Abstract—In order to reconstruct the three-dimensional (3D) point cloud of a draped fabric based on a two-dimensional fabric drape projection, the three-dimensional point clouds of the draped fabrics were scanned with a self-built 3D scanning device. A resampling method based on local linear embedding (LLE) was used to represent different 3D point clouds with the same point number and point sequence. Principal Component Analysis (PCA) was used to reduce the dimension of the resampled 3D point clouds. With PCA, a completed resampled point cloud could be represented with a signature of length fifty-seven. At last, a regression model with a two- dimensional (2D) fabric drape projection as input was constructed and trained to predict the signature of length fifty-seven. With the predicted signature, the 3D point cloud of a draped fabric could be reconstructed. The result shows that all resampled 3D point clouds of draped fabric have the same point number and point sequence. The errors between the reconstructed 3D point clouds and the ground truth are all within 6.92 mm.
Index Terms—Draped fabric, 3D triangular mesh, 2D projection, PCA, deep learning.
Zhicai Yu and Haoyang Xie are with the College of Textiles, Donghua University, Shanghai 201620, China (e-mail: zhicai-yu@qq.com; xie_haoyang@qq.com).
Yueqi Zhong is with College of Textiles, Donghua University, Shanghai 201620, China and the Key Laboratory of Textile Science & Technology of Ministry of Education, College of Textiles, Donghua University, Shanghai, 201620, China (corresponding author; e-mail: zhyq@dhu.edu.cn).
Cite: Zhicai Yu, Yueqi Zhong, and Haoyang Xie, "Reconstructing the Three-Dimensional Point Cloud of a Draped Fabric Based On a Two-Dimensional Projection," International Journal of Machine Learning and Computing vol. 11, no. 6, pp. 413-417, 2021.
Copyright © 2021 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).