CN114686324A - High-efficient fermenting installation of brew wine, vinegar - Google Patents
High-efficient fermenting installation of brew wine, vinegar Download PDFInfo
- Publication number
- CN114686324A CN114686324A CN202210506870.0A CN202210506870A CN114686324A CN 114686324 A CN114686324 A CN 114686324A CN 202210506870 A CN202210506870 A CN 202210506870A CN 114686324 A CN114686324 A CN 114686324A
- Authority
- CN
- China
- Prior art keywords
- feature
- matrix
- vector
- feature matrix
- eigenvector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12G—WINE; PREPARATION THEREOF; ALCOHOLIC BEVERAGES; PREPARATION OF ALCOHOLIC BEVERAGES NOT PROVIDED FOR IN SUBCLASSES C12C OR C12H
- C12G3/00—Preparation of other alcoholic beverages
- C12G3/02—Preparation of other alcoholic beverages by fermentation
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12J—VINEGAR; PREPARATION OR PURIFICATION THEREOF
- C12J1/00—Vinegar; Preparation or purification thereof
- C12J1/10—Apparatus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Organic Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Genetics & Genomics (AREA)
- Mathematical Optimization (AREA)
- Wood Science & Technology (AREA)
- Zoology (AREA)
- Food Science & Technology (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Distillation Of Fermentation Liquor, Processing Of Alcohols, Vinegar And Beer (AREA)
Abstract
The application relates to the field of intelligent brewing under intelligent manufacturing, and particularly discloses a high-efficiency fermentation device for brewing wine and vinegar, which can intelligently adjust the temperature based on the real-time condition of a brewed object in brewing equipment, so that the adjusted temperature is adapted to the activity requirements of different stages on different microorganisms, and the fermentation efficiency of the brewed wine and vinegar is improved.
Description
Technical Field
The invention relates to the field of intelligent brewing under intelligent manufacturing, in particular to a high-efficiency fermentation device for brewing wine and vinegar.
Background
Brewing is the process of converting starch in cereal grains into wine and is divided into two stages of saccharification and alcoholization. The saccharification stage is that grains and grains are converted into fermentable sugars under the action of pretreatment and various biological enzymes, and the alcoholization stage is that the hydrolyzed sugars are metabolized under the action of microorganisms to produce alcohol. The vinegar brewing process is that high carbohydrate food is decomposed under the action of microbe to produce monosaccharide, disaccharide, organic acid, alcohol and aldehyde, which will produce sour taste or wine taste.
The difference between the temperature and the time of air contact during fermentation is mainly found in the wine-brewing and vinegar-making process, as is well known, the vinegar is brewed firstly, that is, the ethanol in the material liquid can be changed into acetic acid, that is, the acetic acid, by fermenting the material liquid after alcohol fermentation through the acetic acid.
However, the existing brewing and vinegar brewing fermentation depend on physical devices, and the self-adaptive adjustment cannot be made according to the fermentation state, so that the brewing effect is not particularly ideal. For brewing and vinegar brewing, the key element is temperature control, for microorganisms the activity at different temperatures is different, while different microorganisms are required to have different activity at different stages.
Therefore, an intelligent, efficient fermentation apparatus for brewing alcoholic beverages and vinegar is desired, which can adaptively control the temperature based on the real-time state of the fermentation product in the fermentation apparatus to improve the fermentation efficiency of the brewing alcoholic beverages and vinegar.
Disclosure of Invention
The present application is proposed to solve the above-mentioned technical problems. The embodiment of the application provides an efficient fermentation device for brewing wine and vinegar, which excavates a feature distribution representation of local features of an image frame in a monitoring video about a brewed object in a high-dimensional space through a first convolution neural network using a three-dimensional convolution kernel, and respectively extracts high-dimensional implicit associated features of temperature data and smell data of each preset time point in a time sequence dimension through a time sequence encoder model, so that after feature information fusion is carried out, the feature information fusion device is further subjected to reparameterization to obtain a general probability distribution containing a special distribution by interpreting feature values as negative logarithms of univariate differences so as to ensure a special example in a sample, and therefore, the certainty of probability expression of the first feature matrix and the second feature matrix as a whole is improved. Further, the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar can be more effectively improved.
According to an aspect of the present application, there is provided a high efficiency fermentation apparatus for brewing wine and vinegar, comprising:
the sensor unit is used for acquiring temperature data and odor data of a plurality of preset time points in a preset time period through a temperature sensor and an odor sensor which are arranged in a fermentation tank, and acquiring an infrared monitoring video of a brewed product in the preset time period through an infrared camera which is arranged in the fermentation tank;
the three-dimensional convolution coding unit is used for enabling the infrared monitoring video to pass through a first convolution neural network using a three-dimensional convolution kernel so as to obtain a first feature vector;
the first time sequence coding unit is used for enabling the temperature data and the smell data of a plurality of preset time points in the preset time period to pass through a time sequence coder comprising a one-dimensional convolution layer and a full connection layer respectively so as to obtain a second eigenvector and a third eigenvector;
a first joint encoding unit, configured to calculate a product of a transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix;
a second joint encoding unit, configured to calculate a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix;
a reparameterization unit configured to reparameterize the first feature matrix and the second feature matrix to obtain a reparameterized first feature matrix and a reparameterized second feature matrix, wherein the reparameterization of the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean value of feature values of all positions in the first feature matrix or the second feature matrix;
a feature fusion unit for calculating a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classification feature matrix; and
and the regulation and control result generation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating that the temperature should be increased or decreased.
According to another aspect of the present application, a fermentation control method of a high-efficiency fermentation apparatus for brewing wine and vinegar includes:
acquiring temperature data and odor data of a plurality of preset time points in a preset time period by deploying a temperature sensor and an odor sensor in a fermentation tank, and acquiring an infrared monitoring video of a brew in the preset time period by deploying an infrared camera in the fermentation equipment;
passing the infrared monitoring video through a first convolution neural network using a three-dimensional convolution kernel to obtain a first feature vector;
respectively passing the temperature data and the smell data of a plurality of preset time points in the preset time period through a time sequence encoder comprising a one-dimensional convolution layer and a full-connection layer to obtain a second eigenvector and a third eigenvector;
calculating a product of the transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix;
calculating a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix;
re-parameterizing the first feature matrix and the second feature matrix to obtain a re-parameterized first feature matrix and a re-parameterized second feature matrix, wherein the re-parameterizing of the first feature matrix and the second feature matrix is performed based on a log function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean of feature values of all positions in the first feature matrix or the second feature matrix;
calculating a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classified feature matrix; and
and passing the classification characteristic matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating that the temperature should be increased or decreased.
