Intelligent wine aging astringent quality identification method

Abstract

The invention relates to an intelligent wine aging astringent quality identification method. The method comprises the following steps: acquiring to-be-identified wine aging preset parameters; wherein the preset parameters comprise aging environment parameters, sample appearance images, aroma parameters and astringent data; inputting the preset parameters into a preset astringent quality intelligent prediction model to obtain a wine aging astringent quality identification result; wherein the astringent quality intelligent prediction model is constructed based on a machine learning algorithm; the astringent quality intelligent prediction model comprises a parameter quality prediction sub-model, an astringent appearance quality prediction sub-model, an aroma quality prediction sub-model and an astringent chemical quality prediction sub-model. According to the invention, real-time monitoring and dynamic prediction of the whole aging process can be realized, so as to ensure the stability and consistency of astringent quality.

Description

Technical Field

[0001] The present invention relates to the field of intelligent winemaking technology, and in particular to an intelligent wine aging astringency quality identification method. Background Art

[0002] The formation and evolution of astringency during wine aging are affected by many factors, such as grape varieties, brewing technology, aging environment, etc. At present, wine companies mainly rely on manual sensory evaluation and chemical analysis methods to judge astringency quality, which has problems such as strong subjectivity, low efficiency, and inability to monitor in real time. In addition, environmental parameters of the aging process, such as temperature, humidity, and alcohol content, also have an important impact on astringency quality, but it is difficult to achieve accurate control and dynamic adjustment of these parameters with existing technologies. Therefore, there is an urgent need for an intelligent wine aging astringency quality identification method and system, which collects various parameters of the aging process in real time through sensors, combines astringency quality characteristic data, and establishes an astringency quality intelligent prediction model to achieve real-time monitoring and dynamic control of the entire aging process to ensure the stability and consistency of astringency quality. The system needs to solve the integration of data acquisition, transmission, and processing, involving multiple technical fields such as sensor technology, data analysis, and control algorithms, which is of great significance to the realization of intelligence and standardization of the wine aging process. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent wine aging astringency quality identification method to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] An intelligent wine aging astringency quality identification method, comprising:

[0006] Collecting preset parameters of wine aging to be identified; wherein the preset parameters include: aging environment parameters, sample appearance image, aroma parameters and astringency data;

[0007] The preset parameters are input into the preset astringency quality intelligent prediction model to obtain the wine aging astringency quality identification result; wherein the astringency quality intelligent prediction model is constructed based on a machine learning algorithm; the astringency quality intelligent prediction model includes: a parameter quality prediction sub-model, an astringency appearance quality prediction sub-model, an aroma quality prediction sub-model and an astringency chemical quality prediction sub-model.

[0008] Optionally, collecting the aging environment parameters of the wine to be identified includes:

[0009] Installing a plurality of sensors in the wine aging container to collect wine aging environment parameters; wherein the aging environment parameters include: temperature parameters, humidity parameters and alcohol content parameters;

[0010] The aging environmental parameters collected in real time are preprocessed to remove abnormal values ​​and obtain standardized aging environmental parameters.

[0011] Optionally, inputting the preset parameters into a preset astringency quality intelligent prediction model to obtain a wine aging astringency quality recognition result includes:

[0012] Inputting the aging environment parameters into the parameter quality prediction sub-model to obtain parameter quality prediction data;

[0013] Inputting the sample appearance image into the astringency appearance quality prediction sub-model to obtain astringency appearance quality prediction data;

[0014] Inputting the aroma parameter into the aroma quality prediction sub-model to obtain aroma quality prediction data;

[0015] Inputting the astringency data into the astringency chemical quality prediction sub-model to obtain astringency chemical quality prediction data;

[0016] Merging the parameter quality prediction data, the astringent appearance quality prediction data, the aroma quality prediction data, and the astringent chemical quality prediction data;

[0017] Constructing an astringency quality evaluation index system and determining the weight of each index; wherein the indexes include: parameter quality, astringency appearance quality, aroma quality and astringency chemical quality;

[0018] The astringency quality evaluation index system is used to evaluate the fused data to obtain the wine aging astringency quality identification result.

[0019] Optionally, inputting the aging environment parameters into the parameter quality prediction sub-model to obtain parameter quality prediction data comprises:

[0020] Classifying and labeling the historical aging environment parameters;

[0021] Using a machine learning algorithm to train the classified aging environmental parameters to construct the parameter quality prediction sub-model of the wine aging environmental parameters;

[0022] The parameter quality prediction sub-model is used to perform quality prediction analysis on the aging environment parameters collected in real time.

