Environmental historical dose retrieval method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an environment historical dose retrieval method and device, electronic equipment and a storage medium, and relates to the technical field of radiation monitoring. The method comprises the following steps: analyzing and structuring a retrieval request input by a user to generate a query vector; inputting the query vector into a pre-trained hypergraph neural network model to obtain a historical dose record corresponding to the retrieval request; wherein the hypergraph neural network model is obtained by training based on hypergraph sample data, and the hypergraph sample data is obtained by modeling voice data, video data and dose data in a radiation environment, so that a traditional multi-system cross-platform query mode is replaced by model reasoning, the time consumption of manual retrieval and system switching is greatly reduced, and the efficiency is improved. And the response efficiency and the user experience are improved.

Description

Technical Field

[0001] This invention relates to the field of radiation monitoring technology, and more specifically, to a method, apparatus, electronic device, and storage medium for retrieving historical environmental doses. Background Technology

[0002] When nuclear power plant workers enter an area with radiation risk for maintenance or routine inspections, they carry multiple devices, including wireless communication equipment for communicating with the control room or other colleagues, mobile cameras for recording the situation on site, and mobile dose monitors for measuring the radiation dose in the environment.

[0003] These devices continuously generate data during operation, but the data is stored in different systems. For example, call logs are stored in one system (such as a voice electronic device 100), video images are stored in another system (such as a video surveillance platform), and dose data is stored in a third system (such as a dose monitoring system). Although these data record the same time period, the same location, and the same event, they are difficult to quickly "stitch together" for analysis because they come from different devices and are stored in different systems.

[0004] When it's necessary to find the radiation dose of a worker in a specific factory building during a specific time period on a specific day, it requires searching the voice system for call records to confirm the worker's location, finding corresponding video footage to confirm if the worker was in a high-dose area, retrieving dose data from the dosing system to confirm the worker's radiation exposure value, and then manually comparing the time, location, and event to piece together a complete "dose history." This process is not only time-consuming and labor-intensive but also prone to errors, especially when rapid response is required (such as accident investigations and safety assessments). Inefficiency can have serious consequences. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide an environmental historical dose retrieval method, apparatus, electronic device and storage medium.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows: In a first aspect, the present invention provides a method for retrieving historical environmental doses, the method comprising: The system parses and structures the user's search request to generate a query vector. The query vector is input into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request; wherein, the hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data and dose data in the radiation environment.

[0007] Optionally, the method includes the step of training a hypergraph neural network model, wherein the step of training the hypergraph neural network model includes: Acquire call audio data, monitoring video data, and dose monitoring data collected in a radiation environment; The call audio data, monitoring video data, and dose monitoring data are preprocessed, and the preprocessed call audio data, monitoring video data, and dose monitoring data are used to construct hypergraph sample data. Using the hypergraph sample data and a preset loss function for historical dose retrieval, a pre-constructed hypergraph neural network model is trained to obtain a hypergraph neural network model for historical dose retrieval.

[0008] Optionally, the preprocessing includes a first preprocessing, a second preprocessing, and a third preprocessing. The first preprocessing includes format conversion, noise reduction, speech recognition, semantic analysis, and information structuring. The second preprocessing includes format conversion, frame extraction synchronization, image enhancement, target detection, and feature structuring. The third preprocessing includes data cleaning, filtering, and data structuring. The preprocessing of the call audio data, monitoring video data, and dose monitoring data includes: The call audio data is subjected to the first preprocessing to obtain preprocessed call audio data; The surveillance video data is subjected to the second preprocessing to obtain preprocessed surveillance video data; The dose monitoring data is subjected to the third preprocessing to obtain preprocessed dose monitoring data.

[0009] Optionally, the step of constructing hypergraph sample data using preprocessed call audio data, preprocessed monitoring video data, and preprocessed dose monitoring data includes: Each data point in the preprocessed call audio data, preprocessed monitoring video data, and preprocessed dose monitoring data is treated as a hypergraph node, resulting in multiple hypergraph nodes. Hyperedges are constructed for multiple hypergraph nodes, and weights are assigned to each hyperedge based on the node relevance parameters of each constructed hyperedge to obtain the hypergraph sample data.

[0010] Optionally, the step of constructing hyperedges for the plurality of hypergraph nodes includes: Connect hypergraph nodes that are in the same time window, located in the same spatial region, and have semantic relevance to the same hyperedge.

[0011] Optionally, the node relevance parameters of the hyperedge include the time interval, spatial distance, and semantic similarity between the hypergraph nodes on the hyperedge, and the step of assigning weights to each hyperedge based on the node relevance parameters of each constructed hyperedge includes: For each hyperedge, the hyperedge weights are calculated based on the time interval, spatial distance, and semantic similarity between the hypergraph nodes on the hyperedge. The weight of the hyperedge is determined based on the weight of the hyperedge between each hypergraph node on the hyperedge.

[0012] Optionally, the step of training a pre-constructed hypergraph neural network model using the hypergraph sample data and a preset loss function for historical dose retrieval to obtain a hypergraph neural network model for historical dose retrieval includes: The hypergraph sample data is divided into a training set, a validation set, and a test set; The training set is input into a pre-constructed hypergraph neural network model for training, and the model parameters are updated using the preset loss function during the training process. After each training round, the model parameters are adjusted using the validation set to obtain the trained hypergraph neural network model, and the performance of the trained hypergraph neural network model is evaluated using the test set. If the performance evaluation results of the trained hypergraph neural network model meet the preset conditions, the trained hypergraph neural network model will be used as the hypergraph neural network model for historical dose retrieval.

[0013] In a second aspect, the present invention provides an environmental historical dose retrieval device, the device comprising: The generation module is used to parse and structure the user's input search request to generate a query vector; The module is used to input the query vector into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request; wherein, the hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data and dose data in the radiation environment.

[0014] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing machine-executable instructions executable by the processor, the processor executing the machine-executable instructions to implement the environmental historical dose retrieval method described in the first aspect above.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the environmental historical dose retrieval method as described in the first aspect above.

[0016] The environmental historical dose retrieval method, apparatus, electronic device, and storage medium provided in this invention parse and structure user-input retrieval requests to generate query vectors. These query vectors are then input into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request. The hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data, and dose data in the radiation environment. Because this invention constructs hypergraph sample data from speech, video, and dose data and trains a hypergraph neural network model that integrates multimodal features, it replaces the traditional multi-system cross-platform query method with model inference, significantly reducing the time spent on manual retrieval and system switching, and improving response efficiency and user experience.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This figure shows a schematic block diagram of an electronic device provided by an embodiment of the present invention; Figure 2 A flowchart illustrating an environmental historical dose retrieval method provided by an embodiment of the present invention is shown; Figure 3 This diagram illustrates a flowchart of a training hypergraph neural network model according to an embodiment of the present invention. Figure 4 This diagram illustrates a process for constructing hypergraph sample data according to an embodiment of the present invention. Figure 5 The diagram shows a functional unit block diagram of an environmental historical dose retrieval device provided in an embodiment of the present invention.

[0020] Icons: 100 - Electronic device; 110 - Memory; 120 - Processor; 130 - Communication module; 200 - Environmental history dose retrieval device; 201 - Generation module; 202 - Acquisition module. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0024] Please refer to Figure 1 This is a block diagram of an electronic device 100. The electronic device 100 includes a memory 110, a processor 120, and a communication module 130. The memory 110, processor 120, and communication module 130 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0025] The memory 110 is used to store programs or data. The memory 110 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0026] The processor 120 is used to read / write data or programs stored in the memory 110 and to perform corresponding functions.

[0027] The communication module 130 is used to establish a communication connection between the electronic device 100 and other communication terminals through the network, and to send and receive data through the network.

[0028] It should be understood that, Figure 1 The structure shown is only a schematic diagram of the electronic device 100. The electronic device 100 may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0029] Please refer to Figure 2 The environmental historical dose monitoring method provided in this embodiment of the invention includes steps S101 to S102.

[0030] S101: The user-input search request is parsed and structured to generate a query vector.

