基于光谱分析的织物色差检测

Fabric color difference detection based on spectral analysis

为有效检测出织物色差,将主成分分析(PCA)与极限学习机(ELM)相结合的方法应用于织物色差检测。首先,利用光纤光谱仪采集净色织物的光谱反射率数据及其对应的颜色特征值L*、a*、b*,建立PCA-ELM预测模型,获取光谱反射率相关数据与L*、a*、b*值之间的映射关系,避免冗长的公式计算。然后,利用CMC(2∶1)色差公式计算标准样本和测试样本之间的色差,并与光纤光谱仪测量值进行对比。实验结果表明,利用PCA-ELM模型获取的L*、a*、b*值与光谱仪测量值各分量的平均误差为0.116 1、0.174 3、0.204 8。在色差检测实验中,利用CMC(2∶1)色差公式得到的织物色差与光谱仪测量值最大误差为0.66 NBS,平均误差为0.087 5 NBS。

In order to effectively detect the fabric color difference,principal component analysis（ PCA） and extreme learning machine（ ELM） method were applied to the fabric color difference detection. Firstly,the spectral reflectance dataset of pure color fabric and corresponding physical color parameters of L＊,a＊,b＊value were acquired by the optical fiber spectrometer; the mapping relation between the spectral reflectance related dataset and L＊,a＊,b＊values was established by the PCA-ELM prediction model; the lengthy formula calculation was avoided. Then, the color difference between standard samples and test samples was obtained by the CMC（ 2 ∶ 1） color difference formula and compared with those spectrometer measured values. The experimental results showed that the average error of each component between L＊,a＊,b＊value obtained by PCA-ELM model and the optical fiber spectrometer measured values are 0. 116 1,0. 174 3 and 0. 204 8 respectively. In addition,the maximum error and the average error of the color difference between the CMC（ 2 ∶ 1） color difference formula calculation values and the spectrometer measured values are 0. 66 NBS and 0. 087 5 NBS respectively in color difference detection experiment.