Sunday, February 14, 2010

But can you crunch the data?

Interesting post on the “Accelerated Image Processing” blog run by Jason Dale. Jason has been thinking about how camera resolutions are going up, along with the bandwidth from camera to PC, and he got to wondering how we will process all those pixels. At the risk of spoiling the tale he spins, it seems that if we run the emerging very high resolution cameras at their fastest frame rates, we will not have time to actually do much image processing at all.

But this begs the question: how many machine vision applications actually make sustained use of the highest possible speeds? Or to put it another way, do we need high speed in anything other than burst mode?

In my experience, (and I haven’t yet been everywhere and seen everything,) there’s usually some “free” time between the arrival of successive objects to be inspected. This means that while high resolution and high speed are both desirable I haven’t as yet felt the need to combine the two.

But the fact that I haven’t encountered a problem doesn’t mean that it doesn’t exist. Indeed, I think that Jason is on to something with significant implications for the future of machine vision. More pre-processing on the framegrabber? In the camera? It’ll be interesting to see how the technology develops.

2 comments:

Anonymous said...

Processing in the interface hardware: http://www.silicon-software.com/products.html

Unknown said...

The field of Hi-speed target tracking and control requires high frame-rates for one. And speed is always an added advantage.

For eg. Most n00bs don't ever use their computer's processor to its full capacity in their day to day activities.. does it mean that they should be made to stick with P3's ?