[0023] Optionally, inputting the sample appearance image into the astringent appearance quality prediction sub-model to obtain astringent appearance quality prediction data comprises:

[0024] According to the pre-established astringency quality evaluation index system, images of wine samples at several time points during the wine aging process are obtained;

[0025] For the wine sample images obtained at several time points, a color feature extraction algorithm is used to extract the color feature vector of the wine sample; and a texture feature extraction algorithm is used to extract the clarity feature vector of the wine sample;

[0026] The extracted color feature vector and clarity feature vector are input into a pre-trained support vector machine classification model to classify the astringency appearance quality of the wine samples and obtain the astringency appearance quality prediction data.

[0027] Optionally, inputting the aroma parameter into the aroma quality prediction sub-model to obtain aroma quality prediction data comprises:

[0028] Acquiring aroma component data at different time points during the wine aging process, and using electronic nose technology to perform quantitative detection and analysis on the aroma component data to obtain quantitative data of the aroma components;

[0029] Extracting preset key data capable of characterizing the astringency quality characteristics of the wine based on the quantitative data, and constructing an astringency aroma quality characteristics data set;

[0030] Obtaining artificial sensory evaluation result data, wherein the artificial sensory evaluation result data is astringency quality scores obtained by wine tasting experts through sensory evaluation of wine samples with different aging times;

[0031] Comparing and verifying the astringent aroma quality characteristic data set with the manual sensory evaluation result data, calculating the correlation between the two sets of data, and if the correlation is greater than a preset threshold, determining that the astringent aroma quality characteristic data set can effectively characterize the astringency quality of the wine;

[0032] A support vector machine algorithm is used, with the astringent aroma quality characteristic data set as input and the artificial sensory evaluation result data as output, to train and establish the aroma quality prediction sub-model;

[0033] Based on the aroma quality prediction sub-model, the aroma quality data of the wine is predicted according to the aroma component data obtained by electronic nose detection and analysis.

[0034] Optionally, inputting the astringency data into the astringency chemical quality prediction sub-model, and obtaining the astringency chemical quality prediction data comprises:

[0035] The high performance liquid chromatography chemical analysis method is used to determine the content changes of astringent substances during wine aging to obtain astringent chemical quality data; wherein the astringent substances include: tannins and anthocyanins;

[0036] The principal component analysis method was used to comprehensively analyze the relationship between the content of various astringent substances and the astringency quality of wine, and a sub-model for predicting astringency chemical quality was constructed to obtain astringency chemical quality prediction data.

[0037] Optionally, the method further comprises:

[0038] According to the output results of the preset astringency quality intelligent prediction model, a machine learning algorithm is used to intelligently judge the astringency quality of the wine at the end of aging, so as to determine the optimal aging time and release time.

[0039] The beneficial effects of the present invention are:

[0040] The present invention discloses an intelligent control method for the astringency quality of wine. A variety of sensors are installed in the aging container to monitor environmental parameters in real time, and the appearance, aroma and chemical quality data of the wine are obtained by using technologies such as machine vision, electronic nose and chemical analysis. These multi-source heterogeneous data are fused and analyzed to construct an intelligent prediction model for astringency quality. At the same time, a machine learning algorithm is used to intelligently distinguish the astringency quality at the end of aging to determine the best time to ship out of the warehouse. The present invention realizes the full-process intelligent monitoring of the astringency quality of wine, ensures the quality stability and consistency of the finished wine, and provides a new technical solution for the precision and intelligence of the winemaking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0042] Figure 1 The present invention is a flowchart of an intelligent wine aging astringency quality identification method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1This embodiment provides an intelligent wine aging astringency quality identification method, including:

[0046] Collecting preset parameters of wine aging to be identified; wherein the preset parameters include: aging environment parameters, sample appearance image, aroma parameters and astringency data;

[0047] The preset parameters are input into the preset astringency quality intelligent prediction model to obtain the wine aging astringency quality identification result; wherein, the astringency quality intelligent prediction model is constructed based on a machine learning algorithm; the astringency quality intelligent prediction model includes: parameter quality prediction sub-model, astringency appearance quality prediction sub-model, aroma quality prediction sub-model and astringency chemical quality prediction sub-model.

[0048] Furthermore, the aging environment parameters of the wine to be identified are collected including:

[0049] Installing a number of sensors in the wine aging container to collect wine aging environment parameters; the aging environment parameters include: temperature parameters, humidity parameters and alcohol content parameters;

[0050] The aging environmental parameters collected in real time are preprocessed to remove abnormal values ​​and obtain standardized aging environmental parameters.

