00422539

open access

Version of Record

MPEG エイゾウ カラノ カイソウテキ クラスタリング ニヨル ジクウカン インデキシング

Spatio-Temporal Indexing of MPEG Video Based on Hierarchical Clustering

Japanese (日本語)

神戸大学都市安全研究センター研究報告

9

237-246

神戸大学都市安全研究センター

2005-03

2007-04-23

Recently, there have been many studies on analyzing human behavior from its spatial trajectory measured by location sensors such as GPS (Global Positioning System). Although we need location sensors to measure each trajectory, it is useful for us to analyze spatio-temporal data and to know how people act. However, as trajectory data is multi-dimensional data, we need to index such a large amount of data to get useful information easily. In this paper, to solve these problems, first, instead of location sensors, we use motion vectors obtained where movement occurs in each frame. Second, in order to access required data easily, we apply hierarchical clustering to index these regions and to construct binary tree. Spatio-temporal data include time attribute, so we regard time attribute as a restriction for hierarchical clustering. To extract region of moving in each frame and index these data, we can get trajectory easily and exactly.

時空間インデキシング

紀要論文