loading
 
Find circle shapes
Kresimir Majdenic from Croatia  [38 posts]
2 years
Any idea how to find circle shapes in image [Circle detect 01.jpg]. There should be 3 circles [Circle detect 02.jpg].

  
Kresimir Majdenic from Croatia  [38 posts] 2 years
I've found that someone already tried to find solution to similar problem [Antibiogram.jpg], [Antibiogram - reading.jpg] but it has been done with OpenCV (http://stackoverflow.com/questions/25667053/find-overlapping-complex-circles-with-opencv).

Can it be done with RR?

  
Steven Gentner from United States  [260 posts] 2 years
Can you post the original image?

STeven.
Kresimir Majdenic from Croatia  [38 posts] 2 years
Here it is [Image.jpg]

 
Kresimir Majdenic from Croatia  [38 posts] 2 years
Here is the complete image [Image 01.jpg] - ignore background :)

 
Steven Gentner from United States  [260 posts] 2 years
Can I assume you are looking to put a circle around the shiny/wet part around the tablets in the petri dish? I assume knowing where the tablets are is probably not sufficient to estimate this.

Do you have any possibility to play with the lighting? The contrast between the sample area and the background is very faint which is why you are having issues in quantifying that. In the images that you posted wrt to OpenCV the contrast is much better and much easier to segment and thus quantify.

You may need to use backlighting (i.e. shine a light behind the sample) to get a better contrast. Or try tilting the light or adding more light to see if this provides better contrast. Or try putting a white paper underneath the petri dish to see if more reflections helps.

Its possible to get a result using your current image but that will NOT be consistent ... best to try first to improve the contrast.

STeven.
Kresimir Majdenic from Croatia  [38 posts] 2 years
You assume right. I would like to measure diameter of circular area around tablets.
How about this [Image 03.jpg] one?

 
Steven Gentner from United States  [1370 posts] 2 years
Yes, that's *much* better. Attached is a robofile that is one interpretation on how to do this with the resulting annotated image. Note, the sample in the center left ... not sure how that should be interpreted. There is some sign of a circle but perhaps not enough to be quantified. At any rate, this is up to the specialists to decide if that should be ignored or not.

Note, I suspect this technique is very sensitive to lighting ... many more images need to be tested to ensure this technique is stable for the sample set.

The resulting CIRCLE_COORDINATES variable array contains x,y,radius of each circle.

STeven.

 

program.robo
Kresimir Majdenic from Croatia  [38 posts] 2 years
Very nice indeed.
Yes I've realized that lightning is of great importance in this problem and already working on appropriate camera/light design.
I also am not sure how to interpret center left one but regardless of that I think your solution is very elegant.

Once again, thank you very much :)

This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index