[0031] Users can enter a search request containing multiple query conditions through the search interface. These conditions can include staff name, factory name, working hours, working area, photos, call records, etc. For example: "Find the dose information of Zhang San during his work in Factory No. 3 on May 12, 2024", "Display all dose records related to Zhao Wu from June 10 to June 15, 2023", "Retrieve all historical dose data related to 'Main Control Room Instrument Anomaly'".

[0032] The retrieval request is parsed, semantically understood, and structurally transformed to generate query vectors that can be used for model inference.

[0033] S102, input the query vector into the pre-trained hypergraph neural network model to obtain the historical dose record corresponding to the retrieval request.

[0034] The hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data, and dose data in the radiation environment.

[0035] The hypergraph neural network model performs information propagation and reasoning within the hypergraph structure based on the features in the query vector, retrieving the historical dose data that best matches the query conditions.

[0036] The training process of the hypergraph neural network model will be introduced below.

[0037] Please refer to Figure 3 The training process of the hypergraph neural network model includes steps S201 to S203.

[0038] S201, acquire call audio data, monitoring video data, and dose monitoring data collected in the radiation environment.

[0039] The audio data for the call can come from sources such as walkie-talkies worn by staff, handheld phones, voice communication terminals, and fixed communication systems deployed on-site. Real-time audio streams are obtained by interfacing with the communication system.

[0040] The sources of surveillance video data can be mobile cameras (such as GoPros worn by staff or body cameras), fixed cameras deployed in key areas of the factory, and camera equipment carried by drones and robots. Real-time video streams are accessed through a video management system (VMS).

[0041] Dose monitoring data can be sourced from personal dosimeters, fixed environmental dosimeters, and portable dosimetry devices. Dose data is acquired by interfacing with on-site monitoring systems.

[0042] S202, preprocess the call audio data, monitoring video data and dose monitoring data, and use the preprocessed call audio data, preprocessed monitoring video data and preprocessed dose monitoring data to construct hypergraph sample data.

[0043] The preprocessing includes a first preprocessing, a second preprocessing, and a third preprocessing. The first preprocessing includes format conversion, noise reduction, speech recognition, semantic analysis, and information structuring. The second preprocessing includes format conversion, frame extraction synchronization, image enhancement, object detection, and feature structuring. The third preprocessing includes data cleaning, filtering, and data structuring.

[0044] In a possible implementation, the preprocessing of call audio data, monitoring video data, and dose monitoring data can be as follows: performing the first preprocessing on the call audio data to obtain preprocessed call audio data; performing the second preprocessing on the monitoring video data to obtain preprocessed monitoring video data; and performing the third preprocessing on the dose monitoring data to obtain preprocessed dose monitoring data.

[0045] The call audio data is uniformly converted into a standard format (such as WAV, PCM) and a uniform sampling rate (such as 16kHz); speech enhancement algorithms (such as Wiener filtering, spectral subtraction, and deep learning speech enhancement models) are used to remove background noise; the audio is converted into text using the ASR system; semantic analysis is performed on the ASR output text to identify entities such as time, location, people, and events, as well as extract words related to dosage, operation, and anomalies.

[0046] The timestamps of each voice segment in the call audio data are synchronized with NTP time, and multiple voice segments are time-aligned. Semantic tags such as "abnormal operation", "device failure", and "abnormal dosage" are added to the call audio data.

[0047] The preprocessed call audio data is structured text data, including call content text, call keywords, dialogue summary, call time, and caller and called party information.

[0048] The surveillance video data is decoded into a frame sequence. Enhancement processing such as illumination correction, defogging, and noise reduction is performed on the video frame images. Super-resolution reconstruction is performed on low-resolution images. Target detection models (such as YOLOv5 and Faster R-CNN) are used to identify workers, equipment, factory signs, lighting conditions, background changes, and abnormal behaviors from the images.

[0049] Extract metadata such as timestamps, camera IDs, and shooting locations from video frames, and synchronize the timestamps with NTP time to ensure consistency with audio and dose data.

[0050] The dose monitoring data collected from different devices will be standardized to a standard unit, reasonable thresholds will be set, and abnormal data that exceeds the physical reasonable range (such as exceeding 10,000 μSv / h) will be removed. Statistical methods (such as the 3σ principle) or machine learning models will be used to identify abnormal data points, and methods such as moving average, Kalman filtering or wavelet transform will be used to remove noise before feature extraction.

[0051] The preprocessed dose monitoring data is structured text data, including current dose value, cumulative dose, dose change rate, location information, and personnel ID.

[0052] In this embodiment of the invention, all data acquisition devices are connected to the NTP server to ensure that the data acquisition time is consistent. For devices that are not automatically synchronized, manual correction is performed using event matching (such as the "start working" event in the voice and the "enter the scene" screen in the video).

[0053] The preprocessed speech text, video features, and dose data will be used to construct hypergraph nodes and hyperedges to obtain hypergraph sample data, which will be described in detail below.

[0054] Please refer to Figure 4 The process of constructing the hypergraph sample data includes steps S301 to S302.

[0055] S301, each data point in the preprocessed call audio data, preprocessed monitoring video data, and preprocessed dose monitoring data is treated as a hypergraph node, resulting in multiple hypergraph nodes.

[0056] A hypergraph is a graph structure that can effectively model complex relationships between multimodal data. Unlike traditional graph structures where each edge connects only two nodes, a hypergraph's hyperedge can connect multiple nodes.

[0057] After preprocessing, the original audio, video, and dose data are converted into structured, vectorized, time-aligned data units, each of which is a "data point".

[0058] Each data point is abstracted into a hypergraph node, resulting in multiple hypergraph nodes, including audio nodes, video nodes, and dose nodes.

[0059] For each hypergraph node, node features are extracted. Specifically, semantic vectors are extracted using BERT, Whisper, and NLP models to obtain node features for speech nodes. Image embeddings are extracted using ResNet and Vision Transformer to obtain node features for video nodes. Node features for dose nodes are obtained by structuring and encoding dose values, equipment information, personnel information, etc., into vectors.

[0060] Each hypergraph node is mapped to a unified attribute structure, and all three types of nodes are collected according to time range, region range, or task range to form a hypergraph node set. This set will serve as the foundation for subsequent hyperedge construction and hypergraph neural network modeling.

[0061] S302, construct hyperedges for multiple hypergraph nodes, and assign weights to each hyperedge based on the node correlation parameters of each constructed hyperedge to obtain hypergraph sample data.

[0062] In a possible implementation, the process of "constructing a hyperedge for multiple hypergraph nodes" can be: connecting hypergraph nodes that are in the same time window, located in the same spatial region, and have semantic relevance to the same hyperedge.

[0063] In this embodiment of the invention, constructing a hyperedge requires satisfying the following three constraints: Time window consistency: All connected hypergraph nodes must be within the same time window. The time window can be set according to business needs, for example: a short window (e.g., ±5 seconds) for capturing instantaneous events (such as abnormal dialogues and dose changes) and a long window (e.g., ±1 minute) for capturing continuous operations (such as equipment inspection and maintenance processes).

[0064] The timestamps of each hypergraph node are normalized, and a sliding window mechanism is used to group hypergraph nodes that are close in time into the same group. For example, audio node A (timestamp t1), video node B (timestamp t2), and dose node C (timestamp t3) are considered to be in the same time window if |t1-t2|≤Δt and |t1-t3|≤Δt.

[0065] Spatial region consistency: All connected hypergraph nodes must be in the same physical spatial region, such as the same factory, the same room, or near the same equipment.

[0066] The spatial code or region ID is matched with the spatial location information carried by the hypergraph nodes to construct the hyperedge within the region. For example, hypergraph nodes A, B, and C are all located in the same spatial region Plant_3_ControlRoom, which can constitute a hyperedge.

[0067] Semantic relevance: All connected hypergraph nodes must be semantically related, meaning that the events, behaviors, or states they represent are logically connected. Semantic relevance between nodes can be determined by detecting overlapping node keywords, calculating the similarity between node feature vectors, or matching node event labels.

