以XML為基礎實做影像搜尋用的知識庫

An Implement of the Knowledge Database for Image-Retrieval Based on XML

指導教授 : 陳正光、陳雙源  研究生 : 葉榮鑫  機電整合研究所 91年


摘要

  這篇論文係一種利用輸入圖片為基礎來學習知識及獲取訊息的方法,其目的是在於建置一個 ”所見即所知” 的學習環境,此學習環境包括一種可輸入影像的裝置,如附有攝影機或掃描器的手機、PDA、翻譯機、筆記型電腦等裝置。

在本篇論文中主要是以BMP及JPG兩種圖檔格式的圖片為研究對象,知識庫則是以XML為基礎,採用多層分散式的系統架構,並將其XML文件存放至一XML Native Database中。當使用者將圖片上傳至Application Server時,Application Server就能立刻算出該張上傳圖片相關的特徵值,並將所得資料以XQuery方式,透過多層分散式系統架構,向後端的知識庫做查詢,查詢結果將以XSL為基礎,依個人喜好及瀏覽裝置可做不同的檔案型態輸出格式,例如HTML、XML、WML、PDF、SVG等等。而這種『一份文件輸入,多份文件輸出』的架構,才能滿足未來網路資料庫的架構,這也是我們提出以XML作為知識庫的主因。最後則是以植物及風景圖片作為系統的驗證。

關鍵詞:XML、影像搜尋

ABSTRACT

  This thesis developed a system for people to learn knowledge and to get message base on import photograph. For the purpose of convenience, this study seeks to set up a learning system to achieve the goal of ”Seeing is Knowing”. This learning system is comprised with a device to input image, e.g. a mobile phone, PDA, translator, notebook etc, attached with a camera or scanner.

The research subjects are pictures using BMP and JPG files format in this thesis. Knowledge database is based on XML and the architecture adopts multi-tier distributed system, then the system use XML document to deposit a Native XML database. When user upload the picture to Application Server, in a minute, Application Server can work out related eigenvalue for this picture, then use the XQuery mode through multi-tier distributed system to query database at the back. The query result is based on XSL to make different export files format depending on an individual and browse device, e.g. HTML, XML, WML, PDF, SVG etc. The architecture that importing only one document can export multiple style documents expects to meet the requirements of the future network database, thus we take XML as our knowledge database. Finally, we use photograph of plants and landscapes to test and verify the performance of the system.

Keywords:XML, Image search