高光谱成像快速检测壳聚糖涂膜草莓可溶性固形物

为了对壳聚糖涂膜草莓可溶性固形物含量(soluble solids content, SSC)进行快速检测,该文采用高光谱成像仪(400～1 000 nm)对0,0.5%,1%浓度的壳聚糖(chitosan, CTS)涂膜草莓分别储藏1,2,4 d后进行成像,并测量样本SSC。通过分析SSC发现,0.5%和1%壳聚糖涂膜草莓,其SSC随着储藏天数的增加均高于0浓度壳聚糖涂膜草莓,说明了0.5%和1%壳聚糖涂层抑制了草莓中SSC的降低,能够延长草莓的新鲜口味。随后采用蒙特卡罗-偏最小二乘法(monte carlo-partial least squares, MCPLS)对异常样本进行剔除。对剔除异常样本后的光谱数据进行不同预处理,以确定最优的预处理方法。为提高运行速度和降低数据维数,采用竞争性自适应权重取样法(competitive adaptive reweighted sampling, CARS)和连续投影算法(successive projections algorithm, SPA)进行特征波段选择。最后,采用偏最小二乘回归(partial least square regression, PLSR)和支持向量回归(support vector regression, SVR)法建立回归模型。最终结果表明:SPA-SVR模型效果最佳,0浓度的壳聚糖涂膜的草莓,建模集精度Rc2为0.865,预测集精度Rv2为0.835;0.5%浓度的壳聚糖涂膜的草莓,建模集精度Rc2为0.808,预测集精度Rv2为0.799;1%浓度的壳聚糖涂膜的草莓,建模集精度Rc2为0.834,预测集精度Rv2为0.875。对储藏第4天的部分样本图像进行主成分分析(principal component analysis, PCA),结果显示除第二主成分图像(PC2)中有部分噪声影响外,PC1和PC3均能完整反映草莓信息,且PC3图像明显呈现出不同浓度壳聚糖涂膜草莓的褐变程度,说明不同浓度的壳聚糖涂膜也会对草莓货架期产生不同影响。综上说明利用高光谱成像技术可以实现壳聚糖涂膜草莓SSC快速检测,有效指导草莓保鲜处理。

Rapid detection of soluble solids content in strawberry coated with chitosan based on hyperspectral imaging

Strawberries are popular fruit for their tender texture, juice and sweet taste. Prior on shelves, the harvesting and storage have always been the problems due to its fragility as well as susceptibility to rot. Chitosan coating has been widely used in fruit preservation, which can delay the storage time of fruits and has good preservation effect. The quality of chitosan-coated fruits is mostly detected by the typical conventional methods of physical or chemical testing. Since such methods need to deal with a large number of samples, which are time-consuming, laborious and destructive for detecting coated fruits. Therefore, in order to explore the possibility of detecting the soluble solids content (SSC) of strawberry coated with chitosan nondestructively and rapidly, hyperspectral imaging technology was employed to estimate the SSC of strawberry coated with chitosan in this study. Strawberry samples coated with 0, 0.5% and 1% chitosan acetic acid which were stored in 3 periods (1, 2 , 4 d). Outliers were eliminated by monte carlo-partial least squares (MCPLS) method, and the number of outliers was 10, 3 and 5 for the above respect treatments. Sample partitioning based on joint X-Y distance (SPXY) was used to split the data after eliminating outliers. After the partition of sample set, the modeling set contains the maximum and minimum SSC values in the three-concentration data, and the range of SSC values in the calibration set and validation set is large and the partition is reasonable. To find out the best model effect, Savitzky-Golay, baseline correction, De-trending, moving average smoothing (MA), multiplicative scatter correction (MSC) and standard normal variate (SNV) were used to pre-process the spectral data after eliminating the outliers. It was found that the strawberry sample data coated with 0 chitosan acetic acid solution pretreated by MSC had the best effect, while the strawberry sample data coated with 0.5% and 1% chitosan acetic acid solution without pretreatment had the best effect. R2 c was the largest and RMSECV was the smallest. Competitive adaptive reweighted sampling (CARS) and successive projection algorithm (SPA) method were applied to select the effective wavelengths, which were helpful for enhancing computer velocity and reducing data dimension. The number of effective wavelengths selected by CARS and SPA for the three concentrations was 32, 30, 20 and 11, 8, 16, respectively. Finally, partial least square method (PLS) and support vector regression (SVR) were used to build regression models. The final results showed that the PLS regression model was less effective than the SVR model, while the full spectrum data and the data of characteristic bands selected by CARS are less effective in the SVR model, and the SPA-SVR model was the best. The value of R2 c reached to 0.865 for strawberry samples coated with 0 chitosan acetic acid solution, and value of R2 v reached to 0.835; for the strawberries coated with 0.5% chitosan acetic acid solution R2 c was 0.808 and R2 v was 0.799; and the R2 c and R2 v were 0.834 and 0.875 for strawberries coated with 1% chitosan acetic acid solution, respectively. These results validated the applicability of hyperspectral imaging technology on rapid detection of SSC in strawberry coated with chitosan.