Transformer substation operation and maintenance real-time monitoring method, device, equipment and medium

Abstract

The invention relates to the field of substation monitoring, in particular to a substation operation and maintenance real-time monitoring method, device and equipment and a medium. A video stream collected by camera equipment of a transformer substation in real time is obtained; performing preliminary identification on the video stream according to an edge computing device, and if a preliminary identification result is abnormal, determining an object in the video stream and an object state of the object; according to a scene type of a monitoring area corresponding to the camera equipment, judging whether the object state meets a safety requirement of the scene type or not; if the object state does not meet the security requirement of the scene type, determining an alarm level corresponding to the object state; and giving an alarm according to the object state and the alarm level. A double-layer identification mechanism of objects and object states is adopted, the existence of entities is identified, and specific state attributes of the entities are analyzed, so that an identification result is deeply matched with power business safety requirements, and false alarms caused by disjunction of traditional visual identification and business logic are reduced.

Description

Technical Field

[0001] This invention relates to the field of substation monitoring, specifically to a method, device, equipment, and medium for real-time monitoring of substation operation and maintenance. Background Technology

[0002] Currently, safety management at substation operation and maintenance sites mainly relies on manual inspections and traditional security systems, which presents the following problems:

[0003] Manual inspection is inefficient: due to limitations in the number of personnel, workload, and inspection frequency, it is difficult to achieve 24 / 7 uninterrupted monitoring.

[0004] Security systems have limited functionality: traditional video surveillance systems only have image recording capabilities and lack intelligent analysis and real-time alarm capabilities;

[0005] Safety measures are not being implemented effectively: key safety factors such as fence status, sign hanging, and personnel behavior cannot be identified and warned of in real time;

[0006] The ability to fuse multi-source heterogeneous data is weak: it is difficult to achieve unified processing and comprehensive analysis of multi-modal data such as images, videos, infrared, and ultraviolet.

[0007] While there are some existing security systems based on computer vision, their recognition scenarios are limited, their accuracy is low, their response latency is high, and they lack deep integration with power business scenarios, thus failing to meet the high-intensity and high-reliability operation and maintenance monitoring needs of substations. Summary of the Invention

[0008] The purpose of this invention is to provide a method, device, equipment, and medium for real-time monitoring of substation operation and maintenance, which solves the problems of the prior art.

[0009] This invention is achieved through the following technical solution:

[0010] In a first aspect, embodiments of the present invention provide a method for real-time monitoring of substation operation and maintenance, including:

[0011] Acquire real-time video streams captured by camera equipment in the substation;

[0012] The edge computing device performs preliminary identification on the video stream. If the preliminary identification result is abnormal, the objects in the video stream and the object status of the objects are determined.

[0013] Based on the scene type of the monitoring area corresponding to the camera device, determine whether the object status meets the security requirements of the scene type;

[0014] If the object state does not meet the security requirements of the scenario type, then determine the alarm level corresponding to the object state;

[0015] An alarm is triggered based on the object's status and the alarm level.

[0016] Preferably, the step of performing preliminary identification of the video stream based on the edge computing device, and determining the objects in the video stream and the object status of the objects if the preliminary identification result is abnormal, includes:

[0017] Frames are extracted from the video stream, and the extracted image frames are preprocessed.

[0018] A lightweight vision model deployed at the edge is used to infer the preprocessed image frames and output preliminary recognition results containing the target category and initial confidence.

[0019] Based on the preliminary identification results, if the initial confidence level is lower than a first threshold but higher than a second threshold, then the objects in the video stream and the object states of the objects are determined, wherein the first threshold is greater than the second threshold.

[0020] Preferably, determining the objects in the video stream and the object states of the objects includes:

[0021] Extract a continuous image sequence within a certain time period before and after the image frame from the video stream;

[0022] Object recognition is performed based on the continuous image sequence to obtain objects, including people, equipment, animals, and signs;

[0023] For each object, the object state is determined. The object state of a person includes equipment wearing state, behavior state, and location state. The object state of equipment includes switch state, indicator light state, and meter pointer state. The object state of an animal includes animal behavior state and animal location state. The object state of a sign includes sign location state and sign content state.

[0024] Preferably, the step of performing object recognition based on the continuous image sequence to obtain the object includes:

[0025] For each frame of the continuous image sequence, convolutional feature extraction is performed to obtain the depth feature map of each frame.

[0026] The deep feature maps of multiple frames in the sequence are concatenated along the channel dimension or fused using temporal pooling to generate sequence-enhanced feature maps;

[0027] The sequence-enhanced feature map is input into the target detection network, and the target bounding box and the category confidence of the target bounding box are output.

[0028] Based on the continuous image sequence, the targets within the target bounding box are tracked and associated to form multiple short-term target trajectory segments;

[0029] Detection results that exceed the fluctuation threshold or appear less frequently than the frequency threshold in the short-term target trajectory segment are removed, resulting in multiple unknown objects;

[0030] For each unknown object, the features of the unknown object in each frame of the continuous image sequence are aggregated to obtain the temporal aggregated features of the unknown object;

[0031] The temporal aggregation features are input into a preset object classifier to obtain objects including specific categories.

[0032] Preferably, the step of inputting the temporal aggregation features into a preset object classifier to obtain objects including specific categories includes:

[0033] Determine whether the object is classified as a person or an animal; when the object is classified as a person or an animal and is located in a key monitoring area of ​​the substation, acquire thermal imaging data that is spatiotemporally aligned with the video stream; extract thermal feature information based on the thermal imaging data, and verify or correct the object's category based on the thermal feature information; wherein, the key monitoring area includes the perimeter of the perimeter wall, the high-voltage equipment area, the main transformer area, and the cable layer.

