Authors:
Xiaoming Peng
1
and
Mohammed Bennamoun
2
Affiliations:
1
University of Electronic Science and Technology of China and The University of Western Australia, China
;
2
The University of Western Australia, Australia
Keyword(s):
Facial point cloud data, Geometric computing, 3D video.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Computer Vision, Visualization and Computer Graphics
;
Geometric Computing
;
Geometry and Modeling
;
Image-Based Modeling
;
Pattern Recognition
;
Software Engineering
;
Surface Modeling
Abstract:
Point cloud data processing is an important topic in geometric computing. One promising application of point cloud data processing is 3D face recognition. With the recent developments of 3D scanning technology, the emergence in the near future of 3D face recognition from 3D video sequences is eminent. Face tracking is a necessary step before the recognition of a face. In this paper, we propose the integration of a nose tip detection method into the process of tracking the face in a 3D video sequence. The nose tip detection method which does not require training nor does it rely on any particular model, can deal with both frontal and non-frontal poses, and is quite fast. Combined with the Iterative Closest Point (ICP) algorithm and a Kalman filter, the nose-tip-detection-based method achieved robust tracking results on real 3D video sequences. We have also shown that it can be used to coarsely estimate the roll, yaw and pitch angles of the face poses.