3D head pose estimation and tracking using particle filtering and ICP algorithm
Type
Conference PaperKAUST Department
Electrical Engineering ProgramDate
2010Permanent link to this record
http://hdl.handle.net/10754/564256
Metadata
Show full item recordAbstract
This paper addresses the issue of 3D head pose estimation and tracking. Existing approaches generally need huge database, training procedure, manual initialization or use face feature extraction manually extracted. We propose a framework for estimating the 3D head pose in its fine level and tracking it continuously across multiple Degrees of Freedom (DOF) based on ICP and particle filtering. We propose to approach the problem, using 3D computational techniques, by aligning a face model to the 3D dense estimation computed by a stereo vision method, and propose a particle filter algorithm to refine and track the posteriori estimate of the position of the face. This work comes with two contributions: the first concerns the alignment part where we propose an extended ICP algorithm using an anisotropic scale transformation. The second contribution concerns the tracking part. We propose the use of the particle filtering algorithm and propose to constrain the search space using ICP algorithm in the propagation step. The results show that the system is able to fit and track the head properly, and keeps accurate the results on new individuals without a manual adaptation or training. © Springer-Verlag Berlin Heidelberg 2010.Publisher
Springer NatureConference/Event name
6th International Conference on Articulated Motion and Deformable Objects, AMDO 2010ISBN
3642140602; 9783642140600ae974a485f413a2113503eed53cd6c53
10.1007/978-3-642-14061-7_22