基于Gabor特征与支持向量机的字符识别系统研究

利用计算机进行字符自动识别与录入的技术对机器翻译、数据挖掘、人工智能等都有着重要的理论意义和实用价值,基于数字图像处理技术的字符识别是其中的一个重要发展方向。文章重点研究了字符特征提取和匹配识别这两个影响字符识别效果的因素,根据中文字符笔画的方向特点,选择了对图像方向特征敏感的Gabor变换作为特征提取方式,在获取字符的特征向量后,先利用最小距离分类器进行预分类,再利用最小距离分类中产生的候选样本集训练SVM分类器,识别时只需利用候选集分类器依次判决,降低了训练和识别工作量,同时提高了识别效率。实验表明,系统对网站导航字符平均识别率达94%以上,具有一定的理论意义和实用价值。

A Study of Character Recognition System Based on Gabor Feature and Support Vector Machine

The technique that recognizes and inputs characters automatically by using a computer has an important theoretical significance and practical value in many fields,such as machine translation,data mining and artificial intelligence.Character recognition using digital image processing is an important development.The article focuses on character feature extraction and matching identification which influence character recognition effect.According to the property of the Chinese character strokes,Gabor transform which is sensitive to image direction feature is applied to extract the feature.After getting character feature vector,minimum distance classifier presorts first,then candidate sample set produced by minimum distance classifier trains the SVM classifier.Using candidate sample set SVM classifier to judge in turn when recognizing,not only reduces the workload of training and recognition,but also improves recognition effect.Based on this system,the average recognition rate of web navigation characters achieves 94% and higher,the results indicate that this system has certain theoretical and practical value.