[0034] Preferably, determining the object state for each object includes:

[0035] For each identified object, in each frame of the continuous image sequence, visual evidence related to the state of the object is extracted, the visual evidence including at least one of the object's pixel region, appearance features, relative spatial location, and text content;

[0036] The object state is obtained based on the fusion results of multi-frame visual evidence and predefined state determination rules. The state determination rules include at least one of multi-frame voting, frequency statistics, persistence judgment, and optimal frame selection.

[0037] Preferably, determining whether the object's state meets the security requirements of the scene type based on the scene type of the monitoring area corresponding to the camera device includes:

[0038] Based on the unique identifier of the camera device, a preset configuration database is queried to obtain the scene type corresponding to the area monitored by the camera device. The scene type includes high-voltage equipment area, main transformer area, switchyard, capacitor shed, perimeter wall, main control room and cable layer.

[0039] Based on the scenario type, a pre-set scenario type and security rule mapping table is retrieved to obtain a set of mandatory security states and a set of prohibited states bound to the scenario type.

[0040] The object state is logically compared with the retrieved mandatory security state set and prohibited state set. If the object state belongs to the prohibited state set, it is determined that it does not meet the security requirements.

[0041] If the object is a person, and the object's state does not fully cover the states required by the mandatory safety state set, then it is determined that it does not meet the safety requirements.

[0042] Preferably, the step of issuing an alarm based on the object status and the alarm level includes:

[0043] Based on the object status, the alarm level, the current time, and the camera device location information, standardized structured alarm information is generated. The structured alarm information includes at least one of the following: timestamp, geographic location, associated device, event type, alarm level, confidence level, and on-site image and short video clip evidence.

[0044] The structured alarm information is pushed in parallel to the central control station monitoring system, mobile inspection APP and safety supervision management platform through predefined communication interfaces, via API calls or message queue broadcasts.

[0045] Based on the alarm level, a corresponding action is matched and triggered from a preset response strategy library. The action includes at least one of the following: activating on-site audible and visual alarms, triggering network broadcast to drive away the user, automatically generating a maintenance work order, or sending a text message and voice notification to a preset responsible person.

[0046] Secondly, embodiments of the present invention provide a real-time monitoring device for substation operation and maintenance, comprising:

[0047] The acquisition module is used to acquire the video stream collected in real time by the camera equipment in the substation;

[0048] The preliminary identification module is used to perform preliminary identification of the video stream based on the edge computing device. If the preliminary identification result is abnormal, the module determines the objects in the video stream and the object status of the objects.

[0049] The judgment module is used to determine whether the state of the object meets the security requirements of the scene type based on the scene type of the monitoring area corresponding to the camera device;

[0050] The alarm level module is used to determine the alarm level corresponding to the object state if the object state does not meet the security requirements of the scenario type.

[0051] The alarm module is used to issue alarms based on the object status and the alarm level.

[0052] Thirdly, embodiments of the present invention provide an electronic device, including: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method of the first aspect described above.

[0053] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0054] At the data processing level, edge computing devices are used for initial identification of the video stream. Through frame extraction and lightweight model inference, rapid screening of anomalies is achieved. This approach ensures both the timeliness of the system response and establishes an effective preliminary filtering mechanism by setting confidence intervals and judging multi-target relationships.

[0055] At the object recognition level, continuous image sequence analysis technology, combined with temporal feature fusion and target tracking methods, improves the stability of object recognition in complex scenes. The use of trajectory segment analysis and temporal feature aggregation effectively reduces recognition errors caused by the quality of single-frame images, enhancing the system's ability to capture dynamic targets.

[0056] At the state determination level, reliable judgment of the object's state is achieved through the fusion of multi-frame visual evidence and the comprehensive application of various determination rules. This analysis method based on time-series information improves the anti-interference capability of state determination, enabling the system to adapt to on-site environmental challenges such as changes in lighting and temporary occlusion.

[0057] At the security assessment level, the established scenario-based security rule base enables differentiated security management. Through precise mapping between scenario types and security requirements, and dual verification of mandatory requirements and prohibitions, the system can perform targeted security monitoring based on the characteristics of different regions.

[0058] At the alarm handling level, the integrity and timeliness of alarm information are ensured through a standardized alarm information structure and a multi-channel push mechanism. Combined with a tiered response strategy, closed-loop management from incident discovery to handling execution is achieved, improving the systematicness and coordination of operation and maintenance response. Attached Figure Description

[0059] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0060] Figure 1 A flowchart illustrating the real-time monitoring method for substation operation and maintenance provided in this application;

[0061] Figure 2 A schematic diagram illustrating an example of real-time monitoring of substation operation and maintenance provided in this application;

[0062] Figure 3 A schematic diagram illustrating an example of real-time monitoring of substation operation and maintenance provided in this application;

[0063] Figure 4 A schematic diagram of the substation operation and maintenance real-time monitoring device provided in this application;

[0064] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0066] It should be noted that, in this document, 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..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0067] It should be noted that all actions involving the acquisition of signals, information, or data in this invention are carried out in compliance with the relevant data protection laws and regulations of the locality and with authorization from the owner of the relevant device.

[0068] Example 1

[0069] Please see Figure 1 This invention provides a method for real-time monitoring of substation operation and maintenance, comprising:

[0070] S1. Obtain the real-time video stream captured by the camera equipment in the substation;

[0071] Specifically, this step involves continuously acquiring video data from the monitored area using image acquisition devices deployed at the substation site, forming a raw video stream. The video stream refers to a collection of image frames transmitted continuously in a time sequence, with encoding formats conforming to standards such as H.264 or H.265. The resolution is typically no less than 1920×1080, and the frame rate is maintained at 25 to 30 frames per second to meet the temporal continuity requirements of subsequent analysis. In practice, the video stream needs to be transmitted to edge computing nodes via network protocols to provide data input for subsequent processing. This step establishes the data foundation for the monitoring system, ensuring real-time visibility and analyzability of the site conditions.

[0072] S2. Based on the edge computing device, perform preliminary identification of the video stream. If the preliminary identification result is abnormal, determine the objects in the video stream and the object status of the objects.

[0073] Specifically, this step processes the video stream in real time within an edge computing device, using a lightweight visual model to detect and classify objects in image frames. If the identification results contain content that does not conform to preset safety specifications, it is marked as an anomaly. Objects refer to the categories of entities identified from the video, including people, equipment, animals, and signs; object states describe the dynamic or static attributes of the entity, such as whether a person is wearing a safety helmet, whether an equipment switch is in the closed position, whether an animal has entered a restricted area, and whether a sign is hanging. In implementation, a convolutional neural network-based object detection model can be used to extract image features, and non-maximum suppression and confidence thresholding are used to filter and determine valid targets and their states. This step transforms the raw video into structured semantic information, providing input for scenario-based risk assessment.

[0074] S3. Based on the scene type of the monitoring area corresponding to the camera device, determine whether the object status meets the security requirements of the scene type;

[0075] Specifically, this step categorizes the areas covered by the camera equipment into different scenario types based on their physical functions and risk characteristics, such as high-voltage equipment areas, perimeter walls, and control rooms, and predefines a set of safety requirements for each type. Safety requirements include mandatory and prohibited states; for example, personnel in high-voltage areas must wear insulated gloves, and intrusion is prohibited within the perimeter. In implementation, the mapping relationship between scenario types and safety rules is queried, and the object's state is logically matched with the rules. If the object's state is prohibited or does not meet mandatory requirements, it is deemed non-compliant. This step automates the judgment of on-site behavioral compliance, mapping visual perception results to business-level safety states.

[0076] S4. If the object state does not meet the security requirements of the scenario type, then determine the alarm level corresponding to the object state;

[0077] Specifically, this step determines the alarm level based on the severity of the violation, object attributes, and contextual information, using a rule engine or classification model. Alarm levels are typically divided into multiple levels, such as low, medium, high, and emergency, and the determination is based on factors including the type of violation, the identity of the object, the duration, and environmental risks. For example, a person not wearing a safety helmet in a general area might trigger a low-level alarm, while near an electrified area, it might be escalated to a medium-level alarm; an animal intruding into an equipment area could be classified based on the animal's size and activity range. In implementation, the level can be determined by querying a pre-configured alarm policy table or using a multi-factor weighted decision model. This step provides a prioritization basis for subsequent responses, ensuring that resources are focused on handling high-risk events.

[0078] S5. Issue an alarm based on the object status and the alarm level.

[0079] Specifically, this step generates structured alarm information and executes corresponding response actions based on the object status and alarm level output from the preceding steps. The alarm information includes fields such as time, location, event type, level, and image evidence, and is pushed to the centralized control system, mobile terminals, or safety monitoring platform through standardized interfaces. Response actions are dynamically triggered according to the alarm level; for example, a low-level alarm generates a maintenance work order, a medium-level alarm activates an audible and visual alarm and notifies the responsible person, and a high-level alarm triggers a broadcast eviction and voice call. In implementation, message queues can be used to ensure reliable information transmission, and a strategy engine can be used to match the response process. This step completes a closed loop from anomaly detection to the issuance of handling instructions, improving the initiative and systematic nature of substation operation and maintenance management.

[0080] In some implementations, S2 involves performing preliminary identification of the video stream based on the edge computing device. If the preliminary identification result is abnormal, the objects in the video stream and their object states are determined, including:

[0081] Frames are extracted from the video stream, and the extracted image frames are preprocessed.

[0082] A lightweight vision model deployed at the edge is used to infer the preprocessed image frames and output preliminary recognition results containing the target category and initial confidence.

[0083] Based on the preliminary identification results, if the initial confidence level is lower than a first threshold but higher than a second threshold, then the objects in the video stream and the object states of the objects are determined, wherein the first threshold is greater than the second threshold.

[0084] Specifically, in the initial identification process of a video stream in an edge computing device, the video stream is first processed by frame extraction, and the extracted image frames are preprocessed to optimize image quality. Then, a lightweight vision model deployed at the edge is used to infer the preprocessed image frames, outputting an initial identification result containing the target category and initial confidence level. If this result satisfies the condition that the initial confidence level is within a fuzzy range between a first threshold and a second threshold, it indicates that the model has doubts about the identification result.

[0085] In some implementations, determining the objects in the video stream and the object states of the objects includes:

[0086] Extract a continuous image sequence within a certain time period before and after the image frame from the video stream;

[0087] Object recognition is performed based on the continuous image sequence to obtain objects, including people, equipment, animals, and signs;

[0088] For each object, the object state is determined. The object state of a person includes equipment wearing state, behavior state, and location state. The object state of equipment includes switch state, indicator light state, and meter pointer state. The object state of an animal includes animal behavior state and animal location state. The object state of a sign includes sign location state and sign content state.

[0089] Specifically, in determining objects and their states in a video stream, the system first extracts a continuous image sequence of a specific duration before and after the current image frame. This sequence provides dynamic information about the target in the temporal dimension. Object recognition is based on this sequence. By analyzing the consistency of the target's appearance features and the continuity of its motion across multiple frames, entity categories such as people, equipment, animals, and signs are distinguished. For each object category, state determination is completed by fusing multi-dimensional visual evidence from the sequence: for people, the equipment wearing status is determined by the statistical frequency of equipment detection results in the head area, the positional status is determined by the spatial relationship between the movement trajectory and a preset area, and the behavioral status is classified by the posture change sequence; for equipment, the on / off status is determined based on the morphological stability of the components across multiple frames, the indicator light status is determined by the brightness change pattern of the indicator light area, and the clearest frame is selected for meter pointer reading recognition; for animals, the behavioral status is classified based on movement trajectory features, and the animal's positional status is determined based on its positional relationship with key equipment areas; for signs, the sign's positional status is determined by continuously detecting its relative position to the installation reference point, and the sign's content status is extracted using optical character recognition technology. This method improves the stability of state determination in complex scenarios by introducing temporal context analysis, and reduces misjudgments caused by fluctuations in the quality of single-frame images.

