高光谱图像技术快速预测发酵醋醅总酸分布

固态发酵是镇江香醋生产的重要环节之一，直接决定着成品醋的风味和品质。但目前固态发酵的生产控制主要依赖人工经验，难以有效保障镇江香醋的品质。该文分析了总酸（total acid content，TAC）、pH值、含水率在不同阶段的变化规律；采用高光谱图像技术结合联合区间偏最小二乘法（synergy interval partial least squares，siPLS）快速预测固态发酵基质（醋醅）的TAC、pH值和含水率，其最佳模型的相关系数R分别为0.8316、0.9455和0.8503；同时利用主成分分析和逐步多元线性回归模型（stepwise multiple linear regression，SMLR）对醋醅高光谱图像进行分析，研究了总酸在醋醅中的分布情况，以此来快速判断醋醅发酵的均匀性。研究表明，利用高光谱图像技术快速预测醋醅的理化参数及其分布的方法是可行的，结果可为镇江香醋固态发酵的工艺控制提供基础数据和技术手段。

Rapidly detecting total acid distribution of vinegar culture based on hyperspectral imaging technology

Abstract: In China and Southeast Asian countries, the solid-state fermentation (SSF) process is maintained empirically, especially in fed-batch fermentation by layers. In this study, the feasibility of determination of the total acid content (TAC), pH value and moisture content of Zhenjiang balsamic vinegar during SSF process were investigated. Hyperspectral imaging technology (HSIT) was combined with an appropriate multivariate analysis method. A synergy interval partial leastsquare (siPLS) was used to select the efficient spectral subintervals and wavelengths by k-fold cross-validation during the development of model. The performance of the final model was evaluated by use of the root mean square error of cross-validation (RMSECV) and correlation coefficient (Rc) for the calibration set, and verified by use of the root mean square error of prediction (RMSEP) and correlation coefficient (Rp) for the validation set. The changes of TAC, pH value and moisture were obtained by normal standard methods. TAC was constantly increased during fermentation process. The change of pH value was from 3.6 to 3.99, which was appropriate for the growth of the main microbes and can inhibit the growth of other unnecessary bacteria in the fermentation process. The moisture was increased in all stages during SSF process. Besides, the experimental results showed that the optimum siPLS model for TAC was achieved by use of 7 PLS factors, when 4 spectral subintervals were selected by siPLS. The predicted precision of the best model obtained was: RMSECV = 0.625, Rc = 0.8316, RMSEP = 0.773, and Rp = 0.7965. The pH value was achieved using siPLS with 6 PLS factors. The predicted precision of the best model obtained was: RMSECV = 0.0465, Rc = 0.9455, RMSEP=0.0482, and Rp = 0.9321. Besides, the moisture content was achieved using siPLS with 4 PLS factors. The predicted precision of the best model obtained was: RMSECV = 0.2104, Rc = 0.8503, RMSEP = 0.2459, and Rp = 0.8277. Finally, the superior performance of the siPLS model was demonstrated by comparison with two other PLS models. The overall results indicated that HSIT can be successfully used for the measurement of TAC, pH value and moisture content in SSF, and the use of siPLS algorithm is the best means of calibration for modelling. Besides, in order to get the distribution map of TAC, principal component analysis (PCA) and stepwise linear regression (SMLR) were used to calculate the TAC in each pixel points. The distribution map of TAC in vinegar culture was obtained. The homogeneity of SSF also could be determined by hyperspectral imaging.