I would use an iPad to do this and I think its totally possible with the camera on the iPad. I think if you made an app to do it you could pay me a few bucks for the suggestion
Otherwise you can use 2000 sensors, in fact I know of a type of photo-transistor matrix sensor that outputs serially, just cant remember the part number at this writing. If my memory serves me correctly it was inexpensive too.
'Not saying it wouldn't work, either but that would be a lot of light sensors. The led/sensor stick would only require enough sensors for a single stack of cigarettes and could scan each stack in turn in a few minutes. Think of it as a bar code reader for cigarette package stacks. Could even motorize it so it scans each stack automatically and transmits the counts.
This sounds more like an idea/product you could sell to the companies that produce the store fixtures rather than to the retailers. An XY array of sensors built in to the display case would be much easier to read as well.
<Paste>
The concept is to space photodiodes or phototransistors according to center to ceter of predetermined package sizes. If light is present, then there is no package, and if light is not present, then add one to the package count. The idea is to use surface mount discrete packages on top of a PCB, which is then applied with some type of coating, with cutouts around the discrete devices, so that the upper surface will have a flat plane so that items can slide across the sensor surface without damage occuring.
In my current view of things, the sensors will be in a column type fashion, with a ribbon connect on the end to plug into some type of motherboard.
Well, does the idea sound feasible and perhaps profitable?
Almost anything is feasible, and can even be profitable, depending on your sales mode.
A better question is : Is it practical ?
You'd need it rugged and flexible. Something that needs customising to this extent "according to center to ceter of predetermined package sizes", is moving sway from practical.
Also "If light is present, then there is no package" - is this transmissive, or reflective ?
Both have issues, Transmissive is more 'solid' but adds more mechanical constraints
Sometimes the simplest is the best : Most items that are 'row-count-able' could have a dispenser/holder, and you limit the draw-off to one end.
Now, you can count from a single switch/sensor near or at the feed/customer end, and have a 'full' reset switch/sensor at the full end.
That side-steps needing anything like "2000 photodiodes or phototransistors", and the constraints of strip packaging.
I agree that ready wired displays would be the way to go. However, Upto now, and to my knowledge, all of the displays in existence, which are a numerous, are manually counted, but I could be wrong. In fact, who knows, there could be an existing patent or a patent pending. That is a large portion of potential business to simply bypass.
Okay, one of the main reasons behind this concept is the elimination of human intervention in the counting process, to cut payroll expenses for businesses. The use of an iPad would still require human intervention. However, having dabbled in programming quite extensively, I would say that bitmap or other image comparisons could be done on a photo to search for rectangles.
Also "If light is present, then there is no package" - is this transmissive, or reflective ?
Both have issues, Transmissive is more 'solid' but adds more mechanical constraints
I would go for transmissive. All of the reflective sensing that I have seen requires at least a small amount of distance from the object that it is detecting. As for the end of row sensing, that would only give you a theoretical count, instead of an actual inventory count. Packages can be removed and put back, as is the case with cigarettes on a normal basis. Such type of activity would not provide an accurate inventory count.
What about a mechanical device? Spring loaded L bracket that exerts force on a Propeller-controlled Capacitive Load Cell As items are placed in inventory they push the L bracket back which causes force on the load cell.
@jmg
I would go for transmissive. All of the reflective sensing that I have seen requires at least a small amount of distance from the object that it is detecting. As for the end of row sensing, that would only give you a theoretical count, instead of an actual inventory count. Packages can be removed and put back, as is the case with cigarettes on a normal basis. Such type of activity would not provide an accurate inventory count.
If you need to allow for that, quadrature counting allows INC and DEC, so that is 2 sensors at feed end.
- and I'd add one about half way, so you can auto-tolerate thickness, and get some fail-safe.
With quadrature, you would naturally INC on fill, and DEC on empty, and any creepage would be reset by the half/full sensors, so it would not accumulate.
Okay, based upon the feedback that I have received so far, here is my new concept. What about distance measuring with some type of laser device, located at the back of each row.
Rows would have a common frontal starting position and the distance sensors would have a common rear mounting position. Count could be established through a mathematical process on the feedback. Perhaps instead of a laser, a PING sensor could be used like Phil mentioned earlier.
Bruce
EDIT: Hey Parallax, please let me know if you have an instant call for a large amount of PING sensors
Use whatever inexpensive and reliable measurement device that fits the bill. I believe the sell point are smart racks to hold the inventory. Network the racks using a TCP/UDP device like the Spinneret. Inventory is accessible, extensible, and configurable.
Parallax let me know if there is a run on Spinnerets.
