|
The applet to the right illustrates the regions of space that is viewed by
both the left camera and right camera in stereo vision. As can be seen,
a flat object would appear to be conical if the hard pixel coordinates were
used to compute the location of the object. It is for this reason and others
that some
form of fuzzy logic MUST be used in stereo vision.
I used my Madazula program to try to experiment with different approaches at stereo vision. I tried out more than 10 approaches so far. All of the approaches I was using involved trying to make matches pixel-by-pixel, but now I think that matches should be made by the outlines found in the images. I think that if outline matches were made it would considerably increase the speed of computing the depth of surfaces. In fact, each surface would be an outline. I plan on updating my program to C++ and then start working on the outline matching algorithm. I have programmed a slide show using my own uigui library to display some images of the device that I built to take stereo image pictures. I called the device SID (Stereo Imaging Device). Click here for SID Pictures |