基于卷积的蒙文词语定位查找方法研究  

Study on Mongolian Words Locating Method Based on Convolution

作　　者：郑军[1] 那木拉[2] 阿喜达[3] 

机构地区：[1]包头师范学院网络信息中心,内蒙古包头014030 [2]包头师范学院图文信息中心,内蒙古包头014030 [3]包头师范学院教务处,内蒙古包头014030

出　　处：《阴山学刊（自然科学版）》2017年第1期53-55,共3页Yinshan Academic Journal(Natural Science Edition)

基　　金：国家自然科学基金项目(61163025);内蒙古自治区高等学校科学研究项目(NJZY16238)

摘　　要：由于历史文化和技术的原因,有大量的蒙古文文献资料是以图像的形式保存在图书馆的.图像格式的文档资料在进行检索和查阅时费时费力,极不方便.为了解决这个问题,提出了一种基于卷积的蒙古文词语定位查找方法.给出了二维离散傅立叶变换模型、二维离散卷积模型以及二者相互转换的数学模型.应用以上模型,在对蒙文文本图像和目标图像进行图像预处理的基础上,设计了基于卷积的蒙古文词语定位查找的具体实现算法.并对算法进行了仿真,通过快速连通区域查找法展示了定位查找到的蒙古文词语.实验表明基于卷积的蒙古文词语定位查找算法可以有效地解决蒙古文图像文本词语查找定位问题.Due to the history of culture and technology,a large number of Mongolian documents are in the form of images stored in the library. Documents in image format take time and effort in search and retrieval,extremely inconvenient. In order to solve this problem,this paper proposes a method of locating Mongolian words based on convolution. The two- dimensional discrete fourier transformation model,the two- dimensional discrete convolution model and the transformation of the two are given. Applying the above models,based on image preprocessing of Mongolian text image and the target image,we design the algorithm for the Mongolian words locating based on convolution. Finally,the experiments were carried out to verify the algorithm,and through the fast connected region search method,we show the Mongolian words located. Experiments show that the Mongolian words localization search algorithm based on convolution can effectively solve the locating problem of Mongolian image text words.

关 键 词：蒙古文 词语定位查找 卷积 傅里叶变换 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]