A guy and I are designing a trapper's camera, something that you can strap onto a tree.
If motion is detected, the camera starts recording to file. But, you don't want to record the background branches swaying in the breeze.
What method would you use to differentiate between recorded and non-recorded frames of video?
I was thinking of using an artificial neural network, where the training data is gotten from a forced recording. Another method would be to get the large contour areas in the motion detection frame.