Image Processing Techniques for Yarn Characterization

Image processing has become a cornerstone in quantifying yarn properties that were once assessed solely by mechanical testers. Core steps involve image acquisition under controlled illumination, preprocessing routines such as denoising, binarisation and morphological operations, followed by image segmentation to extract the yarn core and protruding fibres. Early approaches relied on static thresholding and graph-cut methods to differentiate background from yarn, while statistical modules quantified parameters such as diameter, hairiness and faults. Multifocus fusion techniques addressed depth-of-field limitations by combining images at varying focal planes to yield sharper fibre edges. More recent pipelines incorporate skeletonisation to trace fibre branches and apply the circular Hough transform to detect individual filament cross-sections. The advent of deep learning has further elevated detection accuracy, enabling real-time classification of irregularities and automated measurement of structural features. Such developments hold global significance for textile manufacturing, as they facilitate quality control, reduce waste and enable new insights into yarn behaviour under processing conditions.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

A novel deep learning framework based on an enhanced YOLOv5 algorithm has been developed for yarn hairiness characterisation. By treating loop and protruding fibres as distinct object classes, the model attains improved precision and recall, boosting mean average precision by up to 12% over standard architectures. A rigorous cross-validation protocol ensures robustness across diverse yarn types.

An advanced algorithm for cross-fiber separation employs image pretreatment and skeleton extraction to pinpoint fibre intersection points. A branch-matching strategy based on slope adjacency reconstructs complete fibre paths, yielding hair length measurements that closely mirror manual counts and outperform conventional photoelectric systems in resolving overlapping hairs.

Automated diameter measurement has been realised through MATLAB-based code that integrates binarisation, morphological filtering and the circular Hough transform. This pipeline delivers rapid and accurate determination of yarn and individual fibre diameters from microscopic images, matching manual measurements while reducing operator intervention.

Image Processing Techniques for Yarn Characterization publication trend

The graph below shows the total number of articles in image processing techniques for yarn characterization across all publications each year (not limited to Nature Index journals).

Technical terms

Yarn hairiness: The population of fibres protruding from the main yarn body, influencing fabric appearance and performance.

Image segmentation: The process of partitioning an image into regions corresponding to objects or features of interest.

Binarisation: Conversion of a greyscale image into a two-level image differentiating foreground from background by thresholding.

Morphological operations: Shape-based transformations (e.g. erosion, dilation) applied to binary images to refine structures.

Circular Hough transform: A feature extraction technique for detecting circular shapes within an image.

Skeletonisation: Reduction of objects in a binary image to a one-pixel-wide representation, preserving topological structure.

Deep learning: A class of machine learning using multilayer neural networks to learn hierarchical representations from data.

YOLOv5: A real-time object detection framework optimised for speed and accuracy through single-stage convolutional neural networks.

References

1. Fusing multifocus images for yarn hairiness measurement. Optical Engineering (2014).
2. An algorithm for cross-fiber separation in yarn hairiness image processing. The Visual Computer (2023).
3. MATLAB Algorithms for Diameter Measurements of Textile Yarns and Fibers through Image Processing Techniques. Materials (2022).
4. A Novel Deep Learning Approach for Yarn Hairiness Characterization Using an Improved YOLOv5 Algorithm. Applied Sciences (2024).