Compared with the prior art, the high-efficiency fermentation device for brewing wine and vinegar provided by the application digs out the feature distribution representation of the local features of the image frames in the monitoring video about the brewed object in a high-dimensional space by using the first convolution neural network of the three-dimensional convolution kernel, and respectively extracts the high-dimensional implicit associated features of the temperature data and the smell data of each preset time point in a time sequence dimension by using the time sequence encoder model, so that after feature information fusion is carried out, the feature information fusion device is further subjected to re-parameterization to obtain a general probability distribution containing a special distribution by interpreting the feature values as the negative logarithm of a univariate difference so as to ensure a special example in a sample, and thus, the certainty of probability expression of the first feature matrix and the second feature matrix as a whole is improved. Further, the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar can be more effectively improved.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally represent like parts or steps.
Fig. 1 is a view of an application scenario of a high-efficiency fermentation device for brewing wine and vinegar according to an embodiment of the application.
Fig. 2A is a block diagram of an efficient fermentation apparatus for brewing wine and vinegar according to an embodiment of the present application.
FIG. 2B is a block diagram of a first timing encoding unit in the high efficiency fermentation apparatus according to an embodiment of the present application.
Fig. 3 is a flowchart of a fermentation control method of the high-efficiency fermentation apparatus for brewing alcohol and vinegar according to the embodiment of the present application.
Fig. 4 is a schematic configuration diagram of a fermentation control method of an efficient fermentation device for brewing wine and vinegar according to an embodiment of the present application.
Detailed Description
Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.
Overview of scenes
As mentioned above, brewing is the process of converting starch in cereal grains into wine, and is divided into two stages of saccharification and alcoholization. The saccharification stage is that grains and grains are converted into fermentable sugars under the action of pretreatment and various biological enzymes, and the alcoholization stage is that the hydrolyzed sugars are metabolized under the action of microorganisms to produce alcohol.
The vinegar brewing process is that high carbohydrate food is decomposed under the action of microbe to produce monosaccharide, disaccharide, organic acid, alcohol and aldehyde, which will produce sour taste or wine taste.
The difference between the temperature and the time of air contact during fermentation is mainly found in the wine-brewing and vinegar-making process, as is well known, the vinegar is brewed firstly, that is, the ethanol in the material liquid can be changed into acetic acid, that is, the acetic acid, by fermenting the material liquid after alcohol fermentation through the acetic acid.
However, the existing brewing and vinegar brewing fermentation depend on physical devices, and the self-adaptive adjustment cannot be made according to the fermentation state, so that the brewing effect is not particularly ideal. That is, for brewing wine and vinegar, the key element is temperature control. In particular, the activity is different for the microorganisms at different temperatures, whereas different stages require different microorganisms to have different activities. Therefore, an efficient fermentation device for brewing wine and vinegar is needed to improve the fermentation efficiency of brewing wine and vinegar.
At present, deep learning and neural networks have been widely applied in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks also exhibit a level close to or even exceeding that of humans in the fields of image classification, object detection, semantic segmentation, text translation, and the like.
In recent years, deep learning and development of neural networks provide solutions and schemes for efficient fermentation of brewed wine and vinegar.
It will be appreciated that since the key element for brewing and vinegar brewing is temperature control. For microorganisms, the activity is different at different temperatures, and different microorganisms are required to have different activities in different stages of fermentation, so that the fermentation efficiency of brewed wine and vinegar can be improved. Meanwhile, considering that the brewed product can generate different odors along with the depth of the brewing process, the odor data is combined to assist the dynamic regulation and control of the temperature, so as to more effectively improve the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar.
Based on the above, the applicant of the present application uses a deep neural network model to extract the associated feature information of the temperature data and the odor data at a plurality of time points in a statistical sense in a time dimension, extract the slightly changing features of the fermentation microorganisms and the fermentation products from the infrared monitoring video, and then dynamically regulate and control the fermentation temperature through the classification of the classifier so as to improve the fermentation efficiency of the brewed wine and vinegar.
Specifically, in the technical scheme of the application, firstly, temperature data and odor data of a plurality of preset time points in a preset time period are obtained through a temperature sensor and an odor sensor which are deployed in fermentation equipment, and an infrared monitoring video of a brew in the preset time period is obtained through an infrared camera which is deployed in the fermentation equipment. In one specific example, the temperature sensor DS18B20 and the gas detector can be used to detect the temperature and odor in the fermentation device, and since there is no visible light in the fermentation device, the infrared camera is selected to collect the infrared monitoring video of the fermented product.
Then, the infrared monitoring video is processed in a first convolution neural network using a three-dimensional convolution kernel, so that the feature distribution representation of local features of image frames in the monitoring video, which relate to a brew, in a high-dimensional space is further excavated, and a first feature vector is obtained. Therefore, the fermentation fine dynamic change characteristics of the fermented product in time sequence can be extracted, and the subsequent temperature can be more accurately adjusted based on the fine dynamic change characteristics.
And the temperature data and the smell data of a plurality of preset time points in a preset time period are respectively passed through a time sequence encoder comprising a one-dimensional convolution layer and a full connection layer to obtain a second eigenvector and a third eigenvector. Therefore, the high-dimensional implicit characteristics of the temperature data and the odor data at each preset time point can be respectively extracted, and the high-dimensional implicit associated characteristics of the temperature data and the odor data at each preset time point can be extracted.
