Rapid Qualitative Analysis of Cashmere and Wool Mixed Content Based on Improved XGBoost and SPA Using NIR Spectroscopy

In the textile industry, the deviation between the actual content of cashmere fibers and the standard content has a significant impact on product pricing. Conventional testing methods are plagued by long cycles and complex procedures. Therefore, how to quickly and accurately measure the blending ratio of fibers has become a key challenge in the field of fiber testing. This paper combines Near Infrared (NIR) spectroscopy technology with an improved Extreme Gradient Boosting algorithm to perform quantitative analysis on wool fiber samples from different origins. After preprocessing the spectra using maximum–minimum normalization, a model is established based on the successive projection algorithm to select wavelengths. Validation results indicate that this model significantly can improve prediction accuracy and reduce operation time, providing a convenient, rapid, and efficient method for distinguishing between cashmere and wool content.