Title:	MULTI-PERSON TRACKING BY DETECTION IN THERMAL CAMERAS
Author(s):	Joon-Young Kwak, June-Hyeok Hong, Byoung-Chul Ko, Jae-Yeal Nam
ISBN:	978-972-8939-89-2
Editors:	Yingcai Xiao
Year:	2013
Edition:	Single
Keywords:	Multi-person tracking, person detection, tracking by detection, random forest, association check.
Type:	Short Paper
First Page:	144
Last Page:	148
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This paper presents a robust multi-person tracking algorithm by person detection using thermal images. The person tracking and detection continuously exchange information during the process each other in order to self-correct and self-update. First, a person is detected by analyzing hotspot regions in the images, and then, tracking is performed on each frame using the random forest classifier as the online learning algorithm. Second, we solve the association problem with our proposed association-check method, which considers both the spatial distance and the likelihood between target regions and detected regions. The proposed algorithm is successfully applied to a few thermal videos, including those with complex backgrounds, and the tracking results demonstrate reasonable performance.
	Go Back