Security and protection monitoring method and device for nuclear power station

Abstract

The invention provides a nuclear power station security monitoring method and device, and the method comprises the steps: monitoring a nuclear power station, so as to obtain the multi-modal data of a moving object; performing first analysis on the multi-modal data to determine whether the behavior of the moving object is an abnormal behavior; under the condition that the behavior of the moving object is determined to be an abnormal behavior, performing second analysis on the multi-modal data, and determining a danger level of the behavior of the moving object; and performing safety protection on the nuclear power station according to the danger level. On one hand, safety protection of the nuclear power station can be automatically and intelligently achieved, the reaction capacity and accuracy of safety protection of the nuclear power station are improved, and human errors and response delay are reduced; and on the other hand, safety protection can be more targeted and hierarchical, so that resources are reasonably allocated, and safe and stable operation of the nuclear power station is effectively maintained.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a method and device for security monitoring of nuclear power plants. Background Technology

[0002] As a key pillar of the modern energy system, nuclear power plants bear the crucial mission of providing a continuous and stable power supply for social development. They support not only the daily needs of countless homes but also serve as the cornerstone for the normal operation of various sectors of society, including industrial production, transportation, and healthcare. Therefore, the sensitivity and safety requirements of their operating environment are extremely stringent. Within the complex systems of a nuclear power plant, even the smallest potential threat can trigger a domino effect, leading to a series of chain reactions and causing incalculable casualties, environmental damage, and economic losses.

[0003] Therefore, efficient, accurate, and comprehensive security monitoring of nuclear power plants to promptly detect and eliminate these potential threats is crucial and urgent. Consequently, improving the intelligence level of monitoring systems is particularly pressing to effectively prevent these potential threats. Intelligent monitoring systems can instantly identify abnormal behavior, leaving no potential safety hazard unchecked, and can respond quickly and take timely countermeasures to minimize losses. This has undoubtedly become a critical and urgent issue to be addressed in the safety management of nuclear power plants. Summary of the Invention

[0004] This disclosure aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, the first objective of this disclosure is to propose a method and device for security monitoring of nuclear power plants. By first performing a first analysis on the acquired multimodal data, abnormal behavior can be quickly screened out, preventing the spread of potential risks and buying time for subsequent processing. After determining the abnormal behavior of the moving object, a second analysis is conducted to determine the hazard level, and corresponding protective measures are taken based on the hazard level. On the one hand, it can automatically and intelligently achieve security protection for nuclear power plants, improve the responsiveness and accuracy of nuclear power plant security protection, and reduce human error and response delays. On the other hand, it can make security protection more targeted and hierarchical, so as to rationally allocate resources and effectively maintain the safe and stable operation of nuclear power plants.

[0006] The second objective of this disclosure is to propose a security monitoring device for nuclear power plants.

[0007] The third objective of this disclosure is to propose an electronic device.

[0008] The fourth objective of this disclosure is to provide a non-transitory computer-readable storage medium storing computer instructions.

[0009] To achieve the above objectives, the first aspect of this disclosure provides a method for security monitoring of a nuclear power plant, the method comprising:

[0010] Monitoring nuclear power plants to obtain multimodal data on moving objects;

[0011] The multimodal data is first analyzed to determine whether the behavior of the moving object is abnormal.

[0012] If the behavior of the moving object is determined to be abnormal, a second analysis is performed on the multimodal data to determine the danger level of the moving object's behavior;

[0013] Based on the stated hazard level, safety measures are implemented for the nuclear power plant.

[0014] To achieve the above objectives, a second aspect of this disclosure provides a security monitoring device for a nuclear power plant, the device comprising:

[0015] The monitoring module is used to monitor the nuclear power plant to acquire multimodal data of moving objects;

[0016] The first analysis module is used to perform a first analysis on the multimodal data to determine whether the behavior of the moving object is abnormal.

[0017] The second analysis module is used to perform a second analysis on the multimodal data when it is determined that the behavior of the moving object is abnormal, and to determine the danger level of the behavior of the moving object.

[0018] The first protection module is used to provide safety protection for the nuclear power plant according to the hazard level.

[0019] To achieve the above objectives, a third aspect of this disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0020] To achieve the above objectives, a fourth aspect of this disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method described in the first aspect.

[0021] The nuclear power plant security monitoring method and apparatus provided in this disclosure monitor the nuclear power plant to acquire multimodal data of moving objects; perform a first analysis on the multimodal data to determine whether the behavior of the moving objects is abnormal; if the behavior of the moving objects is determined to be abnormal, perform a second analysis on the multimodal data to determine the hazard level of the behavior of the moving objects; and implement safety protection measures for the nuclear power plant based on the hazard level. Therefore, by first performing a first analysis on the acquired multimodal data, abnormal behavior can be quickly screened out, preventing the spread of potential risks and buying time for subsequent processing. After determining that the behavior of the moving objects is abnormal, the second analysis is conducted to determine the hazard level, and corresponding protective measures are taken based on the hazard level. On the one hand, this can automatically and intelligently achieve safety protection for the nuclear power plant, improving the responsiveness and accuracy of nuclear power plant safety protection and reducing human error and response delays. On the other hand, it can make safety protection more targeted and hierarchical, facilitating the rational allocation of resources and effectively maintaining the safe and stable operation of the nuclear power plant.

[0022] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0023] The above and / or additional aspects and advantages of this disclosure will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, in which:

[0024] Figure 1 This is a flowchart illustrating a nuclear power plant security monitoring method provided in an embodiment of this disclosure;

[0025] Figure 2 This is a flowchart illustrating another nuclear power plant security monitoring method provided in an embodiment of this disclosure;

[0026] Figure 3 This is a flowchart illustrating another nuclear power plant security monitoring method provided in an embodiment of this disclosure;

[0027] Figure 4 This is a schematic diagram of the structure of a nuclear power plant security monitoring device provided in an embodiment of this disclosure. Detailed Implementation

[0028] Embodiments of this disclosure are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0029] It should be noted that the acquisition, storage, use, and processing of data in this disclosed technical solution comply with the relevant provisions of laws and regulations.

[0030] Currently, in the security system of nuclear power plants, monitoring mainly relies on a combination of fixed cameras and manual surveillance. However, this traditional model has significant drawbacks. It not only has blind spots, making it difficult to comprehensively cover all areas of the nuclear power plant, but also suffers from response delays in the face of emergencies, hindering timely and effective responses. Furthermore, given the extremely high sensitivity and safety requirements of nuclear power plants, any potential threat, even the smallest, can trigger a series of serious consequences, causing incalculable losses. Therefore, improving the intelligence level of the monitoring system to achieve immediate identification and rapid response to abnormal behavior has become a crucial issue in the field of nuclear power plant safety management.

