Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

static tracking filtering #158

Open
nanodust opened this issue Dec 19, 2017 · 0 comments
Open

static tracking filtering #158

nanodust opened this issue Dec 19, 2017 · 0 comments

Comments

@nanodust
Copy link
Collaborator

I'm calibrating (single camera) OPT system - likely will expand to 2 cameras.

The default settings worked well out of the box. Very small tweaks (to distance and detection thresholds via rqt_reconfigure) helped get perfect tracking.

I save the tracks into a local mongoDB for later analysis and correlation with other sensors in the room.

Now, a few days later - the camera has not been moved, but I'm noticing there are spurious detections, a few per minute. I can tune them out by increasing track threshold, but they still sneak in, even after tracking performance has degraded.

spurious_tracks

I write as it occurs to me that the process I'm doing downstream in python (filtering single tracks that are in-place for more than 30 second window within a distance threshold) could easily be done as part of OPT.

for example:
That is, the tracker could have a 'blacklist' <x,y,z>
weight could be increment, even be assigned in RVIZ.
obvbious parameters are 'filterDistance' and 'filterTime'.
that way, the system could learn what is spurious, and auto-filter.
If the goal is to detect people, people tend to move.
The same point in the same place over hours/days/weeks is obviously not of interest.

Alternatively, if there is some means of filtering these (other than background_subtraction, or decreasing sensitivity or track detection time) let me know and I'll try.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant