Radiation level monitoring method and system, electronic equipment and storage medium

Abstract

The invention discloses a radiation level monitoring method and system, electronic equipment and a storage medium. The method comprises the following steps: acquiring first radiation data through a preset radiation monitoring sensor; performing data dimension reduction processing on the first radiation data through a feature dimension reduction model to obtain second radiation data; the feature dimension reduction model comprises an auto-encoder based on boundary smoothing; performing data amplification on the second radiation data through a preset generative adversarial network model to obtain third radiation data; the preset generative adversarial network model comprises a quantum state-based generative adversarial network model; performing model training on a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model; and inputting the to-be-analyzed radiation data into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result. According to the embodiment of the invention, the data privacy risk problem can be effectively relieved, and the accuracy of radiation level monitoring is improved. The method can be widely applied to the technical field of monitoring.

Description

Technical Field

[0001] This application relates to the field of monitoring technologies, and particularly to a radiation level monitoring method, system, electronic device, and storage medium. Background Art

[0002] Radiation monitoring is widely applied in various key industries, such as nuclear power plants, environmental protection, medical fields, etc. In related technologies, radiation monitoring data usually has the characteristics of high dimension and complexity. In traditional radiation monitoring methods, there are often data privacy risks, and the accuracy of radiation monitoring is relatively low.

[0003] In summary, the technical problems existing in related technologies need to be improved. Summary of the Invention

[0004] The main purpose of the embodiments of this application is to propose a radiation level monitoring method, system, electronic device, and storage medium, which can effectively alleviate the data privacy risk problem and effectively improve the accuracy of radiation level monitoring.

[0005] To achieve the above object, on the one hand, an embodiment of this application proposes a radiation level monitoring method, and the method includes the following steps:

[0006] Collect first radiation data through a preset radiation monitoring sensor;

[0007] Perform data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data; wherein, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing;

[0008] Perform data augmentation on the second radiation data through a preset generative adversarial network model to obtain third radiation data; wherein, the preset generative adversarial network model includes a generative adversarial network model based on quantum states;

[0009] Train a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model;

[0010] Input the radiation data to be analyzed into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result.

[0011] In some embodiments, before performing the step of performing data dimensionality reduction processing on the first radiation data through the feature dimensionality reduction model to obtain second radiation data, the method further includes:

[0012] Construct an initial autoencoder;

[0013] Input preset input data into the initial autoencoder to generate reconstructed data through forward propagation;

[0014] Smoothing the reconstructed data through an automatic boundary smoothing algorithm to obtain smoothed output data;

[0015] Calculating a weight coefficient according to the preset input data through a preset variance function;

[0016] Calculating a first loss function according to the smoothed output data, the preset input data, and the weight coefficient;

[0017] Updating the parameters of the initial autoencoder according to the first loss function through a gradient descent algorithm to obtain the feature dimensionality reduction model.

[0018] In some embodiments, the updating the parameters of the initial autoencoder according to the first loss function through a gradient descent algorithm to obtain the feature dimensionality reduction model includes:

[0019] Calculating a first expected learning rate through a convex hull convergence algorithm;

[0020] Calculating a parameter update amount according to the first expected learning rate and the first loss function;

[0021] Updating the parameters of the initial autoencoder according to the parameter update amount to obtain the feature dimensionality reduction model.

[0022] In some embodiments, the data augmentation of the second radiation data through a preset generative adversarial network model to obtain third radiation data includes:

[0023] Constructing an initial generative adversarial network model; wherein, the initial generative adversarial network model includes a generator and a discriminator;

[0024] Receiving random noise through the generator to convert the random noise into quantum state data through a quantum state encoding algorithm;

[0025] Generating preset radiation data according to the quantum state data;

[0026] Calculating a similarity of the preset radiation data through the discriminator according to a similarity metric algorithm to obtain an evaluation result;

[0027] Adjusting the parameters of the initial generative adversarial network model according to the evaluation result to obtain the preset generative adversarial network model.

[0028] In some embodiments, the training of a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model includes:

[0029] Construct a central model and local models; wherein, the local models include a plurality of extreme learning machine models, and the plurality of extreme learning machine models correspond to the third radiation data;

[0030] Train the corresponding extreme learning machine models respectively through the third radiation data to obtain local update parameters;

[0031] Transmit the local update parameters to the central model to perform aggregated update on the central model to obtain target update parameters;

[0032] Transmit the target update parameters to each of the extreme learning machine models respectively to perform parameter update on the extreme learning machine models to obtain the target radiation monitoring model.

[0033] In some embodiments, the training of the corresponding extreme learning machine models respectively through the third radiation data to obtain local update parameters includes:

[0034] Input the third radiation data into the extreme learning machine model for non-linear mapping to obtain intermediate output data;

[0035] Calculate target output data according to the intermediate output data and a weight matrix;

[0036] Calculate loss data according to the target output data through a second loss function to determine weight update data through the loss data; wherein, the second loss function includes a gradient penalty coefficient;

[0037] Perform parameter update on the extreme learning machine model according to the weight update data through a backpropagation algorithm to obtain the local update parameters.

[0038] In some embodiments, after performing the parameter update on the extreme learning machine model according to the weight update data through the backpropagation algorithm to obtain the local update parameters, the method further includes:

[0039] Calculate a second expected learning rate according to the weight update data, the intermediate output data, and a base learning rate;

[0040] Adjust the model training learning rate according to the second expected learning rate.

[0041] To achieve the above object, another aspect of the embodiments of the present application proposes a radiation level monitoring system, the system includes:

[0042] A first module, configured to collect first radiation data through a preset radiation monitoring sensor;

[0043] A second module, configured to perform data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data; wherein, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing.

[0044] A third module, configured to perform data augmentation on the second radiation data through a preset generative adversarial network model to obtain third radiation data; wherein, the preset generative adversarial network model includes a generative adversarial network model based on quantum states.

[0045] A fourth module, configured to perform model training on a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model.

[0046] A fifth module, configured to input radiation data to be analyzed into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result.

[0047] To achieve the above object, another aspect of the embodiments of the present application provides an electronic device, where the electronic device includes:

[0048] At least one processor;

[0049] At least one memory, configured to store at least one program;

[0050] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0051] To achieve the above object, another aspect of the embodiments of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above method is implemented.

