How to build grain color sorting system

[thsiriwuth]

Hi

I am now working on build system for grain color sorting. The system process will be like this

- The grain will be running down a chute in a near continous stream (with multiple lane).

- use the GigE line scan camera and NI vision software to detect the grain defect (black color).

- After the vision system detect the defects which can be random in herizontal ROI.

- I suppose the vision system can send position of each defect in ROI (pixel no. or co-ordinate X).

- Then use the defect position to trigger air nozzle ejector to reject the defects.

I have a lot questions to be clarify before go further, please help to answer my questions.

- Which the NI software and hardware can be use to inspect the color defect ?

- How to get the position or coordinate X of each defect in ROI ?

- Is it possible to send the defect position information to PLC ? and How ?

Best regards,
Siriwut H.

[BruceAmmons · Accepted solution]

How much time do you have between the camera taking the picture and turning on the reject air valve?  It doesn't seem like it would be very long to me.

If it is extremely time critical, I would probably do the system using a Camera Link interface that has a built in FPGA.  NI has one I have used before quite successfully.  I would do the image processing on the FPGA, and then control the air valves from an FPGA.  I would want the entire system to be FPGA and RT to make sure the timing all stays right.  You don't want Windows messing up your timing on something like this.

Once your image processing finds a defect, you just queue a delay and fire the air valve after the delay.  You tweak the delay until you are getting the right timing.  Depending on your air valve system, you might be able to eliminate individual grains.

Bruce

[thsiriwuth · Accepted solution]

Hi Bruce

Thank you very much for your suggestion. I estimate about 20-40 mS from taking picture to reject by valve.

Camera link with FPGA seems interesting. I will have a look and study on it. I have some more question if you could help.

Question1; Can I use Line scan camera for this application ? or area scan ?

Seem the line scan more suitable for this kind of application. Because the line scan camera provide higher scan rate

and when the object move like web stream continuously, The ROI should be thin height and wide that also should better to use

line scan camera. What do you think ? 

Question2; How many I/O chanel for FPGA

For the nozzle ejector, It propably has about 20-40 nozzle with valve for each. FPGA has enough I/O to direct command the valves or not ?

Best regards,

Siriwut H.

[MangeshD · Accepted solution]

Adding to Bruce's suggestion if you want to go for FPGA there are lots of possibilities that you can refer to.

http://www.ni.com/white-paper/11417/en/

http://sine.ni.com/nips/cds/view/p/lang/en/nid/208914

[BruceAmmons · Accepted solution]

Line cameras are definitely the way to go for this project.  You read a line, process the line, then decide if you need to do any reject signal, then loop and do it all again.

Image processing is a little different on an FPGA, because you have to handle all the data in a single stream.

You would need to look at the options to figure out the best combination of input and output.  Usually cRIO works very well for output, because you can pick the module you need for a specific application.  I would need a lot more information about your hardware, and I would need to spend some time looking at all the current options to put together the right hardware for you.  That is work that I would need to charge for, unfortunately.

Bruce

[thsiriwuth]

Hi Everyone

Thank you very much for your recommendation.

Best regards,

Siriwut H.