Further, a product of the transposed vector of the first eigenvector and the second eigenvector is calculated to obtain a first eigenvector matrix. It should be understood that the feature information of each position in the first feature vector is fused with the feature information of each position in the second feature vector to strengthen the encoding of the image along the specific direction of the text to encode the corresponding attribute of the infrared image, so that the implicit association feature information related to the temperature is highlighted, and the accuracy of the dynamic temperature adjustment is facilitated.
Meanwhile, the product of the transposed vector of the first eigenvector and the third eigenvector is calculated to obtain a second eigenvector matrix. Namely, the feature information of each position in the first feature vector and the feature information of each position in the third feature vector are fused, so that the coding of the image can be enhanced along the specific direction of the text to code the corresponding attribute of the infrared image, the hidden associated feature information related to the smell is highlighted, and the accuracy of dynamic temperature adjustment is facilitated.
It should be understood that, since the first feature matrix expresses the time-sequence position-based Response between the odor features and the image semantics and the second feature matrix expresses the time-sequence position-based Response between the temperature features and the image semantics, since the time-sequence position-based Response may generate some Outlier responses (Outlier responses), before fusing the first feature matrix and the second feature matrix, the first feature matrix and the second feature matrix are first re-parameterized, specifically:
mi,jan eigenvalue of each position of the eigenvalue matrix is represented, an
Represents the mean of the eigenvalues of all positions of the eigen matrix.
Thus, the reparameterization obtains a generic probability distribution containing a special distribution by interpreting the eigenvalues as the negative logarithms of the univariate differences to ensure concealment of the special examples in the sample, i.e., the outlier response values, to the perturbation of the distribution as a whole, thus improving the certainty in the probability expression of the first and second eigenmatrices as a whole.
Further, a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix is calculated to obtain a classification feature matrix. Namely, the balance between the first feature matrix with the parameterization and the second feature matrix with the parameterization in the final feature matrix is controlled by calculating the weighted sum of the two, namely, the balance between the image semantic response feature expressing the odor and the image semantic response feature expressing the temperature in the final feature matrix is realized, and then the temperature dynamic adjustment result obtained by the final classification feature matrix can focus on the odor change of the brewed product on the basis of the local dynamic features of the brewed product, so that the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar are improved more effectively.
Based on this, this application has proposed a brew wine, high-efficient fermenting installation of vinegar, it includes: the sensor unit is used for acquiring temperature data and odor data of a plurality of preset time points in a preset time period through a temperature sensor and an odor sensor which are arranged in a fermentation tank, and acquiring an infrared monitoring video of a brewed product in the preset time period through an infrared camera which is arranged in the fermentation tank; the three-dimensional convolution coding unit is used for enabling the infrared monitoring video to pass through a first convolution neural network using a three-dimensional convolution kernel so as to obtain a first feature vector; the first time sequence coding unit is used for enabling the temperature data and the smell data of a plurality of preset time points in the preset time period to pass through a time sequence coder comprising a one-dimensional convolution layer and a full connection layer respectively so as to obtain a second eigenvector and a third eigenvector; a first joint encoding unit, configured to calculate a product of a transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix; a second joint encoding unit, configured to calculate a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix; a reparameterization unit configured to reparameterize the first feature matrix and the second feature matrix to obtain a reparameterized first feature matrix and a reparameterized second feature matrix, wherein the reparameterization of the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean value of feature values of all positions in the first feature matrix or the second feature matrix; a feature fusion unit for calculating a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classification feature matrix; and the regulation and control result generation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating that the temperature should be increased or decreased.
Fig. 1 illustrates an application scenario of a high-efficiency fermentation device for brewing wine and vinegar according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, temperature data and odor data at a plurality of predetermined time points within a predetermined period of time are acquired by a temperature sensor (e.g., T1 as illustrated in fig. 1) and an odor sensor (e.g., T2 as illustrated in fig. 1) disposed within a fermentation tank (e.g., F as illustrated in fig. 1), and an infrared monitoring video of a brew (e.g., B as illustrated in fig. 1) for the predetermined period of time is acquired by an infrared camera (e.g., C as illustrated in fig. 1) disposed within the fermentation tank. Here, the temperature sensor may be DS18B20, and the odor sensor may be a gas detector. Then, the temperature data and the smell data of a plurality of preset time points in the preset time period and the infrared monitoring video of the preset time period are input into a server (for example, a server S as illustrated in figure 1) which is deployed with an efficient fermentation algorithm for brewing wine and vinegar, wherein the server can process the temperature data and the smell data of the plurality of preset time points in the preset time period and the infrared monitoring video of the preset time period by the efficient fermentation algorithm for brewing wine and vinegar to generate a classification result for indicating that the temperature should be increased or the temperature should be decreased.
Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.
Exemplary System
FIG. 2A illustrates a block diagram of an efficient fermentation apparatus for brewing wine and vinegar according to an embodiment of the present application. As shown in fig. 2A, the high efficiency fermentation apparatus 200 for brewing wine and vinegar according to the embodiment of the present application includes: the sensor unit 210 is used for acquiring temperature data and odor data of a plurality of preset time points in a preset time period through a temperature sensor and an odor sensor which are arranged in a fermentation tank, and acquiring an infrared monitoring video of a brewed product in the preset time period through an infrared camera which is arranged in the fermentation tank; a three-dimensional convolution coding unit 220, configured to pass the infrared surveillance video through a first convolution neural network using a three-dimensional convolution kernel to obtain a first feature vector; a first time sequence encoding unit 230, configured to pass the temperature data and the odor data at multiple predetermined time points in the predetermined time period through a time sequence encoder including a one-dimensional convolutional layer and a full-link layer to obtain a second eigenvector and a third eigenvector, respectively; a first joint encoding unit 240, configured to calculate a product of the transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix; a second joint encoding unit 250, configured to calculate a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix; a reparameterization unit 260 configured to reparameterize the first feature matrix and the second feature matrix to obtain a reparameterized first feature matrix and a reparameterized second feature matrix, wherein the reparameterization of the first feature matrix and the second feature matrix is performed based on a logarithm of a function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean of feature values of all positions in the first feature matrix or the second feature matrix; a feature fusion unit 270, configured to calculate a weighted sum per location of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classified feature matrix; and a control result generating unit 280 for passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating that the temperature should be increased or decreased.
Specifically, in the embodiment of the present application, the sensor unit 210 and the three-dimensional convolution coding unit 220 are configured to acquire temperature data and odor data at a plurality of predetermined time points in a predetermined time period by deploying a temperature sensor and an odor sensor in a fermentation tank, acquire an infrared monitoring video of a brewed product in the predetermined time period by deploying an infrared camera in the fermentation tank, and pass the infrared monitoring video through a first convolution neural network using a three-dimensional convolution kernel to obtain a first feature vector. As previously mentioned, it will be appreciated that a key element for brewing wine and vinegar is temperature control. For microorganisms, the activity is different at different temperatures, and different microorganisms are required to have different activities in different stages of fermentation, so that the fermentation efficiency of brewed wine and vinegar can be improved. Meanwhile, considering that the brewed product generates different odors along with the depth of the brewing process, the odor data is combined to assist the dynamic regulation and control of the temperature so as to effectively improve the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar. That is, in the technical solution of the present application, the deep neural network model is used to extract the associated feature information of the temperature data and the odor data at the plurality of time points in the time dimension in a statistical sense, extract the slightly changing features of the fermentation microorganisms and the fermentation products from the infrared monitoring video, and then dynamically regulate and control the temperature of the fermentation through the classification of the classifier so as to improve the fermentation efficiency of the brewed wine and vinegar.
Specifically, in the technical scheme of the application, firstly, temperature data and odor data of a plurality of preset time points in a preset time period are obtained through the temperature sensor and the odor sensor which are deployed in the fermentation tank, and an infrared monitoring video of a brewed product in the preset time period is obtained through an infrared camera which is deployed in the fermentation tank. In one specific example, a temperature sensor DS18B20 and a gas detector can be used to detect the temperature and odor in the fermentation device, and the infrared camera is selected to collect the infrared monitoring video of the fermented product because there is no visible light in the fermentation device.
Then, the acquired infrared monitoring video is processed in a first convolution neural network by using a three-dimensional convolution kernel so as to more deeply dig out a feature distribution representation of local features of image frames in the monitoring video, wherein the local features relate to a brew substance, in a high-dimensional space, and therefore a first feature vector is obtained. Therefore, the fermentation fine dynamic change characteristics of the fermented product in time sequence can be extracted, and the subsequent temperature can be more accurately adjusted based on the fine dynamic change characteristics.
More specifically, in an embodiment of the present application, the three-dimensional convolutional encoding unit includes: processing the infrared surveillance video using a first convolutional neural network of the three-dimensional convolutional kernel with the following formula to generate the first feature vector;
wherein the formula is:
wherein Hj、WjAnd RjRespectively representing the length, width and height of the three-dimensional convolution kernel, m represents the number of (l-1) th layer characteristic diagrams,
is a convolution kernel connected to the mth feature map of the (l-1) layer, bljFor biasing, f (-) represents the activation function.
Specifically, in the embodiment of the present application, the first time-sequence encoding unit 230 is configured to pass the temperature data and the odor data at a plurality of predetermined time points in the predetermined time period through a time-sequence encoder including a one-dimensional convolutional layer and a full-link layer, respectively, to obtain a second eigenvector and a third eigenvector. That is, in the technical solution of the present application, the temperature data and the odor data at a plurality of predetermined time points in the predetermined time period are further subjected to time-dimension encoding processing in a time-series encoder including a one-dimensional convolutional layer and a fully-connected layer, respectively, to obtain a second eigenvector and a third eigenvector. Therefore, not only can the high-dimensional implicit characteristics of the temperature data and the odor data at each preset time point be respectively extracted, but also the high-dimensional implicit associated characteristics between the temperature data and the odor data at each preset time point can be extracted.
More specifically, the first time-sequence encoding unit includes: firstly, arranging temperature data of a plurality of preset time points in the preset time period into a one-dimensional temperature input vector according to a time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the temperature input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
represents a matrix multiplication; performing one-dimensional convolution encoding on the temperature input vector by using a one-dimensional convolution layer of the time sequence encoder according to the following formula so as to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
Then, arranging the odor data of a plurality of preset time points in the preset time period into a one-dimensional temperature input vector according to a time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the smell input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
represents a matrix multiplication; performing one-dimensional convolutional encoding on the smell input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula so as to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
Fig. 2B illustrates a block diagram of a first timing encoding unit in the wearable body fluid monitoring device according to an embodiment of the present application. As shown in fig. 2B, the first timing encoding unit 230 includes: the first arrangement subunit 231 is configured to arrange the temperature data at the plurality of predetermined time points in the predetermined time period into a one-dimensional temperature input vector according to a time dimension. A first full-concatenation subunit 232, configured to perform full-concatenation encoding on the temperature input vector obtained by the first arrangement subunit 231 by using a full-concatenation layer of the time-series encoder according to the following formula to extract a high-dimensional implicit feature of a feature value at each position in the input vector, where the formula is:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
representing a matrix multiplication. A first one-dimensional convolution subunit 233, performing one-dimensional convolution encoding on the temperature input vector obtained by the first arrangement subunit 231 by using the one-dimensional convolution layer of the time-series encoder according to the following formula to extract high-dimensional implicit correlation features between feature values of each position in the input vector, where the formula is:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel. A second arranging subunit 234, configured to arrange the odor data at a plurality of predetermined time points in the predetermined time period into a one-dimensional temperature according to a time dimensionThe vector is input in degrees. A second full-concatenation subunit 235, configured to perform full-concatenation encoding on the odor input vector using the full-concatenation layers of the time-series encoder obtained by the second arrangement subunit 234, by using the following formula to extract high-dimensional implicit features of feature values of each position in the input vector, where the formula is:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
representing a matrix multiplication. A second one-dimensional convolution subunit 236, configured to perform, by using the one-dimensional convolution layer of the time-series encoder obtained by the second arranging subunit 234, one-dimensional convolution encoding on the odor input vector by using the following formula to extract high-dimensional implicit correlation features between feature values of each position in the input vector, where the formula is:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
Specifically, in this embodiment, the first joint encoding unit 240 is configured to calculate a product of the transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix. That is, in the technical solution of the present application, in order to fuse the infrared image feature and the temperature dynamic feature, further, a product of the transposed vector of the first feature vector and the second feature vector is calculated to obtain the first feature matrix. It should be understood that the feature information of each position in the first feature vector is fused with the feature information of each position in the second feature vector to strengthen the encoding of the image along the specific direction of the text to encode the corresponding attribute of the infrared image, so that the implicit association feature information related to the temperature is highlighted, and the accuracy of the dynamic temperature adjustment is facilitated.