[0051] Furthermore, the preset parameters are input into the preset astringency quality intelligent prediction model to obtain the wine aging astringency quality identification results including:

[0052] Input the aging environment parameters into the parameter quality prediction sub-model to obtain parameter quality prediction data;

[0053] Input the sample appearance image into the astringency appearance quality prediction sub-model to obtain astringency appearance quality prediction data;

[0054] Inputting the aroma parameters into the aroma quality prediction sub-model to obtain aroma quality prediction data;

[0055] Inputting the astringency data into the astringency chemical quality prediction sub-model to obtain astringency chemical quality prediction data;

[0056] Fusion of parameter quality prediction data, astringency appearance quality prediction data, aroma quality prediction data, and astringency chemical quality prediction data;

[0057] Construct an astringency quality evaluation index system and determine the weight of each index; the indexes include: parameter quality, astringency appearance quality, aroma quality and astringency chemical quality;

[0058] The astringency quality evaluation index system was used to evaluate the fused data and obtain the astringency quality identification results of wine aging.

[0059] Among them, the parameter quality prediction sub-model, astringent appearance quality prediction sub-model, aroma quality prediction sub-model and astringent chemical quality prediction sub-model are all constructed using support vector machines.

[0060] Furthermore, the aging environment parameters are input into the parameter quality prediction sub-model to obtain parameter quality prediction data including:

[0061] Classify and label historical aging environmental parameters;

[0062] The classified aging environmental parameters were trained using machine learning algorithms to construct a parameter quality prediction sub-model for wine aging environmental parameters.

[0063] The parameter quality prediction sub-model is used to conduct quality prediction analysis on the aging environment parameters collected in real time.

[0064] Specifically, in this embodiment, by installing multiple sensors such as temperature, humidity, and alcohol content in the wine aging container, various parameter data of the aging environment are collected in real time, and the data is uploaded to the cloud server through a wireless transmission module.

[0065] Obtain real-time data collected by temperature sensors, humidity sensors, and alcohol sensors installed in wine aging containers;

[0066] Preprocess the real-time data, remove outliers, and obtain standardized aging environment parameter data;

[0067] Upload the aging environment parameter data to the cloud server through the wireless transmission module;

[0068] After receiving the aging environment parameter data, the cloud server stores it in a database and classifies and labels the aging environment parameter data;

[0069] Use machine learning algorithms to train the stored historical aging environmental parameter data and establish a prediction model for wine aging environmental parameters and quality;

[0070] According to the prediction model, the quality prediction analysis is carried out on the newly collected and uploaded aging environment parameter data;

[0071] If the quality prediction analysis results show that the current aging environment parameters cause the wine quality to deteriorate, an early warning message will be sent to the winery managers, and a recommended plan for optimizing the aging environment will be given.

[0072] Specifically, SHT31 temperature and humidity sensors and MQ-3 alcohol sensors are installed in the wine aging container, and data is collected every 5 minutes. The collected data is preprocessed by the Kalman filter algorithm, and outliers exceeding 20% ​​of the normal range are eliminated to obtain normalized temperature, humidity and alcohol data. These data are uploaded to the Alibaba Cloud server through the Wi-Fi module, stored in the MySQL database, and classified and labeled according to the type of wine and aging time. The cloud uses the random forest algorithm to train historical aging data, and establishes a wine quality prediction model based on temperature, humidity and alcohol, with an accuracy of 95%. When new aging data is uploaded, the system automatically performs quality prediction analysis. If the prediction results show that the current aging environment may cause the quality of the wine to deteriorate, such as temperature below 10℃ or above 20℃, humidity below 50% or above 80%, and alcohol content below 10% or above 15%, the system will issue an early warning to the winery managers through SMS and email, and give suggestions for adjusting temperature and humidity or controlling fermentation progress according to the degree of deviation to ensure stable wine quality.

[0073] Furthermore, the sample appearance image is input into the astringency appearance quality prediction sub-model to obtain astringency appearance quality prediction data including:

[0074] According to the pre-established astringency quality evaluation index system, images of wine samples at several time points during the wine aging process are obtained;

[0075] For the wine sample images obtained at several time points, a color feature extraction algorithm is used to extract the color feature vector of the wine sample; and a texture feature extraction algorithm is used to extract the clarity feature vector of the wine sample;

[0076] The extracted color feature vector and clarity feature vector are input into a pre-trained support vector machine classification model to classify the astringency appearance quality of the wine samples and obtain the astringency appearance quality prediction data.

[0077] Specifically, in this embodiment, first, according to the astringency quality evaluation index system, six time points, namely, the 1st day, the 10th day, the 20th day, the 30th day, the 40th day, and the 50th day, are selected during the wine aging process, and wine sample images are collected at each time point. Then, the collected wine sample images are subjected to median filtering denoising and histogram equalization enhancement processing to obtain pre-processed high-quality wine sample images. Next, for the pre-processed wine sample images, the HSV color space component statistical algorithm is used to extract 36-dimensional color feature vectors; at the same time, the gray level co-occurrence matrix algorithm is used to extract 64-dimensional texture feature vectors including contrast, correlation, energy, entropy, etc., which are used to characterize the clarity of the wine sample. The extracted color feature vector and clarity feature vector are cascaded to form a 100-dimensional astringency appearance feature vector, which is input into the support vector machine classification model pre-trained with 10,000 wine sample images to classify the astringency appearance quality of the wine sample. If the classification accuracy exceeds 95%, the astringency appearance quality data of the wine sample at the corresponding time point is recorded and saved. Finally, the astringency appearance quality data of the whole wine aging process, i.e., 6 time points in 50 days, were obtained, and the change curve of astringency appearance quality with aging time was drawn. Through the analysis of the change curve, it was found that the astringency appearance quality of the wine reached the best on the 30th day of aging, so the 30th day was determined to be the best aging time for the wine. When the aging time reached the 30th day, the system automatically output a prompt message, indicating that the wine aging process was over and entered the next stage of wine clarification.

[0078] The texture feature extraction algorithm is used to extract the clarity feature vector of the wine sample.

[0079] The image data of the wine sample is obtained and preprocessed, including image enhancement and image segmentation. The gray-level co-occurrence matrix algorithm is used to extract the texture features of the image and obtain the feature vector reflecting the clarity of the wine. According to the preset clarity threshold, it is judged whether the extracted feature vector meets the clarity requirement. If the feature vector meets the clarity requirement, the wine sample is marked as qualified clarity. If the feature vector does not meet the clarity requirement, the wine sample is marked as unqualified clarity. Statistical analysis is performed on the wine samples marked as qualified and unqualified clarity to obtain the clarity qualified rate. According to the clarity qualified rate, the overall clarity level of the wine sample is determined to provide a reference for wine quality assessment.

[0080] Specifically, a high-resolution camera is used to obtain image data of wine samples. The image resolution is set to 1920×1080 and the color depth is 24 bits. Then the acquired image data is preprocessed and the image is enhanced using the histogram equalization algorithm to improve the contrast of the image. The image is segmented using the Otsu threshold segmentation algorithm to extract the wine liquid area. Then the gray-level co-occurrence matrix algorithm is used to extract the texture features of the image. The gray level is set to 256, the directions are 0°, 45°, 90°, 135°, and the step size is 1. The clarity feature vector of the wine sample is extracted. According to industry standards, the clarity threshold is set to 85, and the extracted feature vector is compared with the threshold. If the feature vector is greater than or equal to 85, the wine sample is marked as qualified in clarity; if the feature vector is less than 85, the wine sample is marked as unqualified in clarity. 100 wine samples were tested for clarity, and statistical analysis showed that 95 samples passed the clarity test, while 5 samples failed the clarity test. The clarity pass rate of this batch of wine samples was calculated to be 95%. Based on the above analysis, it can be judged that the overall clarity level of this batch of wine samples is high, reaching the standard of high-quality wine, and providing an important reference for subsequent wine quality evaluation.

[0081] The color feature extraction algorithm is used to extract the color feature vector of the wine sample.

[0082] Wine image data is obtained based on wine sample collection, and the image is denoised and enhanced by image preprocessing methods to improve image quality. Color feature extraction algorithms such as color histogram and color moment are used to extract color features from preprocessed wine images. Feature selection and dimensionality reduction are performed on the extracted color features to remove redundant and irrelevant features and obtain a compact color feature vector representation. According to the wine quality evaluation standard, supervised learning algorithms such as support vector machine and random forest are used to construct a wine quality analysis model. The color feature vector is input into the trained quality analysis model to predict and evaluate the quality of the wine sample. The performance of the quality analysis model is evaluated by cross-validation and other methods, and the model is optimized and improved according to the evaluation results. The optimized wine quality analysis model is applied to the actual wine quality detection and evaluation tasks to assist artificial wine tasters in quality judgment.

