Eliminate the Round Edges

[Double-Monkey]

        I am working on a program which aims to detect cracks with Vision Assiatant.My target is to get the cracks only after a series of  operation,but there are three round edges that I can't eliminate them.I have uploaded the original images and script file.Does any of you can afford me with a few advice to eliminate the round edges?

        Looking forward to your reply!

[Hatef.fouladi]

there is vi in vision with name of
IMAQ Detect Texture Defect VI
try to use it i think it will be work for you

[Double-Monkey]

       When I just started doing this project,I tried Texture Detection.I spent enough time on this method,But the results were not ideal.Could you please give me some improvements?

[Hatef.fouladi]

indeed your fist step problem is your lighting system
you have to change it
check this link http://www.ni.com/white-paper/6902/en/
after that you can use some morphology function for improve crack feature and then you can solve your problem

[Hatef.fouladi]

indeed your fist step problem is your lighting system
you have to change it
check this link http://www.ni.com/white-paper/6902/en/
after that you can use some morphology function for improve crack feature and then you can solve your problem

[MangeshD]

Hi,

Looking at the images it looks like the processing is easy and detection of these cracks can be achieved using threshold+contour analysis.

But first of all you will need to change the lighting as suggested by Hatef. There are lots of articles on which lighting will be suitable of such crack detection.

For me I guess I will try with diffused lighting since the surface is shiny. Diffuse dome lights are very effective at lighting curved, specular surfaces, commonly found in the automotive industry.

[Double-Monkey]

OK,I will.Thank you very much!

Spoiler

[Double-Monkey]

       I have tried your method,but it can only detect its edges and overlook the crack completely.The results are attached,could you please make some improvements?Thank you very much!

[MangeshD]

Hey,

Looking at the image it seems to be the round edges(black) are in same threshold level as of the crack but if you will look closely the white (round shiny region) is also appears to be evenly bright.

Now you can go with threshold and eliminate method for ignoring the black corner region.

What you have to do...
If the gear is of similar kind and the illumination is same for all the gears you can create a template with the good gear and threshold it where the red ring will appear as extracted region. and then do the absolute difference of threshold template(good) with the threshold cracked image and after that doing a particle filter you can extract that black crack.

Note: We could only suggest you solution for this since for a code solution we will need a good and a cracked image.

[Hatef.fouladi]

could you send the image of your new lighting system
I still doubt that you select best lighting system