More specifically, in an embodiment of the present application, the first joint encoding unit is further configured to: calculating a product of the transposed vector of the first eigenvector and the second eigenvector in a formula to obtain the first eigenvector matrix;
wherein the formula is:
wherein
Denotes the multiplication of vectors, V1Representing said first feature vector, V2Representing the second feature vector in the second set of feature vectors,
a transposed vector representing the first feature vector.
Specifically, in the embodiment of the present application, the second joint encoding unit 250 is configured to calculate a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix. That is, in the technical solution of the present application, in order to fuse the infrared image feature and the odor dynamic feature, further, a product of the transposed vector of the first feature vector and the third feature vector is calculated to obtain the second feature matrix. It should be understood that by fusing the feature information of each position in the first feature vector with the feature information of each position in the third feature vector, the encoding of the image can be enhanced along the specific direction of the text to encode the corresponding attribute of the infrared image, so that the implicit association feature information related to the odor is highlighted, and the accuracy of the dynamic temperature adjustment is facilitated.
Specifically, in this embodiment of the present application, the reparameterizing unit 260 is configured to reparameterize the first feature matrix and the second feature matrix to obtain a reparameterized first feature matrix and a reparameterized second feature matrix, wherein the reparameterizing the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean value of feature values of all positions in the first feature matrix or the second feature matrix. It should be appreciated that since the first feature matrix expresses the time-series position-wise Response between odor features and image semantics and the second feature matrix expresses the time-series position-wise Response between temperature features and image semantics, the first feature matrix and the second feature matrix are first re-parameterized prior to fusing the first feature matrix and the second feature matrix, since the time-series position-wise Response may generate some Outlier responses (Outlier responses). In particular, in this way, the reparameterization obtains a generic probability distribution containing a particular distribution by interpreting the eigenvalues as the negative logarithms of univariate differences to guarantee a particular instantiation in the sample, i.e. the concealment of the outlier response values to the perturbation of the distribution as a whole, thus improving the certainty in the probability expression of the first and second eigenmatrices as a whole.
More specifically, in an embodiment of the present application, the reparameterization unit is further configured to: re-parameterizing the first feature matrix and the second feature matrix in the following formula to obtain the re-parameterized first feature matrix and the re-parameterized second feature matrix;
wherein the formula is:
mi,jan eigenvalue of each position of the characteristic matrix is represented, and
represents the mean of the eigenvalues of all positions of the eigen matrix.
Specifically, in the embodiment of the present application, the feature fusion unit 270 and the adjustment result generation unit 280 are configured to calculate a weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix according to a position to obtain a classification feature matrix, and the classification feature matrix is passed through a classifier to obtain a classification result, where the classification result is used to indicate that the temperature should be increased or decreased. That is, in the technical solution of the present application, further, a weighted sum by location of the reparameterized first feature matrix and the reparameterized second feature matrix is calculated to obtain the classification feature matrix. Namely, by calculating the weighted sum of the two, the balance between the re-parameterized first feature matrix and the re-parameterized second feature matrix in the final feature matrix is controlled, namely, the balance between the image semantic response feature expressing the odor and the image semantic response feature expressing the temperature in the final feature matrix is realized, so that the temperature dynamic adjustment result obtained by the final classification feature matrix can focus on the odor change of the brewed product on the basis of the local dynamic feature including the brewed product, and the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar are more effectively improved. Then, the classification characteristic matrix is passed through a classifier to obtain a classification result indicating that the temperature should be increased or decreased.
More specifically, in this embodiment of the present application, the regulation result generating unit is further configured to: the classifier processes the classification feature matrix to generate the classification result according to the following formula: softmax { (W)n,Bn):…:(W1,B1) L project (F), where project (F) denotes the projection of the classification feature matrix as a vector, W1To WnAs a weight matrix for each fully connected layer, B1To BnA bias matrix representing the layers of the fully connected layer.
In summary, the efficient fermentation apparatus 200 for brewing wine and vinegar based on the embodiment of the present application is illustrated, which excavates a feature distribution representation of local features about a brew in an image frame in a surveillance video in a high-dimensional space by using a first convolution neural network of a three-dimensional convolution kernel, and extracts high-dimensional implicit correlation features of temperature data and odor data at each predetermined time point in a time-series dimension respectively by a time-series encoder model, so that after feature information fusion, it is further re-parameterized to obtain a generic probability distribution containing a specific distribution by interpreting the feature values as negative logarithms of univariate differences to guarantee a specific example in a sample, thus improving certainty in probability expression of the first feature matrix and the second feature matrix as a whole. In this way, the temperature is intelligently adjusted based on the real-time condition of the brew substance in the brewing device, so that the adjusted temperature is adapted to the activity requirements of different stages for different microorganisms, thereby improving the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar.
As described above, the efficient fermentation apparatus 200 for brewing wine and vinegar according to the embodiment of the present application may be implemented in various terminal devices, such as a server for an efficient fermentation algorithm for brewing wine and vinegar, and the like. In one example, the high efficiency fermentation apparatus 200 for brewing wine and vinegar according to the embodiments of the present application may be integrated into a fermentation tank as one software module and/or hardware module. For example, the high efficiency fermentation device 200 for brewing wine, vinegar may be a software module in the operating system of the fermentor, or may be an application developed for the fermentor; of course, the high efficiency wine and vinegar brewing fermentor 200 may also be one of many hardware modules of the fermentor.
