Fire false alarm elimination method and system based on multispectral characteristics and environmental parameters

Abstract

The invention discloses a fire false alarm elimination method and system based on multispectral features and environmental parameters, and relates to the technical field of fire monitoring. The method comprises the following steps: synchronously acquiring multispectral image data, environment data and power station operation data in a monitoring area; according to the real-time sun position information and preset photovoltaic module geometric data, a fire behavior interference area formed by sunlight mirror reflection is calculated; removing the fire behavior interference area from the first suspected fire behavior area to obtain a second suspected fire behavior area; and finally, performing fire behavior judgment on the second suspected fire behavior area by combining whether the power station operation data is abnormal or not. According to the method, dynamic glare interference is actively eliminated by constructing the physical model, and independent electrical data is introduced for cross validation, so that the fire false alarm rate is effectively reduced, the fire monitoring accuracy is improved, and high-precision and reliable monitoring of the fire of the photovoltaic power station under the complex illumination condition is realized.

Description

Technical Field

[0001] This application relates to the field of fire monitoring technology, and in particular to a method and system for eliminating false fire alarms based on multispectral features and environmental parameters. Background Technology

[0002] In recent years, in order to maximize solar energy utilization efficiency and power generation, large-scale ground-mounted photovoltaic power plants have increasingly adopted single-axis or dual-axis solar tracking systems. These tracking systems can drive rows of photovoltaic module arrays to rotate slowly according to the real-time position of the sun, maintaining the optimal angle for receiving sunlight at all times, thus improving the overall power generation efficiency of the power plant. However, this also places higher demands on the safe operation and maintenance of the power plant, especially the automated monitoring of fire situations.

[0003] To meet these requirements, current fire monitoring technologies typically rely on thermal infrared imaging systems deployed at fixed locations. This technology continuously captures thermal imaging video streams from photovoltaic arrays and uses image processing algorithms to identify potential abnormal hotspots. Considering the possibility of transient thermal interference in the environment, some solutions incorporate "temporal stability" or "dwelling time" judgment logic through algorithms. That is, after identifying a high-temperature area, the system does not immediately trigger an alarm but continuously tracks the area. Only when the high-temperature point persists at the same physical location for more than a preset time threshold (e.g., tens of seconds) will the system determine it as a stable hotspot caused by equipment malfunction, thereby triggering a fire warning.

[0004] However, in the actual operation of tracking photovoltaic power plants, as the solar tracking bracket rotates continuously, the incident angle between sunlight and the surface of the photovoltaic modules constantly changes. This causes glare points formed by specular reflection (which appear as extremely high-temperature areas in thermal infrared images) to not appear in a fixed position at a specific time, as they do in stationary power plants. Instead, they slowly "drift" across the module surface. When the drift speed is slow, the residence time of the glare point in the same area may exceed a preset time threshold, causing the system to identify the glare point as a hot spot of equipment failure, thus generating a false fire alarm and reducing the accuracy and reliability of fire monitoring. Summary of the Invention

[0005] This application provides a method and system for eliminating false fire alarms based on multispectral features and environmental parameters, which can be used to avoid false fire alarms caused by environmental interference and improve the accuracy of fire monitoring in photovoltaic power plants.

[0006] Firstly, this application provides a method for eliminating false fire alarms based on multispectral features and environmental parameters, applied to a fire monitoring system. The method includes: simultaneously acquiring multispectral image data, environmental data, and power plant operation data within the monitoring area. The multispectral image data includes at least visible light images, near-infrared images, and thermal infrared images. The environmental data includes at least real-time solar position information. The power plant operation data includes the electrical parameters of each photovoltaic module. The method further includes: extracting spectral feature data from the multispectral image data and performing spectral complementarity verification based on the spectral feature data to obtain spectral complementarity verification results. These results are used to identify image areas that meet preset fire characteristic conditions. Based on the spectral complementarity verification results, the corresponding image area in the multispectral image data is determined as a first suspected fire area. Based on the real-time solar position information and preset photovoltaic module geometric data, a fire interference area is determined, which is a specular reflection area generated by sunlight on the photovoltaic module. The fire interference area is removed from the first suspected fire area to obtain a second suspected fire area. When the power plant operation data corresponding to the second suspected fire area is abnormal, it is determined that a fire exists within the second suspected fire area.

