Using Visual C++ to detect object and color?

[yol]

Hi ppl,

Can I use Visual C++ to com out with a source code to detect object like circle triangular round or square or cube either in 2D or 3D?

Can I detect the color of the object also? Both is using a input from a stereo vison of a camera.

Thanks for your help..

[jonnin]

You can do almost anything that can be done with c++. The major exception is its not scripted so you cannot embed the source in a web page for example (or similar script/portable applications) -- because you cannot expect the typical PC user to have a compiler installed, for one thing.


C++ itself does not have any graphics, you either have to make your own or use existing ones. There are plenty of great graphics libraries out there, with visual, you might use directx / direct3d etc or opengl.

Stereo vision is extremely complex, and C++ is a very rich, complex language. If you do not know the language, it can take many months to become skilled enough to express complex ideas in it. Stereo vision is of limited use, really, you can never really know an object is a cube. Just take a few objects from your dest to prove this: pick up a book, and look at its cover from an angle where you cannot see its "depth" (cannot see the sides). Is it a cube? A rectangle? A plane (zero depth?)? You cannot tell. Pick up a soda can, look at its bottom and you cannot see the sides, is it a sphere? Your brain tells you the answer by context (you recognize a soda can and the brain fills in its probable shape for you) but from the image alone, its not so simple. The farther apart your cameras are, the better your answer will be *to a point* but the harder it will be to correlate the images to make sense of them. THere is a happy medium where they are far enough apart to do both tasks, but the results are generally less than stellar and tend to only really work well in a very heavily controlled environment (always looking at the same sort of things from the same angle and distance etc).




Color on the other hand is very easy to get (from a programming standpoint only). There is no detection, you simply know the color of whatever pixel you are looking at. However, for an object from a camera with lighting and all that, you only know the percieved color and even that can be obscured by shadows and such --- but you can get a sense of its color and, with some smarts, detect shadows or excessive lighting (reflections, washout, etc). I don't know if you can ever get the true color of the object but you can get close enough. For example, if the sun is shining on a shiny, bright yellow cube, it may appear white in the video streams and therefore you will never see the yellow.