Only 13% of the globally produced textile fibers are recycled, leading to fiber production having a large environmental impact. While many new technologies are emerging for the recycling of textiles, very little automated sorting capacity exists. This is a key problem, as fiber-to-fiber recycling requires material and colour sorting of the waste textiles. While textile material sorting has been done using NIR sensors, little literature exists on the subject of material sorting of waste textiles. Existing systems are experimental and based on fixed area sampling to determine the colours, which poses a limitation in performance for multi coloured textiles. This paper proposes a machine vision system for colour sorting of multicoloured waste textiles. The proposed system consists of three parts: (1) an object detection algorithm to define the boundary of the entire textile. (2) a discretization and colour determination algorithm which divides the textile area into a number of cells, and attributes a colour to each cell using colour descriptors. (3) a categorization algorithm which uses a descision tree to sort the textiles based on their colour distribution, using only a few sorting criteria. The proposed system was tested on a set of pre-sorted textiles and it was able to reach an accuracy of 88.3% with a mean computational time of 0.77 seconds for each textile. While a fixed area sampling method had an accuracy of 74.0%, it is concluded that the proposed system performs better, especially for multi coloured textiles.
|Conference||PETRA '23: 16th International Conference on PErvasive Technologies Related to Assistive Environments|
|Period||05/07/2023 → 07/07/2023|
- Colour detection
- Machine vision
- Object detection
- Textile sorting