光學智能分選紡織的設備系列-電控平台建構

Abstract

在快時尚浪潮下，全球每年產生的巨量紡織廢料已對環境造成沉重負擔。有效的回收再利用是實現循環經濟的關鍵，然而，第一步的「精準分選」卻是當前產業面臨的最大瓶頸。傳統人工分選方式不僅成本高昂、效率低下，其穩定性與準確率也難以應對龐大的回收量。為此，自動化分選成為必然趨勢，但現有技術多僅能處理塑料分選，普遍缺乏能同時整合處理紡織品「材質種類」與「顏色」兩大關鍵特徵的智能平台。有鑑於此，本研究旨在開發一套基於光學智能的紡織品雙模組自動化分選系統，以解決紡織品回收流程中，材質與顏色難以被同時精準辨識的難題。本研究設計並實作了一套整合光學感測與機器學習的自動化分選系統。此系統包含兩大核心模組：(1) 材質辨識模組，採用近紅外光譜 (NIR) 技術，以分析不同材質的光譜特徵；(2) 顏色分選模組，運用高解析度相機與HSV色彩辨識，對織物顏色進行精確分類。透過深度學習模型對大量樣本進行訓練，使系統能快速且準確地進行特徵的判讀與分選。本研究成功開發出一套能同時、快速且精準辨識紡織品材質與顏色的自動化分選解決方案。此系統不僅有效克服了傳統人工分選高成本、低效率的困境，更彌補了現有自動化技術在紡織品處理上的缺口。研究成果為快時尚廢棄物的回收再利用鏈條提供了關鍵的技術支持，對推動紡織產業的循環經濟轉型具有顯著的學術價值與高度的產業應用潛力。Driven by the fast fashion trend, the massive volume of textile waste generated globally each year has placed a significant burden on theenvironment. Effective recycling and reuse are pivotal for achieving a circular economy; however, the initial step of precise sorting remains the primary bottleneck for the industry. Traditional manual sorting methods are not only costly and inefficient but also lack the stability and accuracy to handle the vast quantities of collected textiles. Consequently, automated sorting has become an inevitable trend. Yet, existing technologies predominantly focus on plastic sorting and generally lack an intelligent platform capable of simultaneously processing the two key features of textiles: material composition and color. This study, therefore, aims to develop a dual-module automated textile sorting system based on optical intelligence to address the challenge of accurately and concurrently identifying both fabric material and color.This research designs and implements an automated sorting system that integrates optical sensing and machine learning. The system comprises two core modules: (1) a material identification module, which utilizes Near-Infrared (NIR) spectroscopy to analyze the spectral characteristics of different materials, and (2) a color sorting module, which employs a high-resolution camera and HSV color space identification to accurately classify fabric colors. By training machine learning models on a large dataset of samples, the system is enabled to perform rapid and accurate feature interpretation and sorting.This study successfully develops an automated sorting solution capable of simultaneously, rapidly, and accurately identifying the material and color of textiles. This system not only overcomes the challenges of high costs and low efficiency associated with traditional manual sorting but also fills the gap in existing automation technologies for textile processing. The research findings provide critical technological support for the recycling and reuse chain of fast fashion waste, possessing significant academic value and high potential for industrial application in promoting the circular economy transition of the textile industry.