05-08-2014 05:33 AM
My application requires fairly quick repetitive image acquisition and processing (<100ms per cycle). Kind of "realtime" processing, I'd prefer not doing post image processing.
Currently the image processing takes ~200ms per cycle (mostly grey image pattern matching).
~150ms of this time is taken by loading the template images from disk or cloning the template images (using the visionimage flatten/unflatten).
The reason for reloading or cloning is that the Algorithms.MatchPattern2 disposes of the template image during execution i.e. I must reload or clone template images every acquisition/processing cycle.
Ideally, at configuration I would like to provide 1x set of template images to the processing application which is reused (without cloning) every processing cycle allowing for a ~50ms process time.
Any fixes or workaounds?
05-10-2014 10:01 AM
Hi Nemith,
My understanding has been that none of the vision functions dispose of the image in memory unless explicitly told to do so. Does contradict what you have experienced so far?
Also, for your benefit, I would recommend posting your question on our Machine Vision forums too, where you may find more people answering: http://forums.ni.com/t5/Machine-Vision/bd-p/200.