0

More Vision Algorithms

Kyle Brummans 9 months ago in Software / PC-DMIS updated by neil.kay 6 days ago 10

I've seen competing software that has no problem with edges I have problems with in PC°DMIS.


Whenever I'm on a vision project, I always struggle to get edge parameters correct until lots of trial.


Nearest Nominal and Matching Edge are sometimes very amazing and work quite well, but often there are program parameters and changes that are not feasible with those tools (at least not with out lots of work, meaning if I quote for the struggle I lose the work anyway).


Can we develop a plan for implementing better algorithms?  I can, and possibly the rest of the community, can submit difficult tasks to measure and submit them for review.

Hand drilled prototype templates in MDF, 3mm hole 1.75 millemeters out of position.  That is [Dark] to [light] center of the circle to outside dominant edge.  Usually the most useful for this project.  

Things like,

-'ignore first edge' or ignore first 'x#' of edges could be helpful

-remeasure with different algorithms

-locate relative center, then remeasure

If I could automatically: measure the circle as a blob, then set that as the origin of the circle to re-measure at nearest nominal then I would probably cut out 90% of my guess work and trial and error on this job.

Measuring it out of focus got me close to a specified edge result.  Unfortunately using specified edge, I need to re-specify each edge based on edge quality and verify each hole. 

Hi Kyle

If it is possible, email me the images without any overlays and I will review in our simulator

Cheers

Neil

Neil,

I've since completed these jobs, and did not save any raw-image captures.  -Next time-

Hope you have a great weekend!