Rexwagon;319946 wrote:
looks like someone spent way to much money on a paintball gun setup.
I dont get it. Why you would spend that much on a paintball gun. You cant use it to hunt, drive, or live in.

Do you have any idea how much of the shit we use everyday got invented just like this thing did? Someone had an interesting idea, the knowledge to carry it out and the tenacity to finish it and voila these kids create the targeting and acquisition systems on the next generation of UAVs.

Rexwagon;319946 wrote:
looks like someone spent way to much money on a paintball gun setup.
I dont get it. Why you would spend that much on a paintball gun. You cant use it to hunt, drive, or live in.

remove paintball gun, insert AR15!...Duh, Winning...

some of the algorithms are similar to what you'd see in MPEG compression, i.e. detecting an actor moving across a fixed scene. But when you allow for the reference frame to move (i.e. the camera pivots) that makes it trickier.

there are colleges in the US where EE and CS students do sentry gun software as part of their degree.

There are obviously other ways to expand on this. I.e. multiple vision systems that talk to each other. You saw the test of trying to distract the sentry platform by throwing an object and then running the other way. With two communicating turrets they could each take a target to make that type of distraction less successful. You can also setup overlapping zones of fire to make cover and evasion less effective. I.e. if there are two turrets along a line, an adversary cannot do anything but retreat along a perpendicular path to the sentries without reducing his relative motion w.r.t. one or the other. Any other direction the adversary moved would make him relatively more stationary in terms of arc of fire w.r.t. one of the turrets. This would of course highly increase the hit percentage of that sentry unit.

I've always thought it would be interesting to setup these up just as plain old cameras, perhaps with maglites. I think people -- including prowlers or drunks -- will be VERY freaked out if a machine starts tracking them with a spotlight, autonomously.

Plus, that gives you a way to test the effectiveness and understand the operational characteristics of the tracking and sensor package in a non-threatening way. Then you swap out the maglite for a response system and you're g-t-g