ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Does somebody know what the tracking method in people_tracking_filter is?

asked 2014-12-31 03:45:19 -0500

scopus gravatar image

updated 2015-01-02 01:06:05 -0500

Hi, all. I am learning this people package by tracking legs. However, because there are no comments in the source code, I am confused about what kind of tracking method this package uses.

Is it a two-level tracking method? The low-level is by kalman filter, and the high-level is by particle filter. Each particle from particle filter is updated by kalman filter, i.e. a joint probablilistic data association filter(JPDAF)?

Thanks! @David Lu

edit retag flag offensive close merge delete

1 Answer

Sort by » oldest newest most voted

answered 2015-01-03 13:42:33 -0500

David Lu gravatar image

The approach is in fact two levels. The "machine-learning" part outputs its best estimates for each time slice of where the legs are, and a kalman filter takes the successive readings and determines a smoothed estimate over time where the legs are.

edit flag offensive delete link more


Thank you! I know that the "machine-learning" part is used for detecting the legs. The kalman fitler is used for tracking the detected legs, right? However, I think maybe JPDAF is a better method for tracking people for a mobile robot.

scopus gravatar image scopus  ( 2015-01-03 19:39:14 -0500 )edit

That's entirely possible.

David Lu gravatar image David Lu  ( 2015-01-05 07:25:29 -0500 )edit

Question Tools

1 follower


Asked: 2014-12-31 03:45:19 -0500

Seen: 360 times

Last updated: Jan 03 '15