Open AccessPerformance of the pilling evaluation method based on the technique of DFF
Author(s) -
Yu Liu,
Rongwu Wang,
Jinfeng Zhou
Publication year - 2017
Publication title -
industria textilă
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.281
H-Index - 14
ISSN - 1222-5347
DOI - 10.35530/it.068.01.1295
Subject(s) - support vector machine , artificial intelligence , classifier (uml) , computer science , grid , hyperparameter optimization , pattern recognition (psychology) , image (mathematics) , focus (optics) , computer vision , mathematics , physics , geometry , optics
In previous work, we reconstructed the depth image of fabric based on the method of Depth from Focus (DFF) andsegmented pills and fuzz from fabric background. Work in this paper was performed using the segmented image. Here,we demonstrate the prediction operation of the pilling evaluation using a large set of fabric samples. The support vectormachine (SVM) was applied to build the classifier machine by learning from existing data. The grid search method wasused to select the optimal parameter values. The study found that the best prediction accuracy can reach 90.75%,indicating the extracted pilling features from depth image can predict the pilling grade well.