基于PCA+KNN和kernal-PCA+KNN算法的废旧纺织物鉴别  

Identification of Waste Textiles Based on PCA+KNN and kernel-PCA+KNN Algorithms

作　　者：李宁宁 刘正东 王海滨 韩熹 李文霞[1] LI Ning-ning;LIU Zheng-dong;WANG Hai-bin;HAN Xi;LI Wen-xia(School of Materials Design and Engineering,Beijing Institute of Fashion Technology,Beijing 100029,China;School of Fashion Art and Engineering,Beijing Institute of Fashion Technology,Beijing 100029,China;Jifa Group Dyeing and Finishing Factory,Qingdao 266200,China;Beijing Weichuang Yingtu Technology Co.,Ltd.,Beijing 100070,China)

机构地区：[1]北京服装学院材料设计与工程学院,北京100029 [2]北京服装学院服装艺术与工程学院,北京100029 [3]即发集团染整厂,山东青岛266200 [4]北京伟创英图科技有限公司,北京100070

出　　处：《分析测试学报》2024年第7期1039-1045,共7页Journal of Instrumental Analysis

基　　金：中国纺织工业联合会“纺织之光”应用基础研究项目(J202204);研究生教改“新工科”背景下纺织科学与工程创新实践中心建设(NHFZ20230202)。

摘　　要：该研究采集了15类废旧纺织物的4 998张近红外谱图,以7∶3的比例分为训练集和验证集,并分别采用主成分分析(PCA)与核主成分分析(kernal-PCA)两种不同降维方法对数据进行降维,并选用余弦相似度(cosine)核作为kernal-PCA的最佳核函数,最后分别将PCA和kernal-PCA降维处理后的数据进行k-近邻算法(KNN)训练。结果表明,kernal-PCA+KNN的模型准确率(95.17%)优于PCA+KNN模型的准确率(92.34%)。研究表明,kernal-PCA+KNN算法可以实现15类废旧纺织物识别准确率的提升,为废旧纺织物在线近红外自动分拣提供有力的技术支撑。The study collected 4998 near infrared spectra of 15 types of waste textiles,which were divided into a training set and a validation set in a ratio of 7∶3,and the data were downscaled using two different downscaling methods,namely principal component analysis(PCA)and kernal principal component analysis(kernal-PCA),respectively,and the cosine similarity(cosine)kernel was se⁃lected as the best kernel function for kernal-PCA.Finally the PCA and kernal-PCA dimensionality reduction processed data are trained by k-nearest neighbour algorithm(KNN)respectively.The re⁃sults show that the model accuracy of kernal-PCA+KNN(95.17%)is better than that of PCA+KNN model(92.34%).The study shows that the kernal-PCA+KNN algorithm can achieve the improve⁃ment of the recognition accuracy of 15 types of waste textiles,and provide a strong technical support for the online near infrared automatic sorting of waste textiles.

关 键 词：废旧纺织物 主成分分析(PCA) 核主成分分析(kernel-PCA) k-近邻算法(KNN) 分类识别 

分 类 号：O657.3[理学—分析化学] TS106[理学—化学]