Electric fire hazard dynamic monitoring and alarm system based on big data analysis

Abstract

This invention discloses a dynamic monitoring and alarm system for electrical fire hazards based on big data analysis, belonging to the field of electrical safety monitoring and intelligent early warning technology. It includes modules for visual space modeling and interference prediction, temporal thermal anomaly extraction, structural contour recognition and consistency judgment, temperature disturbance assessment and credibility grading, composite credibility modeling and hotspot correction, and dynamic scoring and identification model update. The visual space modeling and interference prediction module acquires the field-of-view projection model of the thermal imaging monitoring area and establishes a high-risk pixel distribution model for reflection interference based on camera installation parameters and spatial geometric features. This invention constructs an intelligent identification closed-loop system from perception to decision-making, integrating spatial modeling, thermal behavior analysis, and dynamic credibility assessment to achieve accurate identification and continuous adjustment of false hotspots, effectively avoiding false alarms and missed alarms, and significantly improving the accuracy and system stability of electrical fire early warning.

Description

Technical Field

[0001] This invention relates to the field of electrical safety monitoring and intelligent early warning technology, specifically to a dynamic monitoring and alarm system for electrical fire hazards based on big data analysis. Background Technology

[0002] The "Dynamic Monitoring and Alarm System for Electrical Fire Hazards Based on Big Data Analysis" refers to an electrical safety management system that integrates multi-source sensing and data acquisition, data mining and analysis, and intelligent early warning strategies. This system collects operational data in real time by deploying various sensors, such as those for temperature, current, voltage, arcing, and residual current, on power distribution lines, electrical terminals, and electrical equipment. This data is then uploaded to a central platform for big data processing and feature analysis. The system utilizes data mining and machine learning algorithms to dynamically identify potential fire risks such as overload, poor contact, aging lines, and insulation degradation. It also constructs a hazard evolution trend model based on historical operating patterns, enabling early detection and severity assessment of fire hazards. Once the monitored data exceeds safety thresholds or a high-risk trend is predicted, the system automatically activates a tiered alarm mechanism, sending audible, visual, SMS, or cloud-based alarm signals to users or the management system. Simultaneously, it can trigger a power-off protection module, effectively preventing electrical fires and improving the inherent safety level of the electrical system.

[0003] The existing technology has the following shortcomings:

[0004] In some high-risk electrical environments, to accurately identify electrical fire hazards, systems often integrate thermal imaging modules as an auxiliary monitoring method to capture the heat distribution status of key areas such as power distribution lines, terminals, and busbar nodes in real time. However, in actual deployment, the monitored areas often contain exposed metal components (such as aluminum alloy casings and copper busbar ends) and complex cable laying structures, which can easily lead to situations where there are highly reflective metal surfaces or irregular obstructions (such as cable tags, plastic insulation sheets, rodents, and other foreign objects) within the camera's field of view. When external heat sources (such as lighting, sunlight, or high-temperature equipment) are reflected from metal surfaces, or when obstructions enter the field of view due to their own high temperature, the thermal imaging device is prone to misjudgment, incorrectly identifying non-heating areas as high-temperature hotspots. At the same time, if foreign objects obstruct key nodes where there is an actual abnormal temperature rise, the system may miss the real risk points, leading to monitoring failure.

[0005] This type of optical misjudgment behavior is characterized by its strong concealment, occasional triggering conditions, and high difficulty in error correction. Once a false alarm is detected, it may cause operational interference such as false disconnection or false alarm. If a false alarm is missed, it may lead to abnormal and continuous heating of electrical connections without timely intervention, which may cause fire accidents and seriously threaten the safety of the system and the reliability of the early warning mechanism.

[0006] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a dynamic monitoring and alarm system for electrical fire hazards based on big data analysis. It constructs an intelligent identification closed-loop system from perception to decision-making, integrating spatial modeling, thermal behavior analysis, and dynamic reliability assessment to achieve accurate identification and continuous adjustment of false hotspots, effectively avoid false alarms and missed alarms, and significantly improve the accuracy of electrical fire early warning and system stability, thereby solving the problems in the background technology mentioned above.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a dynamic monitoring and alarm system for electrical fire hazards based on big data analysis, comprising a visual space modeling and interference prediction module, a time-series thermal anomaly extraction module, a structural contour recognition and consistency judgment module, a temperature disturbance assessment and credibility grading module, a composite credibility modeling and hotspot correction module, and a dynamic scoring and identification model update module.

[0009] The spatial modeling and interference prediction module obtains the field-of-view projection model of the thermal imaging monitoring area and establishes a high-risk pixel distribution model for reflection interference based on camera installation parameters and spatial geometric features.

[0010] The time-series thermal anomaly extraction module, based on the high-risk pixel distribution model of reflection interference, collects multi-period infrared images, counts the temperature change amplitude, and extracts potential hotspot areas with unstable fluctuation characteristics.

[0011] The structural contour recognition and consistency judgment module acquires visible light images corresponding to potential hotspot areas, extracts structural contour information, and determines whether it has a morphological structure consistent with the real heat source.

[0012] The temperature disturbance assessment and credibility grading module performs pixel-level temperature disturbance response analysis based on the morphological consistency results, assesses the stability of temperature changes, and obtains the credibility level range.

[0013] Based on the credibility level range, reflection interference and occlusion features are integrated to construct a composite thermal credibility map, correct the potential hotspot judgment results and dynamically adjust the level;

[0014] The dynamic scoring and identification model update module establishes an anomaly scoring mechanism based on the time-varying trend of the composite heat credibility map and updates the pseudo-hotspot identification learning model to improve the accuracy and immunity to non-real heat sources.

[0015] Preferably, the steps for obtaining the physical field-of-view projection model of the thermal imaging monitoring area and establishing a high-risk pixel distribution model for reflection interference are as follows:

[0016] The installation height, tilt angle, azimuth angle, focal length and infrared imaging resolution parameters of the camera are collected, and combined with the power distribution equipment layout drawings and on-site survey data, a three-dimensional spatial scene model of the thermal imaging monitoring area is constructed.

[0017] Based on this three-dimensional spatial scene model, the surfaces of copper busbar terminals, aluminum alloy shells, cable brackets, and bolted connection components were identified, and their emissivity, reflectivity, and heat capacity parameters were collected to evaluate their reflection behavior under the action of typical environmental heat sources.

[0018] The reflection behavior is mapped onto the image plane, the image pixels corresponding to the high reflectivity areas are labeled, and a weighted analysis is performed in combination with lighting conditions, operating temperature and background thermal interference to form a reflection risk heatmap with a confidence score.

[0019] By training and comparing the reflection risk heat map with real thermal imaging data collected in multiple types of locations, the pixel risk level threshold and distribution weight are dynamically adjusted to generate a high-reflectivity pixel distribution model that adapts to different scenarios.

[0020] Preferably, the steps for acquiring multi-period infrared images and extracting potential hotspot regions with unstable fluctuation characteristics based on a high-risk pixel distribution model for reflection interference are as follows:

[0021] Set the image acquisition cycle and time window, acquire multi-cycle infrared image sequences, and perform image preprocessing to correct thermal drift and normalize the background temperature;

[0022] High-risk pixel regions are extracted based on the image coordinate system, a pixel time series temperature matrix is ​​constructed, and the temperature fluctuation amplitude, change gradient and range are calculated.

[0023] A fluctuation discrimination function is constructed based on temperature fluctuation characteristics to filter out abnormal pixels with drastic fluctuations or trend reversals.

[0024] Spatial clustering algorithm is applied to abnormal pixels to merge regions, extract continuous potential hotspot regions, and assign fluctuation confidence scores for subsequent analysis.

[0025] Preferably, the steps for acquiring visible light images of potential hotspot areas and determining whether they possess the same morphological structure as the actual heat source are as follows:

[0026] A pixel-level correspondence between infrared and visible light images is established through dual-light fusion imaging calibration, mapping the pixel coordinates of potential hotspot areas in the infrared image to the visible light image;

[0027] A local window is cropped from the mapped image, edge image features are extracted using an edge detection algorithm, and structural boundary information is obtained using a contour extraction method.

[0028] Based on the direction, position centroid, and shape matching degree of the structural outline, the spatial morphological consistency between the outline and the hot spot region is analyzed.

[0029] The results of the structural morphology consistency assessment are scored and integrated with the temperature fluctuation characteristics of potential hotspot areas as input for credibility assessment.

[0030] Preferably, the steps for performing pixel-level temperature perturbation response analysis on potential hotspot areas and obtaining the confidence level range are as follows:

[0031] Based on multi-period infrared image data, a temperature time series of each pixel in a potential hotspot area is constructed to fully record the temperature change process.

[0032] Perturbation sensitivity analysis is performed on the temperature sequence of each pixel to identify short-term fluctuations, detect gradient direction consistency, and extract thermal inertial features;

[0033] By aggregating the perturbation response characteristics of each pixel within the region, a temperature stability evaluation index system is constructed to form a regional response stability score.

[0034] Based on the stability score, potential hotspot areas are divided into different confidence level ranges, which serve as the basis for subsequent processing strategies and model updates.

[0035] Preferably, the steps for constructing a composite heat map and dynamically adjusting the level of potential hotspot regions are as follows:

[0036] Using the credibility level range as the initial evaluation result, we extract the reflection interference features and occlusion features corresponding to potential hotspot areas.

[0037] The reflection interference characteristics, occlusion characteristics, and temperature response stability characteristics are standardized to construct a multi-dimensional information vector.

[0038] A weighted function or hierarchical scoring mechanism is used to map multidimensional information to a unified credibility scoring range and generate a composite heat credibility map.

[0039] Based on the level change trends in the map and real-time data, the credibility level of potential hotspot areas is dynamically adjusted to achieve continuous correction of the judgment results.

[0040] Preferably, the steps of establishing a dynamic anomaly scoring mechanism and updating the pseudo-hotspot identification learning model based on the continuous change trend of the composite heat credibility map in the time dimension include:

[0041] Extract the credibility score sequence of potential hotspot areas over multiple consecutive time periods and construct the credibility evolution curve over time;

[0042] Trend analysis is performed on the change curve to identify its stability, abrupt inflection points, amplitude range, and length of stable segments, and to determine whether there are dynamic characteristics of abnormal credibility fluctuations in hotspot areas.

[0043] An anomaly scoring mechanism is constructed based on the above time behavior characteristics. Weights are set for the scoring items, and an anomaly score value is generated for each hotspot area in the current period.

[0044] Hotspot areas with abnormal scores exceeding a set threshold are marked as update samples for the learning model. Image, structure, thermal response, and interference information are extracted as input features to incrementally train and optimize the parameters of the pseudo-hotspot identification learning model, thereby improving the identification accuracy and anti-interference ability of non-real heat sources.

[0045] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0046] This invention constructs a closed-loop intelligent identification mechanism spanning the entire chain from the perception layer to the decision-making layer, encompassing spatial modeling, thermal behavior analysis, structural contour recognition, disturbance response assessment, multi-dimensional credibility modeling, and dynamic learning optimization. In particular, by introducing credibility levels, composite heat maps, and time trend scoring mechanisms, the system possesses the ability to continuously perceive and dynamically adjust the thermal behavior of hotspot areas. It can identify false hotspots caused by non-real heat sources in real time, effectively avoiding false alarms leading to accidental shutdowns or operational interruptions, as well as electrical fire hazards caused by missed alarms. This significantly improves the stability, robustness, and reliability of the early warning system. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0048] Figure 1 This is a schematic diagram of the module of the electrical fire hazard dynamic monitoring and alarm system based on big data analysis of the present invention. Detailed Implementation

[0049] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0050] This invention provides, for example Figure 1The electrical fire hazard dynamic monitoring and alarm system based on big data analysis shown includes a visual space modeling and interference prediction module, a time-series thermal anomaly extraction module, a structural contour recognition and consistency judgment module, a temperature disturbance assessment and credibility grading module, a composite credibility modeling and hotspot correction module, and a composite credibility modeling and hotspot correction module.

[0051] The visual space modeling and interference prediction module obtains the physical field-of-view projection model of the thermal imaging monitoring area and establishes a high-risk pixel distribution model for reflection interference based on the camera's installation parameters and the spatial geometric characteristics of the monitoring area.

[0052] To accurately identify and intervene in thermal imaging misjudgments during electrical fire hazard monitoring, the following steps are used to obtain the physical field-of-view projection model of the thermal imaging monitoring area, and based on this, a high-risk pixel distribution model for reflection interference is constructed. The specific steps are as follows:

[0053] Based on the physical installation parameters of the thermal imaging cameras in the monitored locations, information such as installation height, tilt angle, azimuth angle, focal length, and infrared imaging resolution is collected. Combined with on-site power distribution equipment layout drawings and actual survey data, the geometric boundary range of the camera's imaging area is determined. The spatial relationship between the camera and the monitored object is reconstructed using 3D geometric modeling, forming a projectable scene framework of the monitored area. This scene framework should fully describe the relative position, angle, and visible boundary between the camera's viewing cone and the surrounding environment, thus providing a basis for subsequent spatial calibration of reflection interference areas.

[0054] By combining geometric modeling information within the camera's field of view, surfaces in the monitored area that may exhibit high reflectivity are identified, including but not limited to copper busbar terminals, aluminum alloy casings, cable supports, and bolted connections. Material parameters (such as emissivity, reflectivity, and thermal properties) of these components are collected. Through simulation and analysis based on infrared imaging principles, the infrared reflection behavior of these objects under typical environmental heat sources (such as lighting sources, direct sunlight, and radiant heat from operating equipment surfaces) is evaluated. Combining the field-of-view projection mapping relationship, these surface areas with potential infrared reflection characteristics are projected onto the thermal imaging image plane, and pixel areas that may cause interference are initially marked.

[0055] Based on the aforementioned image projection relationship, a pixel-level spatial distribution model of high-reflectivity risk areas is constructed. Specifically, in the two-dimensional coordinate system of the thermal image, the image regions corresponding to the surfaces of highly reflective objects are calibrated to form a reflection risk marker map. Environmental parameters, including on-site lighting conditions, equipment operating temperature, and background thermal interference, are further superimposed for multi-factor weighted analysis. This analysis is used to estimate the probability intensity value of each pixel in the high-reflectivity area causing enhanced thermal reflection signals under different scenarios. Finally, a dynamically updated "high-risk pixel heatmap for reflection interference" is formed on the image plane, which assigns a confidence rating to each potential interfering pixel to indicate its likelihood of causing misjudgment in actual operation.

[0056] To improve the accuracy and versatility of this high-risk reflection interference distribution model in practical applications, the heat map model was adapted and trained with standardized scenarios of different location types (such as power distribution rooms, switching stations, and industrial plants). By comparing and validating real thermal imaging data from multiple different installation scenarios, the judgment threshold and distribution weight of pixel risk levels were continuously adjusted, improving its ability to accurately characterize the potential for misjudgment in high-reflectivity areas under complex structures. This process not only enhances the robustness of pixel spatial risk modeling but also provides reliable basic data support for subsequent hotspot identification credibility determination.

[0057] The purpose of this step is to provide a precise field-of-view mapping foundation corresponding to the actual physical space for the thermal imaging monitoring system. Based on this, it enables explicit spatial calibration and pre-identification of areas at risk of infrared reflection interference, thereby reducing the probability of misjudgments and omissions in monitoring electrical fire hazards. In actual engineering environments, thermal imaging cameras are often fixedly installed at specific angles and heights. Monitoring objects such as distribution cabinets, busbars, and copper busbar connectors have complex three-dimensional structural characteristics. Without precise modeling of the camera's field of view and the monitoring area, it is impossible to determine which image pixels correspond to which actual spatial locations, and thus, it is impossible to determine whether hotspots in the image are caused by actual heating behavior.

[0058] Therefore, this step first establishes a mapping relationship between thermal image pixels and the real three-dimensional spatial structure by acquiring the camera's installation parameters (such as pitch angle, field of view, focal length, and height) and the spatial geometric features of the monitoring site. After completing the mapping model construction, physical surfaces with high reflectivity and high risk of misjudgment (such as exposed metal surfaces, mirrored cable tags, etc.) within the monitoring area can be further identified and projected into the thermal image to form a high-risk pixel distribution model for reflection interference. This model can serve as a pre-interference filtering mechanism for subsequent identification of abnormal hotspots in infrared images, improving the contextual awareness capability of image analysis.

[0059] This step establishes a precise and quantifiable mapping between physical space and image data, enabling the system to "predict" which areas are more prone to misjudgment during the deployment phase. This allows for highly reliable intelligent monitoring, early warning, and anomaly detection. This is crucial for the practicality and accuracy of electrical fire early warning systems.

[0060] The temporal thermal anomaly extraction module, based on the high-risk pixel distribution model of reflection interference, continuously collects multi-cycle infrared images, counts the temperature change amplitude of each pixel, and extracts potential hotspot areas with unstable temperature fluctuation characteristics.

[0061] Based on the established high-risk pixel distribution model for reflection interference, multi-period infrared image acquisition and fluctuation feature extraction are performed, specifically including the following steps:

[0062] Determine the image acquisition cycle and time window parameters. Considering the thermal inertia characteristics of the electrical equipment's operating conditions, set an appropriate image acquisition frequency and time series length, acquiring a complete sequence of infrared image frames within each cycle. The image acquisition process should cover typical operating phases, including scenarios such as load increase, stable operation, and load fluctuation, ensuring data representativeness and temporal diversity. For each image frame, a standard image preprocessing workflow must be performed, including pixel thermal drift correction, background temperature mean normalization, and boundary noise filtering, to ensure that pixel temperature changes are realistic and comparable in subsequent analysis.

[0063] For each image frame, based on pixel-level spatial positioning relationships, the image region overlapping with the high-risk distribution model of reflection interference is extracted. All pixels within this region are numbered, and a time-series temperature matrix is ​​established. During this process, a two-dimensional data structure of temperature changes of pixels over time should be formed based on the image coordinate system. Subsequently, for each pixel, its corresponding temperature change trajectory is constructed, and statistical quantities such as temperature fluctuation amplitude, gradient, range, and coefficient of variation are calculated within the observation period to characterize the dynamic temperature features of that pixel.

[0064] Based on temperature statistics, an unstable fluctuation feature discrimination function is constructed. This function can set multi-dimensional discrimination conditions according to the feature value range of known pseudo-hotspots in historical data. For example, it can set the temperature fluctuation range to be greater than a certain threshold, the gradient change to show multiple directional reversals, and the short-term peak deviation from the background temperature mean to be greater than a set standard deviation multiple. Pixels that meet any one or more of the above conditions can be judged to have unstable fluctuation features. After completing the discrimination at the pixel level, a spatial clustering algorithm (such as based on connected component analysis or hot neighborhood aggregation) is further introduced to merge the regions of the pixels judged as abnormal, extracting potential hotspot regions in patches to eliminate false judgments caused by isolated points and improve the stability and structural recognizability of hotspot regions.

[0065] The unstable fluctuation feature discrimination function is a logical judgment mechanism used to identify whether pixels or regions in thermal imaging images exhibit abnormal fluctuations in temperature over time. Its core function is to distinguish between pseudo-hotspots (those experiencing non-continuous temperature rises due to external environmental influences) and stable temperature rises caused by actual heating of electrical components. This function does not directly determine the temperature level but focuses on analyzing whether the pixel temperature change pattern over time exhibits phenomena such as drastic fluctuations, irregular changes, short-period oscillations, or random reversals. When constructing this function, existing stability judgment logic in time series analysis methods can be utilized, such as moving average difference judgment, range variation assessment, and trend direction abrupt change detection, combining multiple temperature change behavior indicators as input conditions for judgment. When a pixel or region exhibits multiple sudden temperature rises, random fluctuations, or a lack of stable upward trends within a set time window, a "fluctuating and unstable" label can be triggered according to preset rules. This discrimination method can preemptively screen out pseudo-hotspots caused by occasional factors such as light reflection, ventilation fluctuations, or shading disturbances from many areas that may have thermal anomalies, improving the accuracy and computational efficiency of subsequent analysis.

[0066] For pixels identified as anomalous, region merging is performed. Common spatial clustering algorithms include connected component labeling (CBI), neighborhood density-based clustering methods (such as DBSCAN), region growing algorithms, and thermal neighborhood aggregation rules (such as those based on thermal gradient consistency and spatial proximity). The core objective of these methods is to combine adjacent anomalous pixels with similar thermal characteristics into hotspot regions with spatial continuity and consistent thermal behavior. In practice, connected component labeling can be used first to traverse all pixels identified as anomalous in the 2D thermal image. Pixels that are directly adjacent to each other in space (e.g., vertical, horizontal, or eight-neighborhood) and whose temperature differences are within a tolerable range are grouped to form continuous regions. For regions with drastic changes in thermal characteristics or blurred boundaries, region growing can be further introduced. Starting from the core high-temperature pixel, the region grows outward to surrounding pixels that meet the growing conditions (e.g., decreasing temperature but not below a certain threshold, consistent temperature gradient direction), forming more natural hotspot boundaries. When processing images with complex backgrounds or numerous scattered points, the DBSCAN clustering algorithm can be combined. By defining spatial distance thresholds between pixels and the minimum number of outlier pixels in the neighborhood, densely clustered pixel groups can be identified, automatically removing isolated points and noisy pixels from the heatmap. The synergistic application of these spatial clustering methods not only improves the structural integrity of potential hotspot areas but also enhances their spatial correspondence with real physical heat sources, providing a stable and clear image foundation for subsequent structure recognition and credibility analysis.

[0067] The extracted potential hotspot areas are characterized, numbered, and archived for subsequent analysis in structural consistency discrimination and disturbance response assessment. Simultaneously, based on features such as the proportion of high-risk pixels, overall temperature fluctuation trends, and consistency indicators within each potential hotspot area, an initial fluctuation confidence score is assigned to each area to support subsequent confidence level modeling. This process not only achieves preliminary identification and filtering of false hotspots against the background of reflection interference but also provides a data foundation for tracking the dynamic evolution of hotspots during dynamic monitoring, exhibiting good real-time performance, discriminative ability, and scalability.

[0068] The purpose of this step is to identify, from a temporal perspective, pixel regions with unstable thermal behavior and abnormal changes in multi-cycle infrared images within image areas known to potentially experience reflection interference. This allows for the early screening of potential false hotspots, providing a data foundation for subsequent structural analysis and reliability assessment. Traditional infrared image analysis often relies on single-frame images for temperature judgment, frequently misinterpreting thermal signals caused by incidental factors such as short-term sunlight reflection, transient equipment startup, or localized occlusion disturbances as actual abnormal heating, leading to false alarms or interference with linkage control mechanisms. However, by continuously acquiring multi-cycle infrared images and establishing temperature time series for pixels within each high-risk area, dynamic behavior modeling can be performed from multiple perspectives, including the magnitude of changes, trend direction, and fluctuation frequency. In particular, when a pixel region exhibits frequent sudden increases and decreases, short-term oscillations, and inconsistent thermal trends over multiple cycles, it is often a false signal caused by environmental factors rather than continuous electrical heating. This step effectively enhances the identification of false hotspots by utilizing time-domain information, reduces reliance on single-frame temperature thresholds, and provides target areas for subsequent discrimination mechanisms such as structural contour recognition and disturbance response analysis. This transforms the entire hotspot identification process from static threshold judgment to dynamic behavior recognition, improving the accuracy, stability, and anti-interference capabilities of identifying electrical fire hazards.

[0069] The structural contour recognition and consistency judgment module acquires the corresponding visible light image based on the temperature fluctuation characteristics of potential hotspot areas, extracts structural contour information, and judges whether the potential hotspot areas have a morphological structure consistent with the real heat source.

[0070] To accurately determine whether the previously extracted potential hotspot areas are caused by actual heat sources and to avoid misjudgments due to non-heat source factors such as reflection and occlusion, the corresponding visible light image is further obtained based on the temperature fluctuation characteristics of the area, and structural contour information extraction and morphological consistency discrimination are performed. Specifically, the following steps are included:

[0071] Establish a spatial correspondence between potential hotspot regions and visible light images. Since infrared and visible light images have different imaging principles and field-of-view parameters, dual-light fusion imaging calibration technology is required. This involves considering the camera's installation position, focal length differences, lens distortion parameters, and calibration point coordinates to achieve pixel-level alignment between the infrared and visible light images. After alignment, the pixel coordinates of identified potential hotspot regions in the infrared image can be accurately mapped to the corresponding visible light image, ensuring that the subsequently extracted structural information is spatially consistent with the hotspot region and avoiding misjudgments caused by misalignment analysis.

[0072] In the mapped visible light image, a local image window corresponding to each potential hotspot region is extracted, and its structural contour information is extracted using image processing methods such as edge detection, contour extraction, and gray-level gradient analysis. During the extraction process, priority should be given to identifying geometric contour shapes with typical electrical structural features, such as regularly arranged terminals, clearly defined cable insulation layers, long copper busbars, and square busbar interfaces. If there are obvious abnormal contour features around the hotspot region, such as artificial labels, insulation sheets, bracket corners, or foreign object remnants, these should be marked as potential obstructions to aid in subsequent morphological consistency assessment.

[0073] To extract the structural contour information of potential hotspot regions from visible light images, the following specific steps can be taken:

[0074] Based on the registration relationship between infrared and visible light images, the pixel coordinates of potential hotspot regions in the infrared image are mapped to the visible light image to determine their corresponding image window regions. The mapping relationship can be established through calibration point matching, binocular coordinate transformation, or affine transformation to ensure accurate spatial correspondence.

[0075] The grayscale image is extracted from the mapped local image window, and edge enhancement or noise reduction filtering (such as Gaussian blur, median filtering, etc.) is applied to eliminate texture interference. Then, edge detection algorithms, such as Canny, Sobel, or Prewitt, are used to extract the locations of strong gradient changes in the image to obtain a preliminary edge map.

[0076] Based on the extracted edge map, contour extraction algorithms (such as the findContours method in OpenCV) are used to extract closed or continuous structural boundary contours. The contours are then filtered by shape analysis methods (such as contour area, aspect ratio, boundary length, etc.) to identify object structures with regular geometric features, such as electrical components such as cables, terminals, and copper busbars.

[0077] By calculating parameters such as the orientation, centroid, and shape matching of the contours, it is determined whether these structural contours match the spatial distribution of the infrared hotspot region. If the contour structure highly overlaps with the hotspot location and has the same shape, it is considered to have the support of a real heat source structure; otherwise, it is judged as insufficient structural consistency, which may be a false heat source caused by reflection or occlusion.

[0078] By following the steps above, the understanding of the physical structure of potential hotspots can be enhanced by using visible light images, without relying on the structural information of the infrared images themselves, thus effectively improving the accuracy and intelligence of false hotspot identification.

[0079] A morphological consistency discrimination rule is constructed, which compares and analyzes the structural contour information extracted from visible light images with the hotspot shapes of potential hotspot regions in infrared images. The discrimination rule can comprehensively judge from multiple dimensions such as contour shape, edge closure degree, directional trend, and position centroid matching degree. For example, if the infrared hotspot is a strip-shaped heating area, and there is a metal conductor contour with the same direction and length at the corresponding position in the visible light image, it can be judged as having high morphological consistency; conversely, if the infrared hotspot is located at the edge of a structure but there is no obvious object contour in the visible light, or the contour is curved, irregular, or partially occluded, it can be judged that the morphological consistency of the area is poor, which may be due to reflection interference or occlusion misjudgment.

[0080] The above discrimination results are transformed into a structural morphology consistency level score, which is then fused with the dynamic instability index obtained from the previous temperature fluctuation characteristic analysis of the hotspot area to form a multi-dimensional discrimination basis, providing structural evidence for subsequent credibility assessment and false hotspot identification. Especially in environments with complex conditions and frequent structural occlusion, this structural contour analysis can not only compensate for the deficiencies in infrared image temperature data, but also further confirm whether the potential hotspot area has a spatial correspondence with the actual heat-generating component, thereby improving the accuracy, robustness, and physical interpretability of the overall identification process.

[0081] This step aims to visually verify the physical structure of potential hotspot areas by introducing visible light images. It determines whether the area possesses the geometric shape and structural characteristics expected of a genuine electrical heat source, thereby improving the accuracy and reliability of thermal imaging monitoring in complex environments. While infrared thermal imaging can directly reflect the temperature distribution on an object's surface, it has inherent limitations in recognizing reflection interference, occlusion misleading information, and structural background. Especially in environments with high-reflectivity metal components or frequent structural obstructions, thermal imaging devices can easily misinterpret short-term high-temperature signals reflected from external heat sources such as sunlight and lamplight as electrical fault hotspots. Introducing visible light images supplements the missing structural semantic information in infrared images, particularly in identifying whether hotspot locations correspond to actual conductors, terminals, busbars, load devices, and other electrical components capable of generating heat. If a potential hotspot completely overlaps and matches these typical structures in the visible light image, it indicates the presence of a genuine heat source and is highly reliable. Conversely, if there is no clear structural entity at the hotspot location, or if abnormal structures such as labels, foreign objects, or gaps are present, it is highly likely to be a false hotspot. Therefore, this step not only establishes a spatial semantic connection between image temperature anomalies and physical structures, but also realizes authenticity identification based on visual features, which is an important means to improve the anti-interference capability and discrimination accuracy of electrical fire hazard monitoring systems.

[0082] The temperature disturbance assessment and credibility grading module, based on the morphological consistency judgment results of structural contour information, performs pixel-level temperature disturbance response analysis on potential hotspot areas, evaluates the temperature change stability of potential hotspot areas, and obtains the credibility level range of potential hotspot areas.

[0083] To further enhance the ability to accurately determine the authenticity of potential hotspot areas, after determining the morphological consistency of structural contour information, pixel-level temperature perturbation response analysis is performed on the potential hotspot area based on the determination results to evaluate its temperature change stability and obtain a confidence level range accordingly. The specific steps include:

[0084] Potential hotspot areas that have passed the initial screening for structural contour consistency are selected, and temperature response models for all pixels within these areas are established. The construction of these response models is based on previously acquired multi-period infrared image data, creating a temperature time series for each pixel within the hotspot area. On the time axis, this series covers multiple operating conditions, including equipment startup, stable operation, and load fluctuations, ensuring sufficient dynamic information in the data. The core of this step is to preserve the complete behavioral process of temperature changes without simplifying or averaging the series, in order to restore the true thermal response characteristics of the pixels.

[0085] Perturbation sensitivity analysis is performed on the temperature response sequence of each pixel within the hotspot region, specifically including short-term fluctuation identification, gradient direction consistency detection, and thermal inertia feature extraction. Short-term fluctuation identification is used to detect local peak-valley changes in the temperature curve to determine whether there are frequent sudden changes; gradient direction consistency detection focuses on whether the trend of pixel temperature change is consistent with that of neighboring pixels. If there are temperature trends with opposite directions and opposite periods, it often indicates the presence of interference factors; thermal inertia feature extraction is used to quantify the slowness or lag of pixel temperature response to environmental changes. The temperature rise of real heat sources is usually relatively smooth and the response is stable, while the changes caused by reflection or occlusion are more drastic.

[0086] By comprehensively analyzing the perturbation response characteristics of all pixels within a region, a regional-level temperature stability assessment index system is constructed. This index system should include statistical quantities such as average fluctuation amplitude, inter-pixel trend synergy, and maximum perturbation frequency. Through weighted aggregation, a set of assessment values ​​reflecting the response stability of the entire potential hotspot region is formed. The assessment process should fully consider whether the regional structural contour morphology and thermal behavior are synergistic. If the regional boundary is clear, the temperature change trend is consistent, and the thermal response of each pixel is stable, the stability level is high. Conversely, if the thermal response of pixels within the region is drastically differentiated and there is no obvious central heat concentration trend, it indicates the presence of a disturbance source within the region, and the stability level is low.

[0087] Based on the stability assessment results, each potential hotspot area is divided into different credibility level intervals. The credibility level division can be further divided into multiple segments, such as highly credible, moderately credible, low-credible, and suspicious areas. Different levels correspond to different subsequent processing strategies. For example, highly credible hotspot areas can be included in the alarm mechanism as early warning candidates, while low-credible or suspicious areas require further data accumulation or the introduction of other auxiliary verification methods for re-evaluation. This credibility level not only serves as the output of a single identification but also as the input basis for subsequent dynamic scoring mechanisms and learning model adjustments, providing data support for achieving long-term stable operation and improving risk adaptability.

[0088] The purpose of this step is to further introduce dynamic response analysis of the thermal behavior of potential hotspot areas based on structural morphology assessment. Through pixel-level temperature perturbation detection, it quantifies the stability of temperature changes within the area, thus providing evidence of the area's authenticity based on its thermal behavior and ultimately assigning it a credibility level. This analysis differs from simple temperature level judgment and structural matching. It examines whether the temperature changes of each hotspot area over a continuous period exhibit regularity, continuity, and consistency to determine whether the area is caused by a real electrical heat source. Real heat sources are typically accompanied by physical characteristics such as load, contact, and resistance, and their temperature change process exhibits obvious thermal inertia, a heating path, and a stable conduction trend. False hotspots, such as reflected light sources, partial obstructions, and foreign object attachments, exhibit unstable characteristics such as abrupt changes, irregular oscillations, or significant deviations from the thermal behavior of adjacent areas during temperature changes. By comprehensively evaluating parameters such as the sensitivity of these pixel perturbations, the synergy of their change directions, and the amplitude of their periodic changes, we can not only identify areas with elevated surface temperatures but lacking physical heat sources, but also eliminate some transient anomalies caused by external interference. This provides a more accurate, dynamic, and reliable basis for hotspot identification. Finally, by dividing the reliability level range, hotspots can be clearly categorized into different levels such as high-risk, high-reliability, and low-reliability, enabling differentiated alarms, intelligent linkage control, and serving as the basic data input for subsequent model training. This step is a crucial link in the deep integration of spatial structure and temporal response data, and is a core technology for shifting from "seeing temperature" to "understanding temperature behavior," contributing to the construction of a more intelligent, robust, and effective electrical fire hazard identification mechanism.

[0089] The composite credibility modeling and hotspot correction module integrates the reflection interference characteristics and occlusion characteristics of potential hotspot areas according to their credibility level range, constructs a composite hot credibility map containing multi-dimensional interference information and response information, and corrects and dynamically adjusts the judgment results of potential hotspot areas based on the map.

[0090] To make a more comprehensive, dynamic, and adjustable assessment of the identification results of potential hotspot areas, based on the previously obtained confidence level ranges, the reflection interference characteristics and occlusion characteristics of the area are further integrated to construct a composite thermal confidence map containing multi-dimensional interference information and temperature response information. This map is then used as the basis for correcting and dynamically adjusting the hotspot determination results. The process specifically includes the following steps:

[0091] Based on the obtained confidence level range of potential hotspot areas, this level serves as the foundational weighting layer for the analysis. It is used as the initial evaluation result for subsequent cross-validation and reinforcement correction of interference factors. For each hotspot area, reflection interference characteristics are extracted from previous image processing records. These characteristics include whether the area is covered by high-reflectivity materials, whether it is near the boundary of a modeled high-risk reflection pixel distribution, and whether unnatural spikes appear in the thermal image. Furthermore, the presence of obstructions or foreign objects in the visible light image is assessed, such as cable tags, plastic sheets, dust, insects, or man-made deposits. These typically manifest as irregular outlines, blurred shapes, or interruptions in thermal signal transmission.

[0092] The aforementioned reflection interference features and occlusion features are quantized using a standardized method and then fused with the temperature response stability features of potential hotspot regions to construct a region-level multidimensional information vector. This vector includes, but is not limited to, dimensions such as temperature fluctuation intensity, response direction consistency, inter-pixel gradient co-occurrence, occlusion feature score, and reflection probability score. To ensure balanced weights and semantic relevance among the dimensions, a weighting function or hierarchical scoring mechanism can be introduced to map the multidimensional features to a unified credibility score range. This score will represent the overall thermal credibility of each hotspot region under the current environment and historical behavior.

[0093] In this process, the introduction of weighting functions or hierarchical scoring mechanisms effectively integrates heterogeneous features from multiple dimensions (such as temperature fluctuation intensity, thermal gradient consistency, reflection interference probability, and occlusion morphology influence), providing a unified measurement basis within the same credibility assessment framework and reasonably reflecting the importance of each feature in determining the true hotspot. Because different features have inconsistent numerical ranges, physical meanings, and degrees of influence on the results, direct addition or averaging can lead to information distortion or dilution of key features. Therefore, credibility mapping can be achieved through the following steps:

[0094] First, each feature is normalized to unify its numerical range, for example, by linearly scaling all feature values ​​to between 0 and 1 to eliminate the influence of dimensions. Second, a weight coefficient is assigned to each feature. The weight can be determined based on experience, statistical analysis, or machine learning training results, reflecting the importance of the feature's contribution to the recognition results in historical data. Third, a weighting function is constructed, summing the products of all normalized feature values ​​and their corresponding weights to obtain an initial confidence score between 0 and 1. If there is a hierarchical dependency between features (e.g., occlusion influence depends on reflection probability), a hierarchical scoring mechanism can be introduced to aggregate features hierarchically, such as first forming two sub-scores, "perceptual interference score" and "thermal behavior score," and then merging them into a unified total score. Finally, this total score is mapped to a discrete confidence level range (e.g., 0.8 is medium confidence, 0.4-0.6 is low confidence, and below 0.4 is suspicious), achieving a unified conversion of multi-dimensional information into a confidence judgment standard. Through this mechanism, not only can the judgment logic be quantified, but the evaluation weights can also be adaptively adjusted according to different environments and data characteristics, improving the scientific rigor and robustness of thermal confidence judgment.

[0095] Based on the fusion scoring results, a composite thermal credibility map is generated on a two-dimensional thermal image coordinate system. Each potential hotspot region is assigned a credibility level color block according to its multidimensional score, forming a continuous credibility distribution map at the image level. Different regions in the map represent their credibility levels with different colors or grayscale values; for example, high-credibility regions are marked with dark red, medium-credibility regions with orange, and low-credibility or interference risk regions with light colors or transparency. This map not only clearly visualizes the credibility levels of each hotspot within the entire monitoring range but also supports comparison and judgment between hotspots, trend tracking, and response priority ranking.

[0096] Based on the trend of level changes in the composite thermal reliability map and combined with real-time data updates, the judgment results for each potential hotspot area are corrected and dynamically adjusted. When the area's reliability score increases (e.g., interference characteristics are eliminated or the temperature response tends to stabilize), its reliability level will be automatically improved; when the interference signal strengthens, obstruction increases, or the thermal response becomes unstable, its reliability level will be automatically reduced or it will be marked as suspicious. This dynamic adjustment mechanism ensures that the hotspot judgment results are continuously optimized with changes in the operating environment and data, avoiding false alarms, missed alarms, or judgment lag caused by static judgment, and significantly improving the intelligence and adaptability of electrical fire hazard identification.

[0097] This step aims to further incorporate interference dimensions from the spatial environment, including reflection interference characteristics and occlusion features, based on the initial credibility assessment of potential hotspot areas. It fuses and judges existing thermal response results from multiple angles and sources, thereby constructing a composite thermal credibility map with dynamic expressive capabilities, spatial structure awareness, and semantic integration. Traditional infrared monitoring systems typically rely solely on temperature thresholds or simple fluctuation characteristics for judgment, neglecting physical interference factors in the monitoring environment, such as highly reflective metals, tag obstruction, cable overlap, or foreign object interference, which can severely impact image authenticity. The innovation of this step lies in extending hotspot credibility judgment beyond thermal behavior itself to a fusion analysis of thermal response and physical environment characteristics. By quantifying and integrating these spatial interference factors, it endows each hotspot area with a more complete semantic understanding and background perception capability. During map construction, the credibility level of each hotspot area is not only affected by temperature stability but also comprehensively adjusted by factors such as its proximity to reflection interference sources, the presence of visible light obstruction, and the ambiguity of structural boundaries, ultimately forming an image map with continuous credibility distribution and logical spatial consistency. Based on this data map, dynamic adjustments to the hotspot identification results can be made. For example, if the credibility of a hotspot shows a downward trend over time, its priority or alarm level will be automatically lowered, and vice versa. This mechanism ensures intelligent control capabilities with "retrospective judgment, updatable results, and adaptable strategies," significantly improving the practicality and reliability of the entire monitoring system in complex electrical environments. It also provides a solid basis for subsequent linkage control, graded response, and model optimization.

[0098] The dynamic scoring and identification model update module establishes a dynamic anomaly scoring mechanism based on the continuous change trend of the composite heat credibility map over time, and updates the pseudo-hotspot identification learning model accordingly to improve the accuracy of identifying non-real heat sources and the ability to resist misjudgment.

[0099] To further improve the accuracy and anti-interference capability of identifying non-real heat sources, based on the completed composite heat source reliability map, a dynamic anomaly scoring mechanism is established by fully utilizing its continuous changing trend over time, and the learning model for pseudo-hotspot identification is continuously optimized accordingly. This process introduces time series behavior analysis and a model adaptive update mechanism to form a dynamic heat source identification process with learning and trend awareness capabilities. Specifically, it includes the following steps:

[0100] For all potential hotspot regions identified in the composite thermal confidence profile, a time-dimensional confidence change sequence is established. During multi-cycle infrared image acquisition, the confidence scores of the corresponding hotspot regions at each time point are arranged chronologically to form a continuous curve reflecting the dynamic evolution of confidence. This curve not only reflects changes in hotspot intensity but also demonstrates the fluctuation process of confidence under the combined influence of factors such as thermal response, occlusion interference, and reflection. By extracting key features of this curve, such as the upward trend, downward slope, abrupt inflection points, and the length of stable segments, it is possible to determine whether the confidence trend of the hotspot region exhibits long-term consistency or short-term drastic fluctuations, thus establishing a preliminary description of anomaly characteristics.

[0101] To determine whether the credibility trend of a hotspot region exhibits long-term consistency or short-term drastic fluctuations, it is necessary to extract the credibility score sequence of that region over multiple consecutive time periods from the composite hotspot credibility map and perform key feature analysis on this sequence. First, observe whether the sequence shows a trend curve of gradual increase, continuous decrease, or basic stability. If the change is smooth and without abrupt inflection points, it indicates that the credibility of the region shows stability over time and possesses long-term consistency characteristics. Second, calculate the slope change value and amplitude range within consecutive time periods. If there are sharp increases or decreases, drastic slope reversals, or significant fluctuations in credibility within a numerical range (such as frequent switching between high and low credibility), it indicates that the region may be affected by sudden interference, reflected light sources, or occlusion factors, exhibiting significant short-term abnormal behavior. Furthermore, the length of the stable segment can be analyzed, i.e., the length of time during which the credibility remains relatively stable within a certain fixed interval. If the stable segment is too short or repeatedly interrupted, it also reflects that the hotspot lacks the reliability supported by a continuous heat source. By comprehensively judging these time-series characteristic indicators, it is possible to effectively identify which hotspot areas have "continuously reliable" characteristics and which belong to "dynamically abnormal" trends, thereby providing data basis and decision support for subsequent dynamic scoring and model updates.

[0102] A dynamic anomaly scoring mechanism is constructed to map the characteristics of credibility change trends into a comprehensive anomaly score. This scoring mechanism should include multiple indicator dimensions, such as trend stability, frequency of short-term anomaly outbreaks, credibility amplitude range, and cyclical change patterns, with weights assigned based on the differences in behavior between real and pseudo-hotspots in historical monitoring data. Hotspot areas are scored at each time point, and the score fluctuations over recent periods are combined to form a self-evolving anomaly scoring trend. Higher scores indicate that the area is more likely to exhibit non-real hotspot characteristics in its dynamic behavior, requiring further attention or adjustment of the identification strategy.

[0103] Based on the abnormal scoring results, hotspot regions with scores exceeding a set threshold are marked as "learning samples to be updated." Their corresponding image patches, structural contours, thermal response features, occlusion labels, and other multi-dimensional data are extracted as input features and added to the training set of the pseudo-hotspot recognition learning model. This model can be updated using supervised learning, incremental learning, or online learning, allowing it to continuously optimize its recognition boundaries under new environments and data-driven conditions, improving its ability to distinguish pseudo-hotspots under complex interference conditions. In particular, for some hotspot regions that initially have high credibility but later deteriorate, the model can actively capture their change process, forming negative samples with dynamic labeling meaning, enhancing the model's adaptability and generalization ability.

[0104] By combining the updated identification model, hotspot areas are re-identified and judged in subsequent monitoring cycles, and the results are fed back into the composite heat credibility map, realizing a new round of credibility analysis and anomaly scoring cycle. The whole process forms a dynamic closed loop: the credibility change trend drives anomaly scoring, anomaly scoring guides model updates, and model updates, in turn, improve the accuracy of credibility judgment, thereby realizing the ability to continuously learn and self-evolve in hotspot identification.

[0105] The purpose of this step is to continuously monitor the temporal evolution of potential hotspot areas in the composite thermal credibility map, capture the dynamic characteristics of credibility changes, construct an anomaly scoring mechanism reflecting their behavioral trends, and continuously optimize the pseudo-hotspot identification learning model based on this scoring mechanism, thereby improving the accuracy and immunity to non-real heat sources. Traditional thermal imaging identification methods mostly use static rules, such as temperature thresholds or single image feature judgments, which are difficult to handle complex scenarios with short-term thermal responses, intermittent reflection interference, or the influence of obstructions. The composite thermal credibility map, by integrating multi-dimensional information such as temperature response, structural morphology, reflection, and obstruction, constructs a complete credibility level evaluation system for each hotspot area. The continuous changes in this map over time reveal the trend characteristics of hotspot credibility as it changes with the external environment and its own thermal behavior. This step extracts the credibility score sequence of hotspot areas over multiple periods, identifies whether it is a continuous increase, a trend towards stability, or anomalies such as frequent fluctuations and sudden drops in credibility, and establishes an anomaly scoring mechanism to quantify the degree of behavioral anomalies from dimensions such as temporal stability, change slope, and fluctuation frequency. The scoring results are used to label potential pseudo-hotspot areas and input as training samples into the pseudo-hotspot identification learning model. By continuously accumulating high-confidence anomaly samples and boundary samples, online optimization and knowledge transfer are achieved, enabling the model to dynamically adapt to changes in interference features under different scenarios, effectively improving the sensitivity and robustness of pseudo-hotspot identification. Ultimately, this mechanism not only frees the identification algorithm from static dependence and enables it to understand trends, but also forms a self-updating and self-evolving identification capability, providing long-term stable and adaptive intelligent support for electrical fire early warning systems.

[0106] The aforementioned dynamic monitoring and alarm scheme for electrical fire hazards based on big data analysis significantly improves the accuracy and anti-interference capability of thermal imaging in complex electrical environments, solving the problems of misjudgment and missed detection caused by factors such as reflection interference and occlusion in existing technologies. This scheme constructs a closed-loop intelligent identification mechanism from the perception layer to the decision-making layer, encompassing spatial modeling, thermal image behavior analysis, structural contour recognition, disturbance response assessment, multi-dimensional credibility modeling, and dynamic learning optimization. In particular, by introducing credibility levels, composite thermal maps, and time trend scoring mechanisms, it possesses the ability to continuously perceive and dynamically adjust the thermal behavior of hotspot areas, enabling real-time identification of false hotspots caused by non-real heat sources. This effectively avoids false alarms leading to accidental disconnections or operational interruptions, as well as electrical fire hazards caused by missed alarms, greatly improving the stability, robustness, and reliability of the early warning system, and possessing broad engineering application prospects and industrialization value.

[0107] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0108] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0109] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0110] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0112] 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.

[0113] In addition, the functional units in the various embodiments of this application 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.

[0114] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0115] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A dynamic monitoring and alarm system for electrical fire hazards based on big data analysis, characterized in that, It includes modules for visual space modeling and disturbance prediction, temporal thermal anomaly extraction, structural contour recognition and consistency judgment, temperature disturbance assessment and credibility grading, composite credibility modeling and hotspot correction, and dynamic scoring and identification model update. The spatial modeling and interference prediction module obtains the field-of-view projection model of the thermal imaging monitoring area and establishes a high-risk pixel distribution model for reflection interference based on camera installation parameters and spatial geometric features. The temporal thermal anomaly extraction module, based on the high-risk pixel distribution model of reflection interference, collects multi-period infrared images, statistically analyzes the temperature change amplitude, and extracts potential hotspot areas with unstable fluctuation characteristics. The structural contour recognition and consistency judgment module acquires visible light images corresponding to potential hotspot areas, extracts structural contour information, and determines whether it has a morphological structure consistent with the real heat source. The temperature disturbance assessment and credibility grading module performs pixel-level temperature disturbance response analysis based on the morphological consistency results, assesses the stability of temperature changes, and obtains the credibility level range. Based on the credibility level range, reflection interference and occlusion features are integrated to construct a composite thermal credibility map, correct the potential hotspot judgment results and dynamically adjust the level; The dynamic scoring and identification model update module establishes an anomaly scoring mechanism based on the time change trend of the composite heat credibility map and updates the pseudo-hotspot identification learning model to improve the accuracy and immunity to non-real heat sources. The steps for establishing a dynamic anomaly scoring mechanism and updating the pseudo-hotspot identification learning model based on the continuous change trend of the composite heat credibility map over time include: Extract the credibility score sequence of potential hotspot areas over multiple consecutive time periods and construct the credibility evolution curve over time; Trend analysis is performed on the change curve to identify its stability, abrupt inflection points, amplitude range, and length of stable segments, and to determine whether there are dynamic characteristics of abnormal credibility fluctuations in hotspot areas. An anomaly scoring mechanism is constructed based on the above time behavior characteristics. Weights are set for the scoring items, and an anomaly score value is generated for each hotspot area in the current period. Hotspot regions with abnormal scores exceeding a set threshold are marked as update samples for the learning model. Image, structure, thermal response, and interference information are extracted as input features to incrementally train and optimize the parameters of the pseudo-hotspot recognition learning model.

2. The electrical fire hazard dynamic monitoring and alarm system based on big data analysis according to claim 1, characterized in that, The steps for obtaining the physical field-of-view projection model of the thermal imaging monitoring area and establishing a high-risk pixel distribution model for reflection interference are as follows: The installation height, tilt angle, azimuth angle, focal length and infrared imaging resolution parameters of the camera are collected, and combined with the power distribution equipment layout drawings and on-site survey data, a three-dimensional spatial scene model of the thermal imaging monitoring area is constructed. Based on this three-dimensional spatial scene model, the surfaces of copper busbar terminals, aluminum alloy shells, cable brackets, and bolted connection components were identified, and their emissivity, reflectivity, and heat capacity parameters were collected to evaluate their reflection behavior under the action of typical environmental heat sources. The reflection behavior is mapped onto the image plane, the image pixels corresponding to the high reflectivity areas are labeled, and a weighted analysis is performed in combination with lighting conditions, operating temperature and background thermal interference to form a reflection risk heatmap with a confidence score. By training and comparing the reflection risk heat map with real thermal imaging data collected in multiple types of locations, the pixel risk level threshold and distribution weight are dynamically adjusted to generate a high-reflectivity pixel distribution model that adapts to different scenarios.

3. The dynamic monitoring and alarm system for electrical fire hazards based on big data analysis according to claim 1, characterized in that, The steps for acquiring multi-period infrared images and extracting potential hotspot regions with unstable fluctuation characteristics based on a high-risk pixel distribution model for reflection interference are as follows: Set the image acquisition cycle and time window, acquire multi-cycle infrared image sequences, and perform image preprocessing to correct thermal drift and normalize the background temperature; High-risk pixel regions are extracted based on the image coordinate system, a pixel time series temperature matrix is ​​constructed, and the temperature fluctuation amplitude, change gradient and range are calculated. A fluctuation discrimination function is constructed based on temperature fluctuation characteristics to filter out abnormal pixels with drastic fluctuations or trend reversals. Spatial clustering algorithm is applied to abnormal pixels to merge regions, extract continuous potential hotspot regions, and assign fluctuation confidence scores for subsequent analysis.

4. The electrical fire hazard dynamic monitoring and alarm system based on big data analysis according to claim 1, characterized in that, The steps for acquiring visible light images of potential hotspot areas and determining whether they possess the same morphological structure as actual heat sources are as follows: A pixel-level correspondence between infrared and visible light images is established through dual-light fusion imaging calibration, mapping the pixel coordinates of potential hotspot areas in the infrared image to the visible light image; A local window is cropped from the mapped image, edge image features are extracted using an edge detection algorithm, and structural boundary information is obtained using a contour extraction method. Based on the direction, position centroid, and shape matching degree of the structural outline, the spatial morphological consistency between the outline and the hot spot region is analyzed. The results of the structural morphology consistency assessment are scored and integrated with the temperature fluctuation characteristics of potential hotspot areas as input for credibility assessment.

5. The electrical fire hazard dynamic monitoring and alarm system based on big data analysis according to claim 1, characterized in that, The steps for performing pixel-level temperature perturbation response analysis on potential hotspot areas and obtaining confidence level ranges are as follows: Based on multi-period infrared image data, a temperature time series of each pixel in a potential hotspot area is constructed to fully record the temperature change process. Perturbation sensitivity analysis is performed on the temperature sequence of each pixel to identify short-term fluctuations, detect gradient direction consistency, and extract thermal inertial features; By aggregating the perturbation response characteristics of each pixel within the region, a temperature stability evaluation index system is constructed to form a regional response stability score. Based on the stability score, potential hotspot areas are divided into different confidence level ranges, which serve as the basis for subsequent processing strategies and model updates.

6. The dynamic monitoring and alarm system for electrical fire hazards based on big data analysis according to claim 1, characterized in that, The steps for constructing a composite heat map and dynamically adjusting the level of potential hotspot regions are as follows: Using the credibility level range as the initial evaluation result, we extract the reflection interference features and occlusion features corresponding to potential hotspot areas. The reflection interference characteristics, occlusion characteristics, and temperature response stability characteristics are standardized to construct a multi-dimensional information vector. A weighted function or hierarchical scoring mechanism is used to map multidimensional information to a unified credibility scoring range and generate a composite heat credibility map. Based on the level change trends in the map and real-time data, the credibility level of potential hotspot areas is dynamically adjusted to achieve continuous correction of the judgment results.

Images