Graduate School, Faculty of Agricultural Sciences, Universidad Austral de Chile, Valdivia 5090000, Chile.
Animal. 2013 Jul;7(7):1219-25. doi: 10.1017/S1751731113000505. Epub 2013 Mar 27.
Rapid and efficient methods to evaluate variables associated with fibre quality are essential in animal breeding programs and fibre trade. Near-infrared reflectance spectroscopy (NIRS) combined with multivariate analysis was evaluated to predict textile quality attributes of alpaca fibre. Raw samples of fibres taken from male and female Huacaya alpacas (n = 291) of different ages and colours were scanned and their visible-near-infrared (NIR; 400 to 2500 nm) reflectance spectra were collected and analysed. Reference analysis of the samples included mean fibre diameter (MFD), standard deviation of fibre diameter (SDFD), coefficient of variation of fibre diameter (CVFD), mean fibre curvature (MFC), standard deviation of fibre curvature (SDFC), comfort factor (CF), spinning fineness (SF) and staple length (SL). Patterns of spectral variation (loadings) were explored by principal component analysis (PCA), where the first four PC's explained 99.97% and the first PC alone 95.58% of spectral variability. Calibration models were developed by modified partial least squares regression, testing different mathematical treatments (derivative order, subtraction gap, smoothing segment) of the spectra, with or without applying spectral correction algorithms (standard normal variate and detrend). Equations were selected through one-out cross-validation according to the proportion of explained variance (R 2CV), root mean square error in cross-validation (RMSECV) and the residual predictive deviation (RPD), which relates the standard deviation of the reference data to RMSECV. The best calibration models were accomplished when using the NIR region (1100 to 2500 nm) for the prediction of MFD and SF, with R 2CV = 0.90 and 0.87; RMSECV = 1.01 and 1.08 μm and RPD = 3.13 and 2.73, respectively. Models for SDFD, CVFD, MFC, SDFC, CF and SL had lower predictive quality with R 2CV < 0.65 and RPD < 1.5. External validation performed for MFD and SF on 91 samples was slightly poorer than cross-validation, with R 2 of 0.86 and 0.82, and standard error of prediction of 1.21 and 1.33 μm, for MFD and SF, respectively. It is concluded that NIRS can be used as an effective technique to select alpacas according to some important textile quality traits such as MFD and SF.
近红外反射光谱（NIRS）结合多元分析被评估为预测羊驼纤维的纺织质量属性。从不同年龄和颜色的雄性和雌性华卡约羊驼（n = 291）中采集纤维的原始样本进行扫描，并收集和分析其可见-近红外（NIR；400 至 2500nm）反射光谱。对样本的参考分析包括平均纤维直径（MFD）、纤维直径标准差（SDFD）、纤维直径变异系数（CVFD）、平均纤维曲率（MFC）、纤维曲率标准差（SDFC）、舒适因子（CF）、纺丝细度（SF）和纤维长度（SL）。通过主成分分析（PCA）探索光谱的变化模式（载荷），前四个 PC 解释了 99.97%的光谱可变性，而第一个 PC 单独解释了 95.58%的光谱可变性。通过偏最小二乘回归（PLSR）建立校准模型，测试了光谱的不同数学处理（导数阶数、减法间隙、平滑段），以及是否应用光谱校正算法（标准正态变量和去趋势）。根据解释方差的比例（R 2CV）、交叉验证中的均方根误差（RMSECV）和剩余预测偏差（RPD），通过单向外验证选择方程，RPD 与参考数据的标准差相关联 RMSECV。当使用 NIR 区域（1100 至 2500nm）预测 MFD 和 SF 时，最佳校准模型完成，R 2CV = 0.90 和 0.87；RMSECV = 1.01 和 1.08μm，RPD = 3.13 和 2.73。MFD 和 SF 的 SDFD、CVFD、MFC、SDFC、CF 和 SL 模型的预测质量较低，R 2CV < 0.65，RPD < 1.5。对 91 个样本的 MFD 和 SF 进行的外部验证稍逊于交叉验证，R 2 分别为 0.86 和 0.82，预测误差的标准误差分别为 1.21 和 1.33μm。结论是，NIRS 可以作为一种有效的技术，根据一些重要的纺织质量特性，如 MFD 和 SF，选择羊驼。
