Creating features from scans

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Creating features from scans

    I've always felt that PC-Dmis never made it real easy to extract pieces of scans to create circles, lines etc. Its always just seemed clunky to me. I was wondering if there is a way to do something like go to construct a cir, select the scan you want to use, then click the rad on the cad and PC-Dmis grabs whatever chunk of the scan is in that selected cad element to create a circle.
    Does such a thing exist? Annotation 2020-01-16 130518.png

  • #2
    I believe you are looking to construct-> Circle -> scan segment, then select the segment data
    I wasn't aware of this until recently either

    Comment


    • #3
      Not to me knowledge.

      The only way I know:
      1) Create the scan
      2) Create a feature set of the scan
      3) Hard code the feature set to look at the points instead of centroid
      4) Create the circle or line or whatever you want out of the hard coded feature set

      Comment


      • DAN_M
        DAN_M commented
        Editing a comment
        Would this work for you??
        Instead of NUMHITS, specifying the exact points you want

        Code:
        SCN1       =FEAT/SET,CARTESIAN
                    THEO/<-1.039,0,-0.026>,<0,0,1>
                    ACTL/<-1.039,0,-0.026>,<0,0,1>
                    CONSTR/SET,BASIC,[b]CIR1.HIT[3..6],CIR2.HIT[2..7][/b],,

      • Matthew D. Hoedeman
        Matthew D. Hoedeman commented
        Editing a comment
        the hits vary as to location, UNLESS to makes it a DEFINED scan. At least, I think DEFINED will make the points collected stay in the same place.

      • louisd
        louisd commented
        Editing a comment
        If your measured radius grows or shrinks, or slides in location, the beginning/ending hits will have to be a different count/value, regardless of the total numhits or the scan strategy (defined/findnoms/master, etc). if you have product variation, the captured hits MUST be able to vary with the product, in order to grab the radius repeatably.

    • #4
      The only reason I wanted it to just look within the window of the cad radius is because we run scan programs on 2 machines, one with a B3C and the other with a DC800 controller. For whatever reason, that makes the machine collect data differently. So say I make a program, and do the segment method to pick the points out of the rad, it wont be the same machine to machine.

      Comment


      • #5
        As louisd says - you use Construct Circle from Scan Segment. Instead of selecting a CAD surface (as you suggest), you click on the *approximate* start and end points for the feature you want to extract:

        CircleFromScanSegment.PNG
        AndersI
        SW support - Hexagon Metrology Nordic AB

        Comment


        • #6
          I fully understand the concept of using scan segment, and I use it quite often, my issue is when the program moves to machines with different style controllers (again b3c and dc800) one machine will generate it perfectly, the other will generate the scan a little differently therefor throwing the constructed feature out of whack.
          I've been messing with point clouds a little now and I'm still trying to fully understand the mechanics of it but it is getting me closer to what I'm after.
          If I do a point cloud operator, and select the area where I want it to look, I think that will solve my issue.... maybe.....So if I'm thinking right the only area that will have data is in the box in the pic.... Annotation 2020-01-17 054914.png

          Comment


          • #7
            If your start and end points are far enough from the 'edge' of the CAD surface, there shouldn't be any risk of getting wrong values just because the hits are placed a little differently. At least (distance between scan points + form tolerance + machine uncertainty) from the 'edge'.

            The box you have drawn in that picture seems to try to go all the way to the points where the curvature changes direction, that may work on the CAD model, but not in reality.
            AndersI
            SW support - Hexagon Metrology Nordic AB

            Comment


            • #8
              You're searching the inflection points on the curve.
              Mathematically, it's just when the second derivate is zero.
              PC-DMIS is unable to do it, but you can obtain the result by observing the vectors evolution (that's the same than math definition !)
              If you consider the first circle as an outter, the second as an inner, the thierd as an outter, you can (I assume that the scan is in Z+ workplane) :
              ASSIGN/V1=SCN1.HIT[2..SCN1.NUMHITS].J/SCN1.HIT[2..SCN1.NUMHITS].I
              ASSIGN/V2=SCN1.HIT[1..SCN1.NUMHITS-1].J/SCN1.HIT[2..SCN1.NUMHITS-1].I
              ASSIGN/V3=V2-V1
              ASSIGN/V4=MAXINDEX(V3)
              You can create a feature set with SCN1.HIT[1..MAXINDEX(V3)-2] for the first circle ("-2" to avoid problems, maybe not necessary)

              Then
              ASSIGN/V1B=SCN1.HIT[MAXINDEX(V3)+1..SCN1.NUMHITS].J/SCN1.HIT[MAXINDEX(V3)+1..SCN1.NUMHITS].I
              ASSIGN/V2B=SCN1.HIT[MAXINDEX(V3)..SCN1.NUMHITS-1].J/SCN1.HIT[MAXINDEX(V3)..SCN1.NUMHITS-1].I
              ASSIGN/V3B=V2B-V1B
              ASSIGN/V4B=MININDEX(ABS(V3B))

              You can create a feature set with SCN1.HIT[MAXINDEX(V3)+2..MAXINDEX(V3)+V4B-2] for the second circle
              (...)

              Only tested offline.
              Hope this help.

              It's friday, happy week-end, all !



              *V1, V2 V1B and V2B are the "slope" of vectors. If you substract one from the other, you obtain the evolution. Searching the max gives a transition from outter to inner, seachin 0 gives the other change (so min(abs()) to search 0)
              Last edited by JEFMAN; 01-17-2020, 08:21 AM.

              Comment


              • JEFMAN
                JEFMAN commented
                Editing a comment
                AndersI : I don't use CAD, but vectors should be read from the CAD, no ? In this case, little defects shouldn't change the result ?

              • AndersI
                AndersI commented
                Editing a comment
                I would never trust reality being smooth enough, and at the 'edges' between CAD-surfaces, the point may have been programmed on the left surface and the real, physical hit may be on the right one. It's the same problem as in/out of tolerance - we have an uncertainty band.

              • 2013_junkie
                2013_junkie commented
                Editing a comment
                I dont typically have cad either. Someone asked me if it was possible to just pick the cad like I said.....I didnt know the answer, hence here we are. I've always felt that pc-dmis was very clunky when trying to extract scan info...

            Related Topics

            Collapse

            Working...
            X