摘要
采用便携式近红外光谱仪分别采集了羊毛/棉、羊毛/马海毛、羊毛/氨纶、羊毛/丝、羊毛/羊绒5种羊毛混纺面料的近红外光谱,利用支持向量机算法(SVM)分别对原始光谱和经归一化预处理后的光谱建立分类模型。选用径向基函数(RBF)作为核函数,并采用网格搜索法(Grid Search)、遗传算法(GA)和粒子群算法(PSO)对惩罚参数c和核函数参数γ进行参数寻优。结果表明:PSO-SVM模型分类结果最理想,模型对训练集整体分类准确率为100%,对验证集的整体分类准确率为94.87%,其中羊毛/棉、羊毛/马海毛、羊毛/氨纶3类面料的分类准确率均为100%,羊毛/羊绒的分类准确率为95%,羊毛/丝的分类准确率相对较低为85%。
In this paper,we collected the near infrared spectra of wool / cotton,wool / mohair,wool / spandex,wool / silk and wool / cashmere blended fabrics using a portable near infrared spectrometer,support vector machine（ SVM） algorithm was used to establish the classification models of both the raw spectra and the normalized preprocessed spectra. The radial basis function（ RBF） was used as the kernel function and the grid search method,genetic algorithm（ GA）,particle swarm optimization（ PSO） were utilized for parameter optimization to obtain the best penalty parameter c and kernel function parameter γwhen establishing classification models. The results showed that the classification results of PSO-SVM model by raw spectra were best,the classification accuracy rate of training set was 100% and the classification accuracy rate of validation set was 94. 87%,among the validation set,the classification accuracy rate ofwool / cotton,wool / mohair,wool / spandex blended fabrics were 100%,the classification accuracy rate of wool / cashmere blended fabrics was 95% and the classification accuracy rate of wool / silk blended fabrics was relatively low with 85%.
出处
《毛纺科技》
CAS
北大核心
2016年第4期1-5,共5页
Wool Textile Journal
关键词
近红外光谱技术
支持向量机
羊毛混纺面料
径向基函数
near infrared spectroscopy
support vector machine
wool blended fabrics
radial basis function
