基于近红外光谱结合波长优选检测单颗葡萄的SSC含量

Determination of SSC content in single grape based on NIR combined with wavelength selection

采用无损检测测定单颗葡萄中可溶性固形物(SSC)含量,获得个体和群体信息,以期指导田间管理、葡萄储存条件设置及满足消费者对葡萄口味的不同需求。采用手持式NIR光谱仪在950~1 650nm波长范围采集葡萄的近红外光谱,采用偏最小二乘(PLS)回归建立葡萄SSC预测模型。为了减少冗余无信息变量,增加模型的预测精度和稳定性,采用无信息变量消除法(UVE)、随机蛙算法(RF)筛选出与葡萄SSC含量相关的重要波长变量。结果表明:RF筛选建立的SSC预测模型优于全光谱PLS和UVE筛选建立的模型。RF-PLS模型的校正集、交叉验证及预测集的R2c、R2cv和R^2p分别为0.960 5,0.933 4,0.930 4,校正均方根误差(RMSEC),交叉验证均方根误差(RMSECV)和预测均方根误差(RMSEP)分别为0.638 2,0.829 9,0.868 8。表明通过波长优选后的,基于便携式近红外光谱在预测单颗葡萄SSC含量的应用上完全可行,有较高的预测精度。

The soluble solid content （SSC）, one ot the important index for evaluating the qualities, in a single grape were measure real-timely and non-destructively in this study, This might help to improve the management of the fruits as well as prolong their storage time, satisfying the needs of different customers. The near infrared （NIR） spectra of grapes was detected by using a hand-held NIR spec- trometer, wavelength ranging from 950 to 1, 650 nm. Based on par- tial least squares （PLS） data, the prediction model for SSC content in grapes was established. In order to reduce redundancy and unimformative variables while increase prediction accuracy and stability of this model, sensitive wavelength variables were selected by unin- formative variable elimination （UVE） and random frog （RF）, re spectively. The results showed that the SSC predictive model estab- lished by RFPLS is better than the ones done by full-spectrum PLS and UVE. R2 , R2v andR2of RFPLS were 0.960 5, 0.933 4 and 0.930 4, and RMSEC, RMSECV and RMSEP of it were 0.638 2, 0.829 9 and 0. 868 8, respectively. Our results showed that the spectra based on portable NIR spectrometer could be successfully ap- plied in prediction of SSC content in a single grape with high predic tion accuracy after wavelength selection for the model.