Acorus calamus storage environment monitoring method and system based on artificial intelligence

Abstract

The present invention discloses a method and system for monitoring the storage environment of Acorus calamus based on artificial intelligence. The method includes storage data collection, storage data optimization, establishment of an Acorus calamus storage environment assessment model and Acorus calamus storage environment monitoring. The present invention belongs to the field of storage monitoring, and specifically refers to a method and system for monitoring the storage environment of Acorus calamus based on artificial intelligence. This solution extracts the deep features of the stored Acorus calamus storage samples by introducing a shared encoder, and more completely replicates the correlation pattern between Acorus calamus storage indicators through covariance calibration and subdomain mean calibration; and adds dynamic weight scheduling to the generator loss, dynamically amplifies the calibration strength of the label according to the state severity score, and improves the sensitivity and credibility to extreme environments; through feature space aggregation quantification, focuses on easily confused states, and introduces a deviation gain term to adaptively adjust the gradient contribution of each category, so that boundary storage samples are easier to capture; thereby improving the reliability of Acorus calamus storage environment monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of storage monitoring, and in particular to an artificial intelligence-based Acorus tatarinowii storage environment monitoring method and system. Background Art

[0002] The Acorus calamus storage environment monitoring method refers to a technical process that uses sensors to collect and pre-process environmental data in real time during storage to evaluate different storage conditions. However, common Acorus calamus storage environment monitoring methods suffer from insufficient Acorus calamus storage samples, narrow coverage, and low sensitivity to extreme storage environments, leading to frequent false positives or missed reports and poor storage monitoring accuracy. Common Acorus calamus storage environment monitoring methods also lack strategies for distinguishing between different types of anomalies, ignore differences within feature spaces, and have poor adaptability to minor environmental changes, resulting in low reliability of Acorus calamus storage environment monitoring. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an artificial intelligence-based Acorus calamus storage environment monitoring method and system. In view of the problems that the general Acorus calamus storage environment monitoring method has insufficient Acorus calamus storage samples, narrow coverage, low sensitivity to extreme storage environments, frequent false alarms or missed alarms, and poor storage monitoring accuracy, this scheme introduces a shared encoder to extract the deep features of the stored Acorus calamus storage samples, and through covariance calibration and subdomain mean calibration, ensures that the synthetic Acorus calamus storage samples are not just numerically flattened, but more completely replicate the correlation pattern between Acorus calamus storage indicators; and adds dynamic weights to the generator loss Scheduling, dynamically amplifying the calibration strength of the label according to the state severity score, improving the sensitivity and credibility to extreme environments; and thus improving the accuracy of subsequent storage monitoring; in view of the fact that general Acorus calamus storage environment monitoring methods lack differentiation strategies for different types of anomalies, ignore the differences in the feature space, and have poor adaptability to minor environmental changes, which leads to the problem of low reliability of Acorus calamus storage environment monitoring, this solution quantifies the feature space aggregation, focuses on easily confused states, and improves the discrimination accuracy; and introduces the deviation gain term to adaptively adjust the gradient contribution of each category, so that boundary storage samples are easier to capture; thereby improving the reliability of Acorus calamus storage environment monitoring.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an artificial intelligence-based Acorus tatarinowii storage environment monitoring method, which includes the following steps:

[0005] Step S1: storing data collection;

[0006] Step S2: storage data optimization;

[0007] Step S3: establishing a storage environment assessment model for Acorus tatarinowii;

[0008] Step S4: monitoring the storage environment of Acorus tatarinowii.

[0009] Furthermore, in step S1, the storage data collection is to collect sensor data of Acorus tatarinowii in a storage state; mark the storage state as a data label; and perform feature engineering processing on the collected data to obtain an initial storage data set.

