Geomedia resource discovery and retrieval project

本研究旨在探讨透过网际网路存取各种地理媒体内容的有效技术。Geomedia内容涉及与地理领域相关的多媒体内容（文本，图片和图像，视频，声音）。为了在分布式环境下有效地搜索基于GEOMEDIA的内容，我们建立了一个元级系统，总结了从GEOMEDIA中提取的Meta数据。这些元数据是分层累积的。本课题的研究目标是建立一个能够根据用户的兴趣方便地访问GEOMEDIA数据库的系统，并在GEOMEDIA数据库的基础上开发了一个元级系统。主要成果有：（1）建立了多尺度表示（2）创建可以检测多维图像数据的任意形状的聚类的新颖聚类方法，（3）构建元数据模型，该元数据模型使得能够制定集成系统所需的元数据，以将可视查询引导到相关GEOMEDIA数据库，（4）建立了基于元数据的数据库选择方法的理论基础;（5）建立了一个可视化查询处理组件，该组件集成了从可视化数据内容中提取的异构特征，结果平台包括一个元服务器组件，该组件包括一个层次元数据库、一个元数据库代理和一个查询管理器。根据用户的反馈，系统的性能得到了改进。查询管理器提取查询中的文本、纹理和颜色等异构特征，以实现元数据的匹配。根据各个特征制定子查询。查询管理器将子查询的结果集成到最终的数据库选择决策中。

This research aimed to investigate effective and efficient technique for accessing to various GEOMEDIA contents via the Internet. Geomedia content is related to Multimedia contents relevant to the geography field (texts, pictures and images, video, sounds). In order to search efficiently inside GEOMEDIA based contents in distributed environment, we built a meta-level system which summarizes the meta data extracted from GEOMEDIA. These metadata are accumulated hierarchically. The target of the research was to built a system which can access to GEOMEDIA data easily according to end-user own interests.The project produced a meta-level system on top of geomedia databases. The major achievements are (1) the creation of multi-scale representation (multidimensional) methods for geographic images, (2) the creation of novel clustering approaches that can detect clusters of arbitrary shape of multidimensional image data, (3) the construction of a metadata model that enables the formulation of metadata that are needed for the integrated system to direct a visual query to relevant GEOMEDIA databases, (4) the creation of the theoretical foundation of database selection approaches based on the metadata, and (5) the creation of a visual query processing component that integrates heterogeneous features extracted from the content of visual data.The result platform includes a metaserver component including a hierarchical metadatabase, a metasearch agent, and a query manager. The performance of the system has been refined based on user's feedbacks. Also the query manager extracts heterogeneous features such as text, texture, and color in the query for suitable matching of the metadata.Subqueries have been formulated based on individual features.The query manager integrates the results from the subqueres into the final database selection decision.