Matching Fabric Based on Wavelet Kernel Signature (WKS) and Drape Indicators

Abstract

The drape performance of the fabric is an important part of the fabric style, as well as a special expression of the structural parameters and mechanical properties. The comprehensive drape performance of fabric can support fabric matching. This study was conducted to match fabric based on 3D triangular meshes of draped fabrics. Firstly, the three-dimensional (3D) point cloud of draped fabric was scanned with a self-built 3D scanning device followed with triangulation. Secondly, three drape indicators, i.e., drape coefficient, the number of fabric drape nodes, and the solid rate of a draped fabric was extracted, as well as the wavelet kernel signatures (WKS) of triangular meshes. The WKS of vertices on the 3D boundary was selected and clustered into different classes with K-Means. With the cluster centers, the WKS of vertices on the boundary was encoded into a vector with a fixed length. Finally, two methods for matching fabrics were proposed based on WKS+DR and a vector consisting of three drape indicators separately. The results show that WKS+DR outperforms the vector consisting of drape indicators in characterizing the drape configuration details of a draped fabric. The accuracy of the method based on WKS+DR could reach 89.6%. The fabrics obtained with this method are close to the matching fabrics in both drape-ability and details of a triangle surface.