Digital Extraction
Deep learning algorithms designed for pixel-level classification allow high-throughput extraction of structural features from fiber images. In linen research, U-Net semantic segmentation is used to separate the flax fiber cross-sections from the background resin in microscopic images. The network architecture consists of a contracting path that captures context and a symmetric expanding path that enables precise localization.
This method allows the automated measurement of individual fiber diameters and lumen sizes.
Computational Analysis
Processing irregular and touching flax fiber cross-sections is a primary challenge for standard image analysis. The U-Net semantic segmentation model resolves this issue by learning the typical boundary shapes and cell wall patterns from training datasets. It generates binary masks that clearly separate individual fibers, even when they are clustered tightly together.
This level of detail is necessary to calculate the cross-sectional area bias and fiber roundness. The resulting datasets provide a detailed representation of the fiber distribution in a yarn strand.
Automated Inspection
Integrating this segmentation model into the quality control workflow accelerates the grading of flax fiber shipments. Fast, automated analysis reduces the reliance on manual measurement and improves the reproducibility of results across different mill laboratories. This consistency supports more reliable fiber blending decisions.
