Position Independent Absolute Head Pose Estimation
Most surveillance cameras have a wide-angle field of view and are
situated
unobtrusively in overhead positions. For this type of application, head
pose estimation is very challenging because of the limitations of the
quality
and resolution of the incoming data. In addition, as a person moves,
their
pose relative to the camera changes, but the desired pose is the
absolute
pose with respect to the room or space. We present a solution to
estimate
absolute coarse head pose for wide-angle overhead cameras by
integrating
3D head position and pose information. The work involves image-based
learning,
pose correction based on 3D position, and real-time multi-camera
integration
of low-resolution imagery. The system can be applied to an active face
cataloguer to obtain the best view of the face for surveillance, to
customer
relationship management to record behavior in retail stores or to
virtual
reality as an input device. All demo videos are in MPEG1 format.
Top left: image fromCamera 1,
Top right: image from Camera 2,
Center left: the gauge shows in red, the relative
pose
for Camera 1 (from Camera 1’s point of view), in green, the virtual
pose,
and in blue the absolute pose.
Center right: the gauge shows in red, the relative
pose
for Camera 2 (from Camera 2’s point of view), in green, the virtual
pose,
and in blue the absolute pose.
Bottom left: the path of the person (in red) ending
with
the estimated absolute head pose (black arrow). Currently the absolute
head pose is identical to the head pose from Camera 1’s point of
view.
1. Demo1 (video
2.2MB)
2. Demo2 (video
1.7MB)
Other Research Areas:
|
Other
projects in