Home Applications

Discussion

Left ArrowBack to discussions page
SebastienSebastien Posts: 219 Handy

Every time I am trying to teach a part with the camera, I am always asking myself what is the best background I can use to teach. I understand that we need good contrast with the part. So my first guess is to look at the subtractive primary colors and make sure that the color of the part and the color of the background are not shared region in the circles making this pallet. So for example I used a cyan background for yellow parts. Or avoid using green on yellow for example.

Thanks to Wikipedia for explaining colors to me again! The time I did some art at school is a bit far behind me!

Any tips you pros have?

Is black and white combination the best?

Comments

  • Annick_MottardAnnick_Mottard Posts: 147 Handy
    @Sebastien you've got a part of the answer! To ensure a good part detection, you need to have a good contrast between the background and the part during the object teaching. In order to have a good color contrast for the Wrist Camera, you need to use two colors that are as far apart horizontally as possible on the HSV cone shown below. Thus, black and white are not colors with a good contrast, since it is basically the same color with a different value. A great example of a good contrast would be blue and yellow, both saturated. 

    Rsultats de recherche dimages pour hsv cone


    Annick Mottard
    Product Expert
    Robotiq
    [email protected] 
  • matthewd92matthewd92 Founding Pro, Tactile Sensor Beta Testers Posts: 1,113 Handy
    @Annick_Mottard I am a little confused by your comment about the black and white.  In most of the vision I have done in the past we always backlit the part and therefore created basically a black/white image by having a backlight that became white on the screen and the part becoming the shadow and therefore black.  We did this because 2D, gray scale cameras use the sharp change in light intensity to very clearly define edges of parts.  

    Are you saying this because the wrist camera is a color camera?  I have not had a chance to play with it yet or any other color cameras honestly.
  • Annick_MottardAnnick_Mottard Posts: 147 Handy
    @matthewd92 our Wrist Camera uses color contrasts, not variations in intensity or value of the same color. While some vision systems might use intensity contrasts - like black vs white like you mentioned - our camera detects color contrasts in order to build object models. Note that, for example, a black part on a white background might work with the Wrist Camera, but we recommend you use colors that have a good contrast between them as explained in my previous comment. This will ensure a better object teaching, resulting in a better part detection from the vision system. 
    Annick Mottard
    Product Expert
    Robotiq
    [email protected] 
  • Michal4Michal4 Posts: 1 Recruit
    Good day, I have a question about background contrast. Yes, the graph shows it well, but what if my parts are from steel material like this on the picture? What do yout think is the best color for background? Now i think something like red or green, I dont know.. 
    Picture -> https://images.homedepot-static.com/productImages/f9a426d3-a939-4261-807c-dc8cfd26f7d2/svn/everbilt-flat-washers-807210-64_1000.jpg
  • JClaytonJClayton Posts: 1 Recruit
    I remember learning this in the training.
    Object Teaching
    For reflective objects, you can spray matte paint on it for the object teaching. For reflective objects or objects with too many features, you can print a drawing of the object, with the correct size, with only the features you wish to teach. You can also select only the area of the part you want to teach.

    Source: http://support.robotiq.com/display/RVS/5.1+Guidelines+on+Object+Teaching 

    So, for your steel part, it may be easiest to print out the CAD or to cover it in matte paint. I was also thinking blue painter's tape over the surface with a yellow background may allow you to get that key contrast.

    Hope this helps!

Sign In or Register to comment.
Left ArrowBack to discussions page