DMCA
Multi-object Motion Pattern Classification for Visual Surveillance and Sports Video Retrieval (2002)
BibTeX
@MISC{Hayashi02multi-objectmotion,
author = {Akira Hayashi and Ryuji Nakashima and Toshihiko Kanbara and Nobuo Suematsu},
title = {Multi-object Motion Pattern Classification for Visual Surveillance and Sports Video Retrieval},
year = {2002}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
This paper presents a method to classify scenes based on motion information. While they use object trajectories or optical flow field as motion information in previous work, we use the instantaneous motions of multiple objects in each image. In order to deal with variable number of objects in a scene, we use moment statistics as features. Our approach is based on clustering, a form of unsupervised learning, and needs little human intervention. Furthermore, the probabilistic model based clustering makes it easy to detect scenes with novel patterns.
Keyphrases
sport video retrieval visual surveillance multi-object motion pattern classification motion information previous work unsupervised learning probabilistic model moment statistic variable number novel pattern little human intervention multiple object object trajectory optical flow field instantaneous motion