[0031] At the technical level, existing nuclear power plant security systems often face difficulties in processing massive amounts of real-time data, struggling to efficiently and accurately identify abnormal behavior. Furthermore, even when anomalies are detected, their response speed and automation levels are far from adequate, which not only limits the efficiency of security work but also weakens the system's ability to respond to emergencies.

[0032] In response to at least one of the above-mentioned problems, this disclosure proposes a method and device for security monitoring of nuclear power plants.

[0033] The following description, with reference to the accompanying drawings, describes a nuclear power plant security monitoring method and apparatus according to embodiments of the present disclosure.

[0034] Figure 1 This is a flowchart illustrating a nuclear power plant security monitoring method provided in an embodiment of this disclosure.

[0035] This disclosure illustrates the example of configuring the nuclear power plant security monitoring method in a nuclear power plant security monitoring device. The charging device can be applied to any electronic device so that the electronic device can perform the nuclear power plant security monitoring method function.

[0036] Among them, electronic devices can be any device with computing capabilities, such as personal computers, mobile terminals, servers (or cloud computing), etc. Mobile terminals can be hardware devices with various operating systems, touch screens and / or displays, such as in-vehicle devices, mobile phones, tablets, personal digital assistants, wearable devices, etc.

[0037] like Figure 1 As shown, the security monitoring method for this nuclear power plant may include the following steps:

[0038] Step 101: Monitor the nuclear power plant to obtain multimodal data of moving objects.

[0039] The term "moving object" can be used to indicate an object whose location has changed. It should be noted that this disclosure does not limit the object category of the moving object; for example, the object category of the moving object can be a person, an animal, a vehicle, etc.

[0040] Multimodal data may include image data (or video data), audio data (or sound data), etc., and this disclosure does not limit it.

[0041] As an example, multimodal sensors installed at nuclear power plants can be used to monitor the plant and obtain multimodal data on moving objects.

[0042] Among them, multimodal sensors may include, but are not limited to: cameras, thermal imagers, sound sensors, etc.

[0043] Step 102: Perform a first analysis on the multimodal data to determine whether the behavior of the moving object is anomalous.

[0044] As an example, behavioral pattern recognition technology can be used based on multimodal data to detect the behavior of moving objects and determine whether their behavior is abnormal. In other words, behavioral pattern recognition technology is used to identify the behavior of moving objects based on their multimodal data to determine whether their behavior is abnormal.

[0045] Optionally, in some embodiments, the following steps can be used to determine whether the behavior of the moving object is abnormal:

[0046] Step 1021: Perform second feature extraction on the multimodal data to obtain the target features.

[0047] The target features may include time-series data features, image features, location information, action patterns, action frequency, action intensity, etc.

[0048] As an example, a feature extraction model can be used to extract features from any modality of multimodal data, obtain the corresponding features, and then concatenate the features corresponding to each modality to obtain the target feature.

[0049] As another example, features can be extracted from any modality of the multimodal data to obtain the corresponding features; multiple first features can be selected from the features corresponding to the multimodal data, and the multiple first features can be fused to obtain the target features.

[0050] In the process of selecting multiple first features from the features corresponding to multimodal data, filtering methods (such as variance thresholding, correlation analysis, etc.) and packaging methods (such as recursive feature elimination) can be used to select multiple first features from the features corresponding to multimodal data. This is to select the most relevant and discriminative feature subset from the features corresponding to multimodal data, so as to reduce the amount of data processing, reduce the complexity of data processing, and reduce computational overhead in subsequent data processing, and improve the accuracy and effectiveness of subsequent behavior category recognition.

[0051] Step 1022: Use a target classifier to determine the behavior category of the moving object based on the target features.

[0052] The target classifier can be, but is not limited to, Support Vector Machine (SVM), neural network, decision tree, etc., and this disclosure does not impose any restrictions on it.

[0053] Among them, behavior category can be used to indicate whether the behavior of the moving object is abnormal.

[0054] In one example, target features can be input into a target classifier, and the behavior category of the moving object can be determined based on the output of the target classifier.

[0055] Step 1023: Determine whether the behavior of the moving object is abnormal based on the behavior category.

[0056] In one example, the behavior of a moving object is determined to be abnormal if the behavior category indicates that the moving object's behavior is abnormal, and vice versa.

[0057] Optionally, in some embodiments, the training process of the target classifier may include the following steps: obtaining a training dataset, wherein the training dataset includes sample behavior features of sample behaviors and corresponding labeled behavior categories; using the target classifier to determine the predicted behavior category of the sample behaviors based on the sample behavior features of the sample behaviors; and training the target classifier according to the difference between the labeled behavior category and the predicted behavior category.

[0058] It should be noted that this disclosure does not limit the number of sample behavioral features in the training dataset.

[0059] Among them, the labeled behavior category can be used to indicate whether the sample behavior is abnormal.

[0060] It is understandable that when there is a difference between the labeled behavior category and the predicted behavior category, it indicates that the accuracy of the target classifier is not high. In order to improve the accuracy and reliability of the target classifier and improve its recognition accuracy, the parameters of the target classifier can be adjusted. That is, in this disclosure, the parameters of the target classifier can be adjusted based on the difference between the labeled behavior category and the predicted behavior category.

[0061] To train the target classifier, one possible approach is to generate a loss value based on the difference between the labeled behavior category and the predicted behavior category. The loss value is positively correlated with the difference; that is, the smaller the difference, the smaller the loss value, and vice versa. Thus, in this disclosure, the parameters in the target classifier can be adjusted based on the loss value to minimize the loss value.

[0062] It should be noted that the above example only uses minimizing the loss value as the termination condition for training the target classifier. In actual applications, other termination conditions can also be set. For example, the termination condition can be that the number of training iterations reaches a set number, or the training duration reaches a set duration, etc. This disclosure does not impose any restrictions on this.

[0063] Therefore, by extracting the second feature from the multimodal data of moving objects to obtain target features, we can more accurately focus on key information, remove redundant interference, and improve the effectiveness of features. Furthermore, by using a target classifier to determine the behavior category of moving objects based on the target features, we can quickly and accurately classify behaviors by leveraging the powerful classification capabilities of the classifier. Finally, we can determine whether the behavior is abnormal based on the behavior category, thereby improving the accuracy and reliability of identifying the behavior of moving objects.

[0064] Optionally, in some embodiments, the multimodal data may be preprocessed before analysis. Preprocessing includes, but is not limited to, noise reduction, data normalization, format conversion, etc., and this disclosure does not impose any limitations on this.

[0065] Optionally, in some embodiments, after determining the behavior category of a moving object, the target features and behavior category of the moving object can be saved accordingly for subsequent data processing.

[0066] Step 103: If the behavior of the moving object is determined to be abnormal, a second analysis is performed on the multimodal data to determine the danger level of the moving object's behavior.

[0067] Optionally, in some embodiments, abnormal behavior features of at least one pre-set benchmark abnormal behavior are obtained; for any benchmark abnormal behavior, a scoring model is used to determine the behavior score of the moving object relative to the benchmark abnormal behavior based on the target features and the abnormal behavior features of the benchmark abnormal behavior; and the danger level of the moving object's behavior is determined according to the behavior score of the moving object relative to each benchmark abnormal behavior.

[0068] The abnormal behavior characteristics of the baseline abnormal behavior can be pre-set and can be used to determine the behavior of the moving object. It should be noted that this disclosure does not limit the number of baseline abnormal behaviors to which the pre-set abnormal behavior characteristics belong; there can be one or more.

[0069] In this embodiment of the disclosure, for any benchmark anomalous behavior, a scoring model can be used to determine the behavior score of the moving object relative to the benchmark anomalous behavior based on the target features and the anomalous behavior features of the benchmark anomalous behavior. In one example, the target features and the anomalous behavior features of the benchmark anomalous behavior can be input into the scoring model, thereby obtaining the behavior score of the moving object relative to the benchmark anomalous behavior in response to the output of the scoring model.

[0070] In this embodiment of the disclosure, the danger level of the moving object's behavior can be determined by evaluating the behavior of the moving object relative to various benchmark abnormal behaviors.

[0071] In one example, the behavior scores of a moving object relative to various baseline anomalous behaviors can be weighted and summed to obtain a comprehensive score. The risk level of the moving object's behavior can then be determined based on this comprehensive score. The comprehensive score can be positively correlated with the risk level; a higher comprehensive score indicates a higher risk level, and vice versa. Alternatively, a target value space can be determined from a pre-defined value space corresponding to risk levels, and the risk level corresponding to this target value space can be used to determine the risk level of the moving object's behavior.

[0072] In another example, for any baseline anomalous behavior, the danger level of the moving object's behavior under that baseline anomalous behavior is determined based on the behavior score of the moving object's behavior relative to that baseline anomalous behavior; and the danger level of the moving object's behavior is determined based on the danger level of the moving object's behavior under each baseline anomalous behavior.

[0073] Among these, the behavior score can be positively correlated with the risk level; the higher the behavior score, the higher the risk level, and vice versa. For example, assuming that the baseline abnormal behaviors include baseline abnormal behavior 1, baseline abnormal behavior 2, and baseline abnormal behavior 3, for any of these baseline abnormal behaviors, the risk level of the moving object's behavior under that baseline abnormal behavior is determined based on the behavior score of the moving object's behavior relative to that baseline abnormal behavior. For example, if the risk level of the moving object's behavior under baseline abnormal behavior 1 is level 1, the risk level of the moving object's behavior under baseline abnormal behavior 2 is level 3, and the risk level of the moving object's behavior under baseline abnormal behavior 3 is level 1, then the risk levels of the moving object's behavior under baseline abnormal behavior 1, under baseline abnormal behavior 2, and under baseline abnormal behavior 3 can be used as the risk level of the moving object's behavior. Alternatively, the highest risk level among the above-mentioned baseline abnormal behaviors can be determined as the risk level of the moving object's behavior.

[0074] Therefore, on the one hand, by pre-setting benchmark abnormal behaviors and their characteristics, a clear and unified standard is provided for the entire assessment process, making the behavioral scoring assessment more objective and accurate, and avoiding biases caused by subjective judgment. On the other hand, using a scoring model to quantitatively analyze the target characteristics and benchmark abnormal behavior characteristics can accurately measure the similarity between the moving object's behavior and various benchmark abnormal behaviors, thereby reasonably determining the behavioral score. Ultimately, these scores can be used to effectively determine the risk level.

[0075] Step 104: Implement safety measures for the nuclear power plant based on the hazard level.

[0076] The nuclear power plant security monitoring method of this disclosure monitors the nuclear power plant to acquire multimodal data of moving objects; performs a first analysis on the multimodal data to determine whether the behavior of the moving objects is abnormal; if the behavior of the moving objects is determined to be abnormal, performs a second analysis on the multimodal data to determine the hazard level of the behavior; and implements safety protection measures for the nuclear power plant based on the hazard level. Therefore, by first performing a first analysis on the acquired multimodal data, abnormal behavior can be quickly screened out, preventing the spread of potential risks and buying time for subsequent processing. After determining that the behavior of the moving objects is abnormal, the second analysis is conducted to determine the hazard level, and corresponding protective measures are taken based on the hazard level. On the one hand, this can automatically and intelligently achieve safety protection for the nuclear power plant, improving the responsiveness and accuracy of nuclear power plant safety protection and reducing human error and response delays. On the other hand, it can make safety protection more targeted and hierarchical, facilitating the rational allocation of resources and effectively maintaining the safe and stable operation of the nuclear power plant.

[0077] This disclosure provides another method for security monitoring in nuclear power plants. Figure 2 This is a flowchart illustrating another nuclear power plant security monitoring method provided in an embodiment of this disclosure.

[0078] like Figure 2 As shown, the security monitoring method for this nuclear power plant may include the following steps:

[0079] Step 201: Monitor the nuclear power plant and obtain video data.

[0080] Optionally, in some embodiments, the video data may include multiple consecutive image frames to be identified.

[0081] In one example, a nuclear power plant can be monitored using cameras to obtain video data.

[0082] Step 202: Perform object recognition on the video data to identify moving objects.

[0083] Optionally, in some embodiments, when the video data includes multiple consecutive image frames to be identified, a first feature extraction is performed on each image frame to be identified to obtain corresponding image features; for any image frame to be identified, based on the image features of the image frame to be identified and the image features of historical image frames associated with the image frame to be identified in the video data, it is determined whether there is a moving object in the image frame to be identified; if it is determined that there is a moving object in the image frame to be identified, the moving object in the image frame to be identified is determined.

[0084] Image features may include, but are not limited to: color histogram, image texture, image edges, shape, etc.

[0085] In order to determine the historical image frames associated with the image frame to be identified in the video data, in one example, for any image frame to be identified, the previous image frame in the video data can be used as the historical image frame of the image frame to be identified.

[0086] In another example, for any image frame to be identified, a set number of image frames in the video data that precede the image frame to be identified in time order can be used as historical image frames of the image frame to be identified. The set number of frames can be preset, such as 20 frames, 50 frames, etc., and this disclosure does not limit this.

[0087] As one possible implementation, when the historical image frame is a single frame, for any image frame to be identified, the inter-frame distance between the image frame to be identified and the historical image frame is determined based on the image features of the image frame to be identified and the image features of the historical image frames associated with the image frame to be identified; based on the inter-frame distance, it is determined whether there is a moving object in the image frame to be identified.

[0088] Inter-frame distance can be used to indicate the degree of difference between the image frame to be identified and historical image frames. Optionally, the inter-frame distance can be positively correlated with the degree of difference, that is, the larger the value of the inter-frame distance, the higher the degree of difference, and vice versa.

[0089] Optionally, in some embodiments, a distance metric algorithm may be used to determine the inter-frame distance between the image frame to be identified and the historical image frames associated with the image frame to be identified, based on the image features of the image frame to be identified and the image features of the historical image frames associated with the image frame to be identified.

[0090] The distance metric algorithm can be, for example, Euclidean distance algorithm, cosine similarity algorithm, etc., and this disclosure does not limit it.

[0091] In this embodiment of the disclosure, the presence of a moving object in the image frame to be identified can be determined based on the inter-frame distance. In one example, when the inter-frame distance is greater than a first preset distance threshold, it indicates that a new object has appeared in the image frame to be identified, or that an object is moving. Therefore, it can be determined that a moving object exists in the image frame to be identified, i.e., the image to be identified displays a moving object. The first preset distance threshold can be pre-set, and this disclosure does not limit its value.

[0092] As another possible implementation, when there are multiple historical image frames, for any image frame to be identified, the inter-frame distance between the image frame to be identified and the corresponding historical image frame is determined based on the image features of the image frame to be identified and the image features of any historical image frame associated with the image frame to be identified; based on each inter-frame distance, it is determined whether there is a moving object in the image frame to be identified.

[0093] In this embodiment of the disclosure, the presence of a moving object in the image frame to be identified can be determined based on the inter-frame distances. In one example, a target inter-frame distance is determined based on the inter-frame distances; and the presence of a moving object in the image frame to be identified is determined based on the target inter-frame distance.

[0094] The target inter-frame distance can be the average, mode, maximum, median, etc. of the inter-frame distances, and this disclosure does not impose any restrictions on it.

[0095] For example, when the distance between target frames is greater than the second set distance threshold, it indicates that a new object has appeared in the image frame to be identified or that the object is moving, and it can be determined that there is a moving object in the image frame to be identified. The second set distance threshold can be preset, and this disclosure does not limit its value.

[0096] Therefore, it is possible to automatically identify whether there is a moving object in the image frame to be identified based on the inter-frame distance between the image frame to be identified and the historical image frames, thereby improving the accuracy and effectiveness of the identification.

[0097] In this embodiment of the disclosure, if it is determined that a moving object exists in the image frame to be identified, the moving object in the image frame to be identified can be determined. In one example, object detection can be performed on the image frame to be identified to obtain an object detection box. Based on the detection box, the object displayed in the detection box area of ​​the image frame to be identified can be extracted, and the object can be identified as a moving object in the image to be identified.

[0098] Therefore, based on the image features of consecutive image frames, it is possible to prioritize the determination of whether there is a moving object in the image frame. Then, only when a moving object is present can the moving object in the image frame be identified, which improves the accuracy and reliability of moving object recognition, reduces the complexity of moving object recognition, and effectively reduces computational overhead.

[0099] Step 203: Acquire multimodal data using a multimodal sensor that matches the location information of the moving object.

[0100] The location information may include, but is not limited to: the coordinates and size of the object detection box that displays the moving object in the corresponding image frame to be identified, the position information of the object detection box relative to other reference points (such as cameras) in the image frame to be identified, the actual geographic spatial location information of the moving object, etc.

[0101] In one example, assuming there are multiple image frames to be identified, each displaying a moving object, the coordinates and size of the object detection box in each image frame are analyzed in chronological order to determine the direction of movement of the moving object. For example, the coordinates of the object detection box can move gradually from the right side of the image to the left side in chronological order, and the size of the object detection box can increase and then decrease, indicating that the moving object gradually approaches the camera from the right side and then moves away from the camera from the left side. At this time, cameras within a set distance (e.g., 100 meters, 150 meters, etc.) in front of the left side of the camera can be used to monitor the moving object.

[0102] Step 204: Perform a first analysis on the multimodal data to determine whether the behavior of the moving object is anomalous.

[0103] Step 205: If the behavior of the moving object is determined to be abnormal, a second analysis is performed on the multimodal data to determine the danger level of the moving object's behavior.

[0104] Step 206: Implement safety measures for the nuclear power plant based on the hazard level.

[0105] It should be noted that the execution process of steps 204 to 206 can refer to the execution process of any embodiment of this disclosure, and will not be described in detail here.

[0106] Optionally, in some embodiments, security strategies can be determined and adopted based on the hazard level to protect the nuclear power plant.

[0107] Therefore, based on the hazard level, corresponding security strategies are adopted to protect nuclear power plants, thereby improving the pertinence and effectiveness of nuclear power plant safety protection and avoiding unreasonable waste of resources in the process of safety protection.

[0108] Optionally, in some embodiments, the security strategy may be: acquiring image data of a moving object; authenticating the moving object based on the image data to obtain a verification result; wherein the verification result is used to indicate whether the moving object is an authorized object; and in response to the verification result indicating that the moving object is not an authorized object, employing an isolation mechanism to provide security protection for the nuclear power plant.

[0109] To acquire image data of a moving object, one example is to call a target sensor (such as a camera) that matches the location information of the moving object, and then collect image data of the moving object through the target sensor.

[0110] To implement the identity verification process for mobile objects, optionally, in some embodiments, facial features of the mobile object in the image data can be extracted to obtain the facial features of the mobile object; for any authorized person's facial features stored in the database, the similarity between the facial features of the mobile object and the facial features of the authorized person is determined, wherein the facial features of the authorized person are the facial features of the corresponding authorized person; based on the similarity between the facial features of the mobile object and the facial features of each authorized person, it is determined whether the mobile object is an authorized person.

[0111] Similarity can be used to measure the degree of similarity between the face of a moving object and the face of an authorized person. In one example, a distance metric algorithm can be used to determine the similarity between the facial features of the moving object and the facial features of the authorized person.

[0112] In this embodiment of the disclosure, whether a moving object is an authorized person can be determined based on the similarity between the facial features of the moving object and the facial features of each authorized person. In one example, when there is a target similarity greater than a first preset threshold among the similarities, the moving object is determined to be an authorized person; when there is no target similarity greater than the first preset threshold among the similarities, the moving object is determined not to be an authorized person.

[0113] The first set threshold can be preset, and this disclosure does not restrict its value.

[0114] In this embodiment of the disclosure, when the verification result indicates that the moving object is not an authorized object, an isolation mechanism can be adopted to protect the nuclear power plant. For example, upon confirming that the moving object is not an authorized object, i.e., an unauthorized person, a physical isolation device (such as an electric gate, telescopic fence, etc.) within a set distance (e.g., 200 meters) of the target sensor can be immediately activated, and monitoring equipment within the isolation zone can be activated simultaneously to closely monitor the unauthorized person's movements. The nearest safety response team will be automatically notified, providing the unauthorized person's exact location and appearance description, maintaining the isolation status until the safety response team arrives and takes over the situation.

[0115] Therefore, identity verification through image data can quickly and accurately identify whether a moving object is an authorized person, providing the first solid line of defense for nuclear power plant safety. When the verification result indicates an unauthorized object, timely isolation mechanisms can effectively prevent it from further entering critical areas of the nuclear power plant, avoiding potential safety threats and ensuring the safety and stability of nuclear power plant personnel, facilities, and the environment.

[0116] The nuclear power plant security monitoring method of this disclosure involves monitoring the nuclear power plant to obtain video data; performing object recognition on the video data to identify moving objects; and acquiring multimodal data using a multimodal sensor that matches the location information of the moving objects. This allows for the priority identification of moving objects, followed by the determination of their multimodal data, improving the accuracy and effectiveness of the acquired multimodal data for subsequent data processing.

[0117] This disclosure provides another method for security monitoring in nuclear power plants. Figure 3 This is a flowchart illustrating another nuclear power plant security monitoring method provided in an embodiment of this disclosure.

[0118] like Figure 3 As shown, based on any of the above embodiments of this disclosure, the nuclear power plant security monitoring method may further include the following steps:

[0119] Step 301: Obtain historical data on historical abnormal events.

[0120] Among them, historical abnormal events refer to unexpected events or behaviors that deviate from normal operation or safety standards during the past operation of nuclear power plants.

[0121] Historical data may include, but is not limited to, the abnormal behavior, occurrence time, duration, etc. of historical abnormal events, and this disclosure does not impose any restrictions on this.

[0122] In one example, historical data of historical anomalies can be saved in advance, so that in subsequent applications, historical data related to historical anomalies can be directly retrieved from the relevant database.

[0123] Step 302: Using a target prediction model based on historical data, the target behavior type and target probability are predicted.

[0124] Among them, the target behavior type can be used to indicate the corresponding abnormal behavior, and the target probability can be used to indicate the probability of the abnormal behavior corresponding to the target behavior type occurring within the future target time period.

[0125] The target duration can be preset, and this disclosure does not restrict its value, such as 1 hour, 2 hours, etc.

[0126] It should be noted that this disclosure does not limit the number of target behavior types predicted; it may be, but is not limited to, one.

[0127] The target prediction model can be, for example, a model based on deep learning technology such as LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), or Transformer architecture, etc., and this disclosure does not limit it.

[0128] As an example, historical data can be input into a target prediction model, thereby obtaining the target behavior type and the target probability corresponding to the target behavior type in response to the output of the target prediction model.

[0129] Step 303: Based on the target behavior type and target probability, implement safety protection for the nuclear power plant.

[0130] In one example, when the target probability corresponding to the target behavior type is greater than the second set threshold, it indicates that there is a high probability that abnormal behavior corresponding to the target behavior type will occur within the target time period in the future. Therefore, safety protection measures corresponding to the target behavior type can be taken to protect the nuclear power plant.

[0131] The second set threshold can be preset, and this disclosure does not restrict its value.

[0132] The nuclear power plant security monitoring method of this disclosure acquires historical data on abnormal events; uses a target prediction model based on the historical data to predict target behavior types and target probabilities; wherein, the target behavior type indicates the corresponding abnormal behavior, and the target probability indicates the likelihood of the abnormal behavior corresponding to the target behavior type occurring within a future target time period; and performs safety protection for the nuclear power plant based on the target behavior type and target probability. Thus, the data-predictive protection mode effectively reduces the probability of accidents, maintains the stable operation of the nuclear power plant, and improves the foresight and accuracy of nuclear power plant safety protection work.

[0133] To clearly illustrate the nuclear power plant security monitoring method disclosed herein, a detailed explanation is provided below with examples.

[0134] In one example, the nuclear power plant security monitoring method is applied to a nuclear power plant security monitoring system for illustration. The nuclear power plant security monitoring system includes a security subsystem, a monitoring subsystem, and a control center. The nuclear power plant security monitoring method may include the following steps:

[0135] Step 1: The security subsystem collects real-time environmental data inside and outside the nuclear power plant. Based on the collected data, it uses pre-trained pattern recognition technology to detect abnormal behavior.

[0136] Specifically, a distributed sensor network can capture environmental data inside and outside the nuclear power plant without omission. The environmental data includes visual data, sound data (or audio data), and radiation level data. The visual data (referred to as video data in this disclosure) is transmitted to a central processing unit via a high-speed network for real-time analysis. The analyzed images of personnel activities are compared frame by frame to identify unexpected moving objects (referred to as moving objects in this disclosure). Based on the location information of the identified unexpected moving objects, multimodal sensors in the vicinity are linked to enhance the monitoring accuracy of specific locations.

[0137] It should be noted that, in order to identify unexpectedly moving objects, the following steps can be used:

[0138] First, relevant video data is collected from the environment, including images of people's activities;

[0139] Secondly, the collected video data is preprocessed, such as denoising, normalization, and format conversion, to eliminate unnecessary interference factors.

[0140] Next, feature extraction is performed on any video frame in the video data (referred to as the image frame to be identified in this disclosure) to obtain the corresponding image features. For example, image features such as color histogram, texture, edge, and shape can be extracted.

[0141] Finally, the presence of unexpected moving objects is detected by comparing the differences between consecutive video frames. For each video frame, the difference between this video frame and the previous frame or a series of previous frames can be calculated, and based on the difference, it can be determined whether there is an unexpected moving object in this video frame. The unexpected moving object can be a newly appearing moving object or an existing object that has undergone unexpected movement.

[0142] It should be noted that, in order to determine the position information of an unexpectedly moving object, the following steps can be used:

[0143] After identifying an unexpected moving object, the nuclear power plant's security monitoring system first acquires and records the object's location information. This location information can include the object's coordinates and size within the video frame, as well as its position relative to other environmental reference points (such as cameras).

[0144] Secondly, based on the location information of unexpectedly moving objects, multimodal sensors (such as thermal imagers and sound sensors) in the nearby area that match the location information can be linked to enhance monitoring.

[0145] Finally, by adjusting the orientation and sensitivity of the multimodal sensor, the relative position of the multimodal sensor and the reference point (i.e., the camera that captures the unexpected moving object) is calculated, so that the multimodal sensor is focused on the area where the unexpected moving object is located, thereby providing more detailed monitoring data (referred to as multimodal data in this disclosure) through the multimodal sensor.