[0010] Furthermore, in step S2, the storage data optimization is to construct a Acorus tatarinowii storage data optimization model based on a conditional generative network, which specifically includes the following contents:

[0011] Step S21: Design of the Acorus calamus storage data optimization model architecture; Under different storage state label conditions, the sensor data generated defines the generator G(z|c) and the discriminator D(x|c); the input includes noise and state label c; base loss Expressed as: ; Among them, G and D are the generator and discriminator respectively; It's expectation. Indicates that x is from the real data distribution The sensor data sampled in represents z from the noise distribution The noise obtained by sampling; The data generated by the discriminator to the generator Under condition c, it is the probability estimate of the real data; c is the state label, representing the different storage states of Acorus tatarinowii;

[0012] Step S22: Feature extraction encoding design; Use shared encoder f(·) to map the original data into deep feature space for distribution calibration; Synthesize data domain ;Measured data domain ; The feature matrix is ​​expressed as: ;in, It is the data in the synthetic data domain, which is the sensor data obtained by the generator; It is the data in the measured data domain, which is the collected sensor data; and are the characteristic matrices of the synthetic data domain data and the measured data domain data respectively; and The first and synthetic data domain data; and The first and Measured data domain data;

[0013] Step S23: Distribution calibration; calculate the covariance matrix for the generated and true feature matrices respectively and , expressed as: ; ; Where T is the transpose operation; I is a column vector whose elements are all 1; Calculate the global calibration loss , expressed as: ; One-hot weight normalization is performed on the stored samples of Acorus tatarinowii under each subdomain condition, expressed as: ;Where, d is the dimension of the sensor data feature vector; is the square of the Frobenius norm; is the normalized weight of the i-th Acorus tatarinowii storage sample under condition c; is the one-hot encoding value of the i-th Acorus tatarinowii storage sample under condition c; measures the mean deviation , expressed as: ;Where, C is the total number of subdomains, corresponding to the total number of storage states; and are the normalized weights of synthetic data and real data respectively; j is the generated data index; It is the feature mapping function that maps the original data to the Hilbert space; is the square of the norm in Hilbert space;

[0014] Step S24: Generator loss; Generator final loss Expressed as: ; is the scheduling weight;

[0015] Step S25: Dynamic weight scheduling; introduce the state severity scoring index, and the weight function is expressed as: ; ; ; Where t is the current training step number and tmax is the maximum training step number; is the extreme amplification factor; is the normalized state severity score; is the status severity score; 、 and are the average temperature value in subdomain c, the mean temperature of all subdomains, and the standard deviation of the temperature of all subdomains; 、 and are the average humidity value in subdomain c, the mean humidity value of all subdomains, and the standard deviation of humidity of all subdomains.

[0016] Furthermore, in step S3, the establishment of the Acorus calamus storage environment assessment model specifically includes the following contents:

[0017] Step S31: Model structure design; the model adopts a five-layer feedforward neural network; uses the Mish activation function, expressed as: ; The forward calculation of the network layer l is expressed as: ; ; ; The output layer Softmax is expressed as: ; Where z is the activation function input and is the linear combination output of the neurons in the Acorus tatarinowii storage environment assessment model; is the tanh function; 、 and are the linear combination output, weight matrix and bias vector of the lth layer neurons respectively; is the neuron output of the l-1 layer; It is a batch normalization technique; It is a regularization technique; and They are the results after batch normalization operation and the output after Dropout operation; is the probability that the storage environment of Acorus calamus belongs to category c; is the linear combination output of the cth neuron in the output layer;

[0018] Step S32: store the loss function design; assign an aggregation index to each class based on the statistical differences within the class, and introduce a deviation gain term and a weight factor Expressed as: ; ; ;in, It is the aggregation index of the c-type Acorus calamus stored samples in the feature space; is the average value of the aggregation index of all categories; and is the exponential parameter; K is the number of stored samples of type c Acorus calamus; and are the i-th Acorus calamus storage sample and the j-th Acorus calamus storage sample respectively; is the correlation of Acorus tatarinowii stored samples, using the Pearson correlation coefficient; for each Acorus tatarinowii stored sample (X, Y) the loss is defined : ; Wherein, X is the stored sample of Acorus tatarinowii; Y is the label of the stored sample of Acorus tatarinowii; is the indicator value of the true label on the cth category;

[0019] Step S33: model training; dividing the storage data set into training set / validation set / test set; using Adam optimizer to update parameters; and finally establishing the Acorus tatarinowii storage environment assessment model.

[0020] Furthermore, in step S4, the Acorus calamus storage environment monitoring is based on the established Acorus calamus storage environment assessment model, and the sensor data of the Acorus calamus in the storage state is collected in real time and input into the Acorus calamus storage environment assessment model, and the storage environment monitoring is performed based on the storage state output by the model.

