论文部分内容阅读
当前图书馆书目信息管理大都是通过硬盘等存储器,对海量图书信息进行处理,但是信息输入的效率远远低于图书信息产生效率,使得书目信息查询准确率降低。面向该种问题,提出基于RBF神经网络的书目光学字符特征分类方法,先对海量书目光学图像进行图像去噪、光学字符区域定位以及字模分割的预处理,再采用基于统计与模糊隶属度的特征提取方法,获取书目光学字符特征图像的三大类特征。将RBF网络作为模式分类器,设计面向书目光学字符分类系统的网络模型,实现海量书目光学字符特征的准确分类。实验结果说明,所提方法的分类效率、分类精度都较高。
At present, the bibliographic information management of the library mainly deals with the mass information of books through hard disk storage, but the efficiency of information input is much lower than that of books, which makes the bibliographic information query accuracy lower. Faced with this kind of problem, this paper proposes a bibliographic optical character feature classification method based on RBF neural network. Firstly, image denoising, optical character region location and word segmentation are preprocessed for the massive bibliographic optical image, and then the features based on statistics and fuzzy membership Extraction method, access to the bibliographic optical character feature image of the three major categories of features. Using RBF network as a pattern classifier, a network model for bibliographic optical character classification system is designed to achieve accurate classification of optical character features of massive bibliographies. Experimental results show that the proposed method has higher classification efficiency and higher classification accuracy.