euhlmann
28-03-2016, 22:58
We're trying to improve our NI Vision based vision. We decided to cut the resolution of our image processing frame from the capture resolution of 640x360 to a smaller resolution of 320x180. We hypothesized that this would make our (currently very CPU-intensive) vision run about 4x faster.
The issue is, the code we use that was generated from Vision Assistant refuses to take a scaled image. Here is our NI Vision Assistant generated file: https://github.com/ligerbots/Stronghold2016Robot/blob/master/src/Vision/ImageProcessing.c
And here (https://github.com/ligerbots/Stronghold2016Robot/blob/master/src/Vision/ImageProcessing.c#L198) is the call that we identified is failing.
The code we used to scale the image before passing it to the generated IVA_ProcessImage() was
imaqScale(processing_frame, camera_captured_frame, 2, 2, IMAQ_SCALE_SMALLER, IMAQ_NO_RECT);
we also tried this instead, and it didn't help
imaqResample(processing_frame, camera_captured_frame, width/2, height/2, IMAQ_BILINEAR_FIXED, IMAQ_NO_RECT);
The issue is that with the scaled image, the call to imaqColorThreshold is causing VisionErrChk to immediately jump to Error and exit the processing pipeline.
A call to imaqGetLastError() right after the failure returns 21.
Unfortunately, 21 is not in the documentation as an error code, and imaqGetErrorText(21) returns nothing.
We did find a solution (capturing from the camera in lower resolution to begin with) but we still want to know what's going wrong here, if anyone is able to give some insight.
The issue is, the code we use that was generated from Vision Assistant refuses to take a scaled image. Here is our NI Vision Assistant generated file: https://github.com/ligerbots/Stronghold2016Robot/blob/master/src/Vision/ImageProcessing.c
And here (https://github.com/ligerbots/Stronghold2016Robot/blob/master/src/Vision/ImageProcessing.c#L198) is the call that we identified is failing.
The code we used to scale the image before passing it to the generated IVA_ProcessImage() was
imaqScale(processing_frame, camera_captured_frame, 2, 2, IMAQ_SCALE_SMALLER, IMAQ_NO_RECT);
we also tried this instead, and it didn't help
imaqResample(processing_frame, camera_captured_frame, width/2, height/2, IMAQ_BILINEAR_FIXED, IMAQ_NO_RECT);
The issue is that with the scaled image, the call to imaqColorThreshold is causing VisionErrChk to immediately jump to Error and exit the processing pipeline.
A call to imaqGetLastError() right after the failure returns 21.
Unfortunately, 21 is not in the documentation as an error code, and imaqGetErrorText(21) returns nothing.
We did find a solution (capturing from the camera in lower resolution to begin with) but we still want to know what's going wrong here, if anyone is able to give some insight.