Speeding up image analysis,

[hammersmith]

I am using a VI to analyse an AVI file showing liquid particle flow

I want to acheive this in real time, by using the video capture card, however the system is far too slow to acheive this

I am using colour matching, and doing some binary operations to identify particles from their background.

im using a pentium 4 3ghz, with gig of ram, but this still runs slow

how can i speed up the vi?

also, i would like to use an overlay function to show the particles in colour, by removing the background using binary operations, i.e. seperating the particles from their background in binary, then retriving the colour image and only displaying pixels that are set to logic one

[AnalogKid2DigitalMan]

What are the specs of your video capture card? Could it be too slow for realtime acquisition and transfer to memory and thus the bottleneck?

Is color needed, can you use colored filters, enhanced lighting, or use narrow band light sources to increase the S/N ratio so software has to do less processing?

Can you do an ROI so that you do not need to process the entire image frame?

Do you have enough pixels/particle for sufficient resolution.

Are you using NI Vision Builder? If so, you can benchmark the execution time of each individual vision function. It is an awesome package to do vision prototyping, fooling around, and generating LV code.

Machine vision is a science and an art with a good measure of experience and luck thrown in.

[hammersmith]

Hi, thanks for your reply

to answer your question, the capture card is definatly not the bottle neck, as i have the same problem when analysing a pre recorded avi file for development purposes.

i am using vision assistant to build the base code

i could use a roi, however i have attempted capturing only a small area, 100 x 200 pixels in an attempt to speed the program up

as far as i can tell, the operations carried out are too processor intensive to be acheived in real time?, really not to sure as i have little experiance with labview

[AnalogKid2DigitalMan]

I don't havs Vision Assistant, but in looking at the description for it at:
http://www.ni.com/pdf/products/us/pg596_600_vision_dev_module.pdf
it includes a Performance benchmarking Tool that will report the speed of each function. I would start there to see what the bottleneck are and to see wha the frame rate is.

Again, I would not overlook other aspects such as lighting and pixel resolution to help improve the S/N ratio so software does not have to spend excessive time processing.

Good luck

[hammersmith]

Hi i have utilised the benchmarking tool within vision assistant and the results have shown that i should be able to atain 200 fps, where as i only require 30 fps

at the moment i am reading the data from an avi file so that i can develop offline, not sure if i am doing this efficiantly, or weather image aquisition from a camera will be faster

[AnalogKid2DigitalMan]

Is not 30fps the standard frame rate of video and most analog cameras?????

[hammersmith]

yes, however, the system will only process an avi file at around 3fps, that is why i need to speed up processing to be able to acheive it in real time, as yet im not sure weather or not live aquisition through a capture card would be any faster, however i doubt it very much

[aljuboori]

Hi, I use Vision Builder AI Version 3.5, I need to make fixable selection for ROI (by mouse), Please , how can use (create region of interest step or any other steps) to do selection by mouse when I execute the inspection Thanks Haider