Mining temporal co-orientation pattern from spatio-temporal databases

Abstract

A spatial co-orientation pattern refers to objects that frequently occur with the same spatial orientation, e.g. left, right, below, etc., among images. In this paper, we introduce temporal co-orientation pattern mining which is the problem of temporal aspects of spatial co-orientation patterns. A temporal co-orientation pattern represents how spatial co-orientation patterns change over time. Temporal co-orientation pattern mining is useful for discovering tactics from play sequences of sports video data, because the most tactic patterns of basketball competition are constituted of such spatial co-orientation patterns in time order. We propose the three-stage approach, which transforms the problem into sequential pattern mining, for mining temporal co-orientation patterns. We experimentally evaluate the performance of the proposed algorithm and analysis the effect of these stages.