Long distance color or contrast sensor

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Hello,
I'm looking for a sensor that I'm starting to think doesn't exist and hoping someone here can prove me wrong or might have another idea.

The application is a spool unwinder/winder (re-winder) machine. Picture a giant cassette tape.
The spools contain shiny cable (steel wire rope) 100% of the time.
90% of the time the spools are blue painted steel. The other 10% are unpainted wood.
The unwinder accepts spools/reels varying in outer diameter from 3ft/1m to 6ft/2m.
Inner diameter (drum/belly diameter) is also variable, exact specs unknown.

I need to detect when the spool in the unwinder is almost empty and stop the machine before it fully unwinds the spool and the loose end of cable goes whipping through the air.

If spool size were consistent I could use ultrasonic distance measurement to detect when the spool is almost empty. I have done this in the past with good results. But this time it isn't.
So my other idea was a color or contrast sensor which will detect the transition from the shiny cable color to the <not shiny cable color>. There are industrial color sensors which would work except for the distance requirement. They all need to be close-up to the object being sensed.
Ex1 - "normal" color sensor, range: 30-40mm
Ex2 - "long range" color sensor: 60mm
Ex3 - "super long range" color sensor: 150mm

The ideal position for the sensor is about 5-6ft (2m) from the spool. So obviously none of those will work.

I found this sensor which has a camera lens mount (arducam style) which seems to indicate that a telephoto lens could be employed to sample color from a distance. But there is a dearth of information provided about it and the more closely I look at it the more confused I am. There is an LED inside the lens mount area, among other components, and nothing is centered at the focal point. I Don't know how you would focus it without an image output to monitor.

I can always default to a machine vision camera if I have to, but I have a strong hunch that there is a simpler solution.
Any ideas?
 

MisterBill2

Joined Jan 23, 2018
18,167
Interesting challenge indeed. I am guessing that the camera will be watching the axis of the unwinding spool and see the difference in contrast as the last layer unwinds. If that works it would be great. Color cameras depend on correct and consistent lighting to deliver consistent images for analysis. Perhaps there is some other scheme that would not require a camera.
The obvious variables are the spool weights and relative speeds, and possibly a difference in reflectivity between the cable and the core of the spool.
The simplest will be a pair of reflected light sensors aiming at the axis of the unwinding spool, so that as long as both see a similar reflected light level it is not the last level of cable, but when the levels are suddenly different the core is being exposed, meaning the last level is unwinding. Reflective sensors and switches are common industrial components. So that may be an option worth investigating.
Another scheme would be to monitor the rotation of the spools, and as the unwinding spool to winding spool ration changed it could provide information about the relative diameters, and predict that the end was approaching. But that would not be nearly as accurate.
If the tare weight of the supply spool was known, or if the quantity of cable was known, then the weight of the supply spool could be a good parameter to measure.
But the optical method does seem best, and the devices would not need to be cameras, just light sensors with a 2 or 3 inch field of view at 5 feet distance.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Interesting challenge indeed. I am guessing that the camera will be watching the axis of the unwinding spool and see the difference in contrast as the last layer unwinds. If that works it would be great. Color cameras depend on correct and consistent lighting to deliver consistent images for analysis. Perhaps there is some other scheme that would not require a camera.
The obvious variables are the spool weights and relative speeds, and possibly a difference in reflectivity between the cable and the core of the spool.
The simplest will be a pair of reflected light sensors aiming at the axis of the unwinding spool, so that as long as both see a similar reflected light level it is not the last level of cable, but when the levels are suddenly different the core is being exposed, meaning the last level is unwinding. Reflective sensors and switches are common industrial components. So that may be an option worth investigating.
Thank you for the suggestion, sounds like a good idea! I will look into reflected light sensors. In my experience these have always been digital sensors with an adjustable threshold but I've never needed or looked for versions which expose the analog signal that the threshold is referenced to. If they exist, I'll find them.

Another scheme would be to monitor the rotation of the spools, and as the unwinding spool to winding spool ration changed it could provide information about the relative diameters, and predict that the end was approaching. But that would not be nearly as accurate.
Absolutely. Given line speed (which is read by a laser) and unwinder RPM (which is read by motor encoder) you can (and the machine does) calculate the diameter of the unwinder with superb accuracy. But unless the operator knows (and remembers to enter) the belly diameter, this data is of little use.

Already the operator is given the length of the cable on the unwinder spool and is expected to set a stop in the counter before it unspools completely, but they don't do it consistently, and the result is dangerous. Anything that requires the operator to know something and/or do something other than watch TikTok while the machine runs, is off the table. Save yourself the time needed to address this. You can't possibly provide any commentary that I haven't already almost blurted out in front of my client. I cannot set policy.

If the tare weight of the supply spool was known, or if the quantity of cable was known, then the weight of the supply spool could be a good parameter to measure.
There is existing weight measuring device in the unspooler and it would be a massive undertaking to retrofit one into it. And then it still would require the operator to input a tare weight.

So the optical method is about the only option I think.
 

BobTPH

Joined Jun 5, 2013
8,807
Sounds like a great opportunity for overkill. Camera and neural net pattern recognition. Train it with all of the types of spools you use.
 

panic mode

Joined Oct 10, 2011
2,715
yeah... i understand stigma... lots of clients are wary of cameras and sensors they do not know how to use, troubleshoot and adjust themselves.

but... why do you need long distance color sensor? why can't you just mount sensor closer to the spool?

there is tons of distance sensors using different type of measurement (laser, sound...) to check diameter of the spool. if spool is preloaded (for example by gravity when spool is on horizontal shaft) then one sensor pointed down at spool will tell you how much is left. using second sensor r below it and looking up would allow you do subtract two distance readings and know diameter of the spool center hole.
 
Last edited:

DickCappels

Joined Aug 21, 2008
10,153
You can change the magnification of the image by replacing it with the appropriate lens. Inserting a simple lens element will result in about an 8% decrease in transmission due to front and rear surface reflection. Multiple elements means more losses.

For color setection you might not need anything more complicated than simple optical system (maybe a pair of opaque plates with apertures in them, a photo transistor, a color filter and an amplifier.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
yeah... i understand stigma... lots of clients are wary of cameras and sensors they do not know how to use, troubleshoot and adjust themselves.

but... why do you need long distance color sensor? why can't you just mount sensor closer to the spool?

there is tons of distance sensors using different type of measurement (laser, sound...) to check diameter of the spool. if spool is preloaded (for example by gravity when spool is on horizontal shaft) then one sensor pointed down at spool will tell you how much is left. using second sensor r below it and looking up would allow you do subtract two distance readings and know diameter of the spool center hole.
Knowing diameter of the spool isn't enough. One spool might have a belly/drum diameter of 500mm and another might be 1000. If both these spools had a build-up of 1020mm then spool "A" has a few thousand meters to go while spool "B" is almost out. In order for diameter measurement to work (which is already calculated accurately based on line speed and RPM) the operator would have to enter the belly diameter and they can't be counted on to do that.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Sounds like a great opportunity for overkill. Camera and neural net pattern recognition. Train it with all of the types of spools you use.
I don't think it would have to be trained across a wide array of spool types. It would just have to know the difference between a drum wrapped in shiny cable and <anything else>. But still, as @panic mode said, cameras, AI, etc. often make end users nervous. And I am a firm believer in K.I.S.S. myself, so MV cameras are the last resort.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Maybe a reading on wire rope spooling which use basically mechanical solutions, could spark some ideas.
Sorry I'm not sure what you mean. I get the impression you have something specific in mind that you're expecting I'll uncover rather quickly if I would just lift a finger to do a little research, but I don't know what it is, and let me assure that I have done loads of research over the years. Wire spooling applications are nothing new to me; I have been working in this industry for over a decade and it is what got me into automation. The only part of this that is new territory for me is the part I asked about: differentiating color from a distance.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
but... why do you need long distance color sensor? why can't you just mount sensor closer to the spool?
Sorry, I neglected to answer this.
The avaliable color sensors that I was able to find, don't read past 150mm. Most are 30mm or less.

For this application the starting diameter might be 2000mm and the ending diameter might be 500mm. So even with the longest distance color sensor, the sensor would have to be moved in towards the belly constantly during production. I could be mounted on a linear slide and powered by a stepper but that is unduly complicated and unacceptably vulnerable. It is not uncommon for loose wraps to be encountered in the middle of a reel, so cable will be slinging around in that area where the sensor would have to be. I wager it wouldn't last a week before it's ripped off its mounts and thrown against the cage.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
You can change the magnification of the image by replacing it with the appropriate lens. Inserting a simple lens element will result in about an 8% decrease in transmission due to front and rear surface reflection. Multiple elements means more losses.
Are you suggesting to slap a lens over a commercially available sensor? I don't know how that would work as they emit their own colored light to read color. Light would go out through the lens and come back through the lens? Or light would go out from its typical location (a few mm away from the photoreceptor) and come back in through the lens? I suspect that all or part of the reason why the commercial sensors don't work past a certain difference is that their colored light aren't bright enough. If that's true then I don't see how a lens will help. Can you please clarify?