[0068] Traverse the hypergraph node set, sort all hypergraph nodes by time, use a sliding window mechanism to divide time slices, divide node subsets according to spatial regions within each time slice, calculate semantic similarity for each pair of nodes in each subset to form semantic clusters, and connect nodes in the same cluster to a hyperedge.

[0069] A hyperedge can connect voice nodes, video nodes, and dose nodes, expressing the collaborative event relationships among the three types of data—voice, video, and dose—within this time period.

[0070] The node relevance parameters of a hyperedge include the time interval, spatial distance, and semantic similarity between each hypergraph node on the hyperedge. In a possible implementation, the process of "assigning weights to each hyperedge based on the node relevance parameters of each constructed hyperedge" can be as follows: for each hyperedge, calculate the hyperedge weights between each hypergraph node on the hyperedge based on the time interval, spatial distance, and semantic similarity between each hypergraph node on the hyperedge; and determine the weight of the hyperedge based on the hyperedge weights between each hypergraph node on the hyperedge.

[0071] In this embodiment of the invention, each hyperedge connects multiple hypergraph nodes (e.g., a voice node + a video node + a dose node). To calculate the weights between nodes, quantization is required based on three dimensions: time interval, spatial distance, and semantic similarity. The time interval refers to the time difference between two hypergraph nodes. The smaller the time interval, the more likely the events corresponding to the two hypergraph nodes are related.

[0072] Spatial distance refers to the distance between the spatial locations of two hypergraph nodes.

[0073] Semantic similarity refers to the degree of similarity between the semantics expressed by two hypergraph nodes, and can be measured using vector similarity (such as cosine similarity).

[0074] The formula for calculating the weight of the hyperedge between two hypergraph nodes is as follows:

[0075] in, It is the weight of the hyperedge between hypergraph nodes 𝑖 and 𝑗. It is the spatial distance between hypergraph nodes 𝑖 and 𝑗. It is the time interval between hypergraph nodes 𝑖 and 𝑗. is the semantic similarity between hypergraph nodes 𝑖 and 𝑗, and 𝛼, 𝛽 and 𝛾 are adjustment parameters.

[0076] A hyperedge contains multiple hypergraph nodes, so the weight of a hyperedge can be obtained by aggregating the weights of hyperedges between all hypergraph node pairs. The weight of a hyperedge can be the average of the weights of hyperedges between all hypergraph node pairs, or it can be the maximum of the weights of hyperedges between all hypergraph node pairs.

[0077] For example, the hyperedge HE_001 connects the following three nodes: voice node A (timestamp t1=10:02:35), video node B (timestamp t2=10:02:36, spatial location=Plant_3_ControlRoom), and dose node C (timestamp t3=10:02:37, dose value=1.5μSv / h, location=Plant_3_ControlRoom).

[0078] The time interval between the node pairs is calculated as follows: =1s, =2s, =1s.

[0079] Since nodes A, B, and C are located in the same region, the spatial distances between the node pairs are respectively =0, =0, =0.

[0080] Calculate the semantic similarity between node pairs, respectively = 0.82 (voice alert "instrument malfunction" vs. video alert "red warning light") = 0.75 (speech vs. increased dose), = 0.78 (video vs. dose increase).

[0081] Assuming α = 0.1, β = 0.05, and γ = 0.5, calculate the weights between nodes, which are respectively =0.819, =0.753, =0.792.

[0082] Calculate the weight of the entire hyperedge: =(0.819 +0.753+0.792) / 3 =0.788.

[0083] S203. Using hypergraph sample data and a preset loss function for historical dose retrieval, the pre-constructed hypergraph neural network model is trained to obtain a hypergraph neural network model for historical dose retrieval.

[0084] To enable message passing mechanisms for information exchange between nodes and to aggregate neighbor node information, this invention designs hypergraph convolutional layers in the hypergraph neural network model to learn node features and their higher-order interactions. The information passing process between nodes can be represented by the following formula:

[0085] in, It is the feature vector of node l in the l-th layer. It is the set of neighbors of node 𝑖. It is the edge feature between nodes 𝑖 and 𝑗. It is a learnable weight matrix. It is an activation function. It is the normalization coefficient.