I know that this subject has been covered once before in another thread of mine, but I am now thinking of the Sharp GP2Y0A21YK0F infrared distance measuring sensor offered by Parallax, at this location:
Yea, that part could have been a potential possible solution, but it is now obsolete.
Mouser shows all Analog sensors as obsolete, suggests either they were unreliable, or someone else made a cheaper version, or a (maybe) patent cease issue ?
The idea seems sound.
Another approach would be an optical ping, using 1ns/foot, which would need maybe 100ps resolve.
Sounds doable via an indirect means. - anyone tried a LCD shutter system ?
That part is obsolete according to Sharp. However, I do like the idea of one sensor at the rear, reading distance somehow, whatever type of sensor it may be. As long as it was accurate and did not need to much peripheral support, it would be a fast and easy setup.
Bruce
EDIT: And as long as the sensor was cheap, it would cut manufacturing as well as client costs
I haven't read all comments so maybe this idea is already mentioned.
Almost EVERY product has a barcode. So scanning the barcodes when they are sold should be enough.
What's the special situation here that makes the solution "read the barcodes when selling" does not work??
best regards
Stefan
The management of physical inventory. Items that can be removed from their display and fail to go through the checkout process. Inventory "shrinkage". Loss prevention. SHOPLIFTING! To name a few.
The management of physical inventory. Items that can be removed from their display and fail to go through the checkout process. Inventory "shrinkage". Loss prevention. SHOPLIFTING! To name a few.
Okay, now that we have the why out of the way, an accurate and viable solution must be provided, so that forumistas can start making money from this concept. I propose a joint endeavor to the forum, so that all Propeller forum members may benefit from the knowledge that may ensue.
Since my last post, I have been researching IR ranging and Ultrasonic ranging. From my research, I am almost convinced that an ultrasonic transducer is the way to go, at least for cigarette pack counting.
How about a laser scanner and a stereo camera? You can get a 3D image and should easily recognize present or missing boxes. Detecting the small wiggle of the laser spot at the gap between the boxes would even allow "auto learning" of the package size.
Okay, so please excuse me for bouncing all over the map, but I now have a new concept, based upon research and suggestions. I now believe that a camera and image processing is the way to go.
In another thread, referring to the CMUcam4, Kye is using a OmniVision camera and interfacing it with the Propeller chip, and I noticed that his board has an SD card reader.
Since I have a vision, I am going to stick with that vision, so this discussion will only apply to cigarette pack counting. Now let's also assume a previous theory, which is:
Rows would have a common frontal starting position and the distance sensors would have a common rear mounting position.
However, instead of having a sensor at the rear, a camera board is placed instead, which contains a camera, a Propeller chip, and an SD card reader. Now let's say that an SD card is installed that contains ten images that represent the ten possible inventory counts. I suppose that image comparisons could be done to achieve an accurate count.
Bruce
EDIT: However, I also noticed that in another Parallax product, they utilize a camera and a laser for ranging. I supoose this could also be used in image comparisons.
EDIT: However, I also noticed that in another Parallax product, they utilize a camera and a laser for ranging. I supoose this could also be used in image comparisons.
You are going to need some illumination anyway, so unless you want to do full image processing, and send full pictures of a packet, a point illumination allows simple triangulate-distance calculation.
Linear sensors are cheap, but usually lack optics, and the more generic cameras have huge volumes on their side.
I see DIgikey mentions OV07690-A20A CMOS VGA 20-CSP3 2200 : $0.93
Another plus for a generic camera, is you can see how the system is thinking, during development, but choose to only seek a dot during operation.
I think you could rotate a VGA level sensor ~45' to get a diagonal render, so both X and Y processing could get higher resolution.
- might as well use all that resolution you are buying
This could match well with a Prop, which would find a Full Image hard, but should be able to auto-seek a DOT(s) on the image plane, in real time.
The best way these days would be RFID. A sticker on each item. No wiring at all. Just a portable reader and transmit to computer. Pass reader near inventory and it will read all rfid stickers at once. We use rfid where I work all the time now. They make small rfid stickers or capsuals. Used in all the stores now for shoplifting. A lot of packaged items allready have them in the package already.
A technology similar to this is in use at my local airport. The entire garage is equipped with sensors for the cars. Red equals car, green no car. One can assess the state of the garage before entry and some additional lights guide people to where the free spaces are. That data also is linked to the parking ticket to charge the right fee.
A technology similar to this is in use at my local airport. The entire garage is equipped with sensors for the cars. Red equals car, green on car. One can assess the state of the garage before entry and some additional lights guide people to where the free spaces are. That data also is linked to the parking ticket to charge the right fee.
Well there you have the grand scheme. Very cool concept. Thanks for sharing.
