By Maurice Rahme
I'm trying to implement an object tracking in LabVIEW. I've written a VI (attached) to detect and locate the centroid of elliptical shapes (because my circular subject can possibly be rotated and appear as an ellipse). However, the detection is not successful on every frame.
For this reason, I would like to use the object detection method I've written as a ROI reference for an object tracker, whereby the object detector runs every 10-20 frames or so to make sure the tracker has not accumulated too much error. Any advice on how I can do this? Thanks!
Additionally, any criticism or advice on my object detection method would be appreciated.
Mostly for fun, I have concocted a way of acquiring video using Labview for linux, to mend for the unavailability of IMAQdx.
The idea is to grab via v4l2, actually through the handler libv4l2, and to display simply via the picture control.
I have tested this for the moment only with webcams at 640x480 on LV11 and 13 (both 32bits) on two different installations of Ubuntu 12.04 64bits, and it works. This means v4l-utils 0.8.6-1ubuntu2, 32bit.
Attached is a proof of the concept, essentially a translation of the Appendix E example to the linuxtv API, saved for LV11. I include also a loose set of tools I built myself in order to parse the relevant headers and construct semiautomatically the necessary typedefs (LV13, no backsave out of lazyness).
I have two questions:
-Is this insane enough to qualify for this forum, or do I have seriously seriously to exaggerate?
-Is someone interested in helping me to expand the exercise, checking portability on other distributions and versions of v4l2, other supported image formats, more robust packaging & so on? Intention would be to release a package under GPL.