基于空间关系的维吾尔文图像关键词检索  

Key words retrieval method of Uygur based on spatial relations

作　　者：徐学斌 阿里木江·阿布迪日依木 朱亚俐[1] 阿力木江·艾沙[3] 库尔班·吾布力[1] XU Xue-bin;Alimujiang·Abudiriyimu;ZHU Ya-li;Alim·Aysa;Kurban·Ubul(College of Information Science and Engineering(School of Cyber Science and Engineering),Xinjiang University,Urumqi 830046,China;Project Service Department,Science and Technology Project Service Center of Xinjiang Uygur Autonomous Region,Urumqi 830002,China;Teachers’Department,Xinjiang University,Urumqi 830046,China)

机构地区：[1]新疆大学信息科学与工程学院(网络空间安全学院),新疆乌鲁木齐830046 [2]新疆维吾尔自治区科技项目服务中心项目服务部,新疆乌鲁木齐830002 [3]新疆大学教师工作部,新疆乌鲁木齐830046

出　　处：《计算机工程与设计》2021年第2期497-503,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61563052、61862021、61363064)。

摘　　要：为提高维吾尔文档图像的检索效率,提出一种基于字符空间关系的关键词检索方法。通过对文档图像进行单词切分,提取切分后单词图像的字符空间位置特征,将提取的特征根据单词的连体段数目存储为多个特征文件,根据输入关键词图像的特征寻找对应的特征文件进行查询。从115张印刷体维吾尔文档图像切分后的24460张单词集中选取10张有丰富含义的关键词图像在单词库中进行检索实验,平均准确率为96.47%,平均召回率达到了93.74%,平均每张单词的查询耗时为0.25 s,验证了该方法在维吾尔文档图像检索中的有效性。To improve the retrieval efficiency of uygur document image,a key word retrieval method based on character space relation was proposed.By word segmentation of document image,character space position features of word image after segmentation were extracted.The extracted features were stored as multiple feature files according to the number of concatenated segments of the word,and query was implemented in the corresponding feature files according to the input image features of keywords.Ten keyword images with rich meanings were selected from the 24460 word sets of 115 printed Uygur documents after image segmentation,and retrieval experiments were conducted in the word database.The average accuracy is 96.47%,the ave-rage recall rate reaches 93.74%,and the average query time per word is only 0.25 s.Experimental results show the effectiveness of this method in the retrieval of Uygur document images.

关 键 词：维吾尔语 单词切分 关键词检索 连体段 空间关系 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]