隨著資訊科技處理視訊能力的進步，且視訊所具有動態及豐富的資訊，已使得視訊 成為有價值的資訊來源，因此如何在視訊資料庫中查詢相似視訊已引起廣泛的興趣。而 要自資料庫中識別視訊的重要方法之一，是如何識別視訊中的物件及物件間所具有的空 間時間關係，我們已提出3D Z-string 視訊知識表示法，利用視訊物件在空間及時間的投 影，建立物件間的空間時間關係，並記錄物件大小及移動速度變化的資訊；而視訊查詢 的主要問題是如何由大量的視訊資訊庫中找出使用者所想要的視訊，因之前所提的3D C-string 相似度視訊查詢，採用精確的相似片斷視訊比較，使得視訊中相似度資訊有所 損失，且其運算的複雜度成本過高，為NP-hard 問題。所以在本研究計畫中，我們計畫 以3D Z-string 來探討新的空間時間相似視訊的查詢方法，以能更容易及更具彈性的由資 料庫中查詢視訊間整體的空間時間相似程度，預期查詢將能提供部份視訊相似時的資 訊，並發展能利用使用者回饋資訊更新查詢結果的機制，以提供更具彈性的查詢方法。 且為顯示提出方法的效率及效能，也預計進行多項和之前所提出方法比較的實驗，並計 畫開發視訊資料庫雛型來實證本計畫所發展的成果。
With the advances of processing video in information technology, videos have been promoted as a valuable information resource. Because of its expressive power, videos are an appropriate medium to show dynamic and compound information. There has been widespread interest in the video database to query the similar videos. To retrieve desired videos from a video database, one of the most important methods for discriminating videos is the perception of the objects and the spatio-temporal relations that exist between the objects in a video. We have proposed the 3D Z-string that used the projections of objects to represent spatial and temporal relations between the objects in a video, and to keep track of the motions and size changes of the objects in a video. The video retrieval problem is concerned with retrieving videos that are relevant to the users’ requests from a large collection of videos. In our proposed 3D C-string similarity retrieval method, the exactly match would lose some similar information, and the similarity retrieval cost of 3D C-string is NP-hard. Accordingly, in this research, we plan to develop a new similarity approach that can be easily applied to an intelligent video database management system to infer integrated spatial and temporal relations similarity between videos. Therefore, the query approach can find the partly matched object sets and provide the refined mechanism to meet users’ requirement from the feedbacks that provides a more flexible way to retrieve similar videos. To show the efficiency and effectiveness of approaches, we will plan a series of experiments to compare approaches with the previously proposed approaches. We also plan to develop a prototype video database management system that supports the methods developing in this research plan.