[0083] Specifically, after the wine samples were collected, the wine image data was first denoised by a Gaussian filter. The standard deviation of the filter was set to 5, which effectively removed the random noise in the image. Then, the histogram equalization technique was used to enhance the contrast of the image and make the color characteristics of the wine more obvious. The color histogram method was used to calculate the frequency distribution of each color channel of the image, and then the color moment algorithm was combined to calculate the first-order, second-order and third-order color moments of the image. These color moments reflect the mean, variance and skewness of the image color. In order to simplify the complexity of the model, principal component analysis (PCA) was used to reduce the dimension of the color features, and the principal component with a cumulative contribution rate of 95% was selected as the color feature vector. In the wine quality evaluation stage, a support vector machine model was constructed, using the radial basis function (RBF) kernel, the kernel function parameter γ was set to 1, and the penalty parameter C was set to 1. This setting is helpful for processing nonlinearly separable data. The color feature vector was input into the model for quality prediction. In order to verify the accuracy of the model, a 10-fold cross-validation method was used to calculate the average accuracy of the model. If the accuracy was lower than 85%, the model parameters were adjusted and retrained until the performance requirements were met. Finally, the tuned model was deployed on the wine production line to perform real-time quality testing on wine samples on the assembly line, assisting sommeliers to quickly and accurately evaluate the quality of the wine.

[0084] Furthermore, the aroma parameters are input into the aroma quality prediction sub-model to obtain aroma quality prediction data including:

[0085] Obtain aroma component data at different time points during wine aging, use electronic nose technology to perform quantitative detection and analysis on the aroma component data, and obtain quantitative data of aroma components;

[0086] Based on the quantitative data, the preset key data that can characterize the astringency quality characteristics of wine are extracted to construct the astringency aroma quality characteristics dataset;

[0087] Obtaining artificial sensory evaluation result data, wherein the artificial sensory evaluation result data is astringency quality scores obtained by wine tasting experts through sensory evaluation of wine samples with different aging times;

[0088] The astringent aroma quality characteristic data set is compared and verified with the manual sensory evaluation result data, and the correlation between the two sets of data is calculated. If the correlation is greater than the preset threshold, it is determined that the astringent aroma quality characteristic data set can effectively characterize the astringency quality of wine;

[0089] The support vector machine algorithm was used to train and establish the aroma quality prediction sub-model, taking the astringent aroma quality characteristic dataset as input and the artificial sensory evaluation result data as output.

[0090] Based on the aroma quality prediction sub-model, the aroma quality data of wine is predicted according to the aroma component data obtained by electronic nose detection and analysis.

[0091] Specifically, in this embodiment, during the wine aging process, the aroma components of the wine sample are first quantitatively detected using the electronic nose technology, for example, the volatile organic compounds of the wine sample at different time points are captured by the sensor array of the electronic nose, and these signals are converted into electrical signals and output as quantitative data of the aroma components. These data are extracted from the key data that can characterize the astringency quality of the wine through a specific algorithm, such as principal component analysis (PCA), and an astringent aroma quality characteristic data set is constructed. At the same time, the result data of the sensory evaluation of the wine samples by wine tasting experts at different aging time points are collected, and these data are usually in the form of astringency quality scores. The correlation between the astringent aroma quality characteristic data set and the manual sensory evaluation result data is calculated by statistical analysis software, such as SPSS. If the correlation coefficient exceeds a preset threshold, such as 8, it indicates that these aroma component data can effectively characterize the astringency quality of the wine. Next, a support vector machine (SVM) algorithm is used, with the astringent aroma quality characteristic data set as input and the manual sensory evaluation result data as output, to establish a wine astringency quality prediction model. Through the cross-validation method, the data set is divided into a training set and a test set, and the model is trained and tested. According to the test results, the model parameters are adjusted, such as the kernel function type or the regularization parameter, until the prediction accuracy of the model reaches the preset accuracy requirements, for example, the prediction accuracy reaches 95%. Finally, the optimized wine astringency quality prediction model is applied to actual wine production. According to the latest aroma component data obtained by electronic nose detection and analysis, the astringency quality of wine is predicted, providing a scientific basis for the optimization of aging technology.

[0092] Furthermore, the astringency data is input into the astringency chemical quality prediction sub-model, and the astringency chemical quality prediction data is obtained, including:

[0093] The high performance liquid chromatography chemical analysis method was used to determine the changes in the content of astringent substances during wine aging to obtain astringent chemical quality data; the astringent substances include tannins and anthocyanins;

[0094] The principal component analysis method was used to comprehensively analyze the relationship between the content of various astringent substances and the astringency quality of wine, and a sub-model for predicting astringency chemical quality was constructed to obtain astringency chemical quality prediction data.

[0095] Specifically, in this embodiment, in order to accurately analyze the content changes of astringent substances during the wine aging process, wine samples are first collected at different stages, and three parallel samples are set for each sample to increase the reliability of the data. The analysis is performed by high performance liquid chromatography. First, the chromatographic conditions are optimized, for example, the optimal wavelength for detecting tannins and anthocyanins is determined to be 280 nanometers, acetonitrile and water are selected as mobile phases, and a gradient elution program is set to gradually increase from 5% acetonitrile to 95% acetonitrile within 30 minutes. Sample pretreatment includes filtering with a 45-micron filter membrane to remove impurities, and then taking an appropriate amount of sample to prepare a test solution. In the chromatographic analysis, the test solution is injected into the chromatograph, and the chromatographic peak area data of each astringent substance is recorded. By preparing standard solutions of tannins and anthocyanins of different concentrations, measuring their chromatographic peak areas, and drawing standard curves, the linear range and detection limit of these substances are determined, for example, the detection limit of tannins is 1 mg / L, and the quantitative limit is 3 mg / L. The chromatographic peak area data of the actual sample is substituted into the standard curve to calculate the actual content of tannins and anthocyanins. Finally, the principal component analysis method was used to input the content data of various astringent substances, analyze their correlation with the astringency quality of wine, and construct an astringency quality prediction model. This model can predict the astringency quality of wine according to the content of astringent substances, thereby providing a scientific basis for optimizing the wine aging process.

[0096] Furthermore, an astringency quality evaluation index system was constructed, and the weights of each index were determined, including:

[0097] First, the historical data of astringency quality were preprocessed, and the Z-score standardization method was used to normalize the data to make the different indicators comparable. Then, the principal component analysis method was used to extract the features of the preprocessed data, and the main astringency quality characteristics were extracted, with the cumulative variance contribution rate reaching more than 85%. On this basis, combined with the astringency quality evaluation standard, an astringency quality evaluation index system including color, aroma, taste, etc. was constructed, and the hierarchical analysis method was used to determine the weight of each indicator. The weights of environment, color, aroma, and taste were 1, 2, 4, and 3, respectively.

[0098] Furthermore, this embodiment also determines whether the current astringency quality reaches a preset threshold based on the monitoring data of the whole aging process and in combination with the astringency quality prediction model. If not, the control system automatically adjusts the aging environment parameters, such as temperature, humidity, alcohol content, etc., to dynamically optimize the astringency quality formation conditions.

[0099] Real-time monitoring data during the aging process, including environmental parameters such as temperature, humidity and alcohol content, are obtained as input to the astringency quality prediction model. The real-time monitoring data is input into the prediction model to obtain the predicted astringency quality value under the current aging environment. It is determined whether the predicted astringency quality value reaches the preset threshold. If it does, the current aging environmental parameters are maintained unchanged and monitoring continues. If the predicted astringency quality value does not reach the preset threshold, the control system is triggered to determine the environmental parameters and adjustment range that need to be adjusted based on the difference between the predicted value and the threshold. The control system automatically adjusts the temperature, humidity and alcohol content of the aging environment according to the determined adjustment parameters to optimize the conditions for the formation of astringency quality. After the adjustment is completed, the adjusted real-time monitoring data is obtained and input into the prediction model again to determine whether the adjusted astringency quality prediction value meets the standard. If the adjusted astringency quality prediction value meets the standard, the adjusted environmental parameters are maintained and monitoring continues; if it does not meet the standard, steps 4-6 are repeated until the astringency quality reaches the preset threshold and dynamic optimization is completed.

[0100] During the aging process, the temperature, humidity, alcohol content and other parameters of the aging environment are collected in real time through the wireless sensor network, and the data is collected every 10 minutes. The collected data is cleaned and standardized through the data preprocessing module to remove outliers and noise data, and the data is converted into a format suitable for input into the prediction model. Using the constructed astringency quality prediction model based on the support vector machine (SVM) algorithm, the standardized environmental parameter data is input into the model, and the predicted astringency quality value under the current environment is calculated by the model. The target threshold of astringency quality is set to 80 points (out of 100 points), and the predicted value is compared with the threshold. If the predicted value is greater than or equal to 80 points, it is considered that the current environmental parameters meet the requirements, and the existing parameter settings are maintained and monitoring continues. If the predicted value is less than 80 points, the automatic adjustment system of environmental parameters is triggered. According to the difference between the predicted value and the threshold, combined with the empirical rules in the expert knowledge base, the fuzzy control algorithm is used to determine the specific parameters that need to be adjusted and the adjustment range. For example, when the predicted value is 75 points, it may be necessary to reduce the temperature by 2°C, increase the humidity by 5%, and maintain the alcohol content unchanged. According to the determined adjustment strategy, the control system automatically adjusts the temperature, humidity and alcohol content of the aging environment through the actuator to optimize the conditions for the formation of astringency quality. Within 10 minutes after the adjustment is completed, the adjusted environmental parameter data is collected again and input into the prediction model for re-prediction. If the adjusted astringency quality prediction value reaches 80 points or more, the adjustment effect is considered ideal, and the adjusted environmental parameters are maintained and monitoring continues; if the predicted value still does not meet the standard, the above adjustment process is repeated, and multiple rounds of iterative optimization are performed until the astringency quality stabilizes at the target level. Through closed-loop control of real-time monitoring, intelligent prediction and dynamic regulation, dynamic optimization of aging environmental parameters and astringency quality is achieved, ensuring the quality stability of the aging process.

