PDA

View Full Version : [Project] What sensor will detect different materials?



flow21
10-20-2008, 02:29 PM
Hello,
I need some help. I looking for a sensor or some other technology that would differentiate different materials (bottles of: glass, plastic, and aluminum) to me specific, and I could take that signal and program it.

Thank You in advance...:happy:

DresnerRobotics
10-20-2008, 02:33 PM
As long as you're looking at the same size materials, a digital scale :p

Adrenalynn
10-20-2008, 02:41 PM
How generic do you need it to be? Can you sample representative objects as training prior to live recognition?

The three generic types you suggest will all transmit light differently.

flow21
10-20-2008, 02:48 PM
Not a bad idea...but the problem is that they will have different sizes, and their surfaces porosity are very similar, so I can't go that route.

flow21
10-20-2008, 02:51 PM
How generic do you need it to be? Can you sample representative objects as training prior to live recognition?

The three generic types you suggest will all transmit light differently.
Yes, I will be able to test the materials prior to the live recognition, but aluminum in particular could have some slight physical deformation.

Rudolph
10-20-2008, 03:40 PM
For some reason I'm reminded of the old (~25 years ago) Coors can tester magnet; "If it sticks, it isn't aluminum".

Adrenalynn
10-20-2008, 03:57 PM
That's really ok though, right?

Aluminum will transmit 0% of the light.
Plastic will transmit clear glass minus some percentage of light
Different colors of glass and plastic will transmit different spectrums of light differently.

Did you see my spectrum analyzer based on the analyzer from Make Magazine? It could be trivially modified to accomplish what I wrote above, and more. It could sort red bottles from green bottles from blue bottles, or shades there-of. It could sort thick glass from thin of the same composition, or clear plastic from milk bottles. And it shouldn't cost more than a few dollars + the cost of the microcontroller.

730

This just fires off a Red, Green, Blue, White, IR, and UV [long wave] diode from a microcontroller. Then on the other side there are a couple different light sensors that register the amount of light they see. You calibrate it first with nothing in the way, then drop the known sample in, hit it again - boom - that's your named sample. From then on, you drop a sample in, hit it, and decide based on some margin of error what the unknown sample is.

Am I missing some requirement?

flow21
10-21-2008, 01:23 PM
Yeah, something like that is what I've been thinking. So, I think I could try that, my only concern will be that some of the plastic and aluminum bottles are going to be clear and hopefully the light different would be sufficient enough to distinct one from the other. Also I think the size might become a problem, but I think could work around that, cause I needed to be as small as possible. So, thank you very much, you've been very helpful.

P.S. Any other suggestion don't hesitate to post them.:D

jes1510
10-21-2008, 01:27 PM
Yeah, something like that is what I've been thinking. So, I think I could try that, my only concern will be that some of the plastic and aluminum bottles are going to be clear and hopefully the light different would be sufficient enough to distinct one from the other. Also I think the size might become a problem, but I think could work around that, cause I needed to be as small as possible. So, thank you very much, you've been very helpful.

P.S. Any other suggestion don't hesitate to post them.:D

I didn't think it was possible to have clear aluminum.

Adrenalynn
10-21-2008, 01:45 PM
Anything's "clear" if it's "thin enough" and the light is "bright enough"

Lottsa empty space betwixt those atoms. ;)

Flow -

go for SMT LEDs? I could build the same thing that fit on a pinhead...

robot maker
12-12-2008, 02:28 PM
since you are detecting different types of materials,you will need different types of sensors to detect each material,best i am looking is web cam and using a image compare software with images i have in my database,so far just idea but did find some great software and a very ideal camera,then add fsr ,heat sensor for cold or hot and maybe IR or color sensor

robot maker
12-12-2008, 02:36 PM
Adrenalynn
any links to your project
i did find one by Eric Rosenthal using a adurino board

http://creative-technology.net/MAKE.html

That's really ok though, right?

Aluminum will transmit 0% of the light.
Plastic will transmit clear glass minus some percentage of light
Different colors of glass and plastic will transmit different spectrums of light differently.

