Detection of Duck Down in Goose Down Using Near Infrared Spectroscopy

Ying Zhou,Huirong Xu,Yibin Ying

DOI: https://doi.org/10.13031/2013.24856

2008-01-01

Abstract:Goose down is more luxurious than duck down, for this reason some factories replace the goose down with the duck down or mix goose down into duck down to gain more profit. Percent of duck down in goose down products is an important index during the process of international trading. The traditional method to distinguish between gooses down and duck down is using microscope by visual appraisal. But there are some problems such as limited samples and low accuracy. By last research, it has been proved that near infrared spectroscopy (NIR) is a useful method to distinguish duck down and goose down even in mixture. In this study, the feasibility of using near-infrared spectroscopy to determine duck down content in goose down was examined. And detectability of NIR used on down was also explored. 105 in all samples were the mixture of different proportion of goose down and duck down, that is 5 samples were made of 100% duck down and 0% goose down, 5 samples were made of 95%duck down, and 5% goose down 5 samples were made of 0% duck down and 100% goose down . The regression model of partial least squares (PLS) was selected for the calibration of all down constituents. Another 50 in all samples were also the mixture of different proportion of goose down and duck down, that is 25 samples were made of 100% duck down and 0% goose down, 25 samples were made of 80% duck down and 20% goose down, NIR method combined with chemometics was used to discriminate between them. If the 2 groups can be classified distinctly, more particular mixture (25 samples made of 90% duck down and 10% goose down,) were added. The process was go on until it couldnt be classified well. Spectrum of each sample was acquired using diffuse reflectance method by NEXUS intelligent FT-IR spectrometer. Chemometrics methods including PCA and DPLS were used to analyze the spectrum attained. Various pre-treatments of experimental data such as MSC and derivative were also used trying to make model better. The result shows that detectability of NIR used on down can be low to 5%, and it reach the detecting level of traditional methods.