colour to grey scale and crack detection

[saillesh]

how to convert a colour image to grey scale. also i want to know how to detect the cracks in the objects from their images

[muks · Accepted solution]

There are 3 planes involved in a color image. Each plane having values 0-255. By extracting any single plane you can convert the image to a grey scale one.

Try extracting the luminance plane (in color plane extraction).

Use vision assistant initially to get an idea.....

[muks]

---

how to detect the cracks in the objects from their images

---

If the crack in really nicely different from the object then applying a proper threshold and removing small particles and going with a particle analysis will be enough.

But we need details.If you can post some images then I can suggest you a method

[sudeepn]

Use the vision assistant to start with ,as it will help you extract the colour planes and (also gives an option to convert your image to a grey image) as well as it will help you learn about other vision utilities which labVIEW offers . You can detect the cracks or edges by setting proper thresholds . Please describe your application in a little detail so that I can help you with a more precise solution.

[wbadry]

Saillesh,

What muks suggested is a very practical approach. I personally implemented it with object filtering to detect cracks in an online ceramic plates inspection.

[Jörn]

to convert a color image to gray scale you can average all three planes or you use different weights for the color planes. Typical weights are

30% red + 59% green + 11% blue

See also http://forums.ni.com/ni/board/message?board.id=170&message.id=422864#M422864

Message Edited by Jörn on 2009-08-12 12.08.2009 11:48 PM

[saillesh]

i wan to detect the cracks in tracks of train .can i use fft for this . if any other soluion please help me

[saillesh]

i have to inspect cracks in the train  tracks . is your method of object filtering is suitable for that ?.

[muks]

Well posting a few sample images will do a lot of help. I mean there is no other way  we can suggest algorithms to use. We can only keep guessing.

[saillesh]

im posting my pictures here  (with and without cracks),could u help me with the algorithms.i have taken fft for this, is there any option to get the strength of fft to compare the images?