[0146] Therefore, by using a distributed sensor network, comprehensive data acquisition is achieved, ensuring that critical information such as visual, sound, and radiation levels inside and outside the nuclear power plant are closely monitored, thus improving the comprehensiveness and reliability of the monitoring system. High-speed network transmission of visual data to the central processing unit ensures real-time data analysis and continuous frame comparison, enabling timely identification of unexpected moving objects, shortening the time interval from data acquisition to abnormal behavior detection, and improving the system's response speed.

[0147] It should be noted that, in order to achieve the detection of abnormal behavior based on pattern recognition technology, the following steps can be adopted:

[0148] 1. Data Acquisition: First, collect relevant multimodal data of unexpected moving objects from the environment. The multimodal data may include images, audio, text, sensor readings, etc.

[0149] 2. Preprocessing: Preprocessing the multimodal data can include denoising, normalization, and format conversion to eliminate unnecessary interference and make the data more suitable for subsequent analysis.

[0150] 3. Feature Extraction: Feature extraction is performed on multimodal data to obtain features for any one modality. It should be noted that these features can reflect the inherent attributes of the data. Features can be time-series data features, image features, location information, action patterns, action frequency, action intensity, etc.

[0151] 4. Feature selection: Select the most relevant and discriminative feature subset from the features of multimodal data (the features in the feature subset are referred to as the first feature in this disclosure) to reduce computational complexity and improve recognition accuracy.

[0152] 5. Behavioral pattern (referred to as behavioral type in this disclosure) prediction: A target classifier is used to predict the behavioral patterns of unexpected moving objects based on a subset of features.

[0153] The target classifier, such as a support vector machine, neural network model, or decision tree, is used to classify data based on extracted features. It should be noted that the target classifier can be trained, tested, and validated before being used to predict the behavior patterns of unexpectedly moving objects.

[0154] The training of the target classifier involves using a labeled training dataset to adjust the parameters of the classifier so that it can correctly distinguish between different categories.

[0155] Testing and validation: Evaluate the performance of the target classifier on an independent test dataset to ensure its generalization ability.

[0156] Step 2: The security subsystem assesses the risk level of abnormal behavior;

[0157] When the behavior of an unexpectedly moving object is identified as anomalous, a pre-trained scoring model (e.g., a regression model or a neural network model) is used to quantify the urgency and potential harm of the anomalous behavior. This model can determine a behavior score of the unexpectedly moving object's anomalous behavior relative to the baseline anomalous behavior based on features from the aforementioned feature subset and the anomalous behavior features of a benchmark anomalous behavior in a multi-dimensional feature library, thereby assessing the severity of the unexpectedly moving object's anomalous behavior.

[0158] Based on the scoring results, the nuclear power plant security monitoring system automatically adjusts the response level of the monitoring subsystems within the system and prepares corresponding intervention measures. If the scoring indicates that the abnormal behavior is of high urgency or high hazard, the system will raise its response level and prepare to take more urgent response strategies.

[0159] Step 3: Security Subsystem Predicts Future Trends

[0160] 1. Employing deep learning techniques, such as LSTM, GRU, or Transformer architectures, models can predict the likelihood of future abnormal behavior by learning behavioral patterns from historical data of anomalous events at nuclear power plants. The prediction results can help provide early warnings and guide real-time decision-making and response.

[0161] 2. Employ time series analysis methods (such as ARIMA (Autoregressive Integrated Moving Average Model), LSTM, etc.) to predict future trends of anomalous behavior. Specifically, time series analysis methods can be used to predict the consequences of future anomalous behavior based on environmental and historical data from nuclear power plants.

[0162] It should be noted that in security systems, by predicting future trends, the output of these algorithms is used to guide real-time decision-making and response, helping control centers to take proactive measures to prevent or mitigate the impact of potential threats.

[0163] Step 4: Generate a real-time risk assessment report

[0164] The security subsystem integrates data from all abnormal events to generate a preliminary risk profile. Based on this profile, it adjusts the risk level to differentiate between emergency and non-emergency situations. Using the determined risk level, it generates a risk assessment report and transmits it to the control center in real time.

[0165] Step 4: Security subsystem response;

[0166] 1. Based on the risk report, the control center sends an alarm to the security subsystem through a preset communication protocol. Upon receiving the alarm, the security subsystem immediately adjusts the camera angle to lock onto the suspicious target.

[0167] The process of sending alarms to the security subsystem via a preset communication protocol further includes: the control center interpreting the assessment report, identifying alarm signals that require priority handling, preparing alarm instructions for confirmed alarm signals using pre-approved communication rules, issuing the prescribed instructions, activating the corresponding security subsystem components, and monitoring the security subsystem's response to ensure that the alarm is executed correctly and that communication remains uninterrupted.

[0168] Optionally, operators or automated systems at the control center can parse real-time risk reports, identify the highest priority alert signals, determine the nature of the alert signals (e.g., intrusion, fire, abnormal activity) and their geographic location information, and construct alarm commands using standardized communication protocols (e.g., OPC-UA, Modbus, TCP / IP). These alarm commands can include specific event identifiers, location coordinates, required response levels, and other key information. Finally, alarm commands can be sent to security subsystem components (e.g., video surveillance, access control, alarms). Simultaneously, the monitoring subsystem can monitor these components to ensure the alarm commands are executed correctly and maintain a continuous communication link to receive feedback.

[0169] The process of adjusting the camera angle to lock onto a suspicious target further includes: the security subsystem decoding the received alarm instructions, extracting the coordinate location information of the suspicious target (i.e., the moving object corresponding to the abnormal behavior), adjusting the camera to match the coordinate location information of the suspicious target to optimize the camera's shooting angle and obtain a clearer on-site image; and using high-resolution imaging technology to capture target details and keep the camera focused until the nature of the target is confirmed or a new instruction is received.

[0170] 2. The security subsystem verifies the identity of suspicious targets.

[0171] High-resolution image data of the locked target is input into the artificial intelligence analysis module for identity verification. If the verification result shows that the suspicious target is an unauthorized person, the isolation mechanism is automatically triggered.

[0172] The identity verification process further includes: extracting features from the suspicious target based on the high-resolution image data of the locked target to obtain suspicious features, which may include, but are not limited to, facial features, clothing color, and behavioral patterns; attempting to match the obtained facial features with the facial features of registered legitimate personnel (referred to as authorized personnel facial features in this disclosure); and assessing the identity status of the suspicious target based on the matching results. If no matching record is found, the suspicious target is identified as an unauthorized visitor.

