Matched Filter Detector for Textile Fiber Classification of Signals with Near-Infrared Spectrum

Currently, the production process in the textile industry and post-consumption in the fashion industry generates the second largest of the world's environmental pollution since only one percentage of waste textiles is recycled back into clothing. For this reason, the development of technology plays an important role for the automatic classification of the waste textiles. This paper aims to develop a matched filter detector for textile fiber in near-infrared (NIR) spectrum whose wavelengths range from 1,350 nm to 2,500 nm. Firstly, the NIR spectrum is obtained from the NeoSpectra-Micro sensor placed one centimeter above the fabric. Then, a sample set of the NIR spectrum and the reference spectrum are enhanced by the 12-norm normalization. Next, the reversed reference spectrum is used to build a matched filter whose input is the sample set of the NIR spectrum. Finally, the filter output at the end sample is seen to be identical to the cross-correlation between the NIR spectrum and the reference spectrum, and it is applied to compare with a threshold to make decision on the qualitative classification. Experimental results show that the proposed classification method using the matched filter detector can realize the automatic classification of textile fiber into three groups: natural fibers, synthetic fibers, and blended fibers. Furthermore, the proposed filter detector can effectively improve the classification accuracy and precision.