FiberFind: Machine learning-based segmentation and identification of individual fibers in µCT images of fibrous media — ICTMS Committee Dep of Applied Maths Australian National University
  1. Schedule
  2. Theme 2: Automated methods for volume segmentation/analysis, e.g., machine learning
  3. FiberFind: Machine learning-based segmentation and identification of individual fibers in µCT images of fibrous media

FiberFind: Machine learning-based segmentation and identification of individual fibers in µCT images of fibrous media (135)

Rolf Westerteiger 1 , Andreas Grießer 1 , Christian Wagner 1 , Andreas Wiegmann 1
  1. Math2Market GmbH, Kaiserslautern, RHEINLAND-PFALZ, Germany

A machine learning-based approach to segment curved fibers in a binarized µCT image is presented. A neural network has been trained to label the center curve of each fiber using synthetic training data generated with the GeoDict software. The analysis of these centerlines provides an analytical representation of individual fibers, allowing for analysis of diameter, orientation distribution, curliness, and more.

Full Paper