[0090] In some embodiments, the step of identifying objects based on the continuous image sequence to obtain objects includes:

[0091] For each frame of the continuous image sequence, convolutional feature extraction is performed to obtain the depth feature map of each frame.

[0092] The deep feature maps of multiple frames in the sequence are concatenated along the channel dimension or fused using temporal pooling to generate sequence-enhanced feature maps;

[0093] The sequence-enhanced feature map is input into the target detection network, and the target bounding box and the category confidence of the target bounding box are output.

[0094] Based on the continuous image sequence, the targets within the target bounding box are tracked and associated to form multiple short-term target trajectory segments;

[0095] Detection results that exceed the fluctuation threshold or appear less frequently than the frequency threshold in the short-term target trajectory segment are removed, resulting in multiple unknown objects;

[0096] For each unknown object, the features of the unknown object in each frame of the continuous image sequence are aggregated to obtain the temporal aggregated features of the unknown object;

[0097] The temporal aggregation features are input into a preset object classifier to obtain objects including specific categories.

[0098] Specifically, in the process of object recognition based on continuous image sequences, convolutional feature extraction is first performed on each frame of the sequence to generate a depth feature map corresponding to each frame. This feature map contains an abstract representation of the spatial structure in the image. Subsequently, the depth feature maps of multiple frames are concatenated along the channel dimension, or temporal pooling is used to fuse cross-frame features to generate a sequence-enhanced feature map that incorporates temporal information. This operation can integrate the apparent changes of the target at different times. After the sequence-enhanced feature map is input into the object detection network, the target bounding box and its class confidence are output, forming a preliminary detection result. Based on the continuous sequence, the target in the bounding box is correlated across frames using a target tracking algorithm to form multiple short-term target trajectory segments. These segments describe the continuity of the target in the temporal dimension. By removing detection results in the short-term target trajectory segments where the class confidence fluctuation exceeds a set fluctuation threshold or the target occurrence frequency is lower than a set frequency threshold, unstable and occasional interfering targets are filtered out, resulting in multiple unknown objects to be confirmed. For each unknown object, its features across frames in the sequence are aggregated to generate temporally consistent aggregated features, which reflect the comprehensive characteristics of the target at multiple time points. Finally, the temporally aggregated features are input into a pre-defined object classifier, and the classification decision determines objects belonging to specific categories. This method improves the stability and reliability of object recognition in complex scenes by fusing temporal information and feature aggregation.

[0099] In some embodiments, inputting the temporal aggregation features into a preset object classifier to obtain objects including specific categories includes:

[0100] Determine whether the object is classified as a person or an animal;

[0101] When the object is classified as a person or an animal, and the object is located in a critical monitoring area of ​​the substation, thermal imaging data spatiotemporally aligned with the video stream is acquired.

[0102] Thermal feature information is extracted based on the thermal imaging data, and the category of the object is verified or corrected according to the thermal feature information; wherein, the key monitoring area includes the perimeter of the wall, the high-voltage equipment area, the main transformer area, and the cable layer.

[0103] Further, acquiring thermal imaging data spatiotemporally aligned with the video stream includes:

[0104] By using a pre-calibrated spatial transformation matrix, the target region in the visible light image is mapped to the thermal imaging coordinate system to generate a thermal imaging ROI region;

[0105] The timestamps of visible light data and thermal imaging data are aligned using the Precision Time Protocol (PTP), with the time error controlled within 50 milliseconds.

[0106] Biological thermal features are extracted from the thermal imaging ROI region. These biological thermal features include temperature distribution features to distinguish temperature gradient patterns between humans and animals; thermal profile geometric features, including the aspect ratio, compactness, and temperature center point of the heat source; and dynamic thermal change features to identify temperature fluctuation patterns specific to organisms.

[0107] Further, the step of verifying or correcting the category of the object based on the thermal feature information includes:

[0108] When visible light classifies an object as a person but thermal imaging does not detect a heat source that matches human characteristics, the object category is corrected to non-biological interference.

[0109] When visible light classifies an object as an animal but thermal imaging shows that the heat source has human characteristics, the object category is corrected to a person.

[0110] When thermal imaging detects multiple adjacent heat sources with different temperature characteristics, the object category is determined to be a person carrying an animal.

[0111] Specifically, thermal imaging data refers to image data reflecting the temperature distribution of an object's surface, acquired by an infrared thermal imager; spatiotemporal alignment refers to the technical process of precisely matching different modal data (visible light and thermal imaging) in spatial coordinates and time dimensions, achieving spatial mapping through pre-calibrated camera and thermal imager extrinsic parameter matrices, and using the PTP precise time protocol to ensure time synchronization error <50 milliseconds; biological thermal characteristics include temperature distribution patterns, thermal profile geometry, and dynamic temperature changes—thermal properties that can distinguish between living and non-living things; non-biological interference refers to false detection targets caused by changes in illumination, floating objects, or equipment vibration. This technical solution achieves accurate identification of biological targets in the complex environment of substations by performing multimodal verification of personnel / animal categories in key areas (perimeter of walls, high-voltage equipment areas, etc.), effectively solving the problem of identification failure of traditional single visible light solutions under adverse conditions such as nighttime and foggy weather.

[0112] In some implementations, determining the object state for each object includes:

[0113] For each identified object, in each frame of the continuous image sequence, visual evidence related to the state of the object is extracted, the visual evidence including at least one of the object's pixel region, appearance features, relative spatial location, and text content;

[0114] The object state is obtained based on the fusion results of multi-frame visual evidence and predefined state determination rules. The state determination rules include at least one of multi-frame voting, frequency statistics, persistence judgment, and optimal frame selection.

[0115] Specifically, when determining the object state of each object, the system extracts state-related visual evidence for each identified object in each frame of a continuous image sequence. This visual evidence includes, but is not limited to: pixel region information of the object, used to analyze the spatial distribution of the target in the image; appearance features, such as visual attributes like color, texture, and shape; relative spatial position, i.e., the positional relationship between the target and other reference objects in the scene; and text content, such as character information appearing in the image. By fusing this multi-frame visual evidence and analyzing it based on preset state determination rules, the stable state of the object is finally determined. The state determination rules can employ a multi-frame voting mechanism, i.e., determining the final state based on the mode of the recognition results across multiple frames; or a frequency statistics method, determining the state by calculating the frequency of a specific state appearing in the sequence; or a temporal persistence judgment, requiring the target state to remain stable across multiple consecutive frames; or an optimal frame selection strategy, selecting the highest quality frame from the sequence as the basis for state determination. This method, by comprehensively utilizing temporal information and multi-dimensional visual features, improves the stability and adaptability of state determination and can effectively address issues such as occlusion, blurring, or lighting changes that may exist in a single frame image.

[0116] Furthermore, for personnel, the determination of their object status includes:

[0117] Equipment wearing status determination: In the continuous image sequence, the frequency of the number of frames in which a specific safety device is detected in the head or hand area of ​​the target person is counted. When the frequency is higher than a first preset threshold, the equipment wearing status is determined to be worn; otherwise, it is not worn.

[0118] Behavioral state determination: Based on the changes in the apparent features of the human target in the continuous image sequence, a temporal behavior recognition model is used for classification to obtain behavioral states including walking, running, falling, or operating.

[0119] Location status determination: Based on the spatial location of the person in the sequence of multiple frames, it is compared with the preset electronic fence. When the frequency of the person appearing in the unauthorized area is higher than the second preset threshold, the location status is determined to be in the unauthorized area.

[0120] For device objects, the determination of their object status includes:

[0121] Switch status determination: In the continuous image sequence, the shape of the switch component is identified in multiple frames, and its switch status is determined to be closed or open through a majority voting mechanism.

[0122] Indicator light status determination: In the continuous image sequence, the brightness change of the indicator light area is detected, and the indicator light status is determined to be on, off, or flashing based on its on / off pattern.

[0123] Meter pointer status determination: In the continuous image sequence, select several frames with clear meter dials for identification, perform consistency verification on the identification results, and take the reading that appears most frequently as the meter pointer status.

[0124] For animal objects, the determination of their object state includes:

[0125] Animal behavior status determination: Based on the appearance features and motion information of the animal target in the continuous image sequence, its behavior status is classified as walking, running, flying, standing still, or biting.

[0126] Animal position status determination: Based on the spatial position of the animal in multiple frames of the sequence, determine whether it has entered the critical equipment area, and thus determine whether the animal's position status is entered or not.

[0127] For signage objects, the determination of their object status includes:

[0128] Identification of sign position status: In the continuous image sequence, the relative relationship between the sign and the preset installation position is detected. If there is a significant shift or disappearance in multiple frames, the sign position status is determined to be detached or displaced.

[0129] Signage content status determination: In the continuous image sequence, select frames with clear signage images for optical character recognition, and use the recognized text content as the signage content status.

[0130] In some implementations, determining whether the object's state meets the security requirements of the scene type based on the scene type of the monitored area corresponding to the camera device includes:

[0131] Based on the unique identifier of the camera device, a preset configuration database is queried to obtain the scene type corresponding to the area monitored by the camera device. The scene type includes high-voltage equipment area, main transformer area, switchyard, capacitor shed, perimeter wall, main control room and cable layer.

[0132] Based on the scenario type, a pre-set scenario type and security rule mapping table is retrieved to obtain a set of mandatory security states and a set of prohibited states bound to the scenario type.

[0133] The object state is logically compared with the retrieved mandatory security state set and prohibited state set. If the object state belongs to the prohibited state set, it is determined that it does not meet the security requirements.

[0134] If the object is a person, and the object's state does not fully cover the states required by the mandatory safety state set, then it is determined that it does not meet the safety requirements.

[0135] Specifically, in the scenario compliance assessment process, the system first queries a pre-configured database based on the unique identifier of the camera device to obtain the scenario type corresponding to the monitored area. Scenario types are typical area categories based on the functional zoning of substations, including physical spaces with different risk characteristics and maintenance requirements, such as high-voltage equipment areas, main transformer areas, switchyards, capacitor sheds, perimeter walls, main control rooms, and cable layers. Subsequently, the system retrieves a pre-configured scenario type-safety rule mapping table based on the determined scenario type to obtain the mandatory safety state set and prohibited state set bound to that scenario type. The mandatory safety state set defines the safety conditions that must be met in the scenario, such as personnel wearing specific protective equipment; the prohibited state set specifies dangerous states that are not allowed in the scenario, such as open flames and illegal intrusion. During the logical comparison phase, the system matches the real-time acquired object status with these two sets: if the object status appears in the prohibited state set, it is directly determined to be non-compliant with safety requirements; if the object is a person, it is further checked whether its object status completely covers all requirements in the mandatory safety state set; if any are missing, it is also determined to be non-compliant. This scenario-based dual verification mechanism can adapt to the safety management needs of different areas of the substation and achieve differentiated safety control.

[0136] In some implementations, the step of issuing an alarm based on the object status and the alarm level includes:

[0137] Based on the object status, the alarm level, the current time, and the camera device location information, standardized structured alarm information is generated. The structured alarm information includes at least one of the following: timestamp, geographic location, associated device, event type, alarm level, confidence level, and on-site image and short video clip evidence.

[0138] The structured alarm information is pushed in parallel to the central control station monitoring system, mobile inspection APP and safety supervision management platform through predefined communication interfaces, via API calls or message queue broadcasts.

[0139] Based on the alarm level, a corresponding action is matched and triggered from a preset response strategy library. The action includes at least one of the following: activating on-site audible and visual alarms, triggering network broadcast to drive away the user, automatically generating a maintenance work order, or sending a text message and voice notification to a preset responsible person.

[0140] Specifically, in the alarm generation and response phase, the system first generates standardized structured alarm information based on the object status, alarm level, current time, and camera location information. This information uses a unified data structure, including timestamps, geographic coordinates, associated device numbers, event type classifications, alarm level identifiers, identification confidence scores, and visual evidence such as on-site images or short video clips. Subsequently, through predefined communication interfaces, using application programming interface calls or message queue broadcasting mechanisms, the structured alarm information is pushed in parallel to the central control station monitoring system, mobile inspection application, and safety supervision and management platform, ensuring data synchronization between multiple systems. Finally, based on the alarm level, the system matches the corresponding handling plan from a pre-set response strategy library and executes coordinated actions, including activating on-site audible and visual alarm devices, triggering network voice broadcasts to drive away occupants, automatically generating maintenance work orders, or sending SMS and voice notifications to pre-defined responsible persons. This standardized processing method achieves standardized transmission and hierarchical response of alarm information, improving the efficiency of responding to substation anomalies through multi-channel collaborative handling.

[0141] The following examples will provide a detailed explanation.

[0142] I. An abnormal movement was captured by the deployed intelligent video surveillance system in the perimeter monitoring area of ​​a 500kV substation.

[0143] Specific processing procedure:

[0144] 1. Data Collection and Preliminary Identification

[0145] Perimeter surveillance cameras continuously collect video streams, and edge computing devices analyze frames at a frequency of 5 frames per second.

[0146] The lightweight vision model detected a "person" target in the frame sequence with an initial confidence level of 78%.

[0147] The system determines that the target is within the electronic fence warning area and triggers an anomaly flag.

[0148] 2. Object status confirmation

[0149] The system extracts 15 consecutive frames from the preceding and following 3 seconds for in-depth analysis;

[0150] The target was identified as "outsiders" through temporal feature fusion.

[0151] Status determination:

[0152] Equipment status: No helmet worn (no helmet detected for 12 consecutive frames);

[0153] Behavioral state: Climbing (climbing actions were identified through posture sequences);

[0154] Location status: Located in the perimeter restricted area (trajectory coordinates fall within the electronic fence alarm area).

[0155] 3. Scenario Compliance Assessment

[0156] The scenario type is "perimeter fence" based on the device ID.

[0157] Retrieved from the security rule base:

[0158] Mandatory security status set: empty set (no personnel activity is allowed in this area);

[0159] Prohibited status set: Personnel appearing in the perimeter restricted area;

[0160] Logical comparison result: The object's state violates a prohibitive rule.

[0161] 4. Alarm Level Determination

[0162] Basic level: Matched as "Advanced" according to the "Perimeter Intrusion" rules;

[0163] Attribute adjustment: The target is an outsider, and the level is maintained as "Advanced";

[0164] Time Limit Correction: Intrusion activity lasted for 8 seconds, confirmation level "High";

[0165] System adjustment: There are currently no other advanced alerts, and the system has been confirmed as "advanced".

[0166] 5. Alarm Execution and Response

[0167] Generate structured alarm information;

[0168] Synchronously pushed to: via RabbitMQ message queue

[0169] Central control station monitoring screen;

[0170] Mobile app for inspection personnel;

[0171] Safety supervision platform.

[0172] Triggering coordinated response:

[0173] Activate the on-site audible and visual alarm device;

[0174] Play a warning broadcast saying "Please leave immediately";

[0175] Send an urgent text message to the security manager;

[0176] Generate an emergency response work order.

[0177] II. Figure 2As shown, in the outdoor area on the southeast side of the 220kV Yinque substation, the No. 7 spherical camera simultaneously detected multiple intrusion anomalies.

[0178] Specific processing procedure:

[0179] 1. Data Collection and Preliminary Identification

[0180] The No. 7 PTZ camera continuously collects video streams, which are then analyzed in real time by edge computing devices.

[0181] The system detected four types of abnormal targets simultaneously:

[0182] Border intrusion (confidence level 0.91)

[0183] Foreign object intrusion (confidence levels 0.93, 0.94, 0.93)

[0184] Animal invasion (confidence level 0.91)

[0185] 2. Object status confirmation

[0186] The system performs state analysis for each type of target separately:

[0187] Boundary intrusion target: Confirmed to be a person crossing the boundary, with the location status indicating that they have crossed the electronic fence;

[0188] Foreign object intrusion target: identified as a floating object (balloon, plastic film, etc.), located in the air above the equipment area;

[0189] Animal intrusion target: identified as a large bird, whose behavior is nesting on the structure and whose location is near the insulator.

[0190] 3. Scenario Compliance Assessment

[0191] Based on the equipment identification 220kV Yinque Substation-Outdoor-Southeast-7-PTZ Camera, the scene type is determined to be an outdoor equipment area.

[0192] Search security rules:

[0193] Prohibited statuses include: {personnel crossing boundaries, floating objects above equipment areas, and structures where animals inhabit}.

[0194] Logical comparison results: All three types of targets violate the prohibitive rules.