[0052] The embodiments of the present application at least include the following beneficial effects: The present application provides a radiation level monitoring method, system, electronic device, and storage medium. This solution collects first radiation data through a preset radiation monitoring sensor, and performs data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data. Among them, in the embodiments of the present invention, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing. By using the autoencoder based on boundary smoothing to perform dimensionality reduction processing on the first radiation data, the redundancy of the data can be effectively reduced, the robustness of the model can be improved, and further the accuracy of radiation level monitoring can be improved. Further, in the embodiments of the present invention, a preset generative adversarial network model is used to perform data augmentation on the second radiation data to obtain third radiation data. Correspondingly, in the embodiments of the present invention, the preset generative adversarial network model includes a generative adversarial network model based on quantum states. By using the generative adversarial network model based on quantum states to generate the third radiation data, the problem of insufficient training samples can be effectively alleviated, enabling the model to learn more complex radiation data distributions, and further improving the generalization ability of the model and effectively improving the accuracy of radiation level monitoring. Then, in the embodiments of the present invention, according to the third radiation data, the preset monitoring model is trained through a federated learning architecture to obtain a target radiation monitoring model. Furthermore, the radiation data to be analyzed is input into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result, realizing precise monitoring of the radiation level. Correspondingly, in the embodiments of the present invention, through the method of federated learning, the problem of data privacy risk can be alleviated, and the data privacy protection level can be effectively improved. It is easy to understand that in the embodiments of the present invention, by using the autoencoder based on boundary smoothing for data dimensionality reduction and the generative adversarial network model based on quantum states for data augmentation, the radiation level monitoring accuracy of the trained target radiation monitoring model can be effectively improved. At the same time, through the training method of the federated learning architecture, the problem of data privacy risk can be effectively alleviated. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a schematic flowchart of the radiation level monitoring method provided by the embodiments of the present invention;

[0054] Figure 2 is a schematic flowchart of the training of the feature dimensionality reduction model provided by the embodiments of the present invention;

[0055] Figure 3 is a schematic diagram of the federated learning architecture provided by the embodiments of the present invention;

[0056] Figure 4 is a schematic structural diagram of the radiation level monitoring system provided by the embodiments of the present invention;

[0057] Figure 5 is a schematic hardware structure diagram of the electronic device provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] To make the objectives, technical solutions, and advantages of this application clearer and more understandable, the following further elaborates on this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining this application and are not used to limit this application. When the following description involves the accompanying drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of this application. They are merely examples of devices and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0059] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if", "when" as used herein may be interpreted as "when...", "while...", or "in response to determining".

[0060] The terms "at least one", "multiple", "each", "any one", etc. used in this application, at least one includes one, two, or more than two, multiple includes two or more than two, each refers to each of the corresponding multiple, and any one refers to any one of the multiple.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0062] Before elaborating in detail on the embodiments of this application, first, some nouns and terms involved in the embodiments of this application are described. The nouns and terms involved in the embodiments of this application are subject to the following explanations.

[0063] Federated learning architecture: It is a distributed machine learning architecture, and its core lies in allowing multiple participants to jointly train a shared machine learning model while maintaining their respective data privacy and localization.

[0064] Generative adversarial network: It is a deep learning model composed of two neural networks, a generator and a discriminator. Among them, the generator is responsible for generating realistic data samples, while the discriminator aims to distinguish between the fake samples generated by the generator and the real data samples.

[0065] Radiation monitoring is widely applied in various key industries, such as nuclear power plants, environmental protection, medical fields, etc. In related technologies, radiation monitoring data usually has the characteristics of high-dimensional and complex. In traditional radiation monitoring methods, there are often data privacy risks, and the accuracy of radiation monitoring is relatively low. Exemplarily, traditional radiation monitoring methods mostly rely on a centralized data processing architecture, that is, transmitting monitoring data to a central server for analysis and processing. This method has relatively large data privacy risks. Especially when dealing with sensitive radiation monitoring data, it may be threatened by problems such as data leakage and hacker attacks. In addition, when the feature dimension is relatively high, traditional radiation data processing methods often adopt simple dimensionality reduction methods or do not perform dimensionality reduction processing, which may lead to information loss or inability to effectively extract important features in the data. At the same time, in actual radiation monitoring, the training data is often insufficient, and the sample quantities of different radiation levels are unbalanced, resulting in the model being unable to generalize well. In addition, because the collected radiation data is often high-dimensional and non-linear complex data, it is easy to lead to relatively low classification accuracy.

[0066] In view of this, in the embodiments of the present application, a radiation level monitoring method, system, electronic device and storage medium are provided. This solution collects first radiation data through a preset radiation monitoring sensor, and performs data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data. Among them, the feature dimensionality reduction model in the embodiments of the present invention includes an autoencoder based on boundary smoothing. Then, the embodiments of the present invention perform data augmentation on the second radiation data through a preset generative adversarial network model to obtain third radiation data. Correspondingly, the preset generative adversarial network model in the embodiments of the present invention includes a generative adversarial network model based on quantum states. Then, the embodiments of the present invention train a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model, and then input the radiation data to be analyzed into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result, realizing accurate monitoring of the radiation level, effectively alleviating the data privacy risk problem, and effectively improving the accuracy of radiation level monitoring.

[0067] The radiation level monitoring method provided by the embodiments of the present application relates to the field of monitoring technologies. The radiation level monitoring method provided by the embodiments of the present application can be applied to a terminal, a server, or software running on a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, etc., but is not limited thereto; the server side can be configured as an independent physical server, or as a server cluster or a distributed system composed of multiple physical servers, or as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing the radiation level monitoring method, etc., but is not limited to the above forms.

[0068] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet-type devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0069] Figure 1 is an optional flowchart of the radiation level monitoring method provided by the embodiments of the present application, Figure 1 The method in may include but is not limited to steps S110 to S150.

[0070] Step S110: Collect first radiation data through a preset radiation monitoring sensor.

[0071] Step S120: Perform data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data. Among them, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing.

[0072] Step S130: Augment the second radiation data through a preset generative adversarial network model to obtain third radiation data. The preset generative adversarial network model includes a generative adversarial network model based on quantum states.

[0073] Step S140: Train the preset monitoring model according to the third radiation data through a federated learning architecture to obtain a target radiation monitoring model.

[0074] Step S150: Input the radiation data to be analyzed into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result.

[0075] During the working process of this specific embodiment, the embodiment of the present invention first collects first radiation data through a preset radiation monitoring sensor. Specifically, in the embodiment of the present invention, each client collects it through various types of radiation monitoring sensors. For example, the preset radiation monitoring sensors in the embodiment of the present invention include gamma-ray sensors, X-ray detectors, α and β particle detectors, etc. Among them, these sensors can monitor the radiation levels of different types in the environment in real time, obtain radiation signals through different physical principles, and then convert them into digital signals for processing to obtain the first radiation data. Among them, the embodiment of the present invention transmits the collected radiation data to a cloud server or a local storage device through a wireless network or a wired network, and the collected data is stored in a standardized format, such as CSV format.

