Hey all, I'm planning on making a depth perception program for my comp. Basicly the possibilities are endless. My question to you guys is: What would you think could allow me to use two webcams (which i interface into my java programs and run both at the same time), pointed at angles based on their viewing angles, to get somewhat of a 3d model?

 

I'm talking the actual algorithm aspect of it, not how to program 3d or get webcam in put.

 

Does anyone know what kind of calculations would let me do such a thing.

 

It could be AMAZING.

 

Looking forward to see what you guys think

The biggest problem with this would be to recognize which points on both pictures do correspond. I wouldn´t know how to do that. You could try an approach with comparing colors (I´d compare then on similarity not on total equality). This might work for objects where distinct faces have an own color each. For a human face I´d not expect it to work (it might do so though but I doubt it). Also keep in mind that certain points are only seen by one of the cameras. You could either ignore them for your 3D pic or extrapolate them.

 

Assuming you have identified points that are seen by both cameras constructing the 3D position of them is fairly easy. For each camera you know that the points lie on a certain line in 3D (a ray cast into space). Take the intersection of the lines (or the point of closest approach to avoid problems with inaccurancies) and you have the corresponding 3D point.

 

While the reconstruction is fairly easy and can be programmed by anyone with a little skill in programming I´d expect the recognition to be very complex. So unless you´re a really good programmer (well, you probably wouldn´t ask for advice if you were) or have fun with experimenting on code and being happy with non-optimal progs I think you´d best search the net for informations on that. Alternatively you could try a programming forum since to my experience there are a lot of competent people around there. I used to post on www.gamedev.net some time ago which is a pretty good forum. They are dedicated to game developement, however, so it´s possible they won´t like your question there.

are you saying i'm a bad programmer??

 

Yeah i was thinking of the neural network becaues it could be pretty good for the point recognition. I recently made a program for my friends neural network to give it a webcam to see with.