how can i process the image to find out the text differences in image.....

[Kalanga]

Hi all,

I am doing one vision based project. In my image four cylinderical part will be taken from the top view.In top of that cylinderical part the the model no will be punched(it is 6 character alphanumeric in two lines).All the four cylinderical parts should be from same model. if any other model mixed up with this means, the model no will vary. How can i findout? any suggestions, i tried pattern matching , geomatric mathcing and all its not working perfectly. the cylinderical part color is gray. the model name is punched on that cylinder. i cannot suppose to post the image here.

Thanks,

Kalanga 

[muks]

---

i cannot suppose to post the image here.

---

Then we can only suggest.You can use Optical character recognition (OCR).It would be helpful If you can give us more detials.

1. Will the orientation of cylinder be random?

2. What will the positional accuracy be?

 3. Can you reject/accept using color of the cylinder alone?

[Kalanga]

Thanks muks,

sorry for that.(image post)

1. Will the orientation of cylinder be random?

Yes, the orientation will be random. the texts punched on top the cylinder can come in any direction.

2. What will the positional accuracy be?

Positional accuracy will be 2-3mm. the dia of cylinderical part around 10mm. 

3. Can you reject/accept using color of the cylinder alone?

No, because all of the cylinders are same color.(gray)

I tried with OCR, but i am not getting correct output.it missed sometimes and give result for wrong name some time.

 Thanks,

Kalanga 

[muks]

Before going to processing images, Please ask your self whether this the best lighting I can use to iluminate the object. May I know what is your lighting?