[0086] To enable the model to dynamically adjust the degree of attention given to different nodes, especially for handling the relationships between multimodal data, this embodiment of the invention introduces a self-attention module into the hypergraph neural network model. Based on the cross-modal attention mechanism, it enhances the fusion effect of information from different sources and improves the overall performance of the model.

[0087] Attention Score It can be calculated using the following formula:

[0088] Where a is a learnable one-dimensional vector, and W is a learnable weight matrix. and These are the feature vectors of nodes 𝑖 and 𝑗, respectively, where || represents the feature vector concatenation operation. The activation function helps to solve the gradient vanishing problem, and N(𝑖) is the set of neighbors of node 𝑖.

[0089] The hypergraph neural network model also incorporates a multimodal fusion layer to comprehensively process different types of features from speech, images, and dose, producing a unified representation.

[0090] In this embodiment of the invention, the loss function of the hypergraph neural network model can be designed according to the task requirements. For classification tasks, cross-entropy loss can be used; for regression tasks, mean squared error or other appropriate loss forms can be selected.

[0091] The loss function 𝐿 can be expressed as:

[0092] in, The true label of sample A. Compared with the predicted value The losses between It is a regularization term for the model parameters. It is the regularization strength.

[0093] Furthermore, hypergraph neural network models can incorporate regularization terms to prevent overfitting. Techniques such as Dropout and L2 regularization can be used to reduce model complexity and improve generalization ability. Simultaneously, other forms of regularization, such as label smoothing, can be explored to further enhance the model's robustness.

[0094] In this embodiment of the invention, the hypergraph neural network model can employ an efficient optimizer such as AdamW, combined with a learning rate scheduling strategy to accelerate convergence. The update rule can be expressed as:

[0095] in, These are the model parameters at time step 𝑡. It's the learning rate. It is the gradient of the loss function with respect to the parameters.

[0096] In a possible implementation, step 203 can be implemented as follows: dividing the hypergraph sample data into a training set, a validation set, and a test set; inputting the training set into a pre-built hypergraph neural network model for training, and updating the model parameters using a preset loss function during the training process; after each training round, adjusting the model parameters using the validation set to obtain the trained hypergraph neural network model, and evaluating the performance of the trained hypergraph neural network model using the test set; if the performance evaluation result of the trained hypergraph neural network model meets the preset conditions, then the trained hypergraph neural network model is used as the hypergraph neural network model for historical dose retrieval.

[0097] In this embodiment of the invention, the training set is used for forward and backward propagation, the model parameters are optimized using a preset loss function, the hyperparameters (such as learning rate and regularization strength) are evaluated and adjusted on the validation set, the model performance is evaluated using the test set, and training is stopped when the model performance meets the preset index or reaches the preset number of training rounds. The trained model is then used for historical dose retrieval.

[0098] To perform the corresponding steps in the above embodiments and various possible methods, an implementation of the environmental history dose retrieval device 200 is given below. Further, please refer to... Figure 5 , Figure 5 This is a functional block diagram of an environmental historical dose retrieval device 200 provided in an embodiment of the present invention. It should be noted that the environmental historical dose retrieval device 200 provided in this embodiment has the same basic principle and technical effects as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments. The environmental historical dose retrieval device 200 includes: The generation module 201 is used to parse and structure the user's input search request to generate a query vector.

[0099] The module 202 is used to input the query vector into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request; wherein, the hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data and dose data in the radiation environment.

[0100] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory shown is either stored in or embedded in the operating system (OS) of the electronic device, and can be used by... Figure 1The processor executes the commands. Meanwhile, the data and program code required to execute these modules can be stored in memory.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0102] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0103] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion 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 several instructions to cause a computer device (which may be a personal computer, electronic device 100, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An environmental history dose retrieval method, characterized by, The method includes: The system parses and structures the user's search request to generate a query vector. The query vector is input into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request; wherein, the hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data and dose data in the radiation environment.