[0076] Exemplarily, the data attributes of the first radiation data collected in the embodiment of the present invention include: timestamp (Ta), used to record the specific time of data collection; sensor location (Pa), used to mark the geographical location of the data collection point; gamma-ray radiation value (Ra), used for the measurement result of gamma rays, with the unit of μSv / h; X-ray radiation value (Xa), used for the measurement result of X-rays, with the unit of μSv / h; α-particle radiation value (α a ), used for the measurement result of α particles, with the unit of Bq; β-particle radiation value (β a ), used for the measurement result of β particles, with the unit of Bq; temperature (Ta), that is, the ambient temperature, with the unit of degrees Celsius; humidity (Ha), that is, the ambient humidity, with the unit of percentage; atmospheric pressure (Pa), that is, the ambient atmospheric pressure, with the unit of hPa; wind speed (Va), that is, the ambient wind speed, with the unit of m / s. In addition, in some embodiments of the present invention, the attributes of the data are usually more than 10, and the number of data attributes may reach dozens or even hundreds.

[0077] Furthermore, in the embodiments of the present invention, the feature dimensionality reduction model is used to perform data dimensionality reduction processing on the first radiation data to obtain the second radiation data. Specifically, in the embodiments of the present invention, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing. Correspondingly, since the data dimension of the first radiation data collected in the embodiments of the present invention is relatively high, it may lead to information loss or inability to effectively extract important features in the data. Therefore, in the embodiments of the present invention, the autoencoder based on boundary smoothing is used to perform dimensionality reduction processing on the high-dimensional data (the first radiation data) to obtain the second radiation data, effectively alleviating the problem of information loss that may be brought by traditional dimensionality reduction methods. At the same time, the boundary smoothing strategy is used to enhance the robustness of the model to outliers and boundary effects, thereby improving the processing ability of the monitoring data. Then, in the embodiments of the present invention, the preset generative adversarial network model is used to perform data augmentation on the second radiation data to obtain the third radiation data. Specifically, in the embodiments of the present invention, the preset generative adversarial network model includes a generative adversarial network model based on quantum states. Since insufficient training data and unbalanced sample numbers are likely to cause the model to be difficult to generalize well. Therefore, in the embodiments of the present invention, the generative adversarial network model based on quantum states is used to perform data augmentation on the second radiation data to obtain the third radiation data, thereby enriching the training samples and effectively improving the generalization ability of the model, enabling it to more accurately identify different types of radiation levels. Among them, the generative adversarial network model based on quantum states is a quantum machine learning model that combines quantum computing and generative adversarial networks. In the embodiments of the present invention, the generative adversarial network model based on quantum states is used to generate high-quality and diverse synthetic radiation data, that is, the third radiation data, thereby effectively alleviating the problem of insufficient training samples in practical applications and enhancing the generalization ability of the model.

[0078] Furthermore, in the embodiments of the present invention, the preset monitoring model is trained according to the third radiation data through a federated learning architecture to obtain the target radiation monitoring model. Specifically, after each client in the embodiments of the present invention collects the corresponding first radiation data through the corresponding preset radiation monitoring sensor, the corresponding third radiation data is obtained through data dimensionality reduction processing and data augmentation. In order to protect the data privacy of each client and alleviate the risk of data leakage, the embodiments of the present invention adopt the federated learning architecture to train the preset monitoring model, independently train the model on each client and only exchange model parameters, thereby effectively improving the data factor protection level. Finally, in the embodiments of the present invention, the radiation data to be analyzed is input into the target radiation monitoring model for radiation level analysis to obtain the radiation level monitoring result. Specifically, the radiation data to be analyzed in the embodiments of the present invention refers to the radiation data that needs to be monitored and analyzed for radiation levels. Correspondingly, in the embodiments of the present invention, the target radiation monitoring model obtained through training is used to perform radiation level analysis on the radiation data to be analyzed to obtain the radiation level monitoring result, such as "low radiation", "medium radiation", and "high radiation", etc.

[0079] It should be noted that before generating samples based on the quantum state-based generative adversarial network algorithm in the embodiments of the present invention, the collected data is first labeled. Among them, the labeling method in the embodiments of the present invention is manual labeling, and the labeled categories include: "low radiation", "medium radiation", and "high radiation", a total of 3 categories.

[0080] In some embodiments of the present invention, before performing data dimensionality reduction processing on the first radiation data through the feature dimensionality reduction model to obtain the second radiation data, the radiation level monitoring method provided by the embodiments of the present invention further includes but is not limited to the following steps:

[0081] Construct an initial autoencoder.

[0082] Input the preset input data into the initial autoencoder to generate reconstructed data through forward propagation.

[0083] Perform smoothing processing on the reconstructed data through the automatic boundary smoothing algorithm to obtain smoothed output data.

[0084] Calculate the weight coefficient according to the preset input data through the preset variance function.

[0085] Calculate the first loss function according to the smoothed output data, the preset input data, and the weight.

[0086] Update the parameters of the initial autoencoder according to the first loss function through the gradient descent algorithm to obtain the feature dimensionality reduction model.

[0087] In this specific embodiment, before performing data dimensionality reduction processing through the feature dimensionality reduction model, the autoencoder based on boundary smoothing is first trained, as Figure 2 shown. Specifically, the embodiments of the present invention first construct an initial autoencoder. For example, the embodiments of the present invention construct an initial autoencoder and initialize the corresponding autoencoder parameters, as shown in the following formula (1):

[0088]

[0089] Among them, in the formula represents the weight of the encoder, represents the bias of the encoder; represents the weight of the decoder, represents the bias of the decoder; randn() represents a random function; d r,in is the dimension of the input layer of the encoder, d r,mid is the dimension of the middle layer of the encoder, d r,out is the dimension of the output layer of the encoder; zeros() is a function for generating a vector of all 0s.

[0090] Next, in the embodiment of the present invention, the preset input data is input into the initial autoencoder to generate reconstructed data through forward propagation, and then the reconstructed data is smoothed by the automatic boundary smoothing algorithm to obtain smoothed output data. The preset input data refers to the data used for training the feature dimensionality reduction model. Correspondingly, in order to handle outliers and boundary effects in the dimensionality reduction process, the embodiment of the present invention adopts an automatic boundary smoothing strategy to enhance the robustness of the model by dynamically adjusting the processing of boundary values, and realizes the smoothing of the output (reconstructed data) of the encoder through a smoothing function, as shown in the following formula (2):

[0091]

[0092] Wherein, in the formula is the output of the encoder, τ r () is the smoothing function, and δ r is the smoothing intensity parameter; θ r is the automatically adjusted threshold parameter, which is used to control the degree of smoothing processing and improve the adaptability to the boundary of the input data.

[0093] It should be noted that in the embodiment of the present invention, the smoothing intensity parameter is determined by an adaptive method based on the local density of the data, and the calculation method is as shown in the following formula (3):

[0094]

[0095] Wherein, in the formula Calculating The kernel density estimate in its neighborhood can accurately process the data in high-density regions and sparse regions, and improve the robustness and adaptability of the overall model; represents the neighborhood; k r () is the kernel density estimation function; mean() is the function of taking the average value; ∈ rea is a preset small constant. For example, in the embodiment of the present invention, ∈ rea is set to 0.001.

