Exctract coordinates of continous fibres from microscopic image

Hello everyone,

I am fully aware, that my request is a little bit offtopic. But I am hoping, that some of your knowledge will be transferable to my problem

Sample image and/or code

Imagelink: https://i.postimg.cc/cHB4yLrD/Circle-Test.png

Background & Analysis goals

I am working with continuous fiber reinforced thermoplastics with a low fiber volume (~33%). Based on the microscopic images I want to retrieve the coordinates of the fibers. I want to feed these coordinates into a CAD program to create a model/accurately model the fibre distribution.

Challenges

I did find the “Hough Circle Transform” Plugin which is able to detect circles and even outputs the coordinates. But the problem is, that the fibres are not “circular” enough due to the polishing process, because edges are chipped off. While I am currently working on better polishing results, I would like to know if you have any ideas how to “repair” the fibres broken of edges with a software?

So basically a pre-processing step, before “finding the edges” ( Hough Circle Transform - ImageJ)

Kind regards
Mike

I assume that it is too offtopic?

Hi @FibreGuy , this is not just a BioImage analysis forum, it’s an image analysis forum so it’s perfectly on topic and thank you for providing an example image.

There are lots of similar segmentation questions on the forum (from blueberries to metal inclusions) and it’s likely what you’re after is some form of machine learning to assist in accurate object detection. The advantage of that being that you can include the chipped edges and cracks as part of the object when doing segmentation.

As a start, you might want to look into weka, ilastik or maybe stardist. If you try them and are having problems getting started, you can always post back including their tags in the post.

Hope that helps.

1 Like