抄録
All textile products are obliged to display the quality in order to protect the benefits of customers. For this reason, it is necessary to display the name of the composition fiber and the mixed ratio. In order to display the components, the type of fiber is identified and mixed ratio is analyzed quantitatively in a manner of the Japanese Industrial Standards (JIS). However, JIS method has many problems. To solve those problems, we study the rapid nondestructive inspection method using infrared spectroscopic measurement. In our method, we measure IR spectrum of Cotton/Polyester mixtures, and analyze the absorption information by Partial Least Squares Regression (PLSR). In this report, we tried to improve the prediction accuracy of mixed ratio by using Moving Window PLSR that determine the effective spectral region for constructing a calibration curve. As a result, the prediction error reduced and samples that satisfying the target precision increased.