[0021] The artificial intelligence-based Acorus calamus storage environment monitoring system provided by the present invention includes a storage data acquisition module, a storage data optimization module, an Acorus calamus storage environment assessment model establishment module and an Acorus calamus storage environment monitoring module;

[0022] The storage data acquisition module collects sensor data of the Acorus tatarinowii in a storage state and constructs an initial storage data set;

[0023] The storage data optimization module generates synthetic data by performing adversarial training based on the initial storage data set and the conditional generative network through global covariance, subdomain mean calibration and dynamic weight scheduling to obtain a storage data set;

[0024] The Acorus calamus storage environment evaluation model establishment module is based on the storage data set and the feedforward neural network, and designs a storage loss function through the aggregation degree difference of the Acorus calamus storage samples in each storage state in the feature space, thereby completing the establishment of the Acorus calamus storage environment evaluation model;

[0025] The Acorus calamus storage environment monitoring module performs storage environment monitoring on sensor data collected in real time when the Acorus calamus is in storage, based on the Acorus calamus storage environment assessment model.

[0026] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0027] (1) In view of the problems of insufficient Acorus calamus storage samples, narrow coverage, low sensitivity to extreme storage environments, frequent false alarms or missed alarms, and poor storage monitoring accuracy in general Acorus calamus storage environment monitoring methods, this scheme introduces a shared encoder to extract the deep features of the stored Acorus calamus storage samples. Through covariance calibration and subdomain mean calibration, it ensures that the synthetic Acorus calamus storage samples are not just numerically flattened, but more completely replicate the correlation pattern between Acorus calamus storage indicators; and adds dynamic weight scheduling to the generator loss, dynamically amplifying the calibration strength of the label according to the state severity score, thereby improving the sensitivity and credibility to extreme environments, and thus improving the accuracy of subsequent storage monitoring.

[0028] (2) In view of the fact that general Acorus calamus storage environment monitoring methods lack strategies for distinguishing different types of anomalies, ignore differences in feature space, and have poor adaptability to minor environmental changes, which leads to low reliability of Acorus calamus storage environment monitoring, this scheme aggregates and measures feature space, focuses on easily confused states, and improves discrimination accuracy; and introduces a bias gain term to adaptively adjust the gradient contribution of each category, making it easier to capture boundary storage samples; thereby improving the reliability of Acorus calamus storage environment monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of the process of the artificial intelligence-based Acorus calamus storage environment monitoring method provided by the present invention;

[0030] Figure 2 A schematic diagram of the artificial intelligence-based Acorus calamus storage environment monitoring system provided by the present invention;

[0031] Figure 3 Schematic diagram of the process of step S2.

[0032] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0034] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0035] Example 1, see Figure 1 The present invention provides an artificial intelligence-based Acorus tatarinowii storage environment monitoring method, which includes the following steps:

[0036] Step S1: Storage data collection: collecting sensor data of Acorus tatarinowii in storage state and constructing an initial storage data set;

[0037] Step S2: Storage data optimization: Based on the initial storage data set and the conditional generative network, adversarial training is performed through global covariance, subdomain mean calibration, and dynamic weight scheduling to generate synthetic data to obtain the storage data set;

[0038] Step S3: Establishing a storage environment evaluation model for Acorus tatarinowii; Based on the storage data set and the feedforward neural network, a storage loss function is designed by the difference in the aggregation degree of Acorus tatarinowii storage samples in the feature space in each storage state, thereby completing the establishment of the storage environment evaluation model for Acorus tatarinowii;

[0039] Step S4: Acorus calamus storage environment monitoring: Based on the Acorus calamus storage environment assessment model, the storage environment monitoring is performed on the sensor data collected in real time when the Acorus calamus is in storage.

[0040] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the storage data collection is to collect sensor data of Acorus calamus in the storage state; the sensor data includes temperature, humidity, O2 / CO2 concentration, and microbial indicators; the storage state is marked as a data label; the storage state includes normal, mild abnormality, moderate abnormality, and severe abnormality; the collected data is subjected to feature engineering processing to obtain an initial storage data set.

