Adapting Object Tracking to a Dynamic Environment

CHAU Duc Phu
17.04.2013 - 14:00
Valid
Object tracking quality usually depends on context features of scene (e.g. scene illumination, density of mobile objects). In order to overcome this limitation, we present a new control approach to adapt the object tracking process to the context variations. In an offline phase, training videos are classified by clustering the contextual features to create context clusters. Each context cluster is then associated to satisfactory tracking parameters. In the online control phase, once a context change is detected, the tracking parameters are tuned using the learned values. This approach brings twocontributions: (1) a classification method of video sequences to learn offline trackingparameters, (2) a new method to tune online tracking parameters using scene context. CHAU Duc Phu is a postdoctoral researcher in STARS team, INRIA Sophia Antipolis, France. He is also a lecturer at Technology Department of Phu Xuan Private University in Hue city, Vietnam. In 2008, he obtained his Master degree with specialization in "Artificial Intelligence and Multimedia” from The University of La Rochelle in France and The Francophone Institute for Computer Science in Vietnam. Since 2009, he has conducted research on adapting object tracking to cope with various video scene contexts. In 2012, he obtained his PhD degree from INRIA Sophia Antipolis, France. CHAU Duc Phu now focuses his research on the robust and online learning-based methods for tracking objects in several hours and days.