eduzhai > Applied Sciences > Engineering >

Kalman Filter Based Multiple Person Head Tracking

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: For multi-target tracking, target representation plays a crucial rule inperformance. State-of-the-art approaches rely on the deep learning-based visualrepresentation that gives an optimal performance at the cost of highcomputational complexity. In this paper, we come up with a simple yet effectivetarget representation for human tracking. Our inspiration comes from the factthat the human body goes through severe deformation and inter intra occlusionover the passage of time. So, instead of tracking the whole body part, arelative rigid organ tracking is selected for tracking the human over anextended period of time. Hence, we followed the tracking-by-detection paradigmand generated the target hypothesis of only the spatial locations of heads inevery frame. After the localization of head location, a Kalman filter with aconstant velocity motion model is instantiated for each target that follows thetemporal evolution of the targets in the scene. For associating the targets inthe consecutive frames, combinatorial optimization is used that associates thecorresponding targets in a greedy fashion. Qualitative results are evaluated onfour challenging video surveillance dataset and promising results has beenachieved.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×