[0007] By adopting the above technical solution, the fire monitoring system simultaneously acquires multispectral image data, environmental data, and power plant operation data. Through spectral complementarity verification, it initially screens out the first suspected fire area that meets fire characteristics. Based on real-time solar position information and preset photovoltaic module geometric data, it accurately determines the fire interference area, realizing the location of false heat sources caused by sunbeam reflection. Finally, by correlating anomalies in the power plant operation data, it confirms the actual fire, reducing false alarms caused by solar glare drifting on the tracking support, thereby improving the accuracy and reliability of fire monitoring in complex dynamic lighting environments.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, spectral complementarity verification is performed based on spectral feature data to obtain spectral complementarity verification results. Specifically, this includes: identifying a first candidate region with preset flame texture features based on spectral feature data of a visible light image; identifying a second candidate region with a brightness higher than a preset smoke brightness threshold based on spectral feature data of a near-infrared image; identifying a third candidate region with a temperature higher than a preset abnormal high temperature threshold based on spectral feature data of a thermal infrared image; calculating the spatial intersection of the first, second, and third candidate regions, and determining the image region corresponding to the spatial intersection as the spectral complementarity verification result.

[0009] By adopting the above technical solution, the fire monitoring system uses a spectral complementarity verification process to select image regions in visible light, near-infrared, and thermal infrared images that respectively satisfy the characteristics of flame texture, smoke brightness, and abnormal high temperature as candidate regions. By calculating the spatial intersection of these three candidate regions, regions that simultaneously satisfy the three strongly correlated characteristics of flame, smoke, and high temperature are selected, while non-fire interference sources that only satisfy some characteristics (such as hot spots from equipment with only high temperature or swaying reflective objects with only texture) are filtered out. This improves the accuracy of fire area identification and enhances the robustness of the fire monitoring scheme.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, after determining the fire interference area based on real-time solar position information and preset photovoltaic module geometric data, the method further includes: extracting interference spectral features corresponding to the fire interference area based on spectral feature data, the interference spectral features including the brightness, texture, and shape stability within the fire interference area; determining whether there is a preset disturbance anomaly in the interference spectral features, and identifying the image area with the disturbance anomaly as the third suspected fire area, the disturbance anomaly referring to a sharp decrease in brightness within a short period of time, or the destruction of the stability of its texture and shape; when there is an anomaly in the power plant operation data corresponding to the third suspected fire area, determining that there is a fire within the third suspected fire area.

[0011] By adopting the above technical solution, the fire monitoring system does not simply ignore the fire interference area, but further monitors the stability of its interference spectral characteristics. When the fire interference area shows preset disturbance anomalies such as a sharp decrease in brightness or a disruption of texture shape stability, the fire monitoring system will identify it as a third suspected fire area, thereby effectively identifying the real fire occurring within the solar glare area. This solves the problem of missed detection when the fire and interference overlap in space, ensuring coverage of all potential fires while efficiently eliminating interference, thus improving the comprehensiveness and reliability of fire monitoring.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the first suspected fire area is removed from the fire interference area to obtain the second suspected fire area. Specifically, this includes: obtaining the first pixel coordinate set corresponding to the first suspected fire area and the second pixel coordinate set corresponding to the fire interference area based on multispectral image data; calculating the spatial difference between the first pixel coordinate set and the second pixel coordinate set; and determining the image area corresponding to the spatial difference as the second suspected fire area.

[0013] By adopting the above technical solution, the fire monitoring system defines the first suspected fire area and the fire interference area as the first pixel coordinate set and the second pixel coordinate set, respectively, and calculates the spatial difference between the two. This achieves the removal of the interference area, ensuring that only pixels that completely overlap with the interference are removed. This avoids the possibility of false removal or interference residue that may be caused by using coarse area calculations, thus ensuring the overall effectiveness of the fire monitoring solution.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, when the power plant operation data corresponding to the second suspected fire area is abnormal, it is determined that there is a fire in the second suspected fire area. Specifically, this includes: obtaining the start timestamp of the second suspected fire area and obtaining the equipment identifiers of the photovoltaic modules within the coverage area of ​​the second suspected fire area; extracting the electrical parameters corresponding to the second suspected fire area based on the equipment identifiers and the power plant operation data; checking whether there are parameter abrupt changes in the electrical parameters within a preset time window adjacent to the start timestamp; when there are parameter abrupt changes, and the deviation between the occurrence time of the parameter abrupt changes and the start timestamp is less than a preset time threshold, confirming that the power plant operation data corresponding to the second suspected fire area is abnormal, and determining that there is a fire in the second suspected fire area.

[0015] By adopting the above technical solution, the fire monitoring system establishes a strict spatiotemporal correlation between visual phenomena and physical faults by comparing the start timestamp of the second suspected fire area with the time point when electrical parameters undergo a sudden change. Considering the physical characteristic of fire causing instantaneous damage to electrical performance, the time deviation between the two events must be less than a preset time threshold. This logic of synchronously verifying image evidence and power plant operation data within a very small time scale improves the authenticity of fire monitoring results and eliminates visual false alarms caused by non-electrical faults.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, before obtaining the device identifier of the photovoltaic modules within the coverage area of ​​the second suspected fire area, the method further includes: obtaining the pixel position of the second suspected fire area in multispectral image data and preset photovoltaic module geometric data, the preset photovoltaic module geometric data including the physical boundary coordinates of the photovoltaic modules in the photovoltaic power station; when the pixel range of the second suspected fire area does not overlap with the preset photovoltaic module geometric data, the second suspected fire area is determined as a non-module suspected fire area; within a preset time window, based on the multispectral image data, extracting the temporal dynamic features of the non-module suspected fire area, the temporal dynamic features including at least thermal radiation intensity value and pixel contour area value; when the temporal dynamic features are greater than a preset fire judgment threshold, it is determined that a fire exists within the non-module suspected fire area.

[0017] By employing the above technical solution, the fire monitoring system distinguishes the location of a fire by determining whether a second suspected fire area overlaps with the preset geometric data of photovoltaic modules. For targets identified as non-module suspected fire areas, the fire monitoring system analyzes their temporal dynamic characteristics, confirming the fire by judging whether their thermal radiation intensity value and pixel outline area value continuously increase over time. This approach can confirm equipment fires using electrical data and environmental fires using dynamic visual features, effectively expanding the monitoring range and scene adaptability of the method.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, obtaining the device identifier of photovoltaic modules within the coverage area of ​​the second suspected fire area specifically includes: setting physical world coordinates based on preset photovoltaic module geometric data, and identifying the positioning pixel coordinates corresponding to the physical world coordinates in multispectral image data, wherein the physical world coordinates represent at least four reference points set according to the physical features on the photovoltaic module; calculating coordinate transformation parameters based on the correspondence between the physical world coordinates and the positioning pixel coordinates, wherein the coordinate transformation parameters are used to convert the pixel coordinates into physical world coordinates; performing coordinate transformation calculation on the pixel coordinates corresponding to the second suspected fire area according to the coordinate transformation parameters to obtain the geographical location coordinates of the second suspected fire area; matching the geographical location coordinates with the preset photovoltaic module geometric data to determine the corresponding photovoltaic module, and obtaining the device identifier of the photovoltaic module.

[0019] By employing the aforementioned technical solution, the fire monitoring system identifies the location pixel coordinates in an image and their corresponding physical world coordinates, calculating precise coordinate transformation parameters to establish a reliable mapping from a two-dimensional image to three-dimensional space. Using these coordinate transformation parameters, the pixel coordinates of the second suspected fire area are converted into geographical location coordinates and matched with preset photovoltaic module geometric data, thereby uniquely identifying the affected photovoltaic modules. This precise positioning capability is a crucial prerequisite for subsequent correlation and querying of power plant operation data, ensuring the targeted nature and accuracy of data verification, and guaranteeing the reliability of high-precision fire confirmation.

