Sunday, March 29, 2009

Software selection: tool capabilities

To conclude this short series of postings about how to select the best software package ...

Perhaps the hardest part of evaluating machine vision software packages is comparing the tools available. Sure, they all have a range of filters, edge detection tools, blob detection and so on, but it’s not until you come to develop an application that you uncover particular weaknesses.

For example, sometimes you might need to rotate a Region of Interest to accommodate the way a part is presented – you’ll find not all packages handle this the same way. Likewise, you may need to mask a complex shape: again you’ll find this is addressed in different ways. OCV and OCR can also seem straightforward, until you come to use them at which point their limitations are revealed.

So how do you pick the ‘best’ package for your needs? Given that you probably don’t want to learn multiple products any more than the boss will want to pay for them, and your crystal ball can’t show you what projects will pop up in the next few years, you’ve one chance to get it right.

First, weigh up the factors we discussed previously: hardware flexibility, pricing and support, and ease of use. When you work with evaluation versions to assess ease of use you’ll also start to learn what functions are available to you, and more importantly, if there are any you need that aren’t present.

Second, I suggest limiting yourself to the widely used packages, and preferably those that are sold to end users as well as developers/integrators. (Most vendors will gladly tell you if theirs is an end user product – if it isn’t they don’t the aggravation of your endless calls to their Help Desk.) This means starting your search with the packages I listed in “Software selection: ease of use.”

That’s about all the advice I have to offer on choosing your vision software. Now go shopping!

No comments: