Hi, I am new here but have been droning for 6 years or so and currently 'researching' before building my own drone for Videography.
I came up with this idea:
Touch screen vt downlink monitor, when object/person is selected, gimbal always points at that object perhaps attempting to keep it framed it centrally whilst still maintaining the stabilizing.
Rather than using cameras built in recognition would need to be standalone/open source. Laser scan in combination with camera?
The field of view would have to match the main camera lens to some extent, so might need to be selectable.
I know this kind of recognition is used in other robotic fields.
Requires some human input to select, unless perhaps you 'show it an image' of the object before hand and it has to find it and lock-on.
OK, now I think about it, this is sounding like it must be something the military have or are working on for weapons guidance.
Not only that, I can imagine a miniaturized version of system on the Google automated car combining laser/visual/map/GPS/IR/sonar etc. so that the drone can fly completely autonomously.
I'm getting carried away and it's late now, I will Google this tomorrow and find out who and what is already using this, just an idea.
Replies
http://diydrones.com/profiles/blogs/fpv-with-mouse-control-only-may...
http://docs.opencv.org/trunk/modules/contrib/doc/facerec/
http://docs.opencv.org/modules/video/doc/motion_analysis_and_object...