Segmented Velodyne scans[1] are tracked by associating centroids[2].
Identities of tracked objects become ambiguous after interactions.
An Event Graph [3] is built to model the interactions.
Appearance information (e.g. from camera) is used to resolve the ambiguity.
Douillard, B., Underwood, J.P., Kuntz, N., Vlaskine, V.,
Quadros, A., Morton, P. & Frenkel, A. "On the
Segmentation of 3D LIDAR Point Clouds", IEEE International
Conference on Robotics and Automation, 2011
Morton, P., Douillard, B. & Underwood, J.P. "An
evaluation of dynamic object tracking with 3D LIDAR",
Australasian Conference on Robotics and Automation, 2011
Torabi, A. & Bilodeau, G.-A. "A Multiple Hypothesis
Tracking Method with Fragmentation Handling", Canadian
Conference on Computer and Robot Vision, 2009