Quote:
Originally Posted by Jim Giacchi
Sure, I probably couldn't use any fancy image processing, but my team can never figure out how to program that anyway...
|
From what I've found, most vision processing we have done has seriously compromised every other machine function, because the processor just isn't powerful enough to handle vision and other things. Which would require us to dedicate a 10th of a second or so to just processing the image, which is far too long. We tried it all three years the cRio has been around, every time we've abandoned it because of resource consumption.
On a related note, the CMUcam that the IFI system commonly used worked fairly well in testing for us this year. We didn't have the weight or need for it, but it did work a lot better than the Axis cam. Why? Because the processor dosen't do any work, the camera just feeds it a few numbers over a serial port.
I think something like the IFI system with a more powerful processor would be just awesome. Just enough more to allow a minimal OS with threads, and enough math for floating-point and a few trig operations per loop, but not enough more for them to fake us into thinking we can do vision in real-time.