[0173] The automatic isolation mechanism further includes: upon confirming that a suspicious target is an unauthorized visitor, immediately activating the surrounding physical isolation devices, simultaneously activating the monitoring equipment in the isolation area, closely monitoring the movement of the unauthorized personnel; automatically notifying the nearest security response team, providing the exact location and appearance description of the unauthorized personnel, maintaining the isolation status until the security response team arrives and takes over the handling.

[0174] Physical barriers, such as motorized gates and retractable fences, are deployed around critical areas of a nuclear power plant and can be quickly closed to prevent unauthorized personnel from taking further action.

[0175] Enhanced surveillance equipment: including additional cameras, thermal imagers, motion sensors, etc., to provide denser surveillance coverage within the isolated area and ensure continuous tracking of the movements of unauthorized personnel.

[0176] Communication and Notification System: Automatically notifies the security response team via preset communication protocols, providing real-time location and description to accelerate the deployment of the response team.

[0177] Maintaining Isolation: The isolation mechanism will remain active until the security team arrives to prevent unauthorized personnel from escaping or causing further damage.

[0178] Security Response Team: Well-trained security personnel equipped with the necessary tools and equipment, capable of responding quickly to notifications, taking over on-site handling, and ensuring that unauthorized personnel are effectively controlled.

[0179] Therefore, facial recognition using high-resolution images and deep learning technology can accurately distinguish between authorized and unauthorized individuals, even in complex environments, effectively reducing false alarm rates. Once an unauthorized visitor is confirmed, the system can quickly initiate isolation measures, reducing potential security threats and protecting the core area of ​​the nuclear power plant from intrusion. By activating additional monitoring equipment and notifying the security team, continuous monitoring of unauthorized personnel is ensured, while providing accurate intelligence to on-site personnel and accelerating response times. The seamless collaboration between the automated isolation mechanism and the human security team leverages the efficiency advantages of technology while ensuring necessary human judgment and intervention, thereby improving the overall level of safety management.

[0180] Understandably, once the isolation mechanism is activated, detailed information about abnormal events can be recorded and the security database updated to provide data support for subsequent analysis. This further includes:

[0181] Once the isolation mechanism is activated, the entire timeline, location, characteristics of unauthorized personnel involved, and response actions of the incident can be recorded. The collected information is integrated into the incident log to form a structured incident report, the security database is updated, and newly occurring incidents are added as reference cases for future analysis and prevention. The data analysis process is automatically triggered to perform pattern recognition and trend prediction on newly added incidents.

[0182] The system automatically records the entire event process, including the timeline, location, characteristics of unauthorized personnel, and detailed steps of the system response, ensuring the completeness and accuracy of the event records and facilitating subsequent event review and analysis. The security database is updated in real time, incorporating newly occurring events as case studies. This not only enriches the database's content but also enhances the system's ability to identify and handle similar events, promoting knowledge accumulation and intelligent evolution.

[0183] Automated data analysis processes can identify event patterns and potential trends. Through deep learning of historical events, the system can predict future safety risks, develop preventative measures in advance, and enhance the overall safety defense capabilities of the nuclear power plant. Based on the results of event analysis and trend prediction, safety strategies and response plans can be adjusted periodically or as needed to ensure that the nuclear power plant's security system keeps pace with the times and effectively responds to the ever-changing threat environment.

[0184] In summary, the nuclear power plant security monitoring method disclosed herein can collect real-time environmental data and personnel activity information from inside and outside the nuclear power plant, and utilize pre-trained pattern recognition technology to detect abnormal behavior. Once an anomaly is detected, the system can quickly activate a deep learning network to analyze the level of potential threats and generate a real-time risk assessment report, thereby achieving an automated alarm and response mechanism. This method can significantly improve the intelligent response capability and accuracy of the nuclear power plant security monitoring system, reduce human error and response delays, effectively prevent unauthorized personnel from entering sensitive areas, and ensure the safe operation of the nuclear power plant. Furthermore, the automatic activation of the isolation mechanism and the recording and updating of event information help to continuously optimize security strategies, providing more comprehensive security for the nuclear power plant.

[0185] To achieve the above embodiments, this disclosure also proposes a security monitoring device for nuclear power plants.

[0186] Figure 4 This is a schematic diagram of the structure of a nuclear power plant security monitoring device provided in an embodiment of this disclosure.

[0187] like Figure 4 As shown, the nuclear power plant security monitoring device 40 includes: a monitoring module 41, a first analysis module 42, a second analysis module 43, and a first protection module 44.

[0188] Among them, the monitoring module 41 is used to monitor the nuclear power plant to obtain multimodal data of moving objects.

[0189] The first analysis module 42 is used to perform a first analysis on the multimodal data to determine whether the behavior of the moving object is abnormal.

[0190] The second analysis module 43 is used to perform a second analysis on the multimodal data to determine the danger level of the moving object's behavior when the behavior of the moving object is determined to be abnormal.

[0191] The first protection module 44 is used to provide safety protection for the nuclear power plant according to the hazard level.

[0192] In one possible implementation of this disclosure, the monitoring module 41 is used to: monitor the nuclear power plant and obtain video data; perform object recognition on the video data to determine moving objects; and acquire multimodal data through a multimodal sensor that matches the location information of the moving objects.

[0193] In one possible implementation of this disclosure, the video data includes multiple consecutive image frames to be identified; the monitoring module 41 is configured to: perform a first feature extraction on each image frame to be identified to obtain corresponding image features; for any image frame to be identified, determine whether there is a moving object in the image frame to be identified based on the image features of the image frame to be identified and the image features of historical image frames associated with the image frame to be identified in the video data; and if it is determined that there is a moving object in the image frame to be identified, determine the moving object in the image frame to be identified.

[0194] In one possible implementation of this disclosure, the monitoring module 41 is configured to: for any image frame to be identified, determine the inter-frame distance between the image frame to be identified and the historical image frames associated with the image frame to be identified, based on the image features of the image frame to be identified and the image features of the historical image frames associated with the image frame to be identified; and determine whether there is a moving object in the image frame to be identified based on the inter-frame distance.

[0195] In one possible implementation of this disclosure, the first analysis module 42 is used to: extract second features from multimodal data to obtain target features; use a target classifier to determine the behavior category of the moving object based on the target features; and determine whether the behavior of the moving object is abnormal based on the behavior category.

[0196] In one possible implementation of this disclosure, the second analysis module 43 is configured to: acquire abnormal behavior features of at least one pre-set benchmark abnormal behavior; for any benchmark abnormal behavior, use a scoring model to determine the behavior score of the moving object relative to the benchmark abnormal behavior based on the target features and the abnormal behavior features of the benchmark abnormal behavior; and determine the danger level of the moving object's behavior based on the behavior scores of the moving object relative to each benchmark abnormal behavior.

[0197] In one possible implementation of this disclosure, the first protection module 44 is used to: determine a security strategy based on the hazard level; and adopt the security strategy to provide security protection for the nuclear power plant.

[0198] In one possible implementation of this disclosure, the first protection module 44 is configured to: acquire image data of a moving object; authenticate the moving object based on the image data to obtain a verification result; wherein the verification result is used to indicate whether the moving object is an authorized object; and in response to the verification result indicating that the moving object is not an authorized object, employ an isolation mechanism to provide security protection for the nuclear power plant.

[0199] In one possible implementation of this disclosure, the nuclear power plant security monitoring device 40 may further include:

[0200] The acquisition module is used to acquire historical data of historical abnormal events.

[0201] The prediction module is used to predict the target behavior type and the corresponding target probability based on historical data using a target prediction model. The target behavior type indicates the corresponding abnormal behavior, and the target probability indicates the likelihood of the abnormal behavior corresponding to the target behavior type occurring within the future target time period.

[0202] The second protection module is used to provide safety protection for nuclear power plants based on the target behavior type and target probability.

[0203] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and will not be repeated here.

[0204] The nuclear power plant security monitoring device of this embodiment monitors the nuclear power plant to acquire multimodal data of moving objects; performs a first analysis on the multimodal data to determine whether the behavior of the moving objects is abnormal; if the behavior of the moving objects is determined to be abnormal, performs a second analysis on the multimodal data to determine the hazard level of the behavior of the moving objects; and implements safety protection measures for the nuclear power plant based on the hazard level. Therefore, by first performing a first analysis on the acquired multimodal data, abnormal behavior can be quickly screened out, preventing the spread of potential risks and buying time for subsequent processing. After determining that the behavior of the moving objects is abnormal, the second analysis is conducted to determine the hazard level, and corresponding protective measures are taken based on the hazard level. On the one hand, it can automatically and intelligently achieve safety protection for the nuclear power plant, improving the responsiveness and accuracy of nuclear power plant safety protection, and reducing human error and response delays. On the other hand, it can make safety protection more targeted and hierarchical, so as to rationally allocate resources and effectively maintain the safe and stable operation of the nuclear power plant.

[0205] To implement the above embodiments, this disclosure also proposes an electronic device, comprising:

[0206] At least one processor; and a memory communicatively connected to said at least one processor; wherein,

[0207] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the aforementioned method.

[0208] To implement the above embodiments, this disclosure also proposes a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the aforementioned methods.

[0209] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0210] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0211] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of this disclosure includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of this disclosure pertain.

[0212] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0213] It should be understood that various parts of this disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0214] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0215] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in the form of a hardware module or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0216] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for security monitoring in a nuclear power plant, characterized in that, The method includes: Monitoring nuclear power plants to obtain multimodal data on moving objects; The multimodal data is first analyzed to determine whether the behavior of the moving object is abnormal. If the behavior of the moving object is determined to be abnormal, a second analysis is performed on the multimodal data to determine the danger level of the moving object's behavior; Based on the stated hazard level, safety measures are implemented for the nuclear power plant.

2. The method according to claim 1, characterized in that, The monitoring of the nuclear power plant to obtain multimodal data of moving objects includes: The nuclear power plant was monitored to obtain video data; Object recognition is performed on the video data to identify the moving object; The multimodal data is acquired using a multimodal sensor that matches the location information of the moving object.

3. The method according to claim 2, characterized in that, The video data includes multiple consecutive image frames to be identified; The step of performing object recognition on the video data to determine the moving object includes: Perform first feature extraction on each of the image frames to be identified to obtain the corresponding image features; For any of the image frames to be identified, based on the image features of the image frame to be identified and the image features of historical image frames associated with the image frame to be identified in the video data, it is determined whether the image frame to be identified contains a moving object. If it is determined that there is a moving object in the image frame to be identified, the moving object in the image frame to be identified is determined.

4. The method according to claim 3, characterized in that, For any given image frame to be identified, determining whether a moving object exists in the image frame to be identified, based on the image features of the image frame to be identified and the image features of historical image frames associated with the image frame to be identified in the video data, includes: For any of the image frames to be identified, the inter-frame distance between the image frame to be identified and the historical image frames associated with the image frame to be identified is determined based on the image features of the image frame to be identified and the image features of the historical image frames associated with the image frame to be identified. Based on the inter-frame distance, it is determined whether there is a moving object in the image frame to be identified.

5. The method according to claim 1, characterized in that, The first analysis of the multimodal data to determine whether the behavior of the moving object is abnormal includes: The target features are obtained by performing a second feature extraction on the multimodal data; A target classifier is used to determine the behavior category of the moving object based on the target features; Based on the behavior category, determine whether the behavior of the moving object is abnormal.

6. The method according to claim 5, characterized in that, The second analysis of the multimodal data to determine the danger level of the moving object's behavior includes: Obtain at least one pre-defined abnormal behavior characteristic based on abnormal behavior; For any of the aforementioned baseline abnormal behaviors, a scoring model is used to determine the behavior score of the moving object relative to the baseline abnormal behavior based on the target features and the abnormal behavior features of the baseline abnormal behavior. The danger level of the moving object's behavior is determined based on the behavior score of the moving object relative to each of the benchmark abnormal behaviors.

7. The method according to claim 1, characterized in that, The safety protection measures for the nuclear power plant based on the aforementioned hazard level include: Based on the stated hazard level, determine the security strategy; Security strategies are adopted to protect the nuclear power plant.

8. The method according to claim 7, characterized in that, The security strategy includes: Obtain the image data of the moving object; Based on the image data, the mobile object is authenticated to obtain an authentication result; wherein the authentication result is used to indicate whether the mobile object is an authorized object; In response to the verification result indicating that the moving object is not an authorized object, an isolation mechanism is adopted to provide security protection for the nuclear power plant.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: Retrieve historical data of historical anomalies; Based on the historical data, a target prediction model is used to predict the target behavior type and the corresponding target probability; wherein, the target behavior type is used to indicate the corresponding abnormal behavior, and the target probability is used to indicate the probability that the abnormal behavior corresponding to the target behavior type will occur within the future target time period; The nuclear power plant is protected based on the target behavior type and the target probability.

10. A security monitoring device for a nuclear power plant, characterized in that, The device includes: The monitoring module is used to monitor the nuclear power plant to acquire multimodal data of moving objects; The first analysis module is used to perform a first analysis on the multimodal data to determine whether the behavior of the moving object is abnormal. The second analysis module is used to perform a second analysis on the multimodal data when it is determined that the behavior of the moving object is abnormal, and to determine the danger level of the behavior of the moving object. The first protection module is used to provide safety protection for the nuclear power plant according to the hazard level.

Images