[0101] Furthermore, the method of this embodiment also includes: based on the output results of the preset astringency quality intelligent prediction model, using a machine learning algorithm to intelligently judge the astringency quality of the wine at the end of aging, and determine the optimal aging time and release time.

[0102] According to the output results of the astringency quality prediction model, the support vector machine algorithm is used to intelligently distinguish the astringency quality of wine during the aging process. Through the learning and training of historical data, the association model between astringency quality and aging time is established to predict the change trend of astringency quality under different aging times. The aging time and astringency quality measurement values ​​of the current wine are obtained, and they are input into the trained support vector machine model to determine whether the current astringency quality reaches the preset quality standard threshold. If the current astringency quality does not meet the standard, continue aging, and obtain the astringency quality measurement values ​​regularly, and repeat the judgment process of step 3; if the current astringency quality has met the standard, the current time is determined to be the optimal aging time for the batch of wine. According to the optimal aging time determined in step 4, the random forest algorithm is used to intelligently classify and identify the astringency quality of different batches of wine, and the astringency quality grade of each batch of wine is obtained. For multiple batches of wine with the same astringency quality grade, by comparing their optimal aging time, the batch with the aging time closest to the average aging time of the quality grade is selected for shipment and sale first. Continue to collect market feedback data, optimize and iterate the astringency quality prediction model and intelligent discrimination algorithm, and continuously improve the stability and consistency of the astringency quality of finished wines.

[0103] More specifically, during the aging process of wine, the historical aging data is first learned and trained by the support vector machine algorithm to establish a correlation model between astringency quality and aging time. For example, by analyzing the relationship between temperature, humidity, alcohol content and other parameters in the historical data and astringency quality, a prediction model is established, which can predict the change trend of the astringency quality of wine under specific environmental parameters. In practical applications, it is assumed that the current wine has been aged for 12 months, and the astringency quality measurement value obtained by real-time monitoring is 5, while the quality standard threshold preset by the model is 0. Since the preset threshold is not reached, the system will automatically extend the aging time and regularly detect the astringency quality. When the astringency quality measurement value reaches or exceeds 0, the system records the aging time at this time, for example, it may be 15 months, and sets this time as the optimal aging time. Next, the random forest algorithm is used to intelligently classify and identify the astringency quality of different batches of wine, and the wine is divided into different quality grades according to the classification results. For wines of the same quality grade, the system will compare the optimal aging time of each batch, and select the batch closest to the average aging time of the quality grade for priority delivery and sales. In this way, it can ensure that each batch of wine shipped out meets the corresponding quality standards, while optimizing inventory management and sales strategies. Throughout the process, the system will continue to collect market feedback and environmental monitoring data, and continuously optimize and iterate the astringency quality prediction model and intelligent discrimination algorithm to improve the quality stability and market competitiveness of wine.

[0104] The above embodiments are only descriptions of the preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the protection scope determined by the claims of the present invention.

Claims

1. An intelligent method for identifying the astringency quality of aged wine, characterized in that: include: Collecting preset parameters of wine aging to be identified; wherein the preset parameters include: aging environment parameters, sample appearance image, aroma parameters and astringency data; The preset parameters are input into the preset astringency quality intelligent prediction model to obtain the wine aging astringency quality identification result; wherein the astringency quality intelligent prediction model is constructed based on a machine learning algorithm; the astringency quality intelligent prediction model includes: a parameter quality prediction sub-model, an astringency appearance quality prediction sub-model, an aroma quality prediction sub-model and an astringency chemical quality prediction sub-model.

