Feasibility Study of Visual Computing and Machine Learning Application for Textile Material Sorting

Aims And Objectives

This project aims to study the feasibility of using VC and ML method applying in Textile Recycle Industry for efficiently manages the post-consumer textile waste. It includes an image-based VC technology for supporting textile waste reuse and resale, and a material identification system for sorting textile materials by using NIR / hyperspectral spectroscopy to support efficiently recycling to reuse the post-consumer textile fibre. The process involved collecting and validating reference samples, apply ML technique to auto recognize the garment type in visual technology, afterward, the sorted garments would be measured and pre-treated by NIR / hyperspectral spectrum, and building up the parameters for spectral patterns calculation for textile recycle process recover the fibre. The main part of the project is to proof of the concept for using VC and ML method for identifying the textile fibre in the recycle process. Figure 1 is the conceptual architecture of VC and ML for textile material sorting. The incoming bulk garment will be input to this vision system for training up the proposed model and database for textile material identification.

Figure 1.

Conceptual architecture of visual computing and machine learning for textile material sorting

Figure 2.

Example of main feature and footprint of the bulk incoming garment