[0041] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S2, the storage data optimization is to build a Acorus tatarinowii storage data optimization model based on a conditional generative network, which specifically includes the following contents:

[0042] Step S21: Design of the model architecture for optimizing the storage data of Acorus tatarinowii; Generate diverse sensor data under different storage state label conditions to make up for the lack of measured Acorus tatarinowii storage samples, define the generator G(z|c) and the discriminator D(x|c); Input includes noise and state label c; base loss Expressed as: ; Among them, G and D are the generator and discriminator respectively; It's expectation. Indicates that x is from the real data distribution The sensor data sampled in represents z from the noise distribution The noise obtained by sampling; The data generated by the discriminator to the generator Under condition c, it is the probability estimate of the real data; c is the state label, representing the different storage states of Acorus tatarinowii;

[0043] Step S22: Feature extraction encoding design; Use shared encoder f(·) to map the original data into deep feature space for distribution calibration; Synthesize data domain ;Measured data domain ; The feature matrix is ​​expressed as: ;in, It is the data in the synthetic data domain, which is the sensor data obtained by the generator; It is the data in the measured data domain, which is the collected sensor data; and are the characteristic matrices of the synthetic data domain data and the measured data domain data respectively; and The first and synthetic data domain data; and The first and Measured data domain data;

[0044] Step S23: Distribution Calibration. In a storage environment, metrics are often not isolated but rather interact with each other in complex ways. By comparing the overall covariance of synthetic and real data features, we ensure that the generator not only brings each metric to the same level but also preserves the correlation patterns between them. To this end, we perform two calibrations: calibrating the overall covariance and calibrating the average performance of each environment state.

[0045] Calculate the covariance matrix for the generated and real feature matrices separately and , expressed as: ; ; Where T is the transpose operation; I is a column vector whose elements are all 1; Calculate the global calibration loss , expressed as: ; One-hot weight normalization is performed on the stored samples of Acorus tatarinowii under each subdomain condition, expressed as: ;Where, d is the dimension of the sensor data feature vector; is the square of the Frobenius norm; is the normalized weight of the i-th Acorus tatarinowii storage sample under condition c; is the one-hot encoding value of the i-th Acorus tatarinowii storage sample under condition c; measures the mean deviation , expressed as: ;Where, C is the total number of subdomains, corresponding to the total number of storage states; and are the normalized weights of synthetic data and real data respectively; j is the generated data index; It is the feature mapping function that maps the original data to the Hilbert space; It is the square of the norm in Hilbert space. Smoothing out large deviations: Global calibration makes the synthesis and reality appear seamless globally, eliminating large-scale style differences. Eliminating small deviations: State-level calibration allows even microscopic differences in each storage condition to be accurately reproduced, improving the model's sensitivity to rare and extreme environments.

[0046] Step S24: Generator loss; Generator final loss Expressed as: ; is the scheduling weight; calibrate for different storage environment subdomains to enhance the quality of Acorus tatarinowii storage samples under extreme or scarce conditions; train the Acorus tatarinowii storage data optimization model: randomly initialize G and D parameters; alternate optimization: fix G, update D, and minimize ; Fix D, update G, minimize ;Get the stored data set;

[0047] Step S25: Dynamic weight scheduling; introduce the state severity scoring index to automatically amplify the extreme condition subdomain calibration strength and improve the credibility of the stored samples of synthetic Acorus tatarinowii under extreme conditions. The weight function is expressed as: ; ; ; Where t is the current training step number and tmax is the maximum training step number; is the extreme amplification factor; is the normalized state severity score; is the status severity score; 、 and are the average temperature value in subdomain c, the mean temperature of all subdomains, and the standard deviation of the temperature of all subdomains; 、 and are the average humidity value in subdomain c, the mean humidity value of all subdomains, and the standard deviation of humidity of all subdomains.

[0048] By performing the above operations, the general Acorus calamus storage environment monitoring method has the problems of insufficient Acorus calamus storage samples, narrow coverage, low sensitivity to extreme storage environments, frequent false alarms or missed alarms, and poor storage monitoring accuracy. This solution introduces a shared encoder to extract the deep features of the stored Acorus calamus storage samples. Through covariance calibration and subdomain mean calibration, it ensures that the synthetic Acorus calamus storage samples are not just numerically flattened, but more completely replicate the correlation pattern between Acorus calamus storage indicators; and adds dynamic weight scheduling to the generator loss, dynamically amplifying the calibration strength of the label according to the state severity score, thereby improving the sensitivity and credibility to extreme environments, and thus the subsequent storage monitoring accuracy.

