New Vision Algorithm for MaxLine/MinLine

Kyle Brummans 5 years ago in Metrology Software / PC-DMIS updated by neil kay 4 years ago 7

Similar function to min_circ and max_inscr of a circle.

Potentially using Std.Dev. and allowing to use points only in +% or -% of Std.Dev.

Ex.  I'm using backlight only on gasket material.  It would be useful if
I could Snap the line to the points that are either a minimum or
maximum (depending on material side) To avoid calculations with debris.  In this example lets assume a minimum edge is what I'm looking for.  I select to eliminate 75% of points on the positive side of material (75% of the standard deviation).  I'm then left with 25% of the points to construct a line out of, of which are ignoring 75% of the imperfections seen on the surface.

This is not an extreme example, however even in this scenario - I would have the ability to push the line closer to material and ignore a larger portion of the rough edges.

It would appear to me that this feature would be quite advantageous when working with additive parts... and not just for vision applications.

Thank you for the suggestion. I am trying to understand the requested functionality in better detail. 

In the example image, the rough edges can be filtered with our Outlier Filtering using both STD Dev and Distance which would have the affect of bringing the line closer to the material. I have attached an example and a description from the Help.

Distance Threshold (Outlier Filter) - This specifies the distance in pixels that a point can be away from nominal before discarding it.

Std Dev. Threshold (Outlier Filter) - This sets the standard deviation that a point needs to be away from the nominal CAD to determine if it's an outlier.

If this does not cover your request,  we can discuss further. My contact details are neil.ryan@hexagon.com

Neil, I'll try to see if I can illustrate some examples and send them your way.

As Neil stated above, this is typically what I do. I adjust the filtering to disregard the outliers. 

No doubt, Barry.  There are applications where I do this as well. 

However, if the edge looks like this - you have two options. (As far as I can tell at the moment.)

  1. Clean the edge (on a lot of surfaces potientialy) which also could remove material.
  2. Get inaccurate results.

I've applied every possible filter/light/edge setting combination I could think of and cannot capture what I am after.  In the above pictorial the Edge would be sampled from Dark To Light (←), the blue dots are located NEXT TO  the white spaces I would want sampled. In this scenario I would call it a Minimum-Material Fit Line (If it were a canned option)  or I would have settings that allowed me to adjust where in the array of points (indexed on a histogram) what percentage and location I want to sample points from. 

I could even be interesting to have it be conditional.

  1. Where if the range was high and std. dev. was high (perhaps based on pixel distance), it could reduce the percentage used.  Ex. a heavily flashed edge as pictured directly above.
  2. Where if the range was high and the std. dev. was low, it would increase the percentage used.  Ex. An edge with minimal flash as pictured initially.

Albeit, the more controls you offer - the less users will understand how to operate them.  It would probably be useful at this point to have two types of autolines.  Suggested use autolines and full control autolines.