Alternatively, in another example, the high-efficiency wine and vinegar brewing fermentation device 200 and the fermentation tank may be separate devices, and the high-efficiency wine and vinegar brewing fermentation device 200 may be connected to the fermentation tank through a wired and/or wireless network and transmit the interactive information according to an agreed data format.
Exemplary method
FIG. 3 is a flow chart showing a fermentation control method of the high-efficiency fermentation apparatus for brewing wine and vinegar. As shown in fig. 3, the method for controlling fermentation of an efficient fermentation apparatus for brewing wine and vinegar according to an embodiment of the present application includes the steps of: s110, acquiring temperature data and odor data of a plurality of preset time points in a preset time period by deploying a temperature sensor and an odor sensor in a fermentation tank, and acquiring an infrared monitoring video of a brewed product in the preset time period by deploying an infrared camera in the fermentation tank; s120, enabling the infrared monitoring video to pass through a first convolution neural network using a three-dimensional convolution kernel to obtain a first feature vector; s130, enabling the temperature data and the smell data of a plurality of preset time points in the preset time period to pass through a time sequence encoder comprising a one-dimensional convolution layer and a full-connection layer respectively to obtain a second eigenvector and a third eigenvector; s140, calculating a product of the transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix; s150, calculating the product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix; s160, re-parameterizing the first feature matrix and the second feature matrix to obtain a re-parameterized first feature matrix and a re-parameterized second feature matrix, wherein the re-parameterizing the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean of feature values of all positions in the first feature matrix or the second feature matrix; s170, calculating a position-weighted sum of the re-parameterized first feature matrix and the re-parameterized second feature matrix to obtain a classification feature matrix; and S180, passing the classification characteristic matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating that the temperature should be increased or decreased.
Fig. 4 illustrates an architecture diagram of a fermentation control method of a high-efficiency fermentation device for brewing wine and vinegar according to an embodiment of the application. As shown in fig. 4, in the network architecture of the fermentation control method of the efficient fermentation apparatus for brewing wine and vinegar, first, the infrared surveillance video (e.g., P1 as illustrated in fig. 4) is passed through a first convolution neural network (e.g., CNN1 as illustrated in fig. 4) using a three-dimensional convolution kernel to obtain a first eigenvector (e.g., VF1 as illustrated in fig. 4); then, passing the temperature data (e.g., Q1 as illustrated in fig. 4) and the smell data (e.g., Q2 as illustrated in fig. 4) at a plurality of predetermined time points within the predetermined time period through a time-sequential encoder (e.g., E as illustrated in fig. 4) including one-dimensional convolutional layers and fully-connected layers to obtain a second eigenvector (e.g., VF2 as illustrated in fig. 4) and a third eigenvector (e.g., VF3 as illustrated in fig. 4), respectively; then, a product of the transposed vector of the first eigenvector and the second eigenvector is calculated to obtain a first eigenvector matrix (e.g., MF1 as illustrated in fig. 4); then, a product of the transposed vector of the first eigenvector and the third eigenvector is calculated to obtain a second eigenvector matrix (e.g., MF2 as illustrated in fig. 4); then, re-parameterizing the first feature matrix and the second feature matrix to obtain a re-parameterized first feature matrix (e.g., M1 as illustrated in fig. 4) and a re-parameterized second feature matrix (e.g., M2 as illustrated in fig. 4); then, computing a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classification feature matrix (e.g., MF as illustrated in fig. 4); and, finally, passing the classification feature matrix through a classifier (e.g., a classifier as illustrated in fig. 4) to obtain a classification result indicating that the temperature should be increased or decreased.
More specifically, in steps S110 and S120, temperature data and odor data at a plurality of predetermined time points within a predetermined time period are acquired by deploying a temperature sensor and an odor sensor in a fermentation tank, and an infrared monitoring video of a brew in the predetermined time period is acquired by deploying an infrared camera in the fermentation tank, and the infrared monitoring video is passed through a first convolution neural network using a three-dimensional convolution kernel to obtain a first feature vector. That is, in the technical solution of the present application, first, temperature data and odor data at a plurality of predetermined time points within a predetermined time period are obtained through the temperature sensor and the odor sensor disposed in the fermentation tank, and an infrared monitoring video of a brew in the predetermined time period is obtained through an infrared camera disposed in the fermentation tank. In one specific example, a temperature sensor DS18B20 and a gas detector can be used to detect the temperature and odor in the fermentation device, and the infrared camera is selected to collect the infrared monitoring video of the fermented product because there is no visible light in the fermentation device.
Then, the acquired infrared monitoring video is processed in a first convolution neural network by using a three-dimensional convolution kernel so as to more deeply dig out a feature distribution representation of local features of image frames in the monitoring video, wherein the local features relate to a brew substance, in a high-dimensional space, and therefore a first feature vector is obtained. Therefore, the fermentation fine dynamic change characteristics of the fermented product in time sequence can be extracted, and the subsequent temperature can be more accurately adjusted based on the fine dynamic change characteristics.
More specifically, in step S130, the temperature data and the smell data at a plurality of predetermined time points within the predetermined period are respectively passed through a time-series encoder including a one-dimensional convolution layer and a full-link layer to obtain a second feature vector and a third feature vector. That is, in the technical solution of the present application, the temperature data and the odor data at a plurality of predetermined time points in the predetermined time period are further subjected to time-dimension encoding processing in a time-series encoder including a one-dimensional convolutional layer and a fully-connected layer, respectively, to obtain a second eigenvector and a third eigenvector. Therefore, not only can the high-dimensional implicit characteristics of the temperature data and the odor data at each preset time point be respectively extracted, but also the high-dimensional implicit associated characteristics between the temperature data and the odor data at each preset time point can be extracted.
More specifically, in step S140, a product of the transposed vector of the first eigenvector and the second eigenvector is calculated to obtain a first eigenvector matrix. That is, in the technical solution of the present application, in order to fuse the infrared image feature and the temperature dynamic feature, further, a product of the transposed vector of the first feature vector and the second feature vector is calculated to obtain the first feature matrix. It should be understood that the feature information of each position in the first feature vector is fused with the feature information of each position in the second feature vector to strengthen the encoding of the image along the specific direction of the text to encode the corresponding attribute of the infrared image, so that the implicit association feature information related to the temperature is highlighted, and the accuracy of the dynamic temperature adjustment is facilitated.
