隨著網際網路和科技的發展，使得在網路中，多媒體資料被大量的建立，影片即是其中之一。然而如何從龐大的影片資料庫中，尋找出所想要的影片片段是很不容易的。一段影片可以視為一連串的畫面，根據畫面中的物件之影片尋取是一有效的方法，然而，在先前的研究中，注重在各個畫面物件的位置，並未考慮到在連續畫面間，物件的移動關係變化。 在連續畫面中，物件可能移動，因物件移動而發生的空間關係變化是極為重要的資訊。本研究提出一索引結構，利用物件間的移動空間關係變化來建立一矩形圖形的索引，物件對在每個片段的移動資訊被記錄在一作為索引結構的矩形中。將許多矩形記錄在一個圖上，基於這些矩形，我們可以來分析影片片段的差異，並過濾和影片不相似的影片片段。最後我們利用 R-tree 來過濾影片資料，以提高影片查詢的效率。 By the growth of the technology and internet, a huge amount of multi-media data, such as video, have been created on the internet. However, searching for the pieces you want from this massive video database can be extremely challenging. A video can be seen as a series of frames. Video search based on the object in the video is an effective method. However, in previous studies, most of work is based on the location of the object. The information about the change of the relationship between objects in the continuous frames is not considered. In continuous frames, objects may move. The information about the change of the spatial relationship between objects due to their movement is important. In this study, we present an index structure. The type of change of spatial relationship between objects is used as the index to create a rectangle diagram. The information of objects for a segment of video is recorded as a rectangle in the index structure. Many rectangle are recorded in the diagram. An then we can analyze the difference of each segment of the video based on their rectangle in the diagram, and find the similar segment of the video to the query video. Finally, we use R-tree to filter video data in order to improve the efficiency of the query in video.
