|
CamShift Object Tracking (Composition by Peter)
This sample composition demonstrates the CamShift object tracking method. In this example an initial Haar detection will be performed to define the tracking area which get passed over to the CamShift tracking. The tracking results are send through a Kalman filter to stabilize and predict the object movement (e.g. when the object is covered or only partly visible).The tracking results will be constantly re-evaluated according to a given threshold and the CamShift parameters automatically adjusted. Still a lot of work in progress and plenty to optimize... The attached CVDetection.plugin will need to be copied to (your user folder)/Library/Graphics/Quartz Composer Plug-Ins first |
Must be great ! Thanks for sharing. Unfortunately it crashes on main thread on my MBP 13", with 10.7.3. Here are the details (I have a cvError on line 9) :
Thanks for the info.
Unfortunately I am unable to reproduce this under 10.7.3 (Quartz Composer 4.5 (141.5).
It seems to crash during the initial Haar detection but I cannot pinpoint just yet why. Do you have any other OpenCV plugins loaded which might interfere?
your example runs fine for me in 10.7.4.
any chance on sharing your source with me ? dustin.oconnor@me.com your cv projects interest me peter. actually so does your blob tracker @benoitlahoz so the same goes for you as well on the blob tracker.
i got this working as per your example which was needed. i like the idea of the haar tracker defining an initial region for the cam shift. are you building a cascade on the fly with the cvdetection ? i have been getting interesting results without cvdetetion as well or at least better results than the cam shift cv example. kudos.