地球信息科学学报 ›› 2011, Vol. 13 ›› Issue (3): 401-408.doi: 10.3724/SP.J.1047.2011.00401

• 遥感技术与地图应用 • 上一篇    下一篇

综合多特征遥感图像智能检索方法的概念设计

戴芹, 刘建波, 刘士彬   

  1. 中国科学院对地观测与数字地球科学中心,北京 100094
  • 收稿日期:2010-04-01 修回日期:2011-05-28 出版日期:2011-06-25 发布日期:2011-06-15
  • 作者简介:戴 芹(1978-),女,博士,副研究员,主要研究方向为遥感信息处理与服务。E-mail:qdai@ceode.ac.cn
  • 基金资助:

    国家自然科学基金项目(40701105)。

The Conceptual Framework Design of Intelligent Multi-feature Remote Sensing Image Retrieval

DAI Qin, LIU Jianbo, LIU Shibin   

  1. Center for Earth Observation and Digital Earth, CAS, Beijing 100086, China
  • Received:2010-04-01 Revised:2011-05-28 Online:2011-06-25 Published:2011-06-15

摘要: 遥感数据作为一种重要的时空数据源,在环境监测、资源管理、灾害预报、重大工程监理、国防安全等众多领域发挥着不可或缺的重要作用。遥感图像检索的效率和精度,直接关系到遥感影像数据应用的广泛性和实时性。本文在对国内外其他领域的图像检索技术、遥感图像检索技术的研究进展及其发展趋势进行综合分析的基础上,针对目前遥感图像检索所存在的主体内容和检索主题难以确定、多特征索引难以建立、多特征难以综合利用、检索系统的智能性低、检索效率低等缺陷问题,结合海量遥感数据的多样性、复杂性、不确定性等特征,提出建立综合多特征遥感图像智能检索方法的研究思路,并分别对遥感图像多特征表达模型和智能检索的概念模型进行理论构建和分析,详细设计了综合多特征遥感图像智能检索的总体方案。另外,对综合多特征遥感图像智能检索方法中的如何构建遥感图像的智能标注方法,发展综合多特征的相似度识别模型,建立遥感图像信息的智能反馈机制等系列关键问题及其发展方向进行分析与讨论。

关键词: 多特征, 智能方法, 遥感图像, 智能检索

Abstract: With the increasingly improved earth observation systems, the rapid development of various observation techniques and computer intelligence technology, more and more research fields such as environmental monitoring, resource management, disaster forecast, major projects management, national security are highly rely on remote sensing information and its high timeliness. Therefore, we are not only facing the most abundant resources of remote sensing data, but also facing the challenges of transforming the data into information with effective information extraction technology. The Content-based Remote Sensing Image Retrieval (CBRSIR) as one effect approach that obtains remote sensing information quickly is receiving increased attention in the remote sensing application community, and it is emerging as a major research direction. Based on reviewing and analyzing the related technology on CBRSIR, the main bottleneck problems in current research scope are summarized in this paper, which are: one scene remote sensing image always has no main theme or obvious content; difficulty in indexing the multi-feature extracted from remote sensing image; it is difficult to apply multiple features synthetically; the current systems always have low intelligence and efficiency. These problems affect the development of CBRSR. Considering the diversity, complexity and mass of remote sensing images, this paper proposed an intelligent multi-feature integrated remote sensing image retrieval model and constructed its framework. Then the key technology of the multi-feature integrated model, the multi-feature description model, the theoretical framework of intelligent extraction methods and processing algorithms of the multi-feature are also proposed in this paper, and the feature research trend on multi-feature selection and dimensionality reduction, the multi-feature integrated similarity measure model, the intelligent feedback model are also prospected. In the end, this paper discussed the significance of construction of the intelligent multi-feature integrated remote sensing image retrieval method and the main feature work.

Key words: intelligent method, multi-feature, remote sensing image, intelligent retrieval