Distribution Study Of Physicochemical Indexes In Zhenjiang Balsamic Vinegar Grains Based On Hyperspectral Imaging Technology

Physicochemical indexes of vinegar grains in acetic acid solid-state fermentation of Zhenjiang balsamic vinegar such as moisture and total acid are important influence factors for growth and metabolism of fermentation microorganism. The general detection methods of physicochemical indexes in vinegar grains such as physicochemical detection and near infrared spectroscopy detection are all single point sampling with discontinuous averages as the results. The results general could not describe the two-dimensional distribution of physicochemical indexes and meet the requirements for monitoring the complex fermentation process. For hyperspectral imaging technology, spectral information and image information can be obtained simultaneously. In view of the advantages of hyperspectral imaging technology, the study used the technology to detect distributions of total acid content, p H value, moisture content and non-volatile acid in vinegar grain samples for acetic acid fermentation. The study obtained the visualization distribution maps and extracted the quantitative feature to describe the distribution of the physicochemical indexes.Firstly, the contents of total acid, p H value, moisture and non-volatile acid of vinegar grains in acetic acid fermentation were detected by physical and chemical tests. The vinegar grain samples in upper layer, middle layer and bottom layer of fermenting pool were collected every two days. The results showed that these physicochemical indexes were all varied with the time in different layers so that the fermentation state could not be reflected by using point sampling methods. Taking total acid content for example, in terms of time, it increased sharply to 4.53 % in the first 9 days, then increased slowly to 6.66 % within the 11 th to the 15 th day, and finally kept constant to 7-8 % after 15 days. While in terms of space, it was highest in upper layer(reached 3.10%) within the 1st to the 5th day, and was highest in middle layer(reached 3.10 %) within the 7th to the 11 th day, and finally was relatively closed between each layers with partial difference within the 13 th to the 19 th day.Secondly, the quantitative prediction models of total acid, p H value, moisture content and non-volatile acid were established based on hyperspectral imaging technique. The average spectra of interesting regions in vinegar culture were extracted and preprocessed by Standard normal variable(SNV). Partial least squares(PLS), back-propagation artificial neural network(BP-ANN) and least squares support vector machines(LS-SVM) model were obtained by combining the chemical values and the optimal characteristic wavelengths which were selected by different methods of variables selection. The best model were obtained by compared the results of the prediction models. The results showed that the best prediction models of total acid, PH value and moisture content were PLS models with 8, 7 and 7 PCs after 21, 74 and 67 characteristic wavelengths being analyzed by principal component, respectively. The best prediction model of non-volatile acid was LS-SVM using 6 PCs after 26 characteristic wavelengths being analyzed by principal component. The related coefficient of four best models were 0.904, 0.8662, 0.8140 and 0.9116, respectively. The aforementioned models indicates the relationship between the spectra of physical and chemical parameters and the chemical values and lay a foundation for the detection of distribution maps.Thirdly, visualization distribution maps of physicochemical indexes of vinegar grains were detected and quantitative features of the distribution were described digitally. Hyperspectral images of the 11 d middle layer vinegar grain samples were collected and physicochemical indexes contents in per pixel regions were predicted. Every physicochemical indexes content was converted into the gray value and the gray map was processed by pseudo color so that be easy to observe the distribution. Distribution maps of physicochemical indexes of vinegar grains were obtained and we were easy to compare the difference before and after overturn the grains by eyes directly.Before overturning the grains, the contents of total acid, p H, moisture and non-volatile acid were concentrated in the 4.8~5.8 %, 3.2~3.6, 64~67 % and 2~2.7 % with high local concentrations and uneven distribution. After overturning the grains, the contents of total acid, p H, moisture and non-volatile acid were concentrated in the 3~6 %, 3~4, 59~69 % and 1.8~2.8 % with a few local concentrations and uniform distribution. In order to enable the computer to capture the distribution characteristics in vinegar physiochemical indexes distribution maps quickly and accurately, the study used digital image technology to extract six relevant statistics from distribution map. The six relevant statistics are mean and variance based, contrast, angle direction of second order moment, entropy and average. The first two statistics are on histogram and the others are based on gray difference. The six relevant statistics were considered as quantitative characteristics of distribution of physicochemical indexes contents. The results showed that the quantitative features were accuracy and reliability to compare the fermentation status of vinegar grains in different regions. The study could give out the visual image of physicochemical index distribution and also describe the distribution characteristics digitally. The quantitative characteristics extraction from distribution maps could help computers catch distribution characteristics of physicochemical indexes in vinegar grains automatically so that to provide technical support for the digital monitoring Zhenjiang vinegar acetic acid fermentation.