2
\$\begingroup\$

I have a surface 40x36 cm. I would like to recognize the shape and weight of specific objects on that surface. Here is the "big picture".

enter image description here

The first idea was to use touchscreen overlay (infrared or something like that) + single weight sensor but I need to recognize shapes (and not single points) so that's not going to work. Second idea was to use something like MTmini but I don't have that much space below my surface - eventually I have 5 cm at max. Third idea (current one) is to use pressure map with 20x20 (or something similar, not aware of sensitivity of the sensors and possibility to extract shape from it) sensor grid. Something like foot mapping sensor in this question.

So the idea is that I have blank surface and if I put rectangular object which weights 1kg I can recognize where it is exactly on the surface, what is it's shape and weight.

Blue overlays are common sizes of objects that will be on that surface. Weight of single object will range from 0 kg to about 5 kg - I need at least 20-50 g precision 0 kg to max. 3 kg (usually around 500 g) - I need at least 50 g precision. Anything larger than 250g will sit on multiple sensors so I think that 0-1kg range per sensor is max (precision 50g - 5%) or better yet 0-500g per sensor with 50g precision is 10%.

My question is, what type of sensors to use (I though about piezoelectric ones?), how/where can I check their "sensitivity"/accuracy and where can I get them (cost)? Also if you have some experience, how dense should be my grid? 10x10? 20x20? 40x40? Probably all multiplexed.

Probably I would connect it all to RaspberryPi (or Arduino?) for calculations and analysis.

\$\endgroup\$
3
  • \$\begingroup\$ I think you've set an impossible goal as far as the accuracy of the weight measurement (~1% of object weight). Especially if the object is resting on multiple sensors, the error goes up with the square root of the number of sensors involved (assuming errors are uncorrelated), so if you've got an object resting on, say 25 sensors, the individual sensor accuracy needs to be 5x better than the overall goal you're shooting for. Load cells with 0.2% accuracy exist, but they aren't cheap. Simple resistive, capacitive or piezo sensors aren't going to have anywhere near the required accuracy. \$\endgroup\$
    – Dave Tweed
    Aug 4, 2013 at 13:12
  • \$\begingroup\$ How far can I go with accuracy with "standard" sensors? How much are sensors with <1% accuracy? Give me some numbers. \$\endgroup\$
    – svenkapudija
    Aug 4, 2013 at 14:37
  • \$\begingroup\$ @DaveTweed: looks like Apple has gotten in better in the mean time ("3D" touch in the iPhone 6S). I'm not sure if it would meet the accuracy required in this question though. Didn't crunch any numbers... nor I do I know the actual accuracy of the iPhone 6S force/pressure sensor[s]. \$\endgroup\$
    – Fizz
    Nov 27, 2015 at 8:15

2 Answers 2

Reset to default
3
\$\begingroup\$

This may be possible using quantum tunneling composites (qtc's) which use tunneling effects to drastically change in conductivity with applied pressure. You could use a grid of multiplexed electrodes on the back to "address" a particular spacial location of the qtc mat and measure its resistance. This would give you a grid of force data that could then be 'blob' tracked similar to image recognition algorithms. Once the 'blob' that is the object has been identified you could sum the forces contained within its boundaries to find the total weight. Similar things have been done in industry (http://www.nanotech-now.com/news.cgi?story_id=47663)

You can acquire qtc mats and pills from various sources around the internet. If you can not find a source of mats, you could use many qtc pills to form a mat.

Here are some sources of qtc pills http://www.mindsetsonline.co.uk/product_info.php?products_id=1144

Accuracy will depend on the dynamic range of your ADC circuitry as their resistance can vary on the order of 10 orders of magnitude depending on the pressure. I can imagine a digital capture circuit that uses feedback that selects the optimal reference voltage or voltage divider resistor. However, I believe your 1% accuracy requirement is far to high most technologies.

\$\endgroup\$
4
  • \$\begingroup\$ Didn't know about that one :) Actually, they have a in-house prototype of what I need. See 2nd video on the link - peratech.com/videos.html Anyone had experience building that kind of stuff? I'll email them but not sure how fast response will be (because I'm not big vendor but a hobbyist). \$\endgroup\$
    – svenkapudija
    Aug 4, 2013 at 18:11
  • \$\begingroup\$ I'm not sure how responsive peratech will be to individuals looking to use their technology. Home building one, though, looks like quite the adventure. \$\endgroup\$
    – Nick Arango
    Aug 4, 2013 at 19:09
  • \$\begingroup\$ Is the resistance inversely proportional to force? It it linear, or more of a 2-state device? \$\endgroup\$
    – Scott Seidman
    Sep 3, 2013 at 23:15
  • \$\begingroup\$ the conductivity increases exponentially with pressure. It is a nonlinear relationship, but it is a simple nonlinear relationship. \$\endgroup\$
    – Nick Arango
    Sep 6, 2013 at 12:46
-1
\$\begingroup\$

Low cost option: use an old scanner table, perform an image acquisition and the identify the shapes (if the scanning light isn't a probl

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.