2. The intelligent wine aging astringency quality identification method according to claim 1 is characterized in that: The aging environment parameters of the wine to be identified include: Installing a plurality of sensors in the wine aging container to collect wine aging environment parameters; wherein the aging environment parameters include: temperature parameters, humidity parameters and alcohol content parameters; The aging environmental parameters collected in real time are preprocessed to remove abnormal values ​​and obtain standardized aging environmental parameters.

3. The intelligent wine aging astringency quality identification method according to claim 1 is characterized in that: Inputting the preset parameters into the preset astringency quality intelligent prediction model to obtain the wine aging astringency quality recognition result includes: Inputting the aging environment parameters into the parameter quality prediction sub-model to obtain parameter quality prediction data; Inputting the sample appearance image into the astringency appearance quality prediction sub-model to obtain astringency appearance quality prediction data; Inputting the aroma parameter into the aroma quality prediction sub-model to obtain aroma quality prediction data; Inputting the astringency data into the astringency chemical quality prediction sub-model to obtain astringency chemical quality prediction data; Merging the parameter quality prediction data, the astringent appearance quality prediction data, the aroma quality prediction data, and the astringent chemical quality prediction data; Constructing an astringency quality evaluation index system and determining the weight of each index; wherein the indexes include: parameter quality, astringency appearance quality, aroma quality and astringency chemical quality; The astringency quality evaluation index system is used to evaluate the fused data to obtain the wine aging astringency quality identification result.

4. The intelligent wine aging astringency quality identification method according to claim 3 is characterized in that: Inputting the aging environment parameters into the parameter quality prediction sub-model to obtain parameter quality prediction data includes: Classifying and labeling the historical aging environment parameters; Using a machine learning algorithm to train the classified aging environmental parameters to construct the parameter quality prediction sub-model of the wine aging environmental parameters; The parameter quality prediction sub-model is used to perform quality prediction analysis on the aging environment parameters collected in real time.

5. The intelligent wine aging astringency quality identification method according to claim 3 is characterized in that: Inputting the sample appearance image into the astringency appearance quality prediction sub-model to obtain astringency appearance quality prediction data comprises: According to the pre-established astringency quality evaluation index system, images of wine samples at several time points during the wine aging process are obtained; For the wine sample images obtained at several time points, a color feature extraction algorithm is used to extract the color feature vector of the wine sample; and a texture feature extraction algorithm is used to extract the clarity feature vector of the wine sample; The extracted color feature vector and clarity feature vector are input into a pre-trained support vector machine classification model to classify the astringency appearance quality of the wine samples and obtain the astringency appearance quality prediction data.

6. The intelligent wine aging astringency quality identification method according to claim 3 is characterized in that: Inputting the aroma parameter into the aroma quality prediction sub-model to obtain aroma quality prediction data includes: Acquiring aroma component data at different time points during the wine aging process, and using electronic nose technology to perform quantitative detection and analysis on the aroma component data to obtain quantitative data of the aroma components; Extracting preset key data capable of characterizing the astringency quality characteristics of the wine based on the quantitative data, and constructing an astringency aroma quality characteristics data set; Obtaining artificial sensory evaluation result data, wherein the artificial sensory evaluation result data is astringency quality scores obtained by wine tasting experts through sensory evaluation of wine samples with different aging times; Comparing and verifying the astringent aroma quality characteristic data set with the manual sensory evaluation result data, calculating the correlation between the two sets of data, and if the correlation is greater than a preset threshold, determining that the astringent aroma quality characteristic data set can effectively characterize the astringency quality of the wine; A support vector machine algorithm is used, with the astringent aroma quality characteristic data set as input and the artificial sensory evaluation result data as output, to train and establish the aroma quality prediction sub-model; Based on the aroma quality prediction sub-model, the aroma quality data of the wine is predicted according to the aroma component data obtained by electronic nose detection and analysis.

7. The intelligent wine aging astringency quality identification method according to claim 3 is characterized in that: Inputting the astringency data into the astringency chemical quality prediction sub-model, and obtaining the astringency chemical quality prediction data comprises: The chemical analysis method of high performance liquid chromatography was used to determine the changes in the content of astringent substances during wine aging and obtain astringent chemical quality data; The astringent substances include: tannins, anthocyanins; The principal component analysis method was used to comprehensively analyze the relationship between the content of various astringent substances and the astringency quality of wine, and a sub-model for predicting astringency chemical quality was constructed to obtain astringency chemical quality prediction data.

8. The intelligent wine aging astringency quality identification method according to claim 1 is characterized in that: The method further comprises: According to the output results of the preset astringency quality intelligent prediction model, a machine learning algorithm is used to intelligently judge the astringency quality of the wine at the end of aging, so as to determine the optimal aging time and release time.

Images