Modeling of fabric motion based on small videos

Abstract

Fabric recognition remains an ongoing challenge caused by complicated nonlinear and anisotropic elastic behavior of textiles. In this paper, we devise a computer vision system to automatically analyze the interaction forces of fabrics from videos. We start developing this textile force model by considering particle advection based on optical flow. Our model, which utilizes optical-flow measures of particle interactions within fabrics, incorporates the latent dirichlet allocation model to furnish textile classifications. We validate the correctness of this model by classifying fabrics from videos into an array of textile categories. Our empirical study demonstrates that the fabric classification system powered by our force model is capable of distinguishing different textile materials recorded in videos. We also demonstrate that the interaction force of the same kind of textiles is equal under various unknown wind forces from videos. Our approach to modeling interaction force of textile materials opens a new door to research on future textile recognition technologies.

KEYWORDS:

• Force model
• optical-flow
• small videos
• fabric recognition

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work is supported in part by the Science Foundation of Hubei under Grant No. 2014CFB764 and Department of Education of the Hubei Province of China under Grant No. Q20131608, and Engineering Research Center of Hubei Province for Clothing Information. Xiao Qin's work is supported by the U.S. National Science Foundation under Grants IIS-1618669, OAC-1642133, and CCF-0845257 (CAREER).