Did you see my spectrum analyzer based on the analyzer from Make Magazine? It could be trivially modified to accomplish what I wrote above, and more. It could sort red bottles from green bottles from blue bottles, or shades there-of. It could sort thick glass from thin of the same composition, or clear plastic from milk bottles. And it shouldn't cost more than a few dollars + the cost of the microcontroller.

730

This just fires off a Red, Green, Blue, White, IR, and UV [long wave] diode from a microcontroller. Then on the other side there are a couple different light sensors that register the amount of light they see. You calibrate it first with nothing in the way, then drop the known sample in, hit it again - boom - that's your named sample. From then on, you drop a sample in, hit it, and decide based on some margin of error what the unknown sample is.

Am I missing some requirement?

Resilient
12-17-2008, 05:02 PM
What if you whang it with a stick and measure the sound it makes? I am thinking something like the little yellow drum machine.

robot maker
12-22-2008, 10:30 PM
i did find a color sensors at sparkfun.com that look good for the make project 4 color detect sensor only need to add IR and UV PHOTODIODE


Adrenalynn
any links to your project
i did find one by Eric Rosenthal using a adurino board

http://creative-technology.net/MAKE.html

Quantum
12-23-2008, 12:32 AM
A's project has the whole spectrum included. Why bother building seperate sensors just follow her's and it's a all in one deal. Then your done.

Adrenalynn
12-23-2008, 12:44 AM
I think if I were doing it again, I'd try an RGB LED +UV +IR. That said, I've never really be able to reproduce color with ultimate reliability from those the way a single specific color LED can.

robot maker
12-23-2008, 12:51 AM
mostly because it all in one sensor with filters and very small
her sensors i thought was from the make magazine design by another person,i could be wrong
but the design uses seperate photo diodes for each color and same can be done with this one without adding all diodes because they are built all ready
$20 only need cpu and you are done except for IR DIODE and UV DIODE ,save wiring up 4 diodes and maybe a better design,also have with out demo board for $10
i like only sensor since only a few parts on board and super tiny to fit on fingers or hand
most likey i am gong to do like always instead of others saying it work work or not as good is to try it and compare it ,also another one for adurino board they have too
http://www.sparkfun.com/commerce/product_info.php?products_id=8663


A's project has the whole spectrum included. Why bother building seperate sensors just follow her's and it's a all in one deal. Then your done.

Quantum
12-23-2008, 01:00 AM
Ok then $20 bucks done deal. Get one before there sold out. 10 left

Adrenalynn
12-23-2008, 01:29 AM
Who "P"'d on Light Emitting Diodes? Just wondering...

My design was loosely based on Mr. Rosenthal's but substantially different. In fact, I was in correspondence with him for awhile.

Unfortunately the filtered designs skew predominant temperature of the light at different wavelengths, different power levels, and different ambient lighting. They're all but worthless for spectranalysis in my experience.

robot maker
12-23-2008, 01:30 AM
not ready for it for awhile working on test-bot
when sold out will get more in latter
thinking getting both designs and test with one adrenalynn posted


Ok then $20 bucks done deal. Get one before there sold out. 10 left

robot maker
12-23-2008, 01:37 AM
have you posted your design yet would lve to see it and do my comparing as i do with all sensors,dont like taking others word for it,the could be right or wrong,its better to test it,i learn this from working in R&d lab before they gave my own department
i love every type of sensor on the market and buying them,that what i love the most shopping ,and heard a new one taste snsor using a type of gland i thinks,i think snot gland,very cool,i think is was servo magazine article


Who "P"'d on Light Emitting Diodes? Just wondering...

My design was loosely based on Mr. Rosenthal's but substantially different. In fact, I was in correspondence with him for awhile.

Unfortunately the filtered designs skew predominant temperature of the light at different wavelengths, different power levels, and different ambient lighting. They're all but worthless for spectranalysis in my experience.

Adrenalynn
12-23-2008, 01:41 AM
>> its better to test it

Test away!

robot maker
12-23-2008, 01:48 AM
still need your design,or you keeping it to your self
it might be a great design hard to tell

>> its better to test it

Test away!