[0096] Next, in the embodiment of the present invention, the weight coefficient is calculated according to the preset input data through the preset equation function. Correspondingly, the weight coefficient of the autoencoder in the embodiment of the present invention is used to adjust the contribution of each dimension feature to the error, and dynamically allocate these weights according to the importance of the features. The calculation method is as shown in the following formula (4):

[0097]

[0098] Wherein, in the formula is the feature variance of the input data of the i-th dimension autoencoder; is the feature of the input data of the i-th dimension autoencoder; ∈ reis a small constant to avoid division by zero; Var() is the variance function.

[0099] Furthermore, in the embodiment of the present invention, a first loss function is calculated based on the smoothed output data, the preset input data, and the weight coefficients, and the initial encoder is updated with parameters according to the first loss function through the gradient descent algorithm, so as to obtain a feature dimension reduction model. Among them, in the training process of the autoencoder, the data is processed in the forward propagation manner, and then the parameters of the autoencoder are updated by using the gradient descent method in the error backpropagation manner. Accordingly, in this process, the calculation method of the first loss function of the autoencoder in the embodiment of the present invention is shown in the following formula (5):

[0100]

[0101] wherein, in the formula, L r is the first loss function of the autoencoder; x r is the input data of the autoencoder, is the output reconstructed by the decoder, Sig() is the Sigmoid activation function; |||| is the L2 norm; is the output of the smoothed encoder; is the weight coefficient of the i-th dimension of the autoencoder; n r is the vector dimension corresponding to the samples input to the autoencoder in the current batch. Accordingly, the output reconstructed by the decoder in the embodiment of the present invention is calculated according to the following formula (6):

[0102]

[0103] Accordingly, the training of the autoencoder in the embodiment of the present invention is completed by continuously optimizing the first loss function in the iterative process, and the calculation formula of the parameter update amount for updating the parameters in each iteration is shown in the following formula (7):

[0104]

[0105] wherein, in the formula, is the update amount of the weight parameter of the encoder; is the update amount of the bias parameter of the encoder; is the update amount of the weight parameter of the decoder; is the update amount of the bias parameter of the decoder; is the learning rate of the autoencoder in the t-th iteration.

[0106] Furthermore, the update method of the autoencoder parameters in the embodiment of the present invention is shown in the following formula (8):

[0107]

[0108] Among them, the symbol ← in the formula represents the parameter update operation.

[0109] It is easy to understand that in the embodiments of the present invention, by repeatedly iterating the above steps until the preset iteration stop condition is met, it means that the model training is completed, and a feature dimensionality reduction model is obtained. For example, the preset iteration stop condition in the embodiments of the present invention is to reach the preset maximum number of iterations. For example, the preset maximum number of iterations is set to 1000 times.

[0110] In some embodiments of the present invention, according to the first loss function, the parameters of the initial autoencoder are updated by the gradient descent algorithm to obtain a feature dimensionality reduction model, including but not limited to the following steps:

[0111] Calculate the first expected learning rate through the convex hull convergence algorithm.

[0112] Calculate the parameter update amount according to the first expected learning rate and the first loss function.

[0113] Update the parameters of the initial autoencoder according to the parameter update amount to obtain a feature dimensionality reduction model.

[0114] In this specific embodiment, in order to make the network converge quickly, the embodiments of the present invention optimize the training process of the autoencoder through the convex hull convergence strategy. Specifically, the embodiments of the present invention first calculate the first expected learning rate through the convex hull convergence algorithm, so as to calculate the parameter update amount according to the first expected learning rate and the first loss function, and then update the parameters of the initial autoencoder according to the parameter update amount to obtain a feature dimensionality reduction model. Correspondingly, in the embodiments of the present invention, by adjusting the learning step size and the weight update strategy, the network is made to converge quickly, and the learning rate is dynamically adjusted according to the data distribution characteristics in the current iteration, as shown in the following formula (9):

[0115]

[0116] Among them, in the formula is the learning rate of the initial autoencoder; is the learning rate of the autoencoder in the t-th iteration; γ r is the learning rate decay factor of the autoencoder, is the average diameter of the convex hull in the t-th iteration, is the diameter of the initial convex hull, and the learning rate decreases as the diameter of the convex hull decreases to prevent instability caused by too large a step size in the later stage of learning.

[0117] Correspondingly, the embodiments of the present invention substitute the calculated first expected learning rate into the above formula (7) to calculate the corresponding parameter update amount, and then update the parameters of the initial autoencoder to obtain a feature dimensionality reduction model.

[0118] In some embodiments of the present invention, data augmentation is performed on the second radiation data through a preset generative adversarial network model to obtain third radiation data, including but not limited to the following steps:

[0119] Construct an initial generative adversarial network model. The initial generative adversarial network model includes a generator and a discriminator.

[0120] The generator receives random noise and converts the random noise into quantum state data through a quantum state encoding algorithm.

[0121] Preset radiation data is generated based on the quantum state data.

[0122] The discriminator performs similarity calculation on the preset radiation data according to a similarity metric algorithm to obtain an evaluation result.

[0123] The initial generative adversarial network model is adjusted according to the evaluation result to obtain a preset generative adversarial network model.

[0124] In this specific embodiment, before performing data augmentation, the present invention embodiment first trains a generative adversarial network model based on quantum states. Specifically, the present invention embodiment first constructs an initial generative adversarial network model. The initial generative adversarial network model in the present invention embodiment includes a generator and a discriminator. Correspondingly, the generator in the present invention embodiment is used to generate synthetic data close to real radiation data, and the discriminator is used to distinguish between the generated data and the real data. For example, the present invention embodiment defines the initialization parameters of the generator G and the discriminator D of the generative adversarial network based on quantum states. Among them, the parameters of the generator G are represented by b G and are initialized to random values, as shown in the following formula (10):

[0125] b G = rand(θ G ) (10)

[0126] wherein, in the formula, θ G represents the parameter dimension of the generator.

[0127] At the same time, the parameters of the discriminator D in the present invention embodiment are represented by b D and are shown in the following formula (11):

[0128] b D = rand(θ D ) (11)

[0129] wherein, in the formula, θ D represents the parameter dimension of the discriminator.

[0130] Next, in the embodiments of the present invention, the generator receives random noise and converts the random noise into quantum state data through a quantum state encoding algorithm. Correspondingly, in the embodiments of the present invention, the generator receives random noise as input and converts these noises into a data representation with high-dimensional features through a quantum state encoding process to enhance the data representation ability based on the principle of quantum computing. For example, in the embodiments of the present invention, the input noise vector is represented by b z and the process based on the quantum state is shown in the following formula (12):

[0131]

[0132] wherein, in the formula represents a quantum state encoding function that encodes classical information onto a quantum state. b z is a noise vector randomly sampled from a Gaussian distribution.