[0020] Secondly, this application provides a fire monitoring system, which includes: one or more processors and a memory; the memory is coupled to one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the fire monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a fire monitoring system, cause the fire monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, this application provides a computer program product that, when run on a fire monitoring system, causes the fire monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the fire monitoring system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. By employing a method that determines the fire interference area based on the real-time solar position information and preset photovoltaic module geometric data, and then removing the fire interference area from the first suspected fire area to obtain the second suspected fire area, and combining this with the power station operation data to determine the fire situation, the method can accurately predict and eliminate dynamic glare formed by the reflection of the sun's surface through a physical model, and cross-validate using independent electrical data. This effectively solves the problem in related technologies where solar glare caused by the rotation of the tracking bracket is misjudged as a stable hot spot, resulting in false fire alarms. Thus, it achieves high-precision monitoring of photovoltaic power station fires under complex dynamic lighting conditions.

[0026] 2. By employing the method of extracting the interference spectral features corresponding to the fire interference area and determining whether the interference spectral features have preset disturbance anomalies to identify the third suspected fire area, it is possible to monitor the optical stability of the solar glare while using it as an interference source. This allows for timely capture of drastic changes in brightness, texture, and other features caused by the superposition of real fires, effectively solving the problem of potentially missing real fires occurring in the area when directly excluding interference areas. Consequently, it achieves comprehensive coverage of fires in overlapping areas while efficiently filtering out interference, thus improving the reliability of monitoring.

[0027] 3. Because it adopts a method that extracts the temporal dynamic features of the non-module suspected fire area when the pixel range of the second suspected fire area does not overlap with the preset photovoltaic module geometric data, and determines the fire based on whether the features exceed the preset fire judgment threshold, it can provide an independent confirmation logic based on visual dynamic development trends for suspected fires occurring in areas outside the photovoltaic modules. This effectively solves the problem that relying solely on power plant operation data for confirmation is insufficient to determine environmental fires (such as grassland or cable fires), thereby expanding the monitoring range from the equipment itself to the entire site environment and building a more comprehensive fire protection capability. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a fire false alarm elimination method based on multispectral features and environmental parameters in an embodiment of this application.

[0029] Figure 2 This is another flowchart illustrating the fire false alarm elimination method based on multispectral features and environmental parameters in the embodiments of this application;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a fire monitoring system in an embodiment of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] The following describes the process of the method provided in this implementation. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a fire false alarm elimination method based on multispectral features and environmental parameters in an embodiment of this application.

[0034] S101. Simultaneously acquire multispectral image data, environmental data, and power plant operation data within the monitoring area. The multispectral image data includes at least visible light images, near-infrared images, and thermal infrared images. The environmental data includes at least real-time solar position information. The power plant operation data includes the electrical parameters of each photovoltaic module.

[0035] Among them, multispectral image data represents a set of image information covering different electromagnetic spectrum bands, collected by imaging devices integrating multiple sensors at the same time or within a very short time interval; environmental data refers to various parameters related to the environmental state of the monitored area, used to assist in the analysis of image data; power plant operation data refers to a set of parameters reflecting the real-time working status of power plant equipment, obtained from the monitoring and data acquisition (SCADA) system or other monitoring platforms of photovoltaic power plants; visible light images refer to images taken in the electromagnetic spectrum range visible to the human eye (approximately 400-760nm), mainly used to capture the color, shape, and texture details of objects; near-infrared images refer to images taken in the near-infrared band (approximately 760-2500nm), which have high reflectivity to smoke particles and are therefore often used for smoke detection; thermal infrared images refer to images generated by detecting the infrared radiation emitted by the object itself (usually in the wavelength range of 8-14μm), whose pixel values ​​are directly related to the surface temperature of the object, used to identify abnormal heat sources; real-time solar position information is used to represent the real-time azimuth and elevation angles of the sun in the celestial coordinate system, which is a key input for determining the position of sunlight reflection.

[0036] When monitoring a fire, the fire monitoring system acquires the necessary information and data for fire assessment from front-end sensors and third-party systems. Specifically, the system uses its integrated multispectral imaging module (containing side-by-side or coaxial visible light, near-infrared, and thermal infrared cameras) to capture images of the monitored area, obtaining registered and aligned multispectral image data. Simultaneously, the system calculates and obtains the current real-time solar position information through its built-in GPS module and attitude sensor, or by connecting to external astronomical algorithm services. Furthermore, the system establishes a communication connection with the photovoltaic power plant's backend data center (e.g., via industrial protocols such as Modbus and OPC UA) to retrieve real-time power plant operation data corresponding to each photovoltaic module within the monitored area. To ensure data validity, "synchronous acquisