基于变量筛选的温州蜜桔品质的光谱快速检测

目的利用可见/近红外光谱技术结合变量筛选算法建立预测模型。方法采集7个不同批次蜜桔样本的漫透射光谱,预处理优化后,以无信息变量消除法(uninformative variable elimination,UVE)、竞争性自适应重加权法(competitive adaptive reweighting sampling,CARS)及其组合(UVE-CARS)共3种策略来进行光谱有效波段的筛选,建立蜜桔可溶性固形物含量(soluble solid content,SSC)的偏最小二乘预测模型(partial least square,PLS)。结果比较全变量模型和3个特征变量模型的预测性能,UVE-CARS-PLS模型取得了最优的检测效果,相比全变量模型,建模变量数减少了96.5%,其预测集相关系数R_P提升至0.732,预测集均方根误差(root-mean-square error,RMSEP)下降至0.873~0Brix。结论结合多重变量选择算法,可以进一步压缩建模变量数,简化模型,提高模型预测精度,实现区域蜜桔品质的光谱快速检测。

Rapid spectral detection of satsuma quality in wenzhou based on variable screening

Objective To rapidly detect the quality of satsuma in Wenzhou, the prediction model was established by using visible-near infrared spectroscopy technology and variable selection algorithms. Methods The diffused transmission spectra of seven different batches of satsumas was collected, and then the spectra were optimized using preprocess methods. After then, three effective strategies were used to identify the informative variables from full spectra bands, namely, uninformative variable elimination (UVE), competitive adaptive weighting (CARS), and its combination (UVE-CARS). It aims to establish a rapid and accurate partial least squares (PLS) prediction model for the soluble solids content (SSC) of satsuma. Results The performance of the all prediction models were compared. The UVE-CARS-PLS model achieved the best detection results. Compared with the full variable model, the number of modeling variables reduced by 96.5%, and the correlation coefficient of prediction set (RP) and root mean square error (RMSEP) reached 0.715 and 0.895 0Brix, respectively. Conclusion Based on the calculated results, it can be concluded that the performance of the prediction model can be significantly improved and the complex of the model was also reduced by using variables selection algorithm. Finally, the quality of satsuma can be detected rapidly using visible / near-infrared spectroscopy technology.