Loading [MathJax]/extensions/TeX/ieee_stixext.js
Tracking interacting targets with laser scanner via on-line supervised learning | IEEE Conference Publication | IEEE Xplore

Tracking interacting targets with laser scanner via on-line supervised learning

Publisher: IEEE

Abstract:

Successful multi-target tracking requires locating the targets and labeling their identities. For the laser based tracking system, the latter becomes significantly more c...View more

Abstract:

Successful multi-target tracking requires locating the targets and labeling their identities. For the laser based tracking system, the latter becomes significantly more challenging when the targets frequently interact with each other. This paper presents a novel on-line supervised learning based method for tracking interacting targets with laser scanner. When the targets do not interact with each other, we collect samples and train a classifier for each target. When the targets are in close proximity, we use these classifiers to assist in tracking. Different evaluations demonstrate that this method has a better tracking performance than previous methods when interactions occur, and can maintain correct tracking under various complex tracking situations.
Date of Conference: 19-23 May 2008
Date Added to IEEE Xplore: 13 June 2008
ISBN Information:
Print ISSN: 1050-4729
Publisher: IEEE
Conference Location: Pasadena, CA, USA

I. Introduction

Multiple targets tracking plays an important role in various applications, such as surveillance, human motion analysis, traffic flow analysis and others. Multi-target tracking is much easier when the targets are distinctive or not interacting with each other because it can be solved by using multiple independent trackers. However, for the targets similar in appearance, obtaining the correct trajectories of them becomes significantly more challenging when they interact. Specifically, for the laser-based human tracking systems [1]–[3] (as shown in Fig. 1), which cannot provide the color information of targets, maintaining the correct tracking seems to be an impossible mission when the well-known “merge/split” condition occurs (as shown in Fig. 2). Hence, the goals of this research are: 1) to devise a new method that will help obtain a better tracking performance with laser scanner than those obtained from previous methods when the interactions occur; 2) to make a new attempt to solve the “merge/split” problem in the laser based multi-target tracking.

References

References is not available for this document.