Golan Levin and Collaborators

Flong Blog + News

Machine vision vs. rice, pineapple, nuts

3 October 2008 / general

Reading the questions posted to the OpenCV (Open-Source Computer Vision) mailing list is one of my small pleasures in life. My favorite inquiries are those which combine charmingly flawed English, an improbable or unexpected problem domain, and the implicit expectation that the writer’s problem is not only solvable as described, but has a well-understood solution that everyone must have surely seen before. (These posts form a welcome relief, in any case, from the seemingly endless chatter of spooks asking how to improve their surveillance techniques.) This past week has offered up another bumper crop of gems, from a gaggle of internationals apparently charged with various tasks in food quality control. I give you:

Fun with Food on the OpenCV List

Subject: pineapple segment
From: sunrainhard
Date: Mon, Sep 29, 2008
Now i am researching on the pineapple project.
At first i must segment the pineapple in image.
Could you give me some advise about pineapple segment or pineapple
model? That’s very nice of you if you do me a favor.


Subject: how can I count the rice on the image?
From: Ilyas
Date: Thu, Oct 2, 2008
I have just starting learning OpenCV. I am more comfortable when I use Matlab but it is time to start OpenCV. I would like to know that anyone knows how to count the rice on the image ?
Best Regards,

OK – I’ve got one more. And to be fair, this person seems to speak English and have a clue or two about machine vision. I just love the humility of the problem.

Subject: Surface Crack Detection
From: rumsdiegeige
Date: Wed, Oct 1, 2008
I’m trying to detect cracks on the surface of nuts using a greyscale image. The cracks are oriented radially and are a little less bright than the normal surface. However, the cracks do not have a certain width. I know this question is quite general but nevertheless I am looking forward to your opinion how to proceed with that detection. I am currently thinking to apply the top-hat transform to the image. My boss said one could use a “radial FFT” – does that exist?

Comments are closed.

« Prev post:
» Next post: