A hidden danger space relation determination method and system based on hierarchical early warning

Abstract

This invention discloses a method and system for determining the spatial relationship of potential hazards based on hierarchical early warning, belonging to the field of computer data processing technology. It includes acquiring historical risk data and real-time environmental data, delineating virtual electronic warning zones, and generating dynamic virtual warning zone data; collecting video stream data and identifying and locating potential hazard targets, generating hazard target identification data, and performing spatial matching analysis in conjunction with the dynamic virtual warning zone data to generate an intrusion determination result; if an intrusion occurs, calculating the three-dimensional spatial clearance distance between the hazard target and the live conductor, and outputting a hierarchical alarm result based on multiple key risk factors; if no intrusion occurs, analyzing and predicting the target movement status and location based on the hazard target identification data to generate an out-of-zone early warning result. This invention employs a data processing method that combines dynamic zoning with quantitative analysis and trend prediction of potential hazard targets both inside and outside the zones, enabling intelligent, precise, and forward-looking risk early warning.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, and in particular to a method and system for determining the spatial relationship of potential hazards based on hierarchical early warning. Background Technology

[0002] In the field of safety monitoring of critical infrastructure such as power, transportation, and construction, using information technology to determine the spatial relationships of potential hazards is a core aspect of ensuring safe production. This type of determination typically involves analyzing the spatial relationships between moving or stationary targets and critical facilities within a monitored environment using data acquisition and processing systems. This analysis identifies dangerous approach behaviors that could lead to safety accidents and triggers corresponding early warning mechanisms. This process is an indispensable part of modern automated safety monitoring systems, and the accuracy and timeliness of its results directly impact the success or failure of risk prevention and control.

[0003] In existing technologies, methods for determining the spatial relationships of potential hazards typically rely on video surveillance and simple image processing algorithms. A typical approach involves manually defining a fixed-boundary electronic fence or restricted area within the monitoring system; the geometry and size of this area remain unchanged once set. The system uses background subtraction or simple target recognition algorithms to detect whether an object has entered this fixed area. Once the pixel coordinates of a target are detected falling within the pre-defined electronic fence polygon, the data processing system classifies it as an intrusion event and triggers a standardized, indiscriminate alarm signal, notifying management personnel to handle the situation.

[0004] However, the aforementioned existing technical solutions have significant technical shortcomings in practical applications. First, the electronic fences they employ are static and cannot adaptively adjust to changes in the site environment, such as severe weather or nighttime construction, which increase the risk level. This results in a lack of scientific rigor and flexibility in setting the warning range. Second, their alarm mechanisms are too simplistic, using the same processing method for all intrusion events. They fail to differentiate between the type of intrusion target, its movement state, and the different levels of danger resulting from its actual proximity to the core hazard source, leading to low value density of alarm information. Finally, this method is entirely passive, only reacting after an intrusion has occurred. The system lacks the ability to predict and warn of potential threats approaching the restricted area but not yet entering. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for determining the spatial relationship of potential hazards based on hierarchical early warning. It employs a data processing approach that dynamically delineates virtual warning zones and combines quantitative analysis and trend prediction of potential hazard targets within and outside these zones. This approach enables intelligent, precise, and forward-looking risk warnings.

[0006] The above objectives can be achieved through the following approach:

[0007] A method for determining the spatial relationship of potential hazards based on hierarchical early warning includes the following steps:

[0008] Acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data to generate dynamic virtual warning zone data;

[0009] By collecting video stream data of the scene area through monitoring equipment, and performing intelligent visual analysis based on the video stream data, the system identifies and locates potential hazards, and generates hazard identification data that includes the target category, target location, and target movement status.

[0010] Based on the dynamic virtual warning zone data and the hazard target identification data, spatial matching analysis is performed to determine whether the hazard target has intruded into the virtual electronic warning zone, and an intrusion determination result is generated.

[0011] In response to the condition that the intrusion determination result is an intrusion, the three-dimensional spatial clearance distance between the potential target and the live conductor is calculated;

[0012] Based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, a quantitative assessment is performed, and a graded alarm result is output.

[0013] In response to the condition that the intrusion determination result is non-intrusion, based on the target's movement status and the target's location, the potential risk level is analyzed and predicted, and an external early warning result is generated.

[0014] Optionally, the generation of dynamic virtual warning zone data specifically includes:

[0015] Obtain historical risk data on the geographic information and frequency of past potential incidents;

[0016] By calling external data service interfaces, real-time environmental data containing weather conditions and terrain change information can be obtained;

[0017] A comprehensive risk score is obtained by combining the historical risk data and the real-time environmental data.

[0018] Based on the comprehensive risk score, the boundary range of the virtual electronic warning zone is dynamically adjusted to generate dynamic virtual warning zone data.

[0019] Optionally, generating hazard target identification data that includes target category, target location, and target movement status specifically involves:

[0020] The video stream data is analyzed using a deep learning model to identify the potential hazards and output their category information.

[0021] A preset target tracking algorithm is activated to lock and track the potential target, determine its position changes in continuous video frames, and generate target position data;

[0022] The motion vector of the target is calculated based on the temporal changes of the target location data, and the target movement state data is parsed from the motion vector.

[0023] The target category information, target location data, and target movement status data are integrated to generate the hazard target identification data.

[0024] Optionally, generating the intrusion determination result specifically includes:

[0025] Unify the boundary coordinates of the dynamic virtual warning zone data with the target location data in the hidden danger target identification data into a global coordinate system;

[0026] Using a preset test algorithm for points within polygons, spatial relationship analysis is performed based on the target location and the boundary of the virtual electronic warning zone to obtain spatial inclusion relationship;

[0027] Based on the spatial inclusion relationship, an intrusion status judgment is performed, and an intrusion judgment result indicating whether the potential target is located inside the virtual electronic warning zone is generated.

[0028] Optionally, in response to the condition that the intrusion determination result is an intrusion, the three-dimensional spatial clearance distance between the potential hazard target and the live conductor is calculated, specifically as follows:

[0029] Obtain the three-dimensional spatial coordinate information of live conductors from a pre-set power facility database;

[0030] Spatial registration and coordinate unification are performed based on the target location data and the three-dimensional spatial coordinate information of the charged conductor to construct the spatial geometric relationship between the target and the conductor;

[0031] Based on the spatial geometric relationship, the spatial straight-line distance between the key points on the surface of the potential hazard target and the live conductor that meets the preset conditions is calculated and used as the three-dimensional spatial clearance distance.

[0032] Optionally, the output of the hierarchical alarm result is specifically as follows:

[0033] By integrating and analyzing the target category, the target movement state, and the spatial geometric relationship, multiple key risk factors are obtained.

[0034] The risk feature vector is obtained by combining the three-dimensional spatial clearance distance with the multiple key risk factors.

[0035] Risk assessment and calculation are performed based on the aforementioned risk feature vector to obtain a quantified risk score;

[0036] Based on the risk score, the risk level is classified, and a graded alarm result including alarm type, risk description and handling suggestions is output.

[0037] Optionally, the generation of the external early warning result specifically includes:

[0038] Calculate the azimuth angle between the velocity vector corresponding to the target's movement state and a direction vector pointing from the target's position to the boundary of the virtual electronic warning zone;

[0039] The movement trend of the potential hazard target is determined based on the azimuth angle.

[0040] When the movement trend is toward the virtual electronic warning zone, a low-level alarm is triggered;

[0041] When the movement trend is stationary or away from the virtual electronic warning zone, an alarm filtering command is triggered, and an external warning result is generated together.

[0042] Optionally, the step of dynamically adjusting the boundary range of the virtual electronic warning zone based on the comprehensive risk score to generate dynamic virtual warning zone data specifically involves:

[0043] Based on the comprehensive risk score, candidate boundaries are calculated.

[0044] The difference between the candidate boundary and the currently effective virtual electronic warning zone boundary is calculated to obtain the relevant difference degree;

[0045] The dynamic virtual warning area data is updated using the candidate boundary only when the correlation difference exceeds a preset change threshold.

[0046] Optionally, the method further includes:

[0047] Calculate the distance from the target location to the boundary of the virtual electronic warning zone;

[0048] Based on the comparison result between the distance and a preset tolerance, the proximity state information is determined;

[0049] The proximity status information is integrated into the generation process of the external early warning results to dynamically optimize the prediction process.

[0050] Based on the same inventive concept, the present invention also provides a hazard spatial relationship determination system based on hierarchical early warning, the system comprising:

[0051] The dynamic warning zone delineation module is used to acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data, generating dynamic virtual warning zone data;

[0052] The hazard visual recognition module is used to collect video stream data of the scene area through monitoring equipment, and perform intelligent visual analysis based on the video stream data to identify and locate hazard targets, and generate hazard target recognition data including target category, target location and target movement status;

[0053] The intrusion determination and analysis module is used to perform spatial matching analysis based on the dynamic virtual warning zone data and the hidden danger target identification data to determine whether the hidden danger target has intruded into the virtual electronic warning zone and generate an intrusion determination result.

[0054] The clearance calculation and analysis module is used to calculate the three-dimensional spatial clearance between the potential target and the live conductor in response to the condition that the intrusion determination result is an intrusion.

[0055] The graded alarm analysis module is used to perform quantitative analysis based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, and output graded alarm results.

[0056] The external risk prediction module is used to analyze and predict the potential risk level based on the target's movement status and location, and generate an external early warning result, in response to the condition that the intrusion determination result is non-intrusion.

[0057] Compared with the prior art, the present invention has the following advantages:

[0058] 1. This invention achieves intelligent and adaptive security monitoring by dynamically defining virtual electronic warning zones. The method comprehensively assesses risks based on historical risk data and real-time environmental data, dynamically adjusting the boundaries of the warning zones. This ensures that the warning range accurately matches the current actual risk level, reducing the rigidity and inaccuracy of traditional fixed electronic fences in the face of dynamically changing environments, improving the targeting of warnings, and reducing the probability of false alarms and missed alarms.

[0059] 2. This invention establishes a differentiated, tiered risk assessment mechanism, improving the precision and effectiveness of alarms. This method can distinguish whether a potential hazard is within or outside the warning zone and initiate different analysis strategies accordingly. For targets within the zone, a three-tiered alarm (high, medium, and low) is output through quantitative assessment of three-dimensional clearance distance and multiple risk factors, providing a basis for precise handling. For targets outside the zone, movement trend analysis is used for prediction, achieving clear identification and differentiated response to risks of different natures and urgency levels.

[0060] 3. This invention achieves a shift in safety monitoring from passive response to proactive prevention by predicting the risks of targets outside the warning zone. The method can analyze the movement status and direction of potential targets outside the warning zone, predict their likelihood of intrusion into the warning zone, and trigger a low-level alarm before the intrusion actually occurs. This proactive early warning capability buys valuable time for safety management, enabling risk control to intervene earlier and placing the safety defense line forward, thereby more effectively reducing the occurrence of accidents. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating a method for determining the spatial relationship of potential hazards based on hierarchical early warning, according to an embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram of the dynamic virtual warning zone boundary adjustment according to an embodiment of the present invention.

[0064] Figure 3 This is a schematic diagram of the hazard target identification and tracking according to an embodiment of the present invention.

[0065] Figure 4 This is a schematic diagram of a hazard spatial relationship determination system based on hierarchical early warning, according to an embodiment of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Reference Figure 1 One embodiment of the present invention proposes a method for determining the spatial relationship of hidden dangers based on hierarchical early warning. It adopts a data processing method that combines dynamically delineating virtual warning zones and quantitatively analyzing and predicting the trends of hidden danger targets inside and outside the zones, which can realize intelligent, accurate and forward-looking risk early warning.

[0068] Includes the following steps:

[0069] Acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data to generate dynamic virtual warning zone data;

[0070] By collecting video stream data of the scene area through monitoring equipment and performing intelligent visual analysis based on the video stream data, the system identifies and locates potential hazards and generates hazard identification data that includes the target category, target location, and target movement status.

[0071] Based on dynamic virtual warning zone data and hazard target identification data, spatial matching analysis is performed to determine whether a hazard target has intruded into the virtual electronic warning zone, and an intrusion determination result is generated.

[0072] In response to the intrusion determination result being an intrusion, the three-dimensional spatial clearance distance between the hazard target and the live conductor is calculated;

[0073] Based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, a quantitative assessment is conducted, and graded alarm results are output.

[0074] In response to the condition that the intrusion determination result is non-intrusion, the potential risk level is analyzed and predicted based on the target's movement status and target location, and an external early warning result is generated.

[0075] Specifically, based on historical experience and real-time environmental factors, a dynamically changing virtual electronic warning zone is proactively and adaptively defined as a spatial benchmark for risk assessment. Simultaneously, intelligent visual analysis technology is used to perceive the real-world scene in real time, accurately identify, locate, and track potential hazards, extracting multi-dimensional attributes such as category, location, and movement status. The core of this method lies in establishing a two-branch logical judgment mechanism, using whether a hazard intrudes into the dynamic warning zone as the dividing point. If the target intrudes into the zone, a refined quantitative assessment procedure is immediately initiated. This procedure calculates the three-dimensional spatial clearance distance between the target and the core hazard source and integrates various inherent risk factors of the target to comprehensively assess the severity level of the current hazard. If the target is outside the zone, a separate forward-looking predictive analysis procedure is initiated. This procedure analyzes the relationship between the target's movement vector and the warning zone's orientation to predict its future threat trend. This entire process achieves seamless integration from the dynamic definition of risk areas and intelligent identification of hazard targets to the differentiated analysis and tiered response of internal and external risks. Through this method, intelligent, refined, and forward-looking management of hazard spatial relationship judgment is realized. Dynamically defined virtual electronic warning zones allow for flexible adjustment of the security monitoring scope based on actual risk levels, improving the targeting and accuracy of early warnings and effectively avoiding the missed and false alarms caused by static zone divisions. Secondly, the dual-branch risk analysis logic clearly distinguishes between immediate hazards and potential developing threats, enabling drastically different response strategies for risks of different natures. For threats within the zone, the tiered alarm results output through multi-factor quantitative assessment provide managers with clear decision-making basis, achieving rational resource allocation and precise incident handling. For threats outside the zone, external warnings are generated through trend prediction, moving the security defense line forward, transforming passive response into proactive intervention, and gaining valuable time windows for eliminating potential risks. Ultimately, this method elevates the traditional single-threshold alarm mode into a comprehensive, intelligent security system capable of fully perceiving, deeply understanding, and proactively predicting risks.

[0076] In some embodiments, generating dynamic virtual warning zone data specifically includes:

[0077] Obtain historical risk data, including geographical information and frequency of occurrence of past potential hazard events;

[0078] By calling external data service interfaces, real-time environmental data containing weather conditions and terrain change information can be obtained;

[0079] A comprehensive risk score is obtained by combining historical risk data and real-time environmental data;

[0080] The boundary range of the virtual electronic warning zone is dynamically adjusted based on the comprehensive risk score, generating dynamic virtual warning zone data.

[0081] Specifically, the first step involves extracting historical risk data from a pre-set risk information database. This database includes the geographical information and frequency of past potential hazard events. It is constructed from structured data collected and recorded over a long period, including the geographical location, frequency, and type of historical events. This data originates from real operation and maintenance records and accident reports, and is integrated through manual entry or system archiving to form a static reference database for comprehensive risk assessment and dynamic zoning. Simultaneously, through pre-set external data service interfaces, such as meteorological or geological monitoring services, current real-time environmental data is actively requested and obtained. This real-time environmental data covers weather conditions in the monitored area, such as wind speed and rainfall, as well as terrain changes that may affect safety. After acquiring both types of data, a comprehensive risk assessment is performed. This assessment aims to transform multi-source, heterogeneous data into a unified quantitative indicator, namely, a comprehensive risk score. The calculation of the comprehensive risk score... The formula is expressed as:

[0082] ;

[0083] in, and These are the weighting coefficients for historical risks and environmental risks, respectively. These two coefficients are pre-set by industry safety experts based on long-term experience or through regression analysis of historical data. Together, they determine the importance of different types of risks in the comprehensive assessment, and the sum of the two is usually normalized to 1. It is a historical risk value quantified based on historical risk data, and its value is determined by analyzing the density and severity of historical potential hazards in a specific area; This is an environmental risk value quantified based on real-time environmental data. This value is obtained by mapping specific weather conditions and terrain changes to a predefined risk level lookup table. After calculating the comprehensive risk score, the boundary of a baseline virtual electronic warning zone is dynamically adjusted based on this score. For example, the higher the score, the greater the outward expansion of the warning zone's radius or polygon boundary; conversely, it may shrink. The adjusted set of warning zone boundary coordinates is then formatted and output, forming dynamic virtual warning zone data that can be used in subsequent steps. Figure 2As shown, a grid-like currently active warning zone and gray candidate warning zones are displayed. Vertices in the currently active warning zone are represented by dots, while vertices in the candidate warning zones are represented by crosses. When the risk increases, such as in severe weather or historically high-risk areas, a larger candidate warning zone is calculated. If the difference between this candidate warning zone and the currently active warning zone reaches a certain threshold, the warning zone will be updated to the new range. By combining historical risk data with real-time environmental data to dynamically generate virtual electronic warning zones, intelligent and adaptive adjustment of the warning zone range is achieved. This technology allows security warnings to move beyond static geofencing and accurately delineate areas with truly high potential risks based on historical experience and current environmental changes. This improves the targeting and accuracy of warnings, avoiding frequent false alarms due to excessively large warning zones or missed risks due to excessively small zones, making security monitoring and risk prevention more forward-looking and effective.

[0084] In some embodiments, hazard target identification data including target category, target location, and target movement status is generated, specifically as follows:

[0085] Utilize deep learning models to analyze video stream data, identify potential hazards, and output their category information;

[0086] A preset target tracking algorithm is activated to lock and track the potential target, determine its position changes in continuous video frames, and generate target position data;

[0087] The motion vector of the target is calculated based on the temporal changes of the target location data, and the target movement state data is extracted from the motion vector.

[0088] By integrating target category information, target location data, and target movement status data, hazard target identification data is generated.

[0089] Specifically, the video stream data is first input frame by frame into a pre-trained deep learning model, which can be a convolutional neural network-based object detection network. This specially trained deep learning model can automatically identify predefined potential hazards from complex video backgrounds, such as construction vehicles, cranes, or drones, and assign a clear target category to each identified hazard, while outputting its two-dimensional image coordinates and bounding box in the current video frame. Next, to ensure continuous observation of the same hazard, a pre-defined target tracking algorithm is activated, such as a tracker based on a combination of Kalman filtering and the Hungarian algorithm. This target tracking algorithm locks onto the first identified hazard and continuously predicts its position in subsequent video frames, associating and matching it with new detection results. This suppresses identity loss due to occlusion or pose changes and generates a series of time-ordered target position data representing the target's trajectory in the scene. Subsequently, the motion vector of the hazard is calculated based on this time-series target position data. The calculation of the motion vector of the hazard is then performed. The formula is expressed as:

[0090] ;

[0091] in, The calculated motion vector represents the target's velocity and direction; The spatial coordinates of the potential hazard target obtained from the target location data at the current time t; The spatial coordinates of the potential hazard target at the previous time t-1; This is the time interval between two moments, which is the reciprocal of the video frame rate. This is achieved by analyzing motion vectors. The magnitude of the target can be used to determine whether the target is in motion. If the magnitude is greater than a preset small threshold, the target's movement state is "moving," otherwise it is "stationary." By analyzing the direction of the motion vector, the target's orientation can be obtained, i.e., the target's movement state data. Finally, the target category information obtained in the previous steps, the real-time updated target location data, and the parsed target movement state data are structurally integrated and packaged into a complete set of hazard target identification data for subsequent risk analysis steps. By using deep learning and target tracking technology, automated, accurate identification and continuous tracking of hazard targets in video surveillance footage are achieved. Figure 3As shown, the image displays the movement trajectory of a potential hazard over a time series, visually representing the target's speed and direction at different time points using multiple velocity vectors. The image also indicates the current target location and its category, illustrating how to extract and integrate multi-dimensional information such as target location, movement status, and category from the video stream. Numerical pairs are labeled next to some velocity vectors, representing their X-axis and Y-axis components in the global coordinate system. This approach not only reduces the inefficiencies, fatigue, and oversights inherent in traditional manual monitoring but also extracts high-value structured information from raw, unstructured video data, including the specific type of the potential hazard, its precise spatiotemporal location, and its dynamic behavioral characteristics. This acquisition of multi-dimensional information provides a solid data foundation for subsequent complex spatial relationship determination, risk trend prediction, and tiered early warning systems. This enables the entire early warning system to react more intelligently and promptly based on a deep understanding of the scenario, improving the intelligence and predictability of risk prevention and control.

[0092] In some embodiments, generating an intrusion determination result includes:

[0093] Unify the boundary coordinates of the dynamic virtual warning zone data with the target location data in the hidden danger target identification data into a global coordinate system;

[0094] Using a pre-defined test algorithm for points within polygons, spatial relationship analysis is performed based on the target location and the boundary of the virtual electronic warning zone to obtain the spatial inclusion relationship;

[0095] Intrusion status is determined based on spatial inclusion relationships, and an intrusion determination result indicating whether the potential target is located inside the virtual electronic warning zone is generated.

[0096] Specifically, the boundary coordinates of the obtained dynamic virtual warning zone data are first converted to the target location data contained in the hazard target identification data, unifying them into a preset global coordinate system. This global coordinate system serves as a unified reference benchmark, such as a geographic coordinate system or an on-site engineering coordinate system, ensuring that spatial data from different sensors and data models can be compared and calculated without bias in the same dimension. After coordinate unification, a preset test algorithm is used to analyze the spatial relationship between the hazard target and the virtual electronic warning zone. Here, the virtual electronic warning zone is defined as a closed polygon by a series of boundary coordinate points in the global coordinate system, while the location of the hazard target is considered a geometric point in this coordinate system. The test algorithm emits a virtual ray from the target location point along an arbitrary fixed direction and counts the number of intersections between this ray and all boundaries of the warning zone polygon, thus determining the spatial inclusion relationship. If the number of intersections is odd, the target point is determined to be inside the polygon; if it is even or zero, it is determined to be outside. Finally, an intrusion status judgment is made based on the obtained spatial inclusion relationship. When the determination result indicates that the target is located within the virtual electronic warning zone, an intrusion determination result indicating "intrusion" is generated; otherwise, a result indicating "non-intrusion" is generated. This intrusion determination result is output as a clear logical signal to determine the direction of subsequent risk analysis. By establishing a unified spatial coordinate benchmark and applying a precise geometric determination algorithm, the spatial relationship between potential hazards and dynamic warning zones is automated and highly accurate. An objective mathematical correlation is established between the visually perceived target location and the abstractly defined risk zone boundary, reducing visual errors, subjective ambiguity, and reaction delays that may exist in traditional manual judgment. More importantly, the generated clear and unambiguous intrusion determination result provides a crucial logical decision-making branch point for the entire early warning system. It can effectively distinguish between two distinct risk scenarios: the immediate danger of intrusion and the potential threat of non-intrusion, and accordingly initiate differentiated analysis and response strategies, improving the systematicness, accuracy, and response efficiency of the entire risk management process.

[0097] In some embodiments, in response to the condition that the intrusion determination result is an intrusion, the three-dimensional spatial clearance distance between the potential target and the live conductor is calculated, specifically as follows:

[0098] Obtain the three-dimensional spatial coordinate information of live conductors from a pre-set power facility database;

[0099] Spatial registration and coordinate unification are performed based on the target location data and the three-dimensional spatial coordinate information of the charged conductor to construct the spatial geometric relationship between the target and the conductor.

[0100] Based on spatial geometric relationships, the straight-line distance between key points on the surface of the potential hazard target and the live conductor that meets the preset conditions is calculated and used as the three-dimensional spatial clearance distance.

[0101] Specifically, the process begins by accessing a pre-defined power facility database to obtain the three-dimensional spatial coordinates of live conductors. This database typically originates from Building Information Modeling (BIM) data or precise geographic information mapping data, providing an accurate geometric path description of the conductor in three-dimensional space, such as a series of continuous coordinate points. Simultaneously, the target location data extracted from the hazard identification data is spatially registered and coordinate unified with the obtained three-dimensional spatial coordinates of the live conductors. This involves transforming both coordinates to the same global coordinate system, ensuring all spatial data have a unified measurement benchmark. After coordinate unification, a precise spatial geometric relationship between the hazard target and the live conductor is established. At this point, the hazard target can be modeled as a set of key points representing its outer surface contour, while the live conductor is considered as one or more continuous line segments in space. Based on this, according to the constructed spatial geometric relationship, a computational geometry algorithm is used to traverse each key point on the surface of the hazard target and calculate the shortest straight-line distance from that point to the spatial line segment of the live conductor, meeting a preset condition. For the calculation of the first key point on the surface of the hazard target... Key points to line segment shortest distance The formula is expressed as:

[0102] ;

[0103] in, The first of the hidden danger target surface One key point; For line segments Up to the first Key points The nearest point; and Representing line segments respectively The starting and ending points are the three-dimensional spatial coordinates of the conductor obtained from the power facility database; and These are the minimum and maximum functions, respectively; t is the projection parameter; This refers to the dot product operation of vectors. Let be the magnitude of the vector. After calculating the shortest distance for all key points, the minimum value among all calculation results is taken, and this minimum value is used as the final output three-dimensional spatial clearance distance. By introducing precise calculation of the three-dimensional spatial clearance distance, a crucial shift has been achieved from two-dimensional plane intrusion judgment to three-dimensional spatial hazard proximity quantification. It no longer merely informs managers that "an object has intruded," but provides a precise and objective physical quantity that directly reflects the actual safety margin between the potential hazard and the live conductor. This quantitative assessment capability reduces the ambiguity and subjectivity of risk judgment, and can accurately distinguish the severity of different intrusion events. For example, even with the same intrusion, the risk level represented by a distance of several meters from the conductor is completely different from that represented by a distance of several centimeters. Therefore, the obtained three-dimensional spatial clearance distance provides the most core and reliable decision-making basis for subsequent implementation of refined graded alarms, improves the accuracy and practicality of safety early warning, and upgrades risk management from simple state monitoring to precise situation assessment.

[0104] In some embodiments, the output of graded alarm results is as follows:

[0105] By integrating and analyzing target categories, target movement status, and spatial geometric relationships, multiple key risk factors are obtained.

[0106] By combining the three-dimensional spatial clearance distance with multiple key risk factors, a risk feature vector is obtained;

[0107] Risk assessment and calculation are performed based on risk feature vectors to obtain a quantitative risk score;

[0108] Based on the risk score, the alarms are classified into three levels: high, medium, and low, and the output includes alarm type, risk description, and handling suggestions.

[0109] Specifically, the process begins by integrating multi-dimensional information to form multiple key risk factors. These factors are refined descriptions of risk source attributes, originating from sources such as target categories directly extracted from hazard target identification data (e.g., large construction machinery carries a higher risk than small drones); target movement states extracted from hazard target identification data (e.g., rapidly moving targets carry a higher risk than stationary targets); and spatial geometric relationships established during the calculation of three-dimensional clearance distances (e.g., a hazard target directly below a live conductor carries a higher risk than one located to its side). Through pre-defined rules or models, this qualitative and quantitative raw information is transformed into a set of standardized numerical risk factors. These rules or models are primarily based on domain expert knowledge and industry safety standards, such as the rule of "azimuth angle less than 90 degrees" used to determine whether a hazard target's movement trend is towards the warning zone, and spatial analysis rules that determine whether a target has entered the warning zone by calculating the geometric inclusion relationship between the target's location point and the warning zone boundary polygon. These are all logical judgment criteria directly set during the design phase based on physical laws and safety regulations. Subsequently, the core indicator calculated in the previous step, namely the three-dimensional spatial clearance distance, is combined with the multiple key risk factors just generated to construct a structured risk feature vector. This risk feature vector is a multi-dimensional array that comprehensively and digitally depicts the entirety of the current risk event. For example, this vector can contain multiple components such as [clearance distance, target category code, movement speed, and the angle between the movement direction and the guide wire], providing complete input for subsequent intelligent assessment. Next, this risk feature vector is input into a pre-trained intelligent risk assessment model for risk assessment. The model calculates the input risk feature vector and finally outputs a single, quantitative risk score. The training process of the intelligent risk assessment model first collects and organizes historical hazard event data, which forms the basis for training the intelligent risk assessment model. Each data record corresponds to a real, recorded potential hazard event, comprising three core parts: hazard target characteristics (a description of the hazard target at the time, such as the specific type of large machinery); spatial relationship data (the quantitative relationship between the target and the transmission line in the event, including three-dimensional spatial clearance, relative position, and movement status); and actual risk outcome (the actual risk level, such as high, medium, or low, assigned by experts based on the event's final impact, such as whether it causes a line trip or requires emergency shutdown). During the model training phase, supervised learning machine learning algorithms, such as support vector machines or gradient boosting decision trees, are used to train this data. The training process is essentially an optimization process, aiming to find an optimal mapping function that can accurately predict the corresponding actual risk outcome based on the input hazard target characteristics and spatial relationship data. Finally, the alarm level is determined and output based on this quantified risk score. Multiple risk score thresholds are preset internally to classify different risk levels.When the risk score exceeds the high-risk threshold, it is classified as a high-level alarm; when the score is between the medium and high-risk thresholds, it is classified as a medium-level alarm; and when it is below the medium-risk threshold, it is classified as a low-level alarm. Once the alarm level is determined, a structured hierarchical alarm result is immediately generated and output. This hierarchical alarm result is not just a simple level label, but an instruction containing rich information. Its content clearly indicates the alarm type, such as "mechanical external damage risk," a risk description of the current situation, such as "the clearance between the excavator boom and the 10kV conductor is seriously insufficient," and provides clear handling suggestions, such as "immediately notify the site to stop work and send personnel to verify," thus forming a complete and executable closed-loop alarm. By constructing a multi-dimensional risk feature vector and introducing an intelligent assessment model, a leap from single distance measurement to multi-factor comprehensive judgment has been achieved. This approach can more accurately and comprehensively assess the true severity of risks and effectively distinguish the differentiated threats posed by different types and dynamic situations of hidden danger targets. It reduces the false alarms or missed alarms that may be caused by relying solely on distance for a "one-size-fits-all" approach, making the alarm system more intelligent and refined. The final structured alarm output, which includes handling suggestions, enhances the decision support value of alarm information, enabling maintenance personnel to quickly and accurately understand the on-site situation and take the most appropriate countermeasures.

[0110] In some embodiments, generating out-of-area early warning results specifically involves:

[0111] Calculate the azimuth angle between the velocity vector corresponding to the target's movement state and a direction vector pointing from the target's position to the boundary of the virtual electronic warning zone;

[0112] The movement trend of the potential hazard target is determined based on the azimuth angle.

[0113] A low-level alarm is triggered when the movement trend is toward the virtual electronic warning zone;

[0114] When the movement trend is stationary or away from the virtual electronic warning zone, an alarm filtering command is triggered, and an external warning result is generated together.

[0115] Specifically, the first step is to predict the potential threat posed by the hazardous target. This process is achieved by analyzing the dynamic relationship between the hazardous target and the virtual electronic warning zone. The velocity vector corresponding to the target's movement state, as well as the real-time target position, is extracted from the generated hazardous target identification data. Simultaneously, based on the boundary coordinates of the target position and the dynamic virtual warning zone data, the nearest point on the polygonal boundary of the warning zone to the target position is calculated, and a direction vector pointing from the target position to this nearest point is constructed. Subsequently, the system calculates the azimuth angle between the velocity vector and the direction vector. (The calculation of the azimuth angle...) The formula is expressed as:

[0116] ;

[0117] in, The velocity vector is the velocity vector extracted from the moving state of the potential hazard target. Its direction and magnitude are determined by the rate of change of the target's position over continuous time. This is the direction vector pointing from the current target location to the nearest point on the boundary of the virtual electronic warning zone; This represents the scalar product of two vectors. Calculate the azimuth angle. Then, the movement trend of the potential hazard target is determined based on this angle value. When the target's movement state is "moving" and the azimuth angle is... When the angle is less than 90 degrees, it can be determined that the movement trend is towards the virtual electronic warning zone, and a low-level alarm will be triggered. Conversely, when the target's movement state is stationary, or its movement trend is away from the virtual electronic warning zone, i.e., the azimuth angle is less than 90 degrees, the alarm will be triggered. When the angle is greater than or equal to 90 degrees, an alarm filtering command is triggered, indicating that no alarm is needed in the current situation. These two different processing results are ultimately integrated and output to form an external early warning result that includes potential risk prediction information. By intelligently analyzing the movement trends of potential hazards outside the virtual electronic warning zone, a fundamental shift from passive response to proactive prediction has been achieved. This endows the early warning system with a forward-looking risk insight capability, enabling it to issue early warnings based on the movement intentions of potential hazards before they actually intrude into the risk area, thus gaining valuable warning and response time for safety management personnel. Simultaneously, through accurate judgment of movement direction, a large number of activities near the warning zone that pose no actual threat, such as targets moving parallel or away, can be effectively filtered out, thereby reducing the false alarm rate, making the early warning information more accurate and practically instructive, and improving the operational efficiency and reliability of the entire monitoring system.

[0118] In some embodiments, the boundary range of the virtual electronic warning zone is dynamically adjusted based on the comprehensive risk score to generate dynamic virtual warning zone data, specifically as follows:

[0119] Candidate boundaries are derived based on comprehensive risk scores.

[0120] The difference between the candidate boundary and the currently effective virtual electronic warning zone boundary is calculated to obtain the relevant difference degree;

[0121] The dynamic virtual warning area data is updated using candidate boundaries only when the relevant difference exceeds the preset change threshold.

[0122] Specifically, firstly, based on the latest calculated comprehensive risk score, a pre-defined function model is used for extrapolation. This function model aims to map the quantified comprehensive risk score to a specific geometric boundary adjustment instruction. This model could be a lookup table or machine learning model built based on expert experience and historical data. Its input is the comprehensive risk score, and its output is a distance increment or scaling factor used to adjust the size of the warning zone. For example, when the comprehensive risk score increases, the model outputs a positive distance increment, indicating that the warning zone boundary needs to expand outward; when the score decreases, it outputs a negative distance increment or a small scaling factor, indicating that the boundary needs to contract inward. Then, this adjustment is applied to a pre-defined baseline warning zone boundary or the currently effective warning zone boundary. Through a geometric buffer algorithm that expands or contracts all boundary segments at equal intervals, or by scaling the entire boundary with the geometric center of the warning zone as the origin, a candidate boundary that theoretically best matches the current risk level is generated. Next, the difference between this candidate boundary and the currently effective virtual electronic warning zone boundary is calculated to quantify the geometric differences between the two, thus obtaining a correlation difference degree. This calculation can be achieved by evaluating the proportion of the non-overlapping area of ​​two polygonal regions to their total joint area. For calculating the relevant dissimilarity... The formula is expressed as:

[0123] ;

[0124] in, The area of ​​the candidate boundary polygon itself; The area of ​​the currently active boundary polygon itself; Let be the area of ​​the intersection of the two polygons; The area of ​​the symmetric difference set representing the region enclosed by the candidate boundary and the currently effective boundary, i.e., the total area of ​​the non-overlapping portions of the two regions; This represents the union area of ​​the two regions, i.e., the total area covered by both regions. Both area values ​​are derived geometrically. This is used to obtain the relevant degree of difference. Then, it is compared with a preset change threshold. Only when the calculated relevance difference is... Only when the change threshold is exceeded is the change in risk level deemed significant enough to require adjustment of the warning zone boundary. In this case, the candidate boundary is used to update the dynamic virtual warning zone data, completing an effective boundary change. If the difference does not exceed the threshold, the currently effective boundary remains unchanged until the next evaluation. By introducing a change threshold control mechanism, an effective stabilizer is added to the adjustment process of the dynamic virtual warning zone. This effectively suppresses high-frequency, invalid oscillations or "flickering" of the warning zone boundary caused by small and frequent fluctuations in real-time environmental data, reducing the resulting waste of computing resources and visual interference to monitoring personnel, and enhancing the stability and reliability of dynamic warning zone delineation.

[0125] In some embodiments, the method further includes:

[0126] Calculate the distance from the target location to the boundary of the virtual electronic warning zone;

[0127] Based on the comparison result between the distance and a preset tolerance, the proximity state information is determined;

[0128] By incorporating near-term status information into the generation process of early warning results outside the region, the prediction process can be dynamically optimized.

[0129] Specifically, the first step is to calculate the distance from the target location to the boundary of the virtual electronic warning zone, where the distance refers to the shortest distance that meets the requirements. Specifically, the location of the potential hazard target is considered a spatial geometric point, and the boundary of the virtual electronic warning zone is considered a closed polygon composed of continuous line segments, both located in a unified global coordinate system. Then, a geometric algorithm is used to calculate the perpendicular distance from this point to all line segments constituting the polygon, or the straight-line distance to a vertex, and the minimum value is taken as the shortest distance. After obtaining this shortest distance, it is compared with a preset value, i.e., a preset tolerance. This preset tolerance defines a buffer zone adjacent to the boundary of the virtual electronic warning zone. If the calculated shortest distance is less than or equal to the preset tolerance, an proximity status information indicating that the target has entered this buffer zone is generated. Finally, this proximity status information is used as a new judgment criterion and integrated into the generation process of the external warning result. This means that when analyzing and predicting potential risk levels, not only the movement trend of the potential hazard target is considered, but also whether it is in a proximity status is assessed simultaneously. For example, when a potential target shows a tendency to move towards the warning zone and is also determined to be in an approaching state, its potential risk level will be dynamically upgraded, potentially triggering a more urgent alert than a regular out-of-area warning. By introducing the calculation of the shortest distance between the target and the warning zone boundary and the determination of its approaching state, a crucial spatial dimension is added to the prediction of out-of-area risks. This compensates for the shortcomings of relying solely on movement trend analysis, enabling not only to predict "who is coming," but also to accurately determine "who is almost there." This precise perception of nearby threats improves the sensitivity and accuracy of the prediction process, allowing the early warning system to respond more promptly and specifically to impending intrusion events, thereby building a more robust and intelligent defense before the risk actually occurs.

[0130] like Figure 4 As shown, in some embodiments, the present invention also provides a hazard spatial relationship determination system based on hierarchical early warning, the system comprising:

[0131] The dynamic warning zone delineation module is used to acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data, generating dynamic virtual warning zone data;

[0132] The hazard visual recognition module is used to collect video stream data of the scene area through monitoring equipment, and perform intelligent visual analysis based on the video stream data to identify and locate hazard targets, and generate hazard target recognition data including target category, target location and target movement status;

[0133] The intrusion determination and analysis module is used to perform spatial matching analysis based on the dynamic virtual warning zone data and the hidden danger target identification data to determine whether the hidden danger target has intruded into the virtual electronic warning zone and generate an intrusion determination result.

[0134] The clearance calculation and analysis module is used to calculate the three-dimensional spatial clearance between the potential target and the live conductor in response to the condition that the intrusion determination result is an intrusion.

[0135] The graded alarm analysis module is used to perform quantitative analysis based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, and output graded alarm results.

[0136] The external risk prediction module is used to analyze and predict the potential risk level based on the target's movement status and location, and generate an external early warning result, in response to the condition that the intrusion determination result is non-intrusion.

[0137] To verify the feasibility of this invention in practice, it was applied to a high-voltage transmission line safety monitoring scenario for a provincial power company. This power company is responsible for maintaining a 220kV high-voltage transmission line that crosses construction sites, farmland, and highways. Traditional manual inspection methods are insufficient for 24 / 7, efficient monitoring of potential hazards such as cranes, excavators, and drones, frequently resulting in power outages caused by external objects getting too close to the live conductors. The company hopes to use the method of this invention to achieve automated, intelligent, and tiered early warning systems for internal and external hazards within the transmission line.

[0138] In this embodiment, the company deployed high-definition video surveillance equipment along a selected 5-kilometer power transmission channel and applied the hazard spatial relationship determination system based on hierarchical early warning proposed in this invention. The system first dynamically generates virtual electronic warning zones through a dynamic warning zone delineation module, combining historical data on wire tangling and contact accidents with real-time meteorological data. Subsequently, a hazard visual recognition module analyzes the surveillance video to identify and track hazard targets such as large machinery and drones. When the intrusion determination analysis module determines that a target has intruded into the warning zone, the net distance calculation analysis module and the hierarchical alarm judgment module perform a quantitative assessment of the hazard level and issue an alarm; when the target has not intruded, the external risk prediction module analyzes its movement trend to predict potential risks.

[0139] To verify the beneficial effects of the present invention, a pilot application test was conducted on the 5-kilometer-long 220kV transmission channel for six months in the first half of 2025, and the operational data of each function were recorded in detail.

[0140] Regarding the generation of dynamic virtual warning zones, after the onset of the frequent thunderstorms and strong winds in March 2025, the system acquired real-time environmental data from an external data service interface, indicating that the average wind speed in the monitored area exceeded level 6 and was accompanied by heavy rainfall. Combining this with historical risk data of objects such as agricultural film and overhead wires being blown away by strong winds in the area, the system increased the comprehensive risk score from the usual 0.4 to 0.7. Based on this score, the system dynamically expanded the boundary of the virtual electronic warning zone by 3 meters from the baseline range. When wind speed fluctuated only slightly, the warning zone remained stable because the calculated boundary correlation difference did not exceed the preset 15% change threshold, effectively reducing the "flickering" problem of the warning zone caused by frequent small fluctuations in environmental data.

[0141] Regarding risk prediction outside the warning zone, at 10:15 AM on April 10, 2025, the system's hazard visual recognition module, using a deep learning model, identified a large excavator approximately 25 meters outside the warning zone boundary. The target tracking algorithm determined it was moving at a speed of approximately 5 km / h. The risk prediction module then calculated the azimuth angle between its velocity vector and the direction vector pointing towards the warning zone boundary to be 35°, determining its movement trend as "moving towards the warning zone." At this time, the system triggered a low-level alarm, generating an out-of-zone warning result and notifying the back-end monitoring center personnel to closely monitor the excavator's movement, thus providing advance notice for risk prevention. At another time, the system identified a truck traveling on a nearby road. Calculations showed its movement trend as "moving away from the warning zone," with an azimuth angle of 120°. The system then triggered an alarm filtering command, preventing false alarms.

[0142] Regarding the tiered alarm system within the zone, at 14:30 on May 22, 2025, a consumer-grade drone flew into the virtual electronic warning zone, and the intrusion judgment and analysis module generated an "intrusion" result. The system responded immediately, and the clearance calculation and analysis module retrieved the pre-stored conductor BIM data. After coordinate unification, it calculated that the three-dimensional spatial clearance between the drone and the 220kV live conductor below was 8.5 meters. The tiered alarm judgment module then constructed a comprehensive risk assessment matrix. Combining this clearance distance (less than the preset 10-meter first-level warning threshold), the target category (drone, high risk factor), and the moving state (hovering), the calculated comprehensive risk level was 78 points, matching it as a "medium-level alarm." The system immediately generated alarm details, including "Risk Description: Drone intrusion, distance to conductor 8.5 meters, risk of collision or signal interference" and "Handling Recommendation: Immediately contact the drone operator or activate countermeasures to drive it away," and pushed it to the mobile terminal of the on-site inspection personnel.

[0143] In another high-risk incident, at 9:47 AM on June 5, 2025, a crane operating at the edge of a warning zone intruded into the warning zone during a turning maneuver. Upon detecting the intrusion, the system quickly calculated that the minimum three-dimensional spatial clearance between the top of the crane boom and the live wire was only 4.2 meters, less than the preset 5-meter level-two hazard distance. In the comprehensive risk assessment, due to the extremely close proximity and the target being a large conductive metal object, the overall risk level soared to 95 points. The system issued a "high-level alarm," automatically triggering the on-site audible and visual alarms within 1.5 seconds. Simultaneously, it pushed the alarm information and on-site video feed to the regional safety supervisor with the highest priority, achieving instantaneous response and powerful intervention in the emergency.

[0144] Data comparison shows that the method of this invention demonstrates significant advantages in power transmission channel safety monitoring. Dynamic warning zones enhance the targeting of risk monitoring; the out-of-zone early warning function advances the detection time of potential risks by an average of 5-10 minutes; the tiered alarm mechanism clearly distinguishes over 90% of non-emergency intrusion events from less than 10% of critical events, allowing resources to be focused on truly high-risk scenarios. The overall alarm accuracy rate reaches 98%, and the false alarm rate is reduced compared to traditional fixed-area alarm methods.

[0145] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0146] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for determining the spatial relationship of potential hazards based on hierarchical early warning, characterized in that, The method includes the following steps: Acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data, generating dynamic virtual warning zone data, specifically as follows: Obtain historical risk data on the geographic information and frequency of past potential incidents; By calling external data service interfaces, real-time environmental data containing weather conditions and terrain change information can be obtained; A comprehensive risk score is obtained by combining the historical risk data and the real-time environmental data. Based on the comprehensive risk score, the boundary range of the virtual electronic warning zone is dynamically adjusted to generate dynamic virtual warning zone data; Video stream data of the scene area is collected by monitoring equipment, and intelligent visual analysis is performed based on the video stream data to identify and locate potential hazards, generating hazard identification data that includes the target category, target location, and target movement status. Specifically: The video stream data is analyzed using a deep learning model to identify the potential hazards and output their category information. A preset target tracking algorithm is activated to lock and track the potential target, determine its position changes in continuous video frames, and generate target position data; The motion vector of the target is calculated based on the temporal changes of the target location data, and the target movement state data is parsed from the motion vector. By integrating the target category information, target location data, and target movement status data, the hazard target identification data is generated; Based on the dynamic virtual warning zone data and the hazard target identification data, spatial matching analysis is performed to determine whether the hazard target has intruded into the virtual electronic warning zone, and an intrusion determination result is generated. In response to the condition that the intrusion determination result is an intrusion, the three-dimensional spatial clearance distance between the potential target and the live conductor is calculated; Based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, a quantitative assessment is performed, and a graded alarm result is output. In response to the condition that the intrusion determination result is non-intrusion, based on the target's movement status and the target's location, the potential risk level is analyzed and predicted, and an external early warning result is generated.

2. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 1, characterized in that, The generation of the intrusion determination result is specifically as follows: Unify the boundary coordinates of the dynamic virtual warning zone data with the target location data in the hidden danger target identification data into a global coordinate system; Using a preset test algorithm for points within polygons, spatial relationship analysis is performed based on the target location and the boundary of the virtual electronic warning zone to obtain spatial inclusion relationship; Based on the spatial inclusion relationship, an intrusion status judgment is performed, and an intrusion judgment result indicating whether the potential target is located inside the virtual electronic warning zone is generated.

3. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 2, characterized in that, In response to the condition that the intrusion determination result is an intrusion, the three-dimensional spatial clearance distance between the potential hazard target and the live conductor is calculated, specifically as follows: Obtain the three-dimensional spatial coordinate information of live conductors from a pre-set power facility database; Spatial registration and coordinate unification are performed based on the target location data and the three-dimensional spatial coordinate information of the charged conductor to construct the spatial geometric relationship between the target and the conductor; Based on the spatial geometric relationship, the spatial straight-line distance between the key points on the surface of the potential hazard target and the live conductor that meets the preset conditions is calculated and used as the three-dimensional spatial clearance distance.

4. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 3, characterized in that, The output of the graded alarm results is as follows: By integrating and analyzing the target category, the target movement state, and the spatial geometric relationship, multiple key risk factors are obtained. The risk feature vector is obtained by combining the three-dimensional spatial clearance distance with the multiple key risk factors. Risk assessment and calculation are performed based on the aforementioned risk feature vector to obtain a quantified risk score; Based on the risk score, the risk level is classified, and a graded alarm result including alarm type, risk description and handling suggestions is output.

5. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 4, characterized in that, The specific details of the early warning results generated outside the generation area are as follows: Calculate the azimuth angle between the velocity vector corresponding to the target's movement state and a direction vector pointing from the target's position to the boundary of the virtual electronic warning zone; The movement trend of the potential hazard target is determined based on the azimuth angle. When the movement trend is toward the virtual electronic warning zone, a low-level alarm is triggered; When the movement trend is stationary or away from the virtual electronic warning zone, an alarm filtering command is triggered, and an external warning result is generated together.

6. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 5, characterized in that, The process of dynamically adjusting the boundary range of the virtual electronic warning zone based on the comprehensive risk score to generate dynamic virtual warning zone data is as follows: Based on the comprehensive risk score, candidate boundaries are calculated. The difference between the candidate boundary and the currently effective virtual electronic warning zone boundary is calculated to obtain the relevant difference degree; The dynamic virtual warning area data is updated using the candidate boundary only when the correlation difference exceeds a preset change threshold.

7. The method for determining the spatial relationship of hidden dangers based on hierarchical early warning as described in claim 6, characterized in that, The method further includes: Calculate the distance from the target location to the boundary of the virtual electronic warning zone; Based on the comparison result between the distance and a preset tolerance, the proximity state information is determined; The proximity status information is integrated into the generation process of the external early warning results to dynamically optimize the prediction process.

8. A hazard spatial relationship determination system based on hierarchical early warning, applied to the hazard spatial relationship determination method based on hierarchical early warning as described in any one of claims 1-7, characterized in that, The system includes: The dynamic warning zone delineation module is used to acquire historical risk data and real-time environmental data, and dynamically delineate a virtual electronic warning zone based on the historical risk data and real-time environmental data, generating dynamic virtual warning zone data; The hazard visual recognition module is used to collect video stream data of the scene area through monitoring equipment, and perform intelligent visual analysis based on the video stream data to identify and locate hazard targets, and generate hazard target recognition data including target category, target location and target movement status; The intrusion determination and analysis module is used to perform spatial matching analysis based on the dynamic virtual warning zone data and the hidden danger target identification data to determine whether the hidden danger target has intruded into the virtual electronic warning zone and generate an intrusion determination result. The clearance calculation and analysis module is used to calculate the three-dimensional spatial clearance between the potential target and the live conductor in response to the condition that the intrusion determination result is an intrusion. The graded alarm analysis module is used to perform quantitative analysis based on the three-dimensional spatial clearance distance and multiple key risk factors extracted from the hazard target identification data, and output graded alarm results. The external risk prediction module is used to analyze and predict the potential risk level based on the target's movement status and location, and generate an external early warning result, in response to the condition that the intrusion determination result is non-intrusion.

Images