標題: | 視訊資料查詢系統的設計與實作 The Design and Implementation of a Video Query System |
作者: | 賴昇志 Sheng-Chih Lai 羅濟群 Chi-Chun Lo 資訊管理研究所 |
關鍵字: | 視訊資料切割;場景變化偵測;視訊資料檢索;視訊資料庫;模糊分群;Video Segmentation;Scene Change Detection;Video Indexing;Video Database;Fuzzy Clustering |
公開日期: | 1999 |
摘要: | 本論文設計並實作效率良好之視訊資料庫系統。主要可分為視訊資料的切割,找出視訊資料中的場景。以及建立視訊資料場景的索引,最後提供使用者檢索查詢。
在視訊資料的前端處理部分可再區分為場景變化偵測方面與視訊資料的索引。利用Fuzzy C-Means模糊分群演算法來判斷是否發生場景變化。在判斷出場景變化的發生點後,可以藉此將影像切割成一段一段的場景,再由各段場景內取出具代表性的關鍵視訊框,提供建立檢索的依據。
檢索建立的部分,依照前一步驟取出的關鍵視訊框,利用物件的時間關係來定義各個關鍵視訊框所代表的數對,以這些數對為基礎來建立索引值。並利用這些索引值計算出各段影片的特徵值。在使用者查詢時利用模數計算的方法來判斷場景是否存在於影片之中。
本論文所實作的查詢系統,運用的演算法有相當良好的效率。經由實驗結果發現,確實有良好的效率。 The system described in this theisis can provide query service for user. This video query system is quite efficient. Primary tasks in the query system are scene change detection on video data to get the position of shots and video indexing and querying. We could provide querying and retrieving videos for users' need. On scene change detection, Fuzzy C-Means algorithm is using for calculating feature data in video data and detecting scenes change. After we know the scene-change position, we could cut video into shots and get key frames from shot for indexing and querying. On indexing and querying, we use object temporal relationship to define triples represented to every key frame. These triples are used to build indexing. Standing values of every video can be got from those indexing value. At the time of user querying, we could use modular expression to know if the shot is existed in video. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT880396022 http://hdl.handle.net/11536/65603 |
Appears in Collections: | Thesis |