[0133] Correspondingly, in the embodiments of the present invention, the quantum state encoding function encodes the classical information b z onto a quantum state and uses quantum gate operations to implement the encoding. For example, in the embodiments of the present invention, the quantum rotation gate R(θ) and the quantum superposition state are used. Let θ be the rotation angle, which can be calculated from the elements of the input vector b z as shown in the following formula (13):

[0134]

[0135] wherein, in the formula is the i-th element of the vector b z and θ i is the rotation angle of the corresponding quantum rotation gate. In the embodiments of the present invention, the input noise vector is randomly sampled from a standard normal distribution, and the rotation angle θ i of each element is set to

[0136] Meanwhile, in the embodiments of the present invention, the quantum state encoding function is a series of quantum rotation operations, as shown in the following formula (14):

[0137]

[0138] wherein, in the formula is a quantum rotation operation.

[0139] Furthermore, the embodiments of the present invention generate preset radiation data based on quantum state data. Among them, in the embodiments of the present invention, the noise based on quantum states will be used to generate synthetic radiation data, which are not only diverse but also of high quality, and can help the model learn more complex data distributions. For example, the vector b encoded by quantum states in the embodiments of the present invention encoded , the generator G generates synthetic data b gen in the following manner as shown in Equation (15):

[0140] b gen = G(b encoded ; b G ) (15)

[0141] Among them, in the formula, G(·; b G ) is a parameterized generator function with the parameter b G .

[0142] Furthermore, the embodiments of the present invention use a discriminator to calculate the similarity of the preset radiation data according to the similarity metric algorithm to obtain an evaluation result. Among them, when evaluating the generated data and the real data, the discriminator in the embodiments of the present invention uses the similarity metric method to more accurately measure the similarity between data, so as to provide more effective gradient feedback to the generator. For example, the embodiments of the present invention use a metric method based on quantum computing to calculate the similarity between the generated data and the real data, as shown in Equation (16) below:

[0143]

[0144] Among them, in the formula, b real represents the real radiation data, is an improved similarity metric function.

[0145] It should be noted that the similarity metric function in the embodiments of the present invention is embodied by calculating the Euclidean distance between the generated data and the real data. Taking the Euclidean distance as an example in the embodiments of the present invention, its calculation method is as shown in Equation (17) below:

[0146]

[0147] Among them, in the formula and respectively represent the i-th elements of the generated data and the real data vectors.

[0148] Accordingly, in the embodiments of the present invention, the generator and the discriminator are cyclically iteratively trained in an adversarial framework. By continuously optimizing the generator to generate more accurate data, while improving the discrimination ability of the discriminator, until the generated data is indistinguishable from the real data. For example, in the process of cyclic iteration, the discriminator D and the generator G in the embodiments of the present invention are optimized through the first loss function, as shown in the following formula (18):

[0149]

[0150] In fact, in the formula, D() is the discriminator function, which can output the probability that the data is real data. is the generator loss, is the discriminator loss.

[0151] During the training process, the embodiments of the present invention regularly evaluate the performance of the model, specifically evaluating the quality and diversity of the generated data. Then, based on the evaluation results, the training strategy is further adjusted, such as adjusting the quantum state encoding parameters, optimizing the similarity measurement method, etc., to ensure that the model can effectively perform data augmentation. For example, the performance evaluation of the model in the embodiments of the present invention is carried out by calculating the distance between the generated data and the real data, as shown in the following formula (19):

[0152] Δ=dist(b gen ,b real ) (19)

[0153] Among them, in the formula, dist(,) is the Euclidean distance metric function, which is used to evaluate the quality and diversity of the generated data.

[0154] Finally, the embodiments of the present invention adjust the parameters of the initial generative adversarial network model according to the evaluation results to obtain a preset generative adversarial network model. Among them, based on the evaluation results in the embodiments of the present invention, the parameters of the generator and the discriminator are adjusted, as shown in the following formula (20):

[0155]

[0156] Among them, in the formula, η is the learning rate, and are the gradients of the first loss functions of the generator and the discriminator respectively. In one embodiment, the learning rate η is set to 0.001.

[0157] At the same time, for the parameter b G of the generator G in the embodiments of the present invention, the gradient update is specified using the chain rule, as shown in the following formula (21):

[0158]

[0159] Among them, in the formula Represents the loss The partial derivative of the generator output is the partial derivative of the generator output with respect to its parameters.

[0160] In addition, for the parameter b of the discriminator D in the embodiments of the present invention D , the gradient update is as shown in the following formula (22):

[0161]

[0162] wherein, in the formula is the loss The partial derivative of the discriminator output represents the partial derivative of the discriminator output with respect to its parameters.

[0163] It is easy to understand that in order to ensure the quality and diversity of the generated data, the model in the embodiments of the present invention needs to perform a sufficient number of training iterations, for example, the number of training iterations is set to 10,000 times.

[0164] In some embodiments of the present invention, the preset monitoring model is trained through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model, including but not limited to the following steps:

[0165] Construct a central model and local models. Among them, the local models include a number of extreme learning machine models, and the number of extreme learning machine models corresponds to the third radiation data.

[0166] Train the corresponding extreme learning machine models respectively through the third radiation data to obtain local update parameters.

[0167] Transmit the local update parameters to the central model to perform aggregation update on the central model to obtain target update parameters.

[0168] Transmit the target update parameters to each extreme learning machine model respectively to perform parameter update on the extreme learning machine models to obtain a target radiation monitoring model.

[0169] In this specific embodiment, the embodiment of the present invention first constructs a central model and local models. Specifically, in the embodiment of the present invention, the local model refers to a model independently trained by each client on its local dataset, and each client is provided with a corresponding local model. Correspondingly, the central model refers to a model shared by all participants (each client). Among them, in the embodiment of the present invention, a local model is first constructed in each client respectively, and a central model is constructed in the corresponding central server. Correspondingly, the local models of each client in the embodiment of the present invention include extreme learning machine models, and each extreme learning machine model corresponds to third radiation data one by one. For example, each client collects first radiation data through a corresponding radiation monitoring sensor, and performs data dimensionality reduction and data augmentation to obtain corresponding third radiation data. Then, the embodiment of the present invention trains the corresponding extreme learning machine model with the third radiation data respectively to obtain local update parameters, and then transmits the local update parameters to the central model to perform aggregated update on the central model to obtain target update parameters. Specifically, the embodiment of the present invention inputs the third radiation data corresponding to each client into the extreme learning machine model for model training to obtain local update parameters after the local model is trained and updated. Then, the embodiment of the present invention transmits the local update parameters corresponding to each client to the central model to aggregate and update the central model. Exemplarily, as Figure 3 shown, in the embodiment of the present invention, the local models are trained on the data of their respective clients, and there is an interaction between the local models and the central model, that is, the update flow of the model parameters. θ fed ′ represents the updated model parameters after local training, and θ fed represents the parameters received from the central model. Correspondingly, the central model aggregates the updates from each local model, that is, the local update parameters. The update of the global model (central model) is the way for the federated learning process to aggregate individual updates to create a new improved global model. After the update of the global model is completed, the finally trained model is obtained. Among them, the model parameters in the embodiment of the present invention to Exchange between the local model and the central model. This exchange method allows the central model to aggregate updates and enables the local model to receive new, aggregated parameters, i.e., the target update parameters. Further, the embodiments of the present invention transmit the target update parameters to each extreme learning machine model respectively to update the parameters of the extreme learning machine model and obtain the target radiation monitoring model. Specifically, in the embodiments of the present invention, through multiple iterations, the local model is trained to obtain local update parameters and send them to the central model to perform global updates, and then the updated parameters (target update parameters) are sent back to the local model to update the parameters of the local model and obtain the target radiation monitoring model. Among them, in the embodiments of the present invention, through the federated learning architecture, data privacy protection can be achieved. During the training process, the original data is not shared between clients or with the central server, but only the model parameters are exchanged for updates, that is, by the way that the central model synchronizes the parameters to each client, θ fed ′ is assigned to each client, effectively alleviating the problem of data privacy risk.

[0170] In some embodiments of the present invention, the corresponding extreme learning machine model is trained with the third radiation data to obtain local update parameters, including but not limited to the following steps:

[0171] Input the third radiation data into the extreme learning machine model for non-linear mapping to obtain intermediate output data.

[0172] Calculate the target output data according to the intermediate output data and the weight matrix.

[0173] Calculate the loss data according to the target output data through the second loss function to determine the weight update data through the loss data. Among them, the second loss function includes a gradient penalty coefficient.

[0174] Update the parameters of the extreme learning machine model according to the weight update data through the backpropagation algorithm to obtain local update parameters.

[0175] In this specific embodiment, the embodiment of the present invention first inputs the third radiation data into the extreme learning machine model for non-linear mapping to obtain intermediate output data, and calculates the target output data based on the intermediate output data and the weight matrix. Specifically, in the embodiment of the present invention, the local models of each client adopt the extreme learning machine algorithm based on gradient penalty as the radiation level monitoring model. Correspondingly, in the example of the present invention, the extreme learning machine includes an input layer, a hidden layer, and an output layer. The input layer receives the collected data, the hidden layer uses an activation function to perform non-linear transformation on the features, and is connected to the output layer through the weight matrix. The output data of the output layer, that is, the target output data, is calculated through the weight matrix and the intermediate output data. Exemplarily, the embodiment of the present invention first initializes the parameters of each extreme learning machine model. The initialization method is random initialization, and the initialized parameters follow a normal distribution with a mean of 0 and a variance of the identity matrix. Then, the input layer of the extreme learning machine model in the embodiment of the present invention receives the input data X r , and the main task of the hidden layer is to perform non-linear mapping on the input data X r to obtain an intermediate output H u , that is, the intermediate output data. For example, the weight of the hidden layer of the extreme learning machine model in the embodiment of the present invention is W u , and the bias is b u , then the output of the hidden layer is expressed as shown in the following formula (23):

[0176] H u = Re(W u X u + b u ) ⊙ A u (23)

[0177] Among them, in the formula, H u is the output of the hidden layer; Re() is the ReLU activation function; ⊙ is element-wise multiplication; A u is the local sensitivity mapping factor.

[0178] Correspondingly, the local sensitivity mapping factor in the embodiment of the present invention enables the model to adaptively enhance the attention to local patterns, avoid the global features dominating the learning process of the model, and thus improve the classification accuracy. The calculation formula is as shown in the following formula (24):

[0179]

[0180] Among them, in the formula, F u (H u ) is the local response function on the output of the hidden layer, such as the local feature response of the local maximum (i.e., max pooling); λ uee is the coefficient for adjusting the local sensitivity, controlling the influence degree of the local features on the activation. For example, in the embodiment of the present invention, λuee Set to 0.01.

[0181] Furthermore, in the embodiment of the present invention, the loss data is calculated from the target output data through the second loss function, so as to determine the weight update data based on the loss, and then the parameters of the extreme learning machine model are updated through the backpropagation algorithm according to the weight update data, obtaining the local update parameters. Specifically, the second loss function in the embodiment of the present invention includes a gradient penalty coefficient. Among them, during the forward propagation process, the gradient calculated each time in the embodiment of the present invention needs to be subjected to gradient penalty, which makes the gradient of the model remain within a reasonable range during each parameter update, so as to alleviate the problem of unstable training caused by too large or too small gradients. Exemplarily, in the first loss function in the embodiment of the present invention, a gradient penalty term (gradient penalty coefficient) is adopted to constrain the first loss function, so that the stability and effectiveness of the gradient can be maintained during each parameter update. The calculation formula is shown in the following formula (25):

[0182]

[0183] Wherein, in the formula, λ u is the first gradient penalty coefficient; β u is the second gradient penalty coefficient; γ u is the third gradient penalty coefficient; is the gradient penalty term; is the gradient of the output of the hidden layer of the extreme learning machine with respect to the weight.

[0184] Exemplarily, after the non-linear transformation of the hidden layer, the final result will be passed to the output layer of the classifier. In the output layer of the extreme learning machine, the input signal is mapped through the Softmax function to generate the classification result, that is, the target output data, as shown in the following formula (26):

[0185]

[0186] Wherein, in the formula

[0187] is the classification result of the extreme learning machine; W u is the weight matrix of the output layer; b u is the bias term of the output layer; H u is the output of the hidden layer; Sof() is the Softmax function, whose function is to map the input linear combination to a probability distribution, and finally obtain the probability of each category; θ u is the coefficient controlling the strength of the quadratic term; (H u ) 2Represents the square term of the output of the hidden layer. By adopting the quadratic term, the fitting ability of the classification model to complex data patterns is enhanced. Especially in high-dimensional radiation monitoring data, it can better capture the non-linear features of the data.

[0188] Then, the embodiment of the present invention calculates the first loss function of the extreme learning machine model. The first loss function of the extreme learning machine model includes a constant error calculation term and a gradient penalty term. Through the gradient penalty term, in each optimization process, the classifier can better learn the deep features in the data and reduce the possible bias in the training process. The calculation formula is as shown in the following formula (27):

[0189]

[0190] Where, in the formula Is the first loss function of the extreme learning machine model; Is the first cross-entropy loss function, which is used to measure the difference between the predicted value and the true label; Is the gradient penalty term; λ u Is the gradient penalty coefficient, which controls the contribution of the gradient penalty to the first loss function; Is the local feature enhancement term; A u (i) is the i-th element of the local sensitivity mapping factor; μ uev Is the weight coefficient of the local feature enhancement term, which is used to control the influence of the enhancement term on the final loss. For example, in the embodiment of the present invention, μ uev Is set to 0.2.

