项目名称: 维吾尔民俗图像的双语自动标注与检索关键技术研究
项目编号: No.61262065
项目类型: 地区科学基金项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 古丽松·那斯尔丁
作者单位: 新疆师范大学
项目金额: 43万元
中文摘要: 为了满足人们从海量图像数据库中用不同语言关键词来快速检索新疆维吾尔民俗图像信息,更全面的了解新疆特色文化的需求,本项目着重于研究维吾尔民俗图像的维汉、维英双语自动标注方法,以及基于标注的图像检索关键技术。拟解决的主要问题包括:(1)根据图像语义,对于维吾尔民俗图像进行分组;(2)为图像自动标注的预处理各环节选择、运用有效的方法,包括对图像的注册、量化、去糟、特征抽取等;(3)研究与实现针对所选类图像的自动标注有效算法。尤其是研究与发现对于图像进行维汉、维英双语自动标注的策略及方法。研究与开发多语义图像的双语标注及基于标注的图像检索方法。以上问题的解决涉及到模式识别、图像处理、机器学习以及数据库管理等知识。特别是对于多语义民俗图像的标注,采用优化现有的多标签分类算法,或提出新的多标签分类算法来实现。该项目的应用范围将包括旅游业,博物馆等地的基于维、汉、英关键词的民俗图像检索及幼儿双语教育等。
中文关键词: 维吾尔民俗乐器图像;;多语义图像;;维;汉;英多语关键词;;多标签分类;;图像自动标注
英文摘要: In order to satisfy people's requirements of retrieving information about Xinjiang Uyghur folk-custom images from massive image databases using different language key words, learning more about special culture of Xinjiang. This project focusing on investigation of effective methods for Uyghur-Chinese, Uyghur-English bilingual automatic Uyghur folk-custom images annotation, and key techniques for annotation based image retrieval. The main problems to solve in the proposed project include (1) grouping the Uyghur folk-custom images based on the semantics of the images; (2) Selection and adoption of effective methods for automatic image annotation pre-processing key steps. These include image registration, vectorization, de-noising and feature extraction (3) Investigation and realization of effective automatic annotation algorithms for annotating selected type of Uyghur folk-custom images with key words. Particularly, investigation and exploration key strategies and methods for Uyghur-Chinese, Uyghur-English bilingual automatic annotations. Investigation and exploration of the bilingual automatic image annotation and annotation based image retrieval methods for the condition that images associated with multiple semantic concepts. The problems to be solved of above involve the knowledge of pattern recognition, image
英文关键词: Uyghur folk music instrumental images;;multi-semantic images;;Uyghur-Chinese-English multi-lingual keywords;;multi-label classification;;automatic image annotation