[0049] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, establishing a storage environment assessment model for Acorus calamus specifically includes the following:

[0050] Step S31: Model structure design; the model uses a five-layer feedforward neural network to capture high-order coupling between temperature and humidity, CO2, and microorganisms;

[0051] Layer dimension BatchNorm activation function Dropout

[0052] Input layer 8——

[0053] Hidden layer 164 ✓ Mish 0.2

[0054] Hidden layer 2128✓Mish0.2

[0055] Hidden layer 3256 ✓ Mish 0.3

[0056] Hidden layer 4128✓Mish0.2

[0057] Hidden layer 564✓Mish0.2

[0058] Output layer C (number of categories) —Softmax—

[0059] The Mish activation function has a good effect on the smooth response to amplitude changes in environmental monitoring, which can be expressed as: ; The forward calculation of the network layer l is expressed as: ; ; ; The output layer Softmax is expressed as: ; Where z is the activation function input and is the linear combination output of the neurons in the Acorus tatarinowii storage environment assessment model; is the tanh function; 、 and are the linear combination output, weight matrix and bias vector of the lth layer neurons respectively; is the neuron output of the l-1 layer; It is a batch normalization technique; It is a regularization technique; and They are the results after batch normalization operation and the output after Dropout operation; is the probability that the storage environment of Acorus calamus belongs to category c; is the linear combination output of the cth neuron in the output layer;

[0060] Step S32: storage loss function design; different aggregation degrees in feature space for different storage states, including: the humidity fluctuation of moldy Acorus calamus storage samples is more concentrated, and the temperature fluctuation of frozen Acorus calamus storage samples is more intense; based on the statistical differences within the class, an aggregation degree index is assigned to each class to alleviate the problem of excessive dominance caused by excessive variance of a certain class; and a deviation gain term is introduced to determine the sensitivity to extreme abnormal conditions and the weight factor Expressed as: ; ; ;in, It is the aggregation index of the c-type Acorus calamus stored samples in the feature space; is the average value of the aggregation index of all categories; and is the exponential parameter; K is the number of stored samples of type c Acorus calamus; and are the i-th Acorus calamus storage sample and the j-th Acorus calamus storage sample respectively; is the correlation of stored samples of Acorus tatarinowii, using the Pearson correlation coefficient; when , the aggregation degree is high and it is easy to be confused with other classes, so the gradient contribution is amplified; when , large internal differences, easy to classify, suppress over-dominance; define the loss for each Acorus tatarinowii storage sample (X, Y) : ; Wherein, X is the stored sample of Acorus tatarinowii; Y is the label of the stored sample of Acorus tatarinowii; is the indicator value of the true label on the cth category;

[0061] Step S33: Model training; divide the storage data set into training set / validation set / test set to ensure the consistent proportions of the four storage states; use the Adam optimizer to update the parameters; verify the model performance based on the overall accuracy; use 5-fold cross-validation on the training set; and finally establish the Acorus calamus storage environment assessment model.

[0062] By performing the above operations, we can address the problem that general Acorus calamus storage environment monitoring methods lack strategies for distinguishing different types of anomalies, ignore differences in feature space, and have poor adaptability to minor environmental changes, which leads to low reliability of Acorus calamus storage environment monitoring. This solution aggregates and measures feature space, focuses on easily confused states, and improves discrimination accuracy. It also introduces a bias gain term to adaptively adjust the gradient contribution of each category, making boundary storage samples easier to capture, thereby improving the reliability of Acorus calamus storage environment monitoring.

[0063] Example 5, see Figure 1This embodiment is based on the above embodiment. In step S4, the Acorus calamus storage environment monitoring is based on the established Acorus calamus storage environment assessment model. The sensor data of the Acorus calamus in the storage state is collected in real time and input into the Acorus calamus storage environment assessment model. The storage environment is monitored based on the storage state output by the model. If the storage state is moderately abnormal or seriously abnormal, an early warning is issued to relevant personnel.