For color setection you might not need anything more complicated than simple optical system (maybe a pair of opaque plates with apertures in them, a photo transistor, a color filter and an amplifier.
Are you suggesting rolling my own camera obscura a-la-da-Vinci? Whatever I implement will have to pass the client's smell test and a cardboard box with pinholes isn't going to cut it. I trust that what you're suggesting would work in principle but if it's something I'm going make myself, well it sounds like an awful lot of work to make something that looks legit.
 

DickCappels

Joined Aug 21, 2008
10,153
Are you suggesting to slap a lens over a commercially available sensor? I don't know how that would work as they emit their own colored light to read color. Light would go out through the lens and come back through the lens? Or light would go out from its typical location (a few mm away from the photoreceptor) and come back in through the lens? I suspect that all or part of the reason why the commercial sensors don't work past a certain difference is that their colored light aren't bright enough. If that's true then I don't see how a lens will help. Can you please clarify?


Are you suggesting rolling my own camera obscura a-la-da-Vinci? Whatever I implement will have to pass the client's smell test and a cardboard box with pinholes isn't going to cut it. I trust that what you're suggesting would work in principle but if it's something I'm going make myself, well it sounds like an awful lot of work to make something that looks legit.
Are you suggesting to slap a lens over a commercially available sensor?
Yes, many color sensors do no emit light. They use RGB or similar (i.e. CIE 1931,etc.)

Are you suggesting rolling my own camera obscura a-la-da-Vinci?
I am not recommening it, only mentiontioning that is can be a similarly simple optical system.

If it is your project, not mine -use whatever you think will do the job at a cost you are willing to pay. No need to complain when a suggestion concludes with "It can be as simple as..." It is your nickle and your show.

You are welcome.
 

MisterBill2

Joined Jan 23, 2018
18,167
To do it with a camera will require a fair amount of image analysis, unless the lighting can be adjusted to provide a great deal of contrast so that it is easy to detect when the cable no longer covers the whole surface of the core. Very similar to the approach oof using reflected light sensors. There are optical switches that also have analog outputs, unfortunately I do not recall the brand.
Probably a conversation with an applications engineer of one of the optical proximity switch companies is in order. That is what they are paid for, providing assistance for tough problems.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Are you suggesting to slap a lens over a commercially available sensor?
Yes, many color sensors do no emit light. They use RGB or similar (i.e. CIE 1931,etc.)

Are you suggesting rolling my own camera obscura a-la-da-Vinci?
I am not recommening it, only mentiontioning that is can be a similarly simple optical system.

If it is your project, not mine -use whatever you think will do the job at a cost you are willing to pay. No need to complain when as suggestion concludes with "It can be as simple as..." It is your nickle and your show.

You are welcome.
I seemed ungrateful. My apologies. I did not intend to come across that way. I didn't mean to complain, just to state the constraints of the project. Thank you for your help and I will try to be more considerate in my replies.
 

Reloadron

Joined Jan 15, 2015
7,501
When I was a kid long, long ago I did a science fair project where I used a LDR and a small light reflecting off a wheel of colors. The idea was my LDR drove an oscillator so different colors reflected different light levels. Much later I used similar to detect a spot on polyethylene web (spools) called the eye spot in packaging machinery. My reflective system I mounted the LDR and light source (incandescent flashlight bulb in a few cubes of plexiglass so I could focus the distance for light reflection. Maybe something along those lines.

Ron
 

MisterBill2

Joined Jan 23, 2018
18,167
Keep in mind that the scheme I proposed depends on a difference in reflrctivity between the spool and the wire rope. An actual camera system can use image analysis to detect the difference between the cable and the smooth surface of the spool core. But none of those parts are cheap, so the project budget becomes a big consideration.

One question not asked is how long does it take to stop the take-up spool from spinning??
 

atferrari

Joined Jan 6, 2004
4,764
Sorry I'm not sure what you mean. I get the impression you have something specific in mind that you're expecting I'll uncover rather quickly if I would just lift a finger to do a little research, but I don't know what it is, and let me assure that I have done loads of research over the years. Wire spooling applications are nothing new to me; I have been working in this industry for over a decade and it is what got me into automation. The only part of this that is new territory for me is the part I asked about: differentiating color from a distance.
Ok clear enough.
Sorry for my post.
 

MisterBill2

Joined Jan 23, 2018
18,167
Just read an article about machine vision, unfortunately the greatest distance was 200mm=20Cm=about 5 inches or so. But even some cheap security cameras do much better than that. So somewhere the machine vision selling folks are missing the longer distance market. I doubt that the spooling system needs super high resolution, just adequate color recognition.
 
Top