LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

Automatic Region Detection

Can some one help me with an algorithm for auto region detection to detect the region marked in the image. This region marked in red wont be in the same location and will be displaced in every single object imaged.

Download All
0 Kudos
Message 1 of 4
(2,460 Views)

Can some one help me with an algorithm for auto region detection to detect the region marked in the image. This region marked in red wont be in the same location and will be displaced in every single object imaged.

0 Kudos
Message 2 of 4
(2,458 Views)

Post what you have and where you got stuck. What have you tried?

 

It's really hard to say what works for your situation, without analyzing a few hundred images covering the entire range of situations to expect. And than starts to sound a lot like work Smiley Tongue.

 

Some ideas (in order of appearance in my mind):

 

+ Detect the large circle with a (local) threshold, and make it smaller to match the inside circle.

+ Use a (local) threshold to detect the stuff in the inner circle. Calculate a bounding box around it (with optional added margins).

+ Use a Hough circle detection.

+ Use object analysis VI's.

+ Use a caliper.

0 Kudos
Message 3 of 4
(2,385 Views)

It is also important what else can be on the image and what can not be.

Do background and image intensities vary? Can you set a constant threshold on intensity?

Does object size vary? Can you set constant size threshold?

How precise and fast do you need to detect the circle?

 

In the most simple case I would threshold by intensity, and get the position and size of the largest blob and check the size threshold. I would say searching for specific shapes (circle detection) is always slower than getting general properties. 

0 Kudos
Message 4 of 4
(2,354 Views)