2
$\begingroup$

I am trying to recognize the wood chips in the image shown below. The ultimate goal is to try and recognize the width and length of each wood chip. Currently, I am using an approach that involves bilateral filtering on a grey-level image + Canny edge detection + Hough line transform.

Results are somewhat iffy - can you recommend other algorithms to try out or is this a problem more suited to a statistical / machine learning approach.

enter image description here

$\endgroup$
2
$\begingroup$

It's a tough problem!

What's the exact meaning of recognize the wood chips in this case please?

Assuming you want to separate the wood from the container, I would propose the following processing pipe-line:

  1. for pre-processing, bilateral filtering is a good choice if it fits in your time budget since it will preserve the edges
  2. detect the enclosing circle by using a generalized Hough transform. If you implement your process with OpenCV, you don't even need to compute the gradient of the image before
  3. inside the container, apply some local contrast enhancement to get information up to the borders
  4. compute the image gradient inside the container, and finally segment it based on the gradient orientation (I can't really guess a better clue for segmenting the different chips inside).

Hope this helps...

$\endgroup$
  • $\begingroup$ I've edited the question to better describe the meaning of "recognize the wood chips". Thanks for your input - i'm considering it. $\endgroup$ – Hartmut Behrens Nov 27 '13 at 13:15
0
$\begingroup$

I think you should try some stable segmentation technique like mean shift segmentation or watershed transform. The codes are available both in Open CV and Matlab.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.