Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Non linear continuous crack detection

Hello

 

I am trying to figure out an effective method for Vision Builder to have an automated system to detect cracks and defects.

We were able to filter out the image to have a consecutive line of pixels.  My question is "How do I make Vision Builder to determine

that the continous pixels are in fact a crack so it can fail the process?"

 

Here is a picture of what I am working with

[IMG]http://img.photobucket.com/albums/v356/stickydiljay/crack.jpg[/IMG]

 

I appreciate any input. Thanks in advance

0 Kudos
Message 1 of 6
(3,013 Views)

Hey Diljay,

 

I believe you can do this in multiple ways:

1- Edge Detection

2- Morphological Analysis, fining particles and their sizes

3- Pattern Matching

 

I hope this helps 

 

Regards,

A. Zaatari

National Instruments
Applications Engineer
0 Kudos
Message 2 of 6
(2,980 Views)

I need to detect the crack in eggs can someone please help me on this 

0 Kudos
Message 3 of 6
(2,549 Views)

As a starting point, I would try using a threshold function to see if you can pick it out sufficiently. You might also explore lighting options to help eliminate shadows, such as using multiple lights at different angles.

--------------------------------------
0 Kudos
Message 4 of 6
(2,514 Views)

sir i tried edge detection and then i tried to remove the crack alone from the egg shell but i couldnt can you tell me how to do it 

herewith i have attached the edge detected image of egg for your reference

0 Kudos
Message 5 of 6
(2,498 Views)

Before you get too far on the algorithm I would spend some time on the image acquisition.  You should try to get some lighting to eliminate shadow and glare and select a background that will be quite different from the egg if possible.  Now you can seperate egg from background and mask the egg.  I would probably follow this with a threshold to get the crack candidates and filter out particles that tont have characteristics of a crack like the thickness of the particle is too wide (I have done this using a danielson on the bianry and refiltering by max distances from edge) .    Would have to study many samples to understand the nature of what defines a 'crack'.

Paul Falkenstein
Coleman Technologies Inc.
CLA, CPI, AIA-Vision
Labview 4.0- 2013, RT, Vision, FPGA
0 Kudos
Message 6 of 6
(2,428 Views)