Vision, cameras, filter , lighting choices

    Hi
    I'm looking for information on camera filters and lighting.
    The robot brand is not really important, not even the camera.
    I know there are many more variables such a thickness of the part, contrast, etc, etc
    The answers I'm looking for a very generic. At least they will put me in the right direction for the proper setup. Maybe somebody can tell me what kind of set up they have on their shop.


    Examples
    What filter and lighting should I use that gives me the best picture to find the parts ?
    1)Opaque blue parts running on a red background.
    2)Clear part on a white background.


    I could name many more combinations, Do you have one that is working ?

    Retired but still helping

    This is an old thread but as i stumbled upon this i might as well answer... the simple answer regarding choice of colors for foreground vs background is take a hsv or hsl color space "cake" and choose color and brightness as opposite to each other as possible.


    This makes filtering and thresholding images far easier.


    And ofc use the applicable colorspace for your setup and application. If you are using a monochrome camera you look at maximizing the diffrence in grayscale.


    Regarding lighting as an extremely short answer. Make it as even as possible, and avoid shadows as far as possible. Normally you just want to get a regular whitish lighting that is powerful enough to outmatch any discrepanicies or circumstances. This so when that one light bulb goes dark, a door is left open, a big dude leans against the safety fence, the sun shines through a specific window on the 15th of may, etc.. your vision application does not implode. In case you really need to accentuate something specific choosing lighting colors to either tone up or down specific details is a good way to go. Lighting and intial setup will always be key. You can get around alot of problems with postprocessing but if you put alittle more effort in making the setup as good as possible you will save alot of hassle.


    As mentioned this is just pure basics with alot of stuff left out. I would suggest you also read up on subjects such as focus distance and aperature settings (especially how aperature affects focus distance, alot of people do not seem aware of this), shutter types and shuttertimes, hdr imaging, stereoscopic lighting, low angle and backlighting, calibration and methods, lenses and chip sizes, etc. Before moving on to postprocessing techniques (a good knowledge of thresholding and color spaces is a must have basis), edge detection (sobel, prewitt, hugh, laplacian of gaussian...) and then studying the actual implementations of different pattern matching tools and so on. And yes they do have ALOT of different quirks that are inherent to the specific tools and brands. As anyone that have tried e.g. patmax on contours that are nested may have noticed.


    There really is not any one goto solution in this field. It is more or less also a certain craftsmanship in every applicayion where every machine vision specialist is an artist.

Advertising from our partners