My latest thoughts on this issue is to revert back to the original plan with a circuit board of phototransistors and 74HC165s. However, I am now leaning toward connecting these boards directly to a parallel port card and letting the PC read the inputs directly and having software identify brand and type through bit location, which can be configured through software.
Bruce
EDIT: Additionally, the software will be able to print current inventory reports as well store comma delimited inventory tallys according to brand and type
Since I never stocked cigarettes, I was really just doing a bunch of guessing. And since I was tired of guessing, I spent some time talking with the friendly manager at the local gas station, during which time I wrote down a brief description of the setup, such as rows, columns, etc... Prior to today, I wondered if the cigarette holding trays were spring loaded, and today I discovered the answer was yes. In addition to that discovery, the manager gave me an old, spring loaded tray to toy around with. YIPPEE
Having thought about this for a couple days, I think I now have a decent plan developing. Considering these trays are spring loaded, I should be able to use some sort of strain guage as mentioned and linked to earlier by Mike G, at which point, he linked to a project of Phils, which can be found here:
Since each tray holds on average 4 - 5 columns of cigarette packs, with each column being spring loaded, I propose placing a strain gauge at the front of each column, all of which are connected to a battery powered propeller which is mounted to the front of the tray for easy battery replacement. Additionally, at the rear of each column, an IR transceiver is secured, which is also connected to the battery powered propeller which is mounted to the front of the tray.
Behind each IR transceiver, of each column, another IR transceiver is placed, all of which are remotely connected to another propeller board for polling tray column counts according to the strain put on the strain gauge. Numerous boards will be connected this way.
Numerous Propeller boards each having a predetermined amount of IR transceivers attached, will be connected to each other for serial communication.
The end of the line of the chained Propellers will be attached to the PC.
I propose placing a strain gauge at the front of each column,
By front, do you mean 'pushed against' by the next-sold packet, or on the front-end of he spring ?
If you measure packet-push-pressure you would need a large area sensor, and the spring pressure will vary from unit to unit.
( so empty and Full points will all differ - some CAL step would be needed )
A front sense would also need to ignore the Up/Down spikes around any removal
Comments
By the way, I did not previously mention thank you. So at this time, I would like to thank everyone who has participated in this discussion.
Bruce
Otherwise you can use 2000 sensors, in fact I know of a type of photo-transistor matrix sensor that outputs serially, just cant remember the part number at this writing. If my memory serves me correctly it was inexpensive too.
This sounds more like an idea/product you could sell to the companies that produce the store fixtures rather than to the retailers. An XY array of sensors built in to the display case would be much easier to read as well.
Almost anything is feasible, and can even be profitable, depending on your sales mode.
A better question is : Is it practical ?
You'd need it rugged and flexible. Something that needs customising to this extent "according to center to ceter of predetermined package sizes", is moving sway from practical.
Also "If light is present, then there is no package" - is this transmissive, or reflective ?
Both have issues, Transmissive is more 'solid' but adds more mechanical constraints
Sometimes the simplest is the best : Most items that are 'row-count-able' could have a dispenser/holder, and you limit the draw-off to one end.
Now, you can count from a single switch/sensor near or at the feed/customer end, and have a 'full' reset switch/sensor at the full end.
That side-steps needing anything like "2000 photodiodes or phototransistors", and the constraints of strip packaging.
I agree that ready wired displays would be the way to go. However, Upto now, and to my knowledge, all of the displays in existence, which are a numerous, are manually counted, but I could be wrong. In fact, who knows, there could be an existing patent or a patent pending. That is a large portion of potential business to simply bypass.
Bruce
Good to have you back at the forum.
Okay, one of the main reasons behind this concept is the elimination of human intervention in the counting process, to cut payroll expenses for businesses. The use of an iPad would still require human intervention. However, having dabbled in programming quite extensively, I would say that bitmap or other image comparisons could be done on a photo to search for rectangles.
Bruce
Anyhow welcome back.
I would go for transmissive. All of the reflective sensing that I have seen requires at least a small amount of distance from the object that it is detecting. As for the end of row sensing, that would only give you a theoretical count, instead of an actual inventory count. Packages can be removed and put back, as is the case with cigarettes on a normal basis. Such type of activity would not provide an accurate inventory count.
Bruce
If you need to allow for that, quadrature counting allows INC and DEC, so that is 2 sensors at feed end.
- and I'd add one about half way, so you can auto-tolerate thickness, and get some fail-safe.
With quadrature, you would naturally INC on fill, and DEC on empty, and any creepage would be reset by the half/full sensors, so it would not accumulate.
Rows would have a common frontal starting position and the distance sensors would have a common rear mounting position. Count could be established through a mathematical process on the feedback. Perhaps instead of a laser, a PING sensor could be used like Phil mentioned earlier.
