Data association problem in tracking
in [DSP]
|
Prev: What is the standard output of a low pass digital filter design program?
Next: Data association problem in tracking
From: ombz on 4 May 2010 09:26 Hi all. I want to track multiple objects on a 2d plane. Now I am having quite some problems getting the data association (labeling) part working correctly, i.e. assigning found objects (measurements) to targets. I think I lack some methodology. First: objects can appear and disappear at any time, i.e. there is a variable number of targets and a variable number of measurements at any time. Second: there can be a non-significant number of false measurements - this has nothing to do with a poorly written detection but rather with the underlying data itself... I've quickly read about "linear assignment problem" and "joint probabilistic data association filter". But if I've correctly understood I cannot use the linear assignment approach since number of measurements and targets are both variable and not equal most of the time. Is there some kind of extension for the linear assignment that I could use in to solve my problem? The JPDAF approach seems quite complicated. I think much too complicated for my problem since the objects to track look all the same, the only thing that differ are position and color, and maybe the velocity of the objects. Could you recommend me some simple method to solve the problem? Or some introductory material to the subject? Thank you very much |