Object reacquisition using invariant appearance model

被引：27

作者：

Kang, JM ^{[1
]}

Cohen, I ^{[1
]}

Medioni, G ^{[1
]}

机构：

[1] Univ So Calif, IRIS, Comp Vis Grp, Los Angeles, CA 90089 USA

来源：

PROCEEDINGS OF THE 17TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION, VOL 4 | 2004年

关键词：

D O I：

10.1109/ICPR.2004.1333883

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We present an approach for reacquisition of detected moving objects. We address the tracking problem by modeling the appearance of the moving region using stochastic models. The appearance of the object is described by multiple models representing spatial distributions of objects' colors and edges. This representation is invariant to 2D rigid and scale transformation. It provides a good description of the object being tracked, and produces an efficient blob similarity measure for tracking. Three different similarity measures are proposed, and compared to show the performance of each model. The proposed appearance model allows to track a large number of moving people with partial and total occlusions and permits to reacquire objects that have been previously tracked. We demonstrate the performance of the system on several real video surveillance sequences.

引用

页码：759 / 762

页数：4