基于遗传算法的波段选择法在组合优化问题上具有很大的搜索优势,适应性很广。文章将该方法应用于西洋梨糖度近红外光谱分析中,探讨数据优化筛选的可行性。光谱经多元散射校正或标准归一化处理后进行波段选择,选择结果与样品中被测成分有关,4个品种洋梨的最佳个体染色体编码有一定共性。分别建立了四种洋梨的GA-PLS模型和全谱模型,早红考密斯、五九香、凯斯凯德和康佛伦斯的GA-PLS建模数据点分别从1557减少到了434,496,310和496。GA-PLS/Fr-PLS模型的预测标准偏差分别为0.428/0.518,0.696/0.694,0.425/0.421和0.567/0.633,其中早红考密斯和康佛伦斯GA-PLS模型的预测精度明显优于全谱模型,而五九香和凯斯凯德的GA-PLS模型与全谱模型相近。结果表明,遗传算法用于PLS建立西洋梨糖度校正模型前的数据优化筛选是可行的,有效提高测量精度,减少建模变量。

An improved genetic algorithm was used to implement an automated wavelength selection procedure for use in building multivariate calibration models based on partial least squares regression (PLS).The region selecting by genetic algorithms (R-SGA) was applied in building calibration model of soluble solid content (SSC) of Western pear,and the numbers of latent variables used to build calibration model were further reduced.The Fourier transform near infrared reflectance (FT-NIR) spectra were processed by GA after MSC or SNV,and four PLS calibration models were built by using the optimal combinations of these sub-regions.Meanwhile,the full region selecting PLS (Fr-PLS) models were developed.The R-SGA models variables were 434,496,310 and 496,for Early Red Comice,Wujiuxiang,Cascade and Kang Buddha,respectively.Despite the complexity of the spectral data,the R-SGA procedure was found to perform well (RMSEP=0.428,0.567 for Early Red Comice and Kang Buddha,respectively),leading to calibration models that significantly outperform those based on full-spectrum analyses (RMSEP=0.518,0.633).The prediction precision of GA-PLS models was similar to that of Fr-PLS for Wujiuxiang and Cascade,with RMSEP of 0.696/0.694 and 0.425/0.421 respectively.This work proved that the R-SGA could find optimal values for several disparate variables associated with the calibration model and that the PLS procedure could be integrated into the objective function driving the optimization.