More specifically, in steps S150 and S160, the product of the transposed vector of the first eigenvector and the third eigenvector is calculated to obtain a second eigenvector matrix. That is, in the technical solution of the present application, in order to fuse the infrared image feature and the odor dynamic feature, further, a product of the transposed vector of the first feature vector and the third feature vector is calculated to obtain the second feature matrix. It should be understood that by fusing the feature information of each position in the first feature vector with the feature information of each position in the third feature vector, the encoding of the image can be enhanced along the specific direction of the text to encode the corresponding attribute of the infrared image, so that the implicit association feature information related to the odor is highlighted, and the accuracy of the dynamic temperature adjustment is facilitated.
More specifically, in step S160, the first feature matrix and the second feature matrix are re-parameterized to obtain a re-parameterized first feature matrix and a re-parameterized second feature matrix, wherein the re-parameterization of the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean value of feature values of all positions in the first feature matrix or the second feature matrix. It should be appreciated that since the first feature matrix expresses time-series position-wise responses between odor features and image semantics and the second feature matrix expresses time-series position-wise responses between temperature features and image semantics, the first and second feature matrices are first re-parameterized prior to fusing the first and second feature matrices, since the time-series position-wise responses may produce some Outlier responses (Outlier responses). In particular, in this way, the reparameterization obtains a generic probability distribution containing a particular distribution by interpreting the eigenvalues as the negative logarithms of univariate differences to guarantee a particular instantiation in the sample, i.e. the concealment of the outlier response values to the perturbation of the distribution as a whole, thus improving the certainty in the probability expression of the first and second eigenmatrices as a whole.
More specifically, in step S170 and step S180, a location-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix is calculated to obtain a classification feature matrix, and the classification feature matrix is passed through a classifier to obtain a classification result indicating that the temperature should be increased or decreased. That is, in the technical solution of the present application, further, a weighted sum by location of the reparameterized first feature matrix and the reparameterized second feature matrix is calculated to obtain the classification feature matrix. Namely, by calculating the weighted sum of the two, the balance between the re-parameterized first feature matrix and the re-parameterized second feature matrix in the final feature matrix is controlled, namely, the balance between the image semantic response feature expressing the odor and the image semantic response feature expressing the temperature in the final feature matrix is realized, so that the temperature dynamic adjustment result obtained by the final classification feature matrix can focus on the odor change of the brewed product on the basis of the local dynamic feature including the brewed product, and the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar are more effectively improved. Then, the classification characteristic matrix is passed through a classifier to obtain a classification result indicating that the temperature should be increased or decreased.
In conclusion, the fermentation control method of the high-efficiency fermentation device for brewing wine and vinegar based on the embodiment of the application is clarified, which excavates a feature distribution representation of local features of image frames in a surveillance video with respect to a brew in a high-dimensional space by using a first convolution neural network of a three-dimensional convolution kernel, and the high-dimensional implicit correlation characteristics of the temperature data and the smell data of each preset time point on the time sequence dimension are respectively extracted through a time sequence encoder model, so that, after feature information fusion, it is further re-parameterized to obtain a generic probability distribution containing a special distribution by interpreting the feature values as the negative logarithm of the univariate difference, to guarantee a particular instantiation in a sample, and in this way, to improve the certainty in the probabilistic expression of the first feature matrix and the second feature matrix as a whole. Further, the fermentation efficiency and the fermentation accuracy of the brewed wine and vinegar can be more effectively improved.
The foregoing describes the general principles of the present application in conjunction with specific embodiments, however, it is noted that the advantages, effects, etc. mentioned in the present application are merely examples and are not limiting, and they should not be considered essential to the various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is for the purpose of illustration and description and is not intended to be limiting, since the foregoing disclosure is not intended to be exhaustive or to limit the disclosure to the precise details disclosed.
The block diagrams of devices, apparatuses, devices, systems referred to in this application are only used as illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by those skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably therewith. As used herein, the words "or" and "refer to, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".
It should also be noted that in the devices, apparatuses, and methods of the present application, each component or step can be decomposed and/or re-combined. These decompositions and/or recombinations are to be considered as equivalents of the present application.
The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, the description is not intended to limit embodiments of the application to the form disclosed herein. While a number of example aspects and embodiments have been discussed above, those of skill in the art will recognize certain variations, modifications, alterations, additions and sub-combinations thereof.
Claims (6)
1. The utility model provides a brew high-efficient fermenting installation of wine, vinegar which characterized in that includes:
the sensor unit is used for acquiring temperature data and odor data of a plurality of preset time points in a preset time period through a temperature sensor and an odor sensor which are arranged in a fermentation tank, and acquiring an infrared monitoring video of a brewed product in the preset time period through an infrared camera which is arranged in the fermentation tank;
the three-dimensional convolution coding unit is used for enabling the infrared monitoring video to pass through a first convolution neural network using a three-dimensional convolution kernel so as to obtain a first feature vector;
the first time sequence coding unit is used for enabling the temperature data and the smell data of a plurality of preset time points in the preset time period to pass through a time sequence coder comprising a one-dimensional convolution layer and a full connection layer respectively so as to obtain a second eigenvector and a third eigenvector;
a first joint encoding unit, configured to calculate a product of a transposed vector of the first eigenvector and the second eigenvector to obtain a first eigenvector matrix;
a second joint encoding unit, configured to calculate a product of the transposed vector of the first eigenvector and the third eigenvector to obtain a second eigenvector matrix;
a reparameterization unit configured to reparameterize the first feature matrix and the second feature matrix to obtain a reparameterized first feature matrix and a reparameterized second feature matrix, wherein the reparameterization of the first feature matrix and the second feature matrix is performed based on a logarithmic function value of a difference between a natural exponent function value raised to a power of a feature value of each position in the first feature matrix or the second feature matrix and a natural exponent function value raised to a power of a mean value of feature values of all positions in the first feature matrix or the second feature matrix;
a feature fusion unit for calculating a position-weighted sum of the reparameterized first feature matrix and the reparameterized second feature matrix to obtain a classification feature matrix; and
and the regulation and control result generation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating that the temperature should be increased or decreased.
