The Emerald Project: Redefining Autonomous Textile Sorting through Robotics and Near-Infrared Spectroscopy

From fashion to pure functionality, textiles shape our everyday lives; however, behind every garment lies a complex system of production, consumption, and waste, with an influence that extends far beyond personal use. At least in EU, textiles represent the fourth highest pressure category for climate change and use of primary raw materials, the fifth for greenhouse gas emissions, and the third for water and land use. In addition, synthetic textiles, such as polyester, release micro-plastic fibers during washing, posing risks to marine life and ecosystems. This work is part of a broader European research project called "Emerald" and an innovative textile sorting system was developed by coupling NIR spectroscopy with robotic manipulation, aiming to provide a scalable and efficient alternative to existing hyper/multispectral camera setups and conveyor-based sorting lines, largely deriving from established technologies used for sorting plastic materials. The presence of the robotic manipulator opens new perspectives for a more autonomous, complete, safe and three-dimensional analysis of textiles. This thesis provides a comprehensive description of the entire system’s development and testing, together with an overview of the state of the art in textile sorting. A central part of the work was devoted to the implementation of the key software components, including a segmentation module for textile identification, the spectral acquisition program (for the NIR spectrometer, operating between 1100 and 2500 nm) and the global HMI (developed entirely in LabVIEW) for the control of the communication of all sensors and vision systems with the six-axis robotic arm, fully programmed in Wincaps III. The thesis also presents the design and construction of a custom interface flange, a key mechanical component also produced through 3D printing, which physically integrates three main elements: a special needle gripper, a laser system used as a distance sensor, and the spectrometer head, connected via optical fiber to the main device. Furthermore, the work analyzes the rule-based algorithms implemented to conduct initial forms of composition characterization and recognition, operating on data collected by directly testing certified samples.