[0191] Next, the embodiment of the present invention performs backpropagation and parameter update of the extreme learning machine model. The calculated error will update the weight parameters of the network through the backpropagation algorithm. By using the backpropagation algorithm to update the parameters of the model, the gradient is calculated using the chain rule. The parameter update method for the extreme learning machine model is as shown in the following formula (28):

[0192]

[0193] Where, in the formula, ΔW u Is the weight update amount of the extreme learning machine model; Δb u Is the bias update amount of the extreme learning machine model; Is the gradient of the first loss function of the extreme learning machine model with respect to the weight; W u (t) and b u (t) are the weight and bias of the t-th iteration respectively; W u (t + 1) and b u (t + 1) are the weight and bias of the (t + 1)-th iteration respectively.

[0194] It should be noted that in the embodiments of the present invention, the above steps are repeatedly iterated until a preset stop iteration condition is met, which indicates that the model training is completed. Correspondingly, the preset stop iteration condition in the embodiments of the present invention is to reach a preset maximum number of iterations. For example, the preset maximum number of iterations is set to 1000 times. Correspondingly, by continuously iterating and updating the weights, the extreme learning machine model can be continuously optimized during the training process and finally achieve the optimal classification effect. Among them, after the extreme learning machine model in the embodiments of the present invention is trained, the radiation level monitoring result is the classification result of the extreme learning machine. For example, in the embodiments of the present invention, the classification categories include: "low radiation", "medium radiation" and "high radiation", a total of 3 categories.

[0195] In some embodiments of the present invention, after performing parameter update on the extreme learning machine model by the backpropagation algorithm according to the weight update data to obtain the local update parameters, the radiation level monitoring method provided by the embodiments of the present invention further includes but is not limited to the following steps:

[0196] Calculate the second expected learning rate according to the weight update data, the intermediate output data and the base learning rate.

[0197] Adjust the model training learning rate according to the second expected learning rate.

[0198] In this specific embodiment, in the embodiments of the present invention, each time the weight is updated, the learning rate is adaptively adjusted so that the model can converge faster. Among them, in the embodiments of the present invention, the second expected learning rate is calculated according to the weight update data, the intermediate output data and the base learning rate to adjust the model training learning rate according to the second expected learning rate. Specifically, the base learning rate in the embodiments of the present invention refers to the base learning rate of the extreme learning machine model, and the weight update data refers to the change amount of the weight. Correspondingly, the weight update of the extreme learning machine model in the embodiments of the present invention follows an adaptive convergence strategy, and the learning rate is automatically adjusted according to the change of the current gradient. For example, when the model training is relatively stable, the learning rate will automatically decrease to accelerate the convergence process, or when there is a large gradient change in the model, the learning rate will be appropriately increased to help the model overcome the local optimal solution and further improve the classification accuracy. Among them, the learning rate calculation formula of the extreme learning machine model in the embodiments of the present invention is shown in the following formula (29):

[0199]

[0200] Among them, in the formula, η u (t + 1) is the learning rate of the extreme learning machine model in the (t + 1)-th iteration; η base is the base learning rate of the extreme learning machine model; γ u is the first adjustment coefficient of the learning rate of the extreme learning machine model, and δ uis the second adjustment coefficient of the learning rate of the extreme learning machine model; ΔW u (t) is the change in weight at the t-th iteration; H u (t) is the output of the hidden layer at the t-th iteration. For example, in the embodiments of the present invention, γ u is set to 0.2, δ u is set to 0.3.

[0201] It is easy to understand that when processing radiation level monitoring data, the model often faces the problem of high-dimensional data and is prone to the phenomenon of gradient disappearance. Therefore, the embodiments of the present invention adopt an adaptive gradient disappearance mitigation strategy to adjust the training step size according to the change of the current gradient, effectively alleviating the gradient disappearance problem and making the training process more stable. The adjustment method is shown in the following formula (30):

[0202]

[0203] where, in the formula, η′ u (t + 1) is the learning rate of the extreme learning machine model at the (t + 1)-th iteration after adjustment; λ uas is the third adjustment coefficient of the learning rate of the extreme learning machine model, μ uas is the fourth adjustment coefficient of the learning rate of the extreme learning machine model; α uas is the fifth adjustment coefficient of the learning rate of the extreme learning machine model; β uas is the sixth adjustment coefficient of the learning rate of the extreme learning machine model. For example, in the embodiments of the present invention, λ uas is set to 0.2, μ uas is set to 0.2, α uas is set to 0.1, β uas is set to 0.1.

[0204] It is easy to understand that in the embodiments of the present invention, by adopting a federated learning architecture, multiple decentralized client models are jointly trained without exchanging local data, thus ensuring data privacy while fully utilizing distributed data for model optimization. Compared with traditional centralized learning methods, the problem of data privacy leakage caused by uploading sensitive radiation monitoring data to a central server is effectively alleviated through the federated learning method, which has significant privacy protection advantages. At the same time, for the high-dimensional features that may exist in radiation monitoring data, the embodiments of the present invention adopt an autoencoder based on boundary smoothing for feature dimensionality reduction. The high-dimensional data is compressed into a low-dimensional representation through the encoder-decoder structure of the autoencoder, and the boundary effect and the influence of outliers are alleviated through smoothing processing, enhancing the robustness of the model to noise, effectively alleviating the problem of high data dimensionality, and at the same time improving the stability and accuracy of data processing. In addition, the embodiments of the present invention alleviate the problem of insufficient training samples through a generative adversarial network based on quantum states. Among them, quantum state encoding utilizes the advantages of quantum computing, can simulate more complex radiation data distributions, and generate diverse high-quality synthetic data, effectively expanding the scale of the training dataset and improving the generalization ability of the model. Especially in the case of scarce samples, the accuracy of the model can be significantly improved. Correspondingly, traditional extreme learning machines are prone to problems such as gradient disappearance or explosion during training, resulting in unstable training. To alleviate this problem, the embodiments of the present invention adopt a gradient penalty term in the first loss function of the extreme learning machine, standardize the range of the gradient, and ensure the stability of the training process, so that the model can maintain higher training efficiency and accuracy when processing complex radiation data. At the same time, the extreme learning machine in the embodiments of the present invention also combines an adaptive learning rate and a local sensitivity mapping mechanism, enabling the model to dynamically adjust the learning rate according to the change of the current gradient, thereby accelerating the convergence process and alleviating the problem of unstable training caused by too large or too small gradients. In addition, local sensitivity mapping can enhance the model's attention to local patterns and avoid being dominated by global features, thereby improving the classification accuracy.

[0205] Please refer to Figure 4 , the embodiments of the present application also provide a radiation level monitoring system, which can implement the above-mentioned radiation level monitoring method. The system includes:

[0206] The first module 210 is used to collect first radiation data through a preset radiation monitoring sensor.