2. The apparatus for efficient fermentation of brewed wine and vinegar according to claim 1, wherein said three-dimensional convolutional encoding unit is further configured to: processing the infrared surveillance video using a first convolutional neural network of the three-dimensional convolutional kernel with the following formula to generate the first feature vector;
wherein the formula is:
wherein Hj、WjAnd RjRespectively representing the length, width and height of the three-dimensional convolution kernel, m represents the number of (l-1) th layer characteristic diagrams,
is the convolution kernel connected to the mth feature map of the (l-1) layer, bljFor biasing, f (-) represents the activation function.
3. The apparatus for efficient wine and vinegar brewing according to claim 2, wherein the first time-series encoding unit is further configured to:
arranging the temperature data of a plurality of preset time points in the preset time period into a one-dimensional temperature input vector according to the time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the temperature input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
represents a matrix multiplication; performing one-dimensional convolution encoding on the temperature input vector by using a one-dimensional convolution layer of the time sequence encoder according to the following formula so as to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel;
arranging the odor data of a plurality of preset time points in the preset time period into a one-dimensional temperature input vector according to the time dimension; fully concatenating the odor input vector using a fully concatenated layer of the temporal encoder to provideAnd extracting high-dimensional implicit characteristics of characteristic values of all positions in the input vector, wherein the formula is as follows:
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
represents a matrix multiplication; performing one-dimensional convolution encoding on the smell input vector by using a one-dimensional convolution layer of the time sequence encoder according to the following formula so as to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
4. The efficient wine and vinegar fermentation device of claim 3, wherein the first joint encoding unit is further configured to: calculating a product of the transposed vector of the first eigenvector and the second eigenvector in a formula to obtain the first eigenvector matrix;
wherein the formula is:
wherein
Representing multiplication of vectors, V1Representing said first feature vector, V2Representing the second feature vector in the second set of feature vectors,
a transposed vector representing the first feature vector.
5. The efficient fermentation device for brewing wine and vinegar of claim 4, wherein the parameterization unit is further configured to: re-parameterizing the first feature matrix and the second feature matrix in the following formula to obtain the re-parameterized first feature matrix and the re-parameterized second feature matrix;
wherein the formula is:
mi,jan eigenvalue of each position of the characteristic matrix is represented, and
represents the mean of the eigenvalues of all positions of the eigen matrix.
6. The high efficiency fermentation apparatus for brewing wine and vinegar of claim 5, wherein the control result generating unit is further configured to: the classifier processes the classification feature matrix to generate the classification result according to the following formula: softmax { (W)n，Bn)：...：(W1，B1) L project (F), where project (F) denotes the projection of the classification feature matrix as a vector, W1To WnAs a weight matrix for each fully connected layer, B1To BnA bias matrix representing the layers of the fully connected layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210506870.0A CN114686324A (en) | 2022-05-09 | 2022-05-09 | High-efficient fermenting installation of brew wine, vinegar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210506870.0A CN114686324A (en) | 2022-05-09 | 2022-05-09 | High-efficient fermenting installation of brew wine, vinegar |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114686324A true CN114686324A (en) | 2022-07-01 |
Family
ID=82145401
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210506870.0A Withdrawn CN114686324A (en) | 2022-05-09 | 2022-05-09 | High-efficient fermenting installation of brew wine, vinegar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114686324A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117535452A (en) * | 2024-01-09 | 2024-02-09 | 延边大学 | On-line monitoring method and system for fungus chaff fermented feed production |
-
2022
- 2022-05-09 CN CN202210506870.0A patent/CN114686324A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117535452A (en) * | 2024-01-09 | 2024-02-09 | 延边大学 | On-line monitoring method and system for fungus chaff fermented feed production |
| CN117535452B (en) * | 2024-01-09 | 2024-03-26 | 延边大学 | On-line monitoring method and system for fungus chaff fermented feed production |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110111399B (en) | Image text generation method based on visual attention | |
| CN110929092B (en) | Multi-event video description method based on dynamic attention mechanism | |
| CN111461307A (en) | General disturbance generation method based on generation countermeasure network | |
| CN110188866B (en) | Feature extraction method based on attention mechanism | |
| CN110084266B (en) | Dynamic emotion recognition method based on audio-visual feature deep fusion | |
| CN115761813A (en) | Intelligent control system and method based on big data analysis | |
| CN115951584A (en) | Temperature control system and method for roller fixation machine | |
| CN115013298B (en) | Real-time performance online monitoring system and monitoring method of sewage pump | |
| CN109218134B (en) | Test case generation system based on neural style migration | |
| CN115482467B (en) | Automatic irrigation system for intelligent gardens | |
| CN111461157A (en) | Self-learning-based cross-modal Hash retrieval method | |
| CN110046252A (en) | A kind of medical textual hierarchy method based on attention mechanism neural network and knowledge mapping | |
| CN114686324A (en) | High-efficient fermenting installation of brew wine, vinegar | |
| CN115235612B (en) | Intelligent fault diagnosis system and method for servo motor | |
| CN114821450B (en) | Laminating machine for processing solar cell panel and control method thereof | |
| CN115984745A (en) | Moisture control method for black garlic fermentation | |
| CN115861887A (en) | Fungus detection method for black garlic | |
| CN116426331A (en) | Intelligent preparation method and system of perfume oil | |
| CN116030018A (en) | Incoming material qualification inspection system and method for door processing | |
| CN117158923A (en) | Remote home-care monitoring method based on meta universe | |
| CN115205788A (en) | Food material quality monitoring system | |
| CN113420815A (en) | Semi-supervised RSDAE nonlinear PLS intermittent process monitoring method | |
| CN117593398A (en) | Remote sensing image generation method based on diffusion model | |
| CN116533529B (en) | Intelligent control method and system for ultrasonic welding PC (polycarbonate) sheet | |
| CN116851856B (en) | Pure waterline cutting processing technology and system thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220701 |
|
| WW01 | Invention patent application withdrawn after publication |