Development and Preliminary Validation of an Automatic and Intelligent System (fiber Mult) for Cross-Sectional Evaluation of Animal Fibers

This paper describes the development and preliminary validation of an automatic system based on artificial intelligence. The fiber cross-sectional and related properties of alpaca, llama, and mohair fibers were investigated and fiber perimeter, area, major and minor axis, medullation, visually objectionable (VOF) and visually nonobjectionable fibers (VNOF), medulla dimensions, and fiber density were measured. For each sample, 8–12 cross-section images were captured. Each image captured on the Fiber Mult contained cross-sections of some 70–300 fibers. To identify VOF, samples from 52 alpaca animal were hand-sorted into VOF and VNOF and both sets were tested with the Fiber Mult. The results for fiber density from the automatic capabilities of the Fiber Mult were compared with results from direct visual counting. The developed methodology enabled measurement of each sample in 60 s, with comparable results to traditional methods. For all the animals, the medullated fibers were coarser and more elliptical than the nonmedullated fibers. The small contribution of the repeatability (≈2%) and reproducibility (≈0.7%) to the total variation, when the Fiber Mult measured different characteristics demonstrated its good precision. The Fiber Mult will have potential application in fiber-textile evaluation and for genetically improving the fiber quality of fleece-bearing animals.