Tracking

Models for low-level tracking are basic image primitives (i.e. blobs) while models for high-level tracking are coherent, structured sets of specific primitives. Complex models are gradually generated from basic models. For instance, the blob corresponding to the subject is detected, then the possibility of segmenting this blob into body parts is tested. If the test result is positive, segmentation is performed and may further continue inside the body parts (i.e. identifying the fingers of a hand, eyes in a face, etc.), thus refining the model as much as possible.

Low-level tracking


RealVideo : MPEG : AVI

High-level tracking

All available information at the time of acquisition, e.g. the calibration and synchronization parameters of the cameras, is used in order to produce the 3D temporal description of a person, for a multi-camera sequence. The high-level description includes lenghts, pose and position.

The parameters of a skeletal representation – the model – of a person are estimated in 3D space. Feedback from the model in each image allows a simple segmentation to isolate feature points on the subject. The invariant elements of the description – the limb lengths – are filtered to produce the final result. An approach based on the extended Kalman filter is used for this purpose.

The 3D description of a person is generally made by fitting a model specific to each individual. The challenge here is to propose a method using a generic model of a person whose invariant elements are estimated by observation. To ensure the robustness of the segmentation to occlusions is also a challenge.


RealVideo : MPEG : AVI

RealVideo : MPEG : AVI

RealVideo : MPEG : AVI

RealVideo : MPEG : AVI

RealVideo : MPEG : AVI (1)

AVI (2)

RealVideo : MPEG : AVI


(1) Hungarian matching on area and position; observation variance of area = 2500 pixels4.
(2) Hungarian matching on area and position, then Hungarian matching on position for resulting matches; observation variance of area = 900 pixels4.

RealVideo coded at 320x240, 220 kbps.
MPEG coded at 352x240.
AVI coded with Microsoft Video 1 at 640x480, 4800 kbps.