Dear all, I'm getting video from a cooling chamber through an axis webcam that has some motion detection triggers set up. Those triggers work flawlessly except that they obviously but wrongly fire when someone turns on or off the lights. I've hooked up a light sensor to a raspberry and now know when the lights were turned on or off and mark these videos.
But there's a huge problem: Being in a cooling chamber, the neon lights take about 40 seconds to settle down when they are switched on after having been turned off for the night. And unfortunately 40 seconds is long enough to walk in there, take some stuff out of the racks closest to the door and vanish again. So I'm looking into using opencv to specifcially scan those videos for trespassers while not getting distracted by the illumination change. I've looked into http://answers.opencv.org/question/29088/motion-detection-false-alarms-from-light-changes/ but don't know if I can transfer this to my problem. Background subtraction seemed a good idea but above post made me reluctant to try it... Anyone came across a similar challenge and found a suitable solution?