Products > Robotics and Automation

Clear Tape Detection

<< < (4/4)

Interesting challenge and solution discussion. Not sure I have much to add....

I was looking at this video and it is impressive. In a practical sense, the commercial sensors may be the way to go although adapting it to your set up could be challenging.

But using one is no fun (for me) and thinking about alternatives is more interesting.

What about three (it could be done with two) still cameras images taken from a fixed location and each with a matched light source.  The positions are the center of where the tape should be and two positions on the periphery where the cardboard should be. A quick analysis of the "glare" (i.e., brightness) of the photos might identify tape vs. no tape quite accurately, and relatively cheaply. Maybe two or three image sensors in lieu of cameras... and if the comparative difference is large enough, simpler light sensors could be tried, but always measuring a delta between two (or three) sensors.

In the video they use the thru beam method which is not possible in the OP case. You can put a reflector on the other side of the tape once you tape the box.

Sounds like the problem has a solution with the aforementioned COTS sensors.

But to solve it if you were trying to invent a solution other than buying / integrating the purpose made options I'd consider:

1: Machine vision on the sides and / or bottom of the package with appropriate lighting and looking for the pattern of the differing gloss or the variance in appearance of the edges vs. the taped middle zone vs. the untaped perihpery zone.  If a human under non-optimum lighting conditions can look at the package from a modest distance (e.g. several meters / a few feet) and easily at a glance be able to reliably see that the package is taped or not then the taping must be distinctive enough in appearance vs. the package that a simple machine vision image processing solution can probably get the right result from either feature extraction or ML.

2: The gloss / reflectivity / polarization options as aforementioned using either a 2D image sensor with controlled lighting or possibly just a 1D / 2D sensor or something like a barcode scanner approach with a point / line light source and some photo receivers.

3: I didn't see anyone mention thickness gague or dielectric or sonic based approaches.  Even though tape is thin it usually has a very prominent steep edge to it which is easily felt as an abrupt height variation over a short distance on the surface no matter how invisible the tape may be in optical contrast.  So therefore it is possible you might have some ultrasound or mechanical "feeler" or optical or capacitive range measurement based options that could pick up the change in distance profile / thickness / dielectric constant / sonic reflectivity over the distance.

Another simpler option which might facilitate a machine vision or scanner based solution -- you asked about clear tape, but perhaps it'd be possible to
consider just not using completely clear tape?

You can order printed tape which is either opaque or clear otherwise e.g.:

Or you could use plain opaque tape with some differing color than the box.

If the machine vision had something easy to look for like a prominent color difference or a pattern of printing (e.g. logo, bar code, whatever) then that would make the taped status easier to determine than almost invisible clear tape.


--- Quote from: PBn on October 03, 2019, 09:09:02 pm ---Anyone have thoughts or other ideas?

--- End quote ---
Replace clear tape with brown tape.

Have you checked to see what it looks like under UV.  I'm constantly surprised by what glows and wot don't.


[0] Message Index

[*] Previous page

There was an error while thanking
Go to full version