[0207] The second module 220 is used to perform data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data. Among them, the feature dimensionality reduction model includes an autoencoder based on boundary smoothing.

[0208] The third module 230 is configured to perform data augmentation on the second radiation data through a preset generative adversarial network model to obtain third radiation data. The preset generative adversarial network model includes a generative adversarial network model based on quantum states.

[0209] The fourth module 240 is configured to train a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model.

[0210] The fifth module 250 is configured to input the radiation data to be analyzed into the target radiation monitoring model for radiation level analysis to obtain a radiation level monitoring result.

[0211] It can be understood that the content in the above method embodiments is applicable to the system embodiments. The functions specifically implemented in the system embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0212] An embodiment of the present application further provides an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above radiation level monitoring method is implemented. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0213] It can be understood that the content in the above method embodiments is applicable to the device embodiments. The functions specifically implemented in the device embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0214] Please refer to Figure 5 , Figure 5 which illustrates the hardware structure of an electronic device according to another embodiment. The electronic device includes:

[0215] A processor 310, which can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is configured to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0216] The memory 320 can be implemented in the form of a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 320 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 320 and are called by the processor 310 to execute the radiation level monitoring method of the embodiments of this application;

[0217] The input / output interface 330 is used to implement information input and output;

[0218] The communication interface 340 is used to implement communication interaction between this device and other devices. It can implement communication through a wired manner (such as USB, network cable, etc.) or through a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0219] The bus 350 transmits information between various components of the device (such as the processor 310, the memory 320, the input / output interface 330, and the communication interface 340);

[0220] Among them, the processor 310, the memory 320, the input / output interface 330, and the communication interface 340 achieve communication connections with each other inside the device through the bus 350.

[0221] The embodiments of this application also provide a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned radiation level monitoring method.

[0222] It can be understood that the content in the above method embodiments is applicable to the embodiments of this storage medium. The functions specifically implemented by the embodiments of this storage medium are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0223] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0224] The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0225] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

[0226] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0227] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0228] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0229] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects and indicates that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (item) of the following" or its similar expression refers to any combination of these items, including any combination of single item (item) or plural items (items). For example, at least one (item) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0230] In several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the above division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of devices or units can be in electrical, mechanical or other forms.

[0231] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0232] In addition, each functional unit in various embodiments of this application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0233] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0234] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings, and thus do not limit the scope of the rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the rights of the embodiments of the present application.

Claims

1. A radiation level monitoring method, characterized in that: The method comprises the following steps: Collecting first radiation data through a preset radiation monitoring sensor; Performing data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data; wherein the feature dimensionality reduction model includes an autoencoder based on boundary smoothing; Amplifying the second radiation data by using a preset generative adversarial network model to obtain third radiation data; wherein the preset generative adversarial network model includes a quantum state-based generative adversarial network model; Performing model training on a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model; The radiation data to be analyzed is input into the target radiation monitoring model to perform radiation level analysis to obtain radiation level monitoring results.

2. The method according to claim 1, characterized in that Before performing the data dimensionality reduction processing on the first radiation data by using the feature dimensionality reduction model to obtain the second radiation data, the method further includes: Build the initial autoencoder; Inputting the preset input data into the initial autoencoder to generate reconstructed data through forward propagation; Smoothing the reconstructed data using an automatic boundary smoothing algorithm to obtain smoothed output data; A weight coefficient is calculated by using a preset variance function according to the preset input data; A first loss function is calculated according to the smoothed output data, the preset input data and the weight coefficient; The parameters of the initial autoencoder are updated through a gradient descent algorithm according to the first loss function to obtain the feature dimensionality reduction model.

3. The method according to claim 2, characterized in that The step of updating the parameters of the initial autoencoder by using a gradient descent algorithm according to the first loss function to obtain the feature dimension reduction model includes: The first expected learning rate is calculated by the convex hull convergence algorithm; Calculate a parameter update amount according to the first expected learning rate and the first loss function; The parameters of the initial autoencoder are updated according to the parameter update amount to obtain the feature dimensionality reduction model.

4. The method according to claim 1, characterized in that The step of performing data amplification on the second radiation data by using a preset generative adversarial network model to obtain third radiation data includes: Constructing an initial generative adversarial network model; wherein the initial generative adversarial network model includes a generator and a discriminator; receiving random noise through the generator to convert the random noise into quantum state data through a quantum state encoding algorithm; Generate preset radiation data according to the quantum state data; The discriminator performs similarity calculation on the preset radiation data according to a similarity measurement algorithm to obtain an evaluation result; The parameters of the initial generative adversarial network model are adjusted according to the evaluation results to obtain the preset generative adversarial network model.

5. The method according to claim 1, characterized in that: The method of training a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model includes: Constructing a central model and a local model; wherein the local model includes a plurality of extreme learning machine models, and the plurality of extreme learning machine models correspond to the third radiation data; Using the third radiation data, respectively train the corresponding extreme learning machine models to obtain local update parameters; Transmitting the local update parameters to the central model to perform aggregate update on the central model to obtain target update parameters; The target update parameters are transmitted to each of the extreme learning machine models respectively to update the parameters of the extreme learning machine model to obtain the target radiation monitoring model.

6. The method according to claim 5, characterized in that The step of training the corresponding extreme learning machine models respectively by using the third radiation data to obtain local update parameters includes: Inputting the third radiation data into the extreme learning machine model for nonlinear mapping to obtain intermediate output data; Calculate target output data according to the intermediate output data and the weight matrix; Calculating loss data through a second loss function according to the target output data to determine weight update data through the loss data; wherein the second loss function includes a gradient penalty coefficient; The parameters of the extreme learning machine model are updated through a back propagation algorithm according to the weight update data to obtain the local update parameters.

7. The method according to claim 6, characterized in that After performing the step of updating the parameters of the extreme learning machine model by a back propagation algorithm according to the weight update data to obtain the local update parameters, the method further includes: Calculate a second expected learning rate according to the weight update data, the intermediate output data and the basic learning rate; The model training learning rate is adjusted according to the second expected learning rate.

8. A radiation level monitoring system, characterized in that: The system comprises: A first module is used to collect first radiation data through a preset radiation monitoring sensor; A second module is used to perform data dimensionality reduction processing on the first radiation data through a feature dimensionality reduction model to obtain second radiation data; wherein the feature dimensionality reduction model includes an autoencoder based on boundary smoothing; A third module is used to perform data amplification on the second radiation data through a preset generative adversarial network model to obtain third radiation data; wherein the preset generative adversarial network model includes a quantum state-based generative adversarial network model; A fourth module is used to perform model training on a preset monitoring model through a federated learning architecture according to the third radiation data to obtain a target radiation monitoring model; The fifth module is used to input the radiation data to be analyzed into the target radiation monitoring model to perform radiation level analysis to obtain radiation level monitoring results.

9. An electronic device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images