,具有很好的应用前景。更多还原

【Abstract】 Chinese characters recognition is a typical pattern recognition problem. The recognition of Chinese characters is of great importance in the competition of our country in IT. Research on Chinese characters recognition can trace back to 1960’s. But in our country, we start the research in 1970’s, and it had great achievements from 1980. The dissertation does a general research on recognition of printing characters, and finally gives an example: Chinese characters recognition system based on USB, in which many techniques are adopted, such as video play and capture, image processing, pattern recognition and so on.In this dissertation we use DirectShow to capture an image from USB camera. After getting an image with character, the paper implements the processing of the image. The processing includes gray transform, binarization, smooth, line segmentation and character segmentation, normalization, thinning and so on. Then we extract features of each character, features includes structure features and statistic features. The paper extracts structure features such as number of cross stroke, rough girding features, periphery features, number of hole, and statistic features such Gabor transform features. To represent the character exactly, we extract a large amount of features, which bring troubles to recognition, therefore, principal component analysis (PCA) is applied to analyze the features, PCA can reduce the number of features effectively. In addition, the dissertation gives a research on how to build a dictionary dynamically and classify method.We realize the Chinese characters recognition system based on USB via techniques mentioned above, the system can recognize characters on images captured by USB camera. The experiments show that the techniques researched are feasible, and have bright foreground.更多还原