Industry: analyze image to check features on a product

imagej

#1

Dear members,

Now for something a bit different, hope you can help.
This is about industrial usage of image analysis software.

I am already using ImageJ for porosity testing and measurements in the die casting industry, and that task is pretty much trivial and flawless.

Now I have to come up with idea for quality control of finished parts. A picture of finished part(s) would be automatically taken and loaded into anlysis software. Software should, I hope so, analyze that image and provide an OK/NOK output.

I attached two images showing how those parts look like, taken at same angle as it will be done if this method becomes operative.
There is only one feature that should be checked (x3): the presence of thread in marked holes. So, if those holes are threaded, output is OK, and if not, output is NOK.

I hope these Images will make it more clear:

Images are .jpg, but those are just for reference.
Please give me some rough guidance on this problem, each word in your reply is much appreciated!

Regards,
Zoran


#2

Good day Zoran,

if the objects are positioned and photographically captured in a reproducible fashion, things are rather easy. I highly recommend to use some sort of template that fixates the object in a defined position and a camera stand with a fixed position with respect to the object.

If you can’t guarantee defined positions, things get much more involved.

Regards

Herbie


#3

Hi Herbie,

yes, both camera and part(s) to be analyzed will be fixed in a defined position. So, absolutely reproducible.
Glad to hear that solution could be simple.

I am also interested to learn the procedure if those positions are not guaranteed, for future use, a very rough guide would be appreciated (just to know where to start learning).

Thank you for your answer!

Regards,
Zoran


#4

Zoran,

I forgot to mention that a third conditionn should be met as well:
The position of the light source should also be fixed with respect to object and camera.

The two posted sample images violate all three conditions:
There is a slight rotation and perhaps even a minor scale difference and the lighting is different.

If you can’t guarantee reproducible conditions I’d first search for the three elliptic surfaces from which the holes start. There are various ways to do this but I know of none that is easy to realize if a highly reliable detection is desired.

Regards

Herbie


#5

Hi Herbie,

Sample images are here just to explain this problem better than my so-so english. Those were taken with smartphone on my desk, just to illustrate things I wish to accomplish. Final assembly will guarantee reproducibility, and will use much better camera and a proper light source.

Thank you,

Regards,
Zoran


#6

Herbie, anyone??? Please!

Regards,
Zoran


#7

Herbie, anyone??? Please!

What?

Please be so kind and post seriously taken and typical unprocessed images in the original TIF- or PNG-format.
No JPG-format though, because JPG introduces artifacts!
(Converting a JPG-compressed image to TIFF- or PNG-format doesn’t make sense.)
You may also post images as Zip-archives or make them accessible via a dropbox-like service.

Without such images it won’t pay off for me to start thinking about any processing.

Regards

Herbie


#8

Oh, sorry!!!

I thought that pictures i included are enough for rough guide for me to start playing with. Sorry for my ignorance and thank you for your help, I will post new images taken in a proper manner as you described!

Regards,
Zoran