Bruce
EDIT: Hey Parallax, please let me know if you have an instant call for a large amount of PING sensors
Parallax let me know if there is a run on Spinnerets.
http://www.parallax.com/Store/Sensors/AllSensors/tabid/760/CategoryID/46/List/0/SortField/0/catpageindex/4/Level/a/ProductID/776/Default.aspx
Now I am wondering if it would be accurate enough.
The idea seems sound.
Another approach would be an optical ping, using 1ns/foot, which would need maybe 100ps resolve.
Sounds doable via an indirect means. - anyone tried a LCD shutter system ?
That part is obsolete according to Sharp. However, I do like the idea of one sensor at the rear, reading distance somehow, whatever type of sensor it may be. As long as it was accurate and did not need to much peripheral support, it would be a fast and easy setup.
Bruce
EDIT: And as long as the sensor was cheap, it would cut manufacturing as well as client costs
Almost EVERY product has a barcode. So scanning the barcodes when they are sold should be enough.
What's the special situation here that makes the solution "read the barcodes when selling" does not work??
best regards
Stefan
Okay, now that we have the why out of the way, an accurate and viable solution must be provided, so that forumistas can start making money from this concept. I propose a joint endeavor to the forum, so that all Propeller forum members may benefit from the knowledge that may ensue.
Since my last post, I have been researching IR ranging and Ultrasonic ranging. From my research, I am almost convinced that an ultrasonic transducer is the way to go, at least for cigarette pack counting.
Bruce
In another thread, referring to the CMUcam4, Kye is using a OmniVision camera and interfacing it with the Propeller chip, and I noticed that his board has an SD card reader.
Since I have a vision, I am going to stick with that vision, so this discussion will only apply to cigarette pack counting. Now let's also assume a previous theory, which is:
Bruce
EDIT: However, I also noticed that in another Parallax product, they utilize a camera and a laser for ranging. I supoose this could also be used in image comparisons.
You are going to need some illumination anyway, so unless you want to do full image processing, and send full pictures of a packet, a point illumination allows simple triangulate-distance calculation.
Linear sensors are cheap, but usually lack optics, and the more generic cameras have huge volumes on their side.
I see DIgikey mentions OV07690-A20A CMOS VGA 20-CSP3 2200 : $0.93
Another plus for a generic camera, is you can see how the system is thinking, during development, but choose to only seek a dot during operation.
I think you could rotate a VGA level sensor ~45' to get a diagonal render, so both X and Y processing could get higher resolution.
- might as well use all that resolution you are buying
This could match well with a Prop, which would find a Full Image hard, but should be able to auto-seek a DOT(s) on the image plane, in real time.
Well there you have the grand scheme. Very cool concept. Thanks for sharing.
Bruce
Bruce
EDIT: Additionally, the software will be able to print current inventory reports as well store comma delimited inventory tallys according to brand and type
Since I never stocked cigarettes, I was really just doing a bunch of guessing. And since I was tired of guessing, I spent some time talking with the friendly manager at the local gas station, during which time I wrote down a brief description of the setup, such as rows, columns, etc... Prior to today, I wondered if the cigarette holding trays were spring loaded, and today I discovered the answer was yes. In addition to that discovery, the manager gave me an old, spring loaded tray to toy around with. YIPPEE
Having thought about this for a couple days, I think I now have a decent plan developing. Considering these trays are spring loaded, I should be able to use some sort of strain guage as mentioned and linked to earlier by Mike G, at which point, he linked to a project of Phils, which can be found here:
- Since each tray holds on average 4 - 5 columns of cigarette packs, with each column being spring loaded, I propose placing a strain gauge at the front of each column, all of which are connected to a battery powered propeller which is mounted to the front of the tray for easy battery replacement. Additionally, at the rear of each column, an IR transceiver is secured, which is also connected to the battery powered propeller which is mounted to the front of the tray.
- Behind each IR transceiver, of each column, another IR transceiver is placed, all of which are remotely connected to another propeller board for polling tray column counts according to the strain put on the strain gauge. Numerous boards will be connected this way.
- Numerous Propeller boards each having a predetermined amount of IR transceivers attached, will be connected to each other for serial communication.
- The end of the line of the chained Propellers will be attached to the PC.
I think that would be the ticket.Bruce
By front, do you mean 'pushed against' by the next-sold packet, or on the front-end of he spring ?
If you measure packet-push-pressure you would need a large area sensor, and the spring pressure will vary from unit to unit.
( so empty and Full points will all differ - some CAL step would be needed )
A front sense would also need to ignore the Up/Down spikes around any removal