So I did this project for myself and Comp Sci AP. It's basically one of those tripod sentry guns that you see in half-life, except it uses Airsoft and a webcam. A little site I made for it:

http://ringmast3r.no-ip.com/sentry/

but more importantly, the demonstration video:

http://youtube.com/watch?v=-CkRyiO6uuE

The video was made at 2:00A before the project was due, and the site was made that day, and the morning before the project was due, so bare with my horrible late-night announcing skills (made up on the spot) and any messups on the site.

Since nobody ever posts to showcases without pics:


Yes, the idea is basically ripped off of Aaron and Ezra's though all of the work is my own (besides API's and stuff), and yes I have clearly stated this fact.

Nice.

That is really impressive bro

haha thats really cool

and

i want one

To guard the banstick.

That was awesome! I was laughing out loud on the part where you get shot...

That reminds me that I forgot to mention: you can hear the servo that pulls the trigger about .1 seconds before it actually fires, which is why it misses me so many times (my leet ducking).

I'm planning on making a much better explaination video that goes through the whole targeting process, because I doubt any of you read through and fully understood that section of the site. That way, colleges may like it more.

You should make one with a real firearm to guard your cookies.

reminds me of that one episode from Macgyver... those UZIs (if memory serves) mounted on the tripods

you shouldn't shoot yourself.

Monoxide and I discussed the many uses this could have.

The most prominent ones being:

• Set it up in a toilet stall
• Set it up down the hall from someone's bedroom door
• Set it up in a closet
• Use it to shoot tennis balls for playing tennis (useful)
• Dodgeball, crank up that sucker and make it throw dodgeballs really hard
• Paintball, most notably woodsball where it could be used for something such as flag guarding.
• And...set it up in the corner of a fatty's kitchen, covering the fridge and their snack stache. Just to torture them.

Hahaha, that's quite impressive! The movement looks a bit jittery...I bet with some better build materials and a bit more time you could get it damn precise.

I think setting it up down the hall outside someone's bedroom door would be pretty effective. Every morning, just to get to school, you'd have to hone your skillz!

Awesomeness! That will keep my cat off the kitchen counter!

(until he learns how to lob grenades of course)

Jeez, that's amazing! How the hell did you get it to detect targets? Send me one!

You should put it outside your door while you're gone or asleep. It may discourage burglars.

Yeah, now I want to make one. :[

I read through your site and my main question is: What happens if you calibrate it in a room with the lights on... and then someone turns the lights off? (or vice versa) Does it just go crazy and start shooting everywhere? (I haven't seen the video)

Also, how many rounds does your SMG carry? If you're a determined ninja like me, you could just throw a pillow into the room on a stick and string and move it around to expend all the ammo and then walk right in and steal the cookies or whatever it's guarding.

I never got the obsession with guns. So I am quite unimpressed!

It's not very hard

After watching Aliens, I want a sentry gun.

Cool stuff!

Gris: In short, yes, if the entire environment changes too drastically, it will go pretty much crazy (it's funny to watch). When I had the camera mounted on the same tripod as the gun, the gun's up and down motion would give it a pretty good shake, causing this to happen. Eventually the thing would be seizuring so violently that it would either unplug its own servos, or come loose from the tripod and fall off. My mom said that was "cruel." The gun is supposed to hold up to 80 bbs IIRC.

Mega: Yah, now, it really isn't that hard. I couldn't really find any good sources at the time, so I made the algorithm up on my own (yes I realize it's basically the same as others). You have to realize when you look at that link though, it's C# .NET, in which you can basically say "Yes, please perform an erode filter" and it will do it (at least by the looks of their code). I plan on simplifying mine, then expanding on it in ways I haven't yet decided.

"B gun's dry"

Yeah this is awesome. Check out this one too where a guy made one from Lego.

I'm thinking I want a sentry gun too. Hmmm, I already have that stripped-down webcam with the motorized pan/tilt.... *pulls out a pad of graph paper and starts drawing*

/me picks up sentry gun

"Oh dear... if I put this back down, I think Alex is gonna tell me that troops are coming..."

Couldn't you just use an infrared camera to get round this?

If he had a spare bit of cash, say, $8000 laying around, he could use a FLIR camera and only target heat sources, and avoid the problem completely...but that's hardly practical, haha. It'd be awesome though.

You should make it predict movement....assuming in practical applications it might be targetting stuff that's farther away than in that video, it might be useful to have it predict movements based on current movements and lead targets and such. I dunno..if you're still looking for ways to improve it...it's pretty nifty as it is.

I've gotta make me one of those. :-P

Seriously, have you though about releasing that code and such under GPL or something? I totally want to make one now.

sum1: While an infrared camera would help the problem, it wouldn't solve it because most light sources are sources of infrared light, and any change would decrease the amount of infrared light available. However, since infrared is already in greyscale, the use of an infrared camera would eliminate the need for a greyscaling filter.

The simple solution is if too much of the current frame is different from the background, make it the background. Also, maybe make it recapture the background after being idle for a certain amount of time (say, a minute), so that something like the sun setting, or other gradual/little changes wouldn't eventually make it crazy.

Bob: Yes heat sensing would be the ultimate solution for deciding if something is alive, and therefore a target. It may attempt to shoot through walls though, so you'd have to use it in conjunction with a normal camera.
It already does simple prediction, which is why in the video you'll see the laser dot fly past me when I jump. It's just a simple thistarget+(lasttarget-thistarget)*ratio=wheretoaim kind of thing.

Spi: The code is in the Image Processing section of the writeup on the site. I'm not really proud of the code, just the results so far. The code has a lot of extra things from earlier, and parts that seem to be working by chance. I have a lot of cleaning up to do on it before I really flaunt it or expand on it.

Ah, I see that, hadn't noticed it before. Pretty cool.

Impressive.