2. The environmental history dose retrieval method of claim 1, wherein, The method includes the step of training a hypergraph neural network model, wherein the step of training the hypergraph neural network model includes: Acquire call audio data, monitoring video data, and dose monitoring data collected in a radiation environment; The call audio data, monitoring video data, and dose monitoring data are preprocessed, and the preprocessed call audio data, monitoring video data, and dose monitoring data are used to construct hypergraph sample data. Using the hypergraph sample data and a preset loss function for historical dose retrieval, a pre-constructed hypergraph neural network model is trained to obtain a hypergraph neural network model for historical dose retrieval.

3. The environmental historical dose retrieval method as described in claim 2, characterized in that, The preprocessing includes a first preprocessing, a second preprocessing, and a third preprocessing. The first preprocessing includes format conversion, noise reduction, speech recognition, semantic analysis, and information structuring. The second preprocessing includes format conversion, frame extraction synchronization, image enhancement, target detection, and feature structuring. The third preprocessing includes data cleaning, filtering, and data structuring. The preprocessing of the call audio data, monitoring video data, and dose monitoring data includes: The call audio data is subjected to the first preprocessing to obtain preprocessed call audio data; The surveillance video data is subjected to the second preprocessing to obtain preprocessed surveillance video data; The dose monitoring data is subjected to the third preprocessing to obtain preprocessed dose monitoring data.

4. The environmental historical dose retrieval method as described in claim 3, characterized in that, The steps for constructing hypergraph sample data using preprocessed call audio data, preprocessed monitoring video data, and preprocessed dose monitoring data include: Each data point in the preprocessed call audio data, preprocessed monitoring video data, and preprocessed dose monitoring data is treated as a hypergraph node, resulting in multiple hypergraph nodes. Hyperedges are constructed for multiple hypergraph nodes, and weights are assigned to each hyperedge based on the node relevance parameters of each constructed hyperedge to obtain the hypergraph sample data.

5. The environmental historical dose retrieval method as described in claim 4, characterized in that, The step of constructing hyperedges for multiple hypergraph nodes includes: Connect hypergraph nodes that are in the same time window, located in the same spatial region, and have semantic relevance to the same hyperedge.

6. The historical dose retrieval method as described in claim 4, characterized in that, The node relevance parameters of the hyperedge include the time interval, spatial distance, and semantic similarity between nodes of each hypergraph on the hyperedge. The step of assigning weights to each hyperedge based on the node relevance parameters of each constructed hyperedge includes: For each hyperedge, the hyperedge weights are calculated based on the time interval, spatial distance, and semantic similarity between the hypergraph nodes on the hyperedge. The weight of the hyperedge is determined based on the weight of the hyperedge between each hypergraph node on the hyperedge.

7. The environmental historical dose retrieval method as described in claim 2, characterized in that, The step of training a pre-constructed hypergraph neural network model using the hypergraph sample data and a preset loss function for historical dose retrieval to obtain a hypergraph neural network model for historical dose retrieval includes: The hypergraph sample data is divided into a training set, a validation set, and a test set; The training set is input into a pre-constructed hypergraph neural network model for training, and the model parameters are updated using the preset loss function during the training process. After each training round, the model parameters are adjusted using the validation set to obtain the trained hypergraph neural network model, and the performance of the trained hypergraph neural network model is evaluated using the test set. If the performance evaluation results of the trained hypergraph neural network model meet the preset conditions, the trained hypergraph neural network model will be used as the hypergraph neural network model for historical dose retrieval.

8. An environmental historical dose retrieval device, characterized in that, The device includes: The generation module is used to parse and structure the user's input search request to generate a query vector; The module is used to input the query vector into a pre-trained hypergraph neural network model to obtain historical dose records corresponding to the retrieval request; wherein, the hypergraph neural network model is trained based on hypergraph sample data, which is modeled from speech data, video data and dose data in the radiation environment.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor to implement the environmental historical dose retrieval method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the environmental historical dose retrieval method as described in any one of claims 1-7.

Images