Hi,
I am just starting Computer Vision, which is why my question will not be very specific. We have at our disposal a set of videos that were recorded from birds' necks; the point is to learn more about behaviour of a certain species. There are in principle three actions to classify:
- Bird is flapping its wings, which characterised by rapid camera movement: flapping example
- Bird is gliding, the camera moves relatively steadily: gliding example
- Bird is in a colony, again no rapid movement, but also amount of blue is much less than in the point above: colony example
On top of that there are also videos that are just messy due to a camera that was fogged up or moved to a position that limits field of view to the bird's feathers (example).
Can you give me some ideas what algorithms from OpenCV I could use to discriminate flapping (1) from gliding (2)? Since we are talking about birds that spend their day at the see, I guess distinction between gliding (2) and colony (3) should be simple by checking histogram content, right?
Any thoughts about finding the instances in which camera was fogged? Histogram again perhaps? Gray tends to dominate there.