[0064] Example 6, see Figure 2 This embodiment is based on the above embodiment. The artificial intelligence-based Acorus calamus storage environment monitoring system provided by the present invention includes a storage data acquisition module, a storage data optimization module, an Acorus calamus storage environment assessment model establishment module and an Acorus calamus storage environment monitoring module;

[0065] The storage data acquisition module collects sensor data of the Acorus tatarinowii in a storage state and constructs an initial storage data set;

[0066] The storage data optimization module generates synthetic data by performing adversarial training based on the initial storage data set and the conditional generative network through global covariance, subdomain mean calibration and dynamic weight scheduling to obtain a storage data set;

[0067] The Acorus calamus storage environment evaluation model establishment module is based on the storage data set and the feedforward neural network, and designs a storage loss function through the aggregation degree difference of the Acorus calamus storage samples in each storage state in the feature space, thereby completing the establishment of the Acorus calamus storage environment evaluation model;

[0068] The Acorus calamus storage environment monitoring module performs storage environment monitoring on sensor data collected in real time when the Acorus calamus is in storage, based on the Acorus calamus storage environment assessment model.

[0069] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0070] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0071] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based method for monitoring the storage environment of Acorus tatarinowii, characterized by: The method comprises the following steps: Step S1: Storage data collection: collecting sensor data of Acorus tatarinowii in storage state and constructing an initial storage data set; Step S2: Storage data optimization: Based on the initial storage data set and the conditional generative network, adversarial training is performed through global covariance, subdomain mean calibration, and dynamic weight scheduling to generate synthetic data to obtain the storage data set; Step S3: Establishing a storage environment evaluation model for Acorus tatarinowii; Based on the storage data set and the feedforward neural network, a storage loss function is designed by the difference in the aggregation degree of Acorus tatarinowii storage samples in the feature space in each storage state, thereby completing the establishment of the storage environment evaluation model for Acorus tatarinowii; Step S4: monitoring the storage environment of Acorus calamus; based on the Acorus calamus storage environment assessment model, performing storage environment monitoring on the sensor data collected in real time when Acorus calamus is in storage; Step S2 includes step S22: feature extraction encoding design; using a shared encoder f(·) to map the original data to a deep feature space for distribution calibration; synthesizing the data domain ;Measured data domain ; The feature matrix is ​​expressed as: ;in, It is the data in the synthetic data domain, which is the sensor data obtained by the generator; It is the data in the measured data domain, which is the collected sensor data; and are the characteristic matrices of the synthetic data domain data and the measured data domain data respectively; and The first and synthetic data domain data; and The first and Measured data domain data.

2. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 1, characterized in that: In step S2, the storage data optimization is to build a Acorus tatarinowii storage data optimization model based on a conditional generative network, which specifically includes the following contents: Step S21: Design of the Acorus calamus storage data optimization model architecture; Under different storage state label conditions, the sensor data generated defines the generator G(z|c) and the discriminator D(x|c); the input includes noise and state label c; base loss Expressed as: ; Among them, G and D are the generator and discriminator respectively; It's expectation. Indicates that x is from the real data distribution The sensor data sampled in represents z from the noise distribution The noise obtained by sampling; The data generated by the discriminator to the generator Under condition c, it is the probability estimate of the real data; c is the state label, representing the different storage states of Acorus tatarinowii; Step S22: feature extraction coding design; Step S23: distribution calibration; Step S24: Generator loss; Generator final loss Expressed as: ; is the scheduling weight; is the global calibration loss; is the measure mean deviation; Step S25: Dynamic weight scheduling.

3. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 2, characterized in that: In step S2, the distribution calibration is to calculate the covariance matrix of the generated and true feature matrices respectively. and , expressed as: ; ; Where T is the transpose operation; I is a column vector whose elements are all 1; Calculate the global calibration loss , expressed as: ; One-hot weight normalization is performed on the stored samples of Acorus tatarinowii under each subdomain condition, expressed as: ;Where, d is the dimension of the sensor data feature vector; is the square of the Frobenius norm; is the normalized weight of the i-th Acorus tatarinowii storage sample under condition c; is the one-hot encoding value of the i-th Acorus tatarinowii storage sample under condition c; measures the mean deviation , expressed as: ;Where, C is the total number of subdomains, corresponding to the total number of storage states; and are the normalized weights of synthetic data and real data respectively; j is the generated data index; It is the feature mapping function that maps the original data to the Hilbert space; is the square of the norm in the Hilbert space.

4. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 3, characterized in that: In step S2, the dynamic weight scheduling is to introduce the state severity scoring index, and the weight function is expressed as: ; ; ; Where t is the current training step number and tmax is the maximum training step number; is the extreme amplification factor; is the normalized state severity score; is the status severity score; 、 and are the average temperature value in subdomain c, the mean temperature of all subdomains, and the standard deviation of the temperature of all subdomains; 、 and are the average humidity value in subdomain c, the mean humidity value of all subdomains, and the standard deviation of humidity of all subdomains.

5. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 4, characterized in that: In step S3, the establishment of the Acorus calamus storage environment assessment model specifically includes the following contents: Step S31: Model structure design; the model adopts a five-layer feedforward neural network; uses the Mish activation function, expressed as: ; The forward calculation of the network layer l is expressed as: ; ; ; The output layer Softmax is expressed as: ; Where z is the activation function input and is the linear combination output of the neurons in the Acorus tatarinowii storage environment assessment model; is the tanh function; 、 and are the linear combination output, weight matrix and bias vector of the lth layer neurons respectively; is the neuron output of the l-1 layer; It is a batch normalization technique; It is a regularization technique; and They are the results after batch normalization operation and the output after Dropout operation; is the probability that the storage environment of Acorus calamus belongs to category c; is the linear combination output of the cth neuron in the output layer; Step S32: storing the loss function design; Step S33: model training; dividing the storage data set into training set / validation set / test set; using Adam optimizer to update parameters; and finally establishing the Acorus tatarinowii storage environment assessment model.

6. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 5, characterized in that: In step S3, the storage loss function is designed based on the statistical differences within the class, assigning an aggregation index to each class, and introducing a deviation gain term and a weight factor Expressed as: ; ; ;in, It is the aggregation index of the c-type Acorus calamus stored samples in the feature space; is the average value of the aggregation index of all categories; and is the exponential parameter; K is the number of stored samples of type c Acorus calamus; and are the i-th Acorus calamus storage sample and the j-th Acorus calamus storage sample respectively; is the correlation of Acorus tatarinowii stored samples, using the Pearson correlation coefficient; for each Acorus tatarinowii stored sample (X, Y) the loss is defined : ; Wherein, X is the stored sample of Acorus tatarinowii; Y is the label of the stored sample of Acorus tatarinowii; is the indicator value of the true label on the cth category.

7. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 6, characterized in that: In step S1, the storage data collection is to collect sensor data of Acorus calamus in a storage state; mark the storage state as a data label; and perform feature engineering processing on the collected data to obtain an initial storage data set.

8. The artificial intelligence-based Acorus calamus storage environment monitoring method according to claim 7, characterized in that: In step S4, the Acorus calamus storage environment monitoring is based on the established Acorus calamus storage environment assessment model, and the sensor data of Acorus calamus in the storage state is collected in real time and input into the Acorus calamus storage environment assessment model, and the storage environment monitoring is performed based on the storage state output by the model.

9. An artificial intelligence-based Acorus calamus storage environment monitoring system, configured to implement the artificial intelligence-based Acorus calamus storage environment monitoring method according to any one of claims 1 to 8, characterized in that: It includes storage data acquisition module, storage data optimization module, Acorus tatarinowii storage environment assessment model establishment module and Acorus tatarinowii storage environment monitoring module; The storage data acquisition module collects sensor data of the Acorus tatarinowii in a storage state and constructs an initial storage data set; The storage data optimization module generates synthetic data by performing adversarial training based on the initial storage data set and the conditional generative network through global covariance, subdomain mean calibration and dynamic weight scheduling to obtain a storage data set; The Acorus calamus storage environment evaluation model establishment module is based on the storage data set and the feedforward neural network, and designs a storage loss function through the aggregation degree difference of the Acorus calamus storage samples in each storage state in the feature space, thereby completing the establishment of the Acorus calamus storage environment evaluation model; The Acorus calamus storage environment monitoring module performs storage environment monitoring on sensor data collected in real time when the Acorus calamus is in storage, based on the Acorus calamus storage environment assessment model.

Images