Live video compression?

[MrWibbles]

We want to have the robot send the camera's image, live, to the Classmate, but we're worried about bandwidth. What algorithms/formats should we use for compression? Is there anything in Labview for it?

Thanks in advance, AdamS

[Joe Ross]

Is there something wrong with the camera image that is sent to the dashboard by the default code?

[Bryscus]

We use WindRiver and C++ and looking through the code there is a member function to set the compression automagically. I'm not sure how well it works though. I know for sure that the camera is slow when it's not compressed (the default code does not appear to compress the image at all). We may get a whopping 5 fps if we're lucky... But look around, there should be a simple way to do it.

- Bryce

[CrazyCarl461]

I don't know a ton about the vision system, but I noticed last year with the analyzer tool that the part of the video processing that took the longest was a VI that converted the jpeg provided by the camera into a character matrix to be used by the image processing VIs. When that was going on we only got a few FPS reliably as well. If you didn't need to image process (only wanted to look at it) I suppose that conversion would never need to happen. I'm not sure if the video it sends to the dashboard is the raw jpeg or the matrix, but either way it might be waiting for that conversion to finish.

Anyone have a better idea of exactly what it does?

[jhersh]

Originally Posted by CrazyCarl461

If you didn't need to image process (only wanted to look at it) I suppose that conversion would never need to happen. I'm not sure if the video it sends to the dashboard is the raw jpeg or the matrix, but either way it might be waiting for that conversion to finish.

The PCVideoServer simply forwards the jpegs that were compressed by the camera. They are not recompressed and they are not sent raw. Your best bet is to adjust the compression / image size that the camera is sending, not trying to recompress on the cRIO.