Automatic Image Processing for Fiber Structures

X. Chen, R. Stoddard
Argonne National Laboratory,
United States

Keywords: fiber, nanofiber, nanotube, alignment, orientation, nano particles


Advances in fabrication and engineering of nanoscale materials such as nanofibers, nanorods, nanotubes, and nanoparticles requires precise morphology characterization by analyzing micrographs from microscopy. Human characterization measurements of parameters such as nanofiber orientation, mean diameter, and diameter standard deviation from scanning electron microscopy (SEM) images can be slow and inaccurate. The presented system uses existing edge detection algorithms with a Hough transform and an appropriate edge paring algorithm to recognize and measure each fiber orientation angle and diameter for assembly characterization. This new system has an intuitive approach of recognizing each fiber for robust performance on nanofiber assemblies with measurement difficulties typical in real-life images such as beads, curved fibers, and varying area density, which have prevented any existing method from being practically adapted. Fiber alignment and diameter standard deviation also show good agreement with manual results. The performance of our new method offers significant improvements over existing methods and has become a dependable tool in our research environment. Our innovative approach of combining edge detection with Hough transform for feature detection and measurements also lays the ground for robust image analyze on a variety of nanostructures such as nanotubes and nanoparticles.