1

I want to count how many peanuts there are in this picture using image segmentation K-MEANS , does anyone know how to go about this?

peanutzzzz

Rody Oldenhuis
  • 37,726
  • 7
  • 50
  • 96
user1802057
  • 125
  • 1
  • 2
  • 4
  • 4
    For future reference, please refrain from phrasing your question like "gimme the codezz plz"; at least show [what you have tried](http://mattgemmell.com/2008/12/08/what-have-you-tried/). – Rody Oldenhuis Nov 06 '12 at 06:24
  • 1
    Take a look [here](http://blogs.mathworks.com/steve/2006/06/02/cell-segmentation/), and [here](http://stackoverflow.com/questions/7989818/how-to-count-number-objects-in-an-image-identify-by-color-in-matlab). – Rody Oldenhuis Nov 06 '12 at 06:30
  • as @RodyOldenhuis says, please, ask specific questions. don't just ask people to make your program! – Ander Biguri Nov 06 '12 at 15:15
  • correlation clustering is a better suited objective function for clustering where the main aim is detecting the underlying number of clusters. – Shai Dec 02 '12 at 18:59

1 Answers1

0

Late answer, but still. I would use the clustering in order to separate the peanuts from the background. But a color thresholding or a dominant color analysis should be enough and give comparable results. Except that I would consider each dark area (shadows) as the background in order to separate the peanuts.

Then you get a binary mask, and you can apply an ultimate eroded in order to estimate the number of peanuts.

FiReTiTi
  • 5,597
  • 12
  • 30
  • 58