Identification of varieties of cashmere by Vis/NIR spectroscopy technology based on PCA-SVM

One mixed algorithm was presented to discriminate cashmere varieties with principal component analysis (PCA) and support vector machine (SVM). Cashmere fiber has such characteristics as threadlike, softness, glossy and high tensile strength. The quality characters and economic value of each breed of cashmere are very different. In order to safeguard the consumer’s rights and guarantee the quality of cashmere product, quickly, efficiently and correctly identify cashmere has significant meaning to the production and transaction of cashmere material. This research adopts Vis/NIRS spectroscopy diffuse techniques to collect the spectral data of cashmere. The near infrared fingerprint of cashmere were acquired by principal component analysis (PCA), followed which support vector machine (SVM) methods were used to further identify the cashmere material. The result of principal component analysis (PCA) indicated that the score map made by the scores of PC1, PC2 and PC3. 10 principal components (PCs) were selected as the input of support vector machine (SVM) based on the reliabilities of PCs of 99.99%. 100 cashmere samples were used for calibration and the remaining 75 cashmere samples were used for validation. A one-against-all multi-class SVM model has been built, the capabilities of SVM with different kernel function were comparative analyzed, and the result showed that SVM possessing the Gaussian kernel function has the best identified capabilities with the accuracy of 100%. This research indicated that the data mining method of PCA-SVM has a good identification effect, and can work as a new method on rapid identification of cashmere material varieties.