[0195] 4. Alarm Level Determination

[0196] Assess each type of intrusion risk separately:

[0197] Border Intrusion (Personnel): Advanced Alert;

[0198] Foreign object intrusion (floating object): Intermediate alarm (may cause short circuit);

[0199] Animal intrusion (birds): Intermediate alarm (may cause equipment malfunction);

[0200] System comprehensive assessment: Multiple risks exist; the original levels of each target will be maintained.

[0201] 5. Alarm Execution and Response

[0202] Generate composite alarm information:

[0203] Alarm time: 16:04:11 on March 5, 2025;

[0204] Alarm location: Southeast outdoor equipment area of ​​220kV Yinque Substation;

[0205] Surveillance equipment: No. 7 dome camera;

[0206] Overall risk level: High (multiple intrusion risks exist);

[0207] Concurrent event details:

[0208] Border Intrusion Alert

[0209] Event type: Personnel boundary intrusion;

[0210] Risk level: High;

[0211] Confidence level: 91%;

[0212] Target characteristics: Personnel crossing the boundary were detected;

[0213] Foreign object intrusion alarm

[0214] Event type: Floating object intrusion;

[0215] Risk level: Medium;

[0216] Confidence level: 93%;

[0217] Target characteristics: Floating objects were detected above the equipment area;

[0218] Animal invasion alarm

[0219] Event type: Bird invasion;

[0220] Risk level: Medium;

[0221] Confidence level: 91%;

[0222] Target characteristics: Identify bird activity in the architectural region;

[0223] On-site evidence: Image sequences containing object intrusion, such as Figure 2 As shown.

[0224] Tiered response and handling:

[0225] For border intrusion: activate the audible and visual alarm to notify security personnel for on-site handling;

[0226] Foreign object intrusion: Generate an inspection work order and arrange for the removal of the foreign object;

[0227] Animal intrusion: Trigger bird deterrent devices and arrange for nest cleanup.

[0228] 3. In the western area of ​​the 10W room of the 220kV Yinque Substation, the No. 26 dome camera detected an abnormal fence status.

[0229] Specific processing procedure:

[0230] 1. Data Collection and Preliminary Identification

[0231] The PTZ camera No. 26 captured the on-site video stream during the inspection.

[0232] Edge computing devices continuously detected "fence collapse" events through video analysis;

[0233] The system recorded two recognition results with confidence levels of 0.95 and 0.97, respectively, both exceeding the preset threshold.

[0234] 2. Object Status Confirmation

[0235] The system continuously analyzes the fence status:

[0236] Object to be identified: Substation protective fence;

[0237] Object state: Downward-falling state;

[0238] Location information: Specific section west of room 10W;

[0239] Through multi-frame verification, it was confirmed that the fence was indeed in a collapsed state and was not a temporary obstruction or light and shadow interference.

[0240] 3. Scenario Compliance Assessment

[0241] Based on the camera's location on the west side of room 10W, the scene type is determined to be an equipment protection zone.

[0242] Search the security rule base:

[0243] Prohibited conditions set: {Protective fence collapsed, perimeter facilities damaged};

[0244] Logical comparison result: The collapsed state of the fence violates safety regulations.

[0245] 4. Alarm Level Determination

[0246] Basic level assessment: Damage to protective facilities falls under the category of medium-level alarm;

[0247] Risk Factor Analysis:

[0248] This could cause people to accidentally enter a dangerous area;

[0249] It may affect the safe operation of the equipment;

[0250] It needs to be repaired promptly;

[0251] It was ultimately confirmed as a medium-level alarm.

[0252] 5. Alarm Execution and Response

[0253] Generate structured alarm information:

[0254] Time: 14:47, March 4, 2025;

[0255] Location: West fenced area of ​​Room 10W;

[0256] Event type: Fence collapsed;

[0257] Alarm level: Medium;

[0258] Confidence level: 0.97;

[0259] On-site evidence: Image sequences containing collapsed fences, such as Figure 3 As shown.

[0260] Implement response measures:

[0261] Automatically generate maintenance work orders and assign maintenance teams;

[0262] Mark dangerous areas on electronic maps;

[0263] Send a notification to the regional manager;

[0264] The event is recorded for subsequent security analysis.

[0265] Example 2

[0266] Please see Figure 4 This invention provides a real-time monitoring device for substation operation and maintenance, comprising:

[0267] The acquisition module 401 is used to acquire the video stream collected in real time by the camera equipment in the substation;

[0268] The preliminary identification module 402 is used to perform preliminary identification on the video stream based on the edge computing device. If the preliminary identification result is abnormal, the object in the video stream and the object status of the object are determined.

[0269] The judgment module 403 is used to determine whether the state of the object meets the security requirements of the scene type based on the scene type of the monitoring area corresponding to the camera device;

[0270] Alarm level module 404 is used to determine the alarm level corresponding to the object state if the object state does not meet the security requirements of the scenario type;

[0271] Alarm module 405 is used to issue alarms based on the object status and the alarm level.

[0272] It should be noted that each module and unit in the substation operation and maintenance real-time monitoring device in this embodiment corresponds one-to-one with each step in the substation operation and maintenance real-time monitoring method in the aforementioned embodiment. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned substation operation and maintenance real-time monitoring method, and will not be repeated here.

[0273] Example 3

[0274] Please see Figure 5 This embodiment provides an electronic device, including at least one processor 501 and a memory 502. Optionally, the device further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0275] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0276] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0277] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0278] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0279] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0280] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0281] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0282] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0283] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0284] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0285] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0286] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0287] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of 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.

[0288] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0289] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. 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. A method for real-time monitoring of substation operation and maintenance, characterized in that, include: Acquire real-time video streams captured by camera equipment in the substation; The edge computing device performs preliminary identification on the video stream. If the preliminary identification result is abnormal, the objects in the video stream and the object status of the objects are determined. Based on the scene type of the monitoring area corresponding to the camera device, determine whether the object status meets the security requirements of the scene type; If the object state does not meet the security requirements of the scenario type, then determine the alarm level corresponding to the object state; An alarm is triggered based on the object's status and the alarm level.

2. The method according to claim 1, characterized in that, The step of performing preliminary identification of the video stream based on the edge computing device, and determining the objects in the video stream and their object states if the preliminary identification result is abnormal, includes: Frames are extracted from the video stream, and the extracted image frames are preprocessed. A lightweight vision model deployed at the edge is used to infer the preprocessed image frames and output preliminary recognition results containing the target category and initial confidence. Based on the preliminary identification results, if the initial confidence level is lower than a first threshold but higher than a second threshold, then the objects in the video stream and the object states of the objects are determined, wherein the first threshold is greater than the second threshold.

3. The method according to claim 2, characterized in that, Determining the objects in the video stream and the object states of the objects includes: Extract a continuous image sequence within a certain time period before and after the image frame from the video stream; Object recognition is performed based on the continuous image sequence to obtain objects, including people, equipment, animals, and signs; For each object, the object state is determined. The object state of a person includes equipment wearing state, behavior state, and location state. The object state of equipment includes switch state, indicator light state, and meter pointer state. The object state of an animal includes animal behavior state and animal location state. The object state of a sign includes sign location state and sign content state.

4. The method according to claim 3, characterized in that, The step of identifying objects based on the continuous image sequence to obtain objects includes: For each frame of the continuous image sequence, convolutional feature extraction is performed to obtain the depth feature map of each frame. The deep feature maps of multiple frames in the sequence are concatenated along the channel dimension or fused using temporal pooling to generate sequence-enhanced feature maps; The sequence-enhanced feature map is input into the target detection network, and the target bounding box and the category confidence of the target bounding box are output. Based on the continuous image sequence, the targets within the target bounding box are tracked and associated to form multiple short-term target trajectory segments; Detection results that exceed the fluctuation threshold or appear less frequently than the frequency threshold in the short-term target trajectory segment are removed, resulting in multiple unknown objects; For each unknown object, the features of the unknown object in each frame of the continuous image sequence are aggregated to obtain the temporal aggregated features of the unknown object; The temporal aggregation features are input into a preset object classifier to obtain objects including specific categories.

5. The method according to claim 4, characterized in that, The step of inputting the temporal aggregation features into a preset object classifier to obtain objects including specific categories includes: Determine whether the object is classified as a person or an animal; When the object is classified as a person or an animal, and the object is located in a critical monitoring area of ​​the substation, thermal imaging data spatiotemporally aligned with the video stream is acquired. Thermal feature information is extracted based on the thermal imaging data, and the category of the object is verified or corrected based on the thermal feature information; The key monitoring areas include the perimeter of the perimeter wall, the high-voltage equipment area, the main transformer area, and the cable layer.

6. The method according to claim 3, characterized in that, Determining the object state for each object includes: For each identified object, in each frame of the continuous image sequence, visual evidence related to the state of the object is extracted, the visual evidence including at least one of the object's pixel region, appearance features, relative spatial location, and text content; The object state is obtained based on the fusion results of multi-frame visual evidence and predefined state determination rules. The state determination rules include at least one of multi-frame voting, frequency statistics, persistence judgment, and optimal frame selection.

7. The method according to claim 1, characterized in that, The step of determining whether the object's state meets the security requirements of the scene type based on the scene type of the monitoring area corresponding to the camera device includes: Based on the unique identifier of the camera device, a preset configuration database is queried to obtain the scene type corresponding to the area monitored by the camera device. The scene type includes high-voltage equipment area, main transformer area, switchyard, capacitor shed, perimeter wall, main control room and cable layer. Based on the scenario type, a pre-set scenario type and security rule mapping table is retrieved to obtain a set of mandatory security states and a set of prohibited states bound to the scenario type. The object state is logically compared with the retrieved mandatory security state set and prohibited state set. If the object state belongs to the prohibited state set, it is determined that it does not meet the security requirements. If the object is a person, and the object's state does not fully cover the states required by the mandatory safety state set, then it is determined that it does not meet the safety requirements.

8. The method according to claim 1, characterized in that, The step of issuing an alarm based on the object status and the alarm level includes: Based on the object status, the alarm level, the current time, and the camera device location information, standardized structured alarm information is generated. The structured alarm information includes at least one of the following: timestamp, geographic location, associated device, event type, alarm level, confidence level, and on-site image and short video clip evidence. The structured alarm information is pushed in parallel to the central control station monitoring system, mobile inspection APP and safety supervision management platform through predefined communication interfaces, via API calls or message queue broadcasts. Based on the alarm level, a corresponding action is matched and triggered from a preset response strategy library. The action includes at least one of the following: activating on-site audible and visual alarms, triggering network broadcast to drive away the user, automatically generating a maintenance work order, or sending a text message and voice notification to a preset responsible person.

9. A real-time monitoring device for substation operation and maintenance, characterized in that, include: The acquisition module is used to acquire the video stream collected in real time by the camera equipment in the substation; The preliminary identification module is used to perform preliminary identification of the video stream based on the edge computing device. If the preliminary identification result is abnormal, the module determines the objects in the video stream and the object status of the objects. The judgment module is used to determine whether the state of the object meets the security requirements of the scene type based on the scene type of the monitoring area corresponding to the camera device; The alarm level module is used to determine the alarm level corresponding to the object state if the object state does not meet the security requirements of the scenario type. The alarm module is used to issue alarms based on the object status and the alarm level.

10. An electronic device, characterized in that, include: At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in any one of claims 1-8.

Images