posted
I'm a computer science/computer engineering student, and I've come up with an idea for my senior project in a couple years. In chapel every day, someone goes up into the rafters, stands on a walkway and takes pictures of the seats. Theoretically, someone looks at these pictures, determines which seats are empty and who they belong to, and thereby takes attendance. It doesn't seem to happen all that often, but such is the theory.
I want to automate it. The idea is to place a digital camera or cameras along this walkway. At certain times of day set by a piece of software on the network, the camera(s) will take pictures from certain positions and angles. These pictures will be sent over the network to a script which will analyze the pictures, determine which seats are empty, and update the online database of absences.
Of course, I'd like to find a pre-existing camera that has the capability to be controlled and take high-res snapshots over a network, then send those pictures back. However, my funding is not limitless, and nothing even remotely close at ANY price has yet been seen. So first question: anyone know if such a thing exists?
Failing that, I need to have the engineering mechanics guys build a mounting for a pre-existing camera. The most likely design for each camera is thus: the camera will be mounted on a motorized squveling assembly. Both the mounting and the camera will be controlled by a small computer, which will require a way to interface with the camera and the network. That computer will store instructions passed to it over the network, and execute them at the programmed times, pointing and shooting. The camera will need to be controllable over its USB or wireless interface, and the computer will need a USB host interface and a network connection, wired or wireless. My best guess right now is a Zaurus SL-5500 with a USB host CF adaptor and a hub to plug the wifi adaptor and camera into. However, most of that is just because those are pre-existing parts that can be modified relatively easily.
What I'd like to know is whether any better pre-existing parts exist that I can't find, OR whether there's a setup that I can build that would work better than this setup. Any other suggestions regarding any aspect of the project are also welcome. No sarcasm about the idea of taking attendance in chapel, though, please. We get enough of that at shcool already.
-------------------- "This is why you people think I'm so unknowable. You don't listen!" - God, "God, the Devil and Bob"
Registered: Mar 1999
| IP: Logged
-------------------- Justice inclines her scales so that wisdom comes at the price of suffering. -Aeschylus, Agamemnon
Registered: Aug 2002
| IP: Logged
Look, they're gonna take attendance anyway, I may as well save time for the people who are wasting hours doing it now. Besides, if you have any better ideas for a senior group project for EECE and EMCH majors that involves enough programming to count for a CS project, please, let me know.
-------------------- "This is why you people think I'm so unknowable. You don't listen!" - God, "God, the Devil and Bob"
Registered: Mar 1999
| IP: Logged
posted
Valerie the Virtual Girlfriend. You will need to program her personality and build the USB penis port. Fun for all except during the trial runs.
-------------------- The philosopher's stone. Those who possess it are no longer bound by the laws of equivalent exchange in alchemy. They gain without sacrifice and create without equal exchange. We searched for it, and we found it.
Registered: Mar 1999
| IP: Logged
posted
How about a program and cameras set up to scan the faces of those entering, logging them, and creating an attendance report.
This would minumized the chance of falling from said rafter while mounting the cameras there....
-------------------- "You are a terrible human, Ritten." Magnus "Urgh, you are a sick sick person..." Austin Powers A leek too, pretty much a negi.....
Registered: Sep 2000
| IP: Logged
No, but are you talking about Megapixel resolution? And your program would be a HUGE pain in the ass. Training the program to reconize the various shapes, colors, positions and perspectives of humans would be, well, really freakin' hard. That's AI stuff. Do you do AI stuff? AI stuff is cool. Where's Catman at? And if it's distinct seats it's maybe easier, but if it's pews, there again you'd likely need a miracle. You'd probably get more consistent and accurate results with a door beam leg counter. Which would probably defeat the purpose of the exercise.
As for equipment there is the above mentioned web-cam which seems an obvious choice. I know that the Phillips ToUCam is a popular hobbyist camera frequently modified by amateur astrophotographers for taking long exposures and various other custom software hacks. The X-10 system has some nice wireless cams, some with infra-red, wide angle, color, b&W, some with positionable mounts, and optional video to web capture hardware/software. But none of that's going to give you megapixels. You could spend $200,000 and get a nice HD Camera for this, but if you did I'd be forced to come down there and punch you in the stomach several times. Which leaves us with digital still cameras. My crappy 2MP digital camera does have the capacity to snap pictures at user definable intervals, but lacks the ability to download these until you're finished which would require regular pulling of the card/connecting the cable. The Nikons tend to be popular with the DigiCam Astrophotgraphy crowd.
It's an interesting idea, but it would be profoundly difficult to implement. There was that MIT thing where the robot watched cars in the parking lot. This isn't that but it's interesting. This is probably the guy to talk to.
Registered: Sep 2000
| IP: Logged
posted
Seriously, you want to automate the process of calculating who's most likely to suffer eternal damnation for lack of regulated worship?
Um... this is a CLASS. You fail to attend class a certain number of times, you fail the class. Same as every other class on campus, just... vastly larger.
Training the program to reconize the various shapes, colors, positions and perspectives of humans would be, well, really freakin' hard
Don't have to. Everyone has an assigned seat, and all seats are purple. If there's a big field of purple where a person should be, that seat is empty, and that person is absent.
Registered: Mar 1999
| IP: Logged
Cartman
just made by the Presbyterian Church
Member # 256
posted
So. Not to discourage you or anything, but 3D image recognition is still only in its infancy - the good folks over in AI #101 have basically been struggling with this (well, minus the attendance issue) problem for twenty years and even they, in all their combined wisdom, haven't cracked the nut yet. I'd start praying for some divine inspiration if I were you. B)
Anyway, what you're proposing would involve enough programming for ten CS projects. To mathematically describe a human face (which would be your first order of business) you'd have to write dedicated software that could convert a bunch of digicam inputs of your fellow worshipper to a wireframe surface and then break up that surface into smaller chunks of triangles (assuming you don't want to invest in a supercomputer to run your program) and then do a vertex-by-vertex comparison with a mesh already in your church's database (to name one popular technique) and correct for mesh deformation (something as simple as a facial expression can throw off your algorithms and fuck up the wrong person's afterlife) and take into account perspective distortions and a host of other pesky optical effects that could spell someone's doom if not filtered out - in short, the level of complexity would be way over your head alone.
Registered: Nov 1999
| IP: Logged
Cartman
just made by the Presbyterian Church
Member # 256
posted
Hmm... on the other hand, if your script only has to differentiate between empty and occupied seats because everyone's seat is predetermined anyway, it's certainly doable.
Registered: Nov 1999
| IP: Logged
posted
Hope you don't have fat people that wear purple, they would be absent while sitting there.
-------------------- "You are a terrible human, Ritten." Magnus "Urgh, you are a sick sick person..." Austin Powers A leek too, pretty much a negi.....
Registered: Sep 2000
| IP: Logged
This technology already exists. Companies use it on injection-molding presses all the time.
If a part fails to drop from the mold, and the mold closes major damage can ocure. Some companies use a system that photographs the empty mold and each cycle compares the picture of the empty mold to the image that the camara sees. It then flashes an alarm and prevents the mold from closing. Some of the advanced ones will tell you which cavity/how many parts didn't drop.
I don't have one here, but look around on the net.
Omega, PM me and I'll help you track this down.
-------------------- Like A Bat Out Of Hell...
Registered: Aug 2001
| IP: Logged
![[Smile]](smile.gif)
![[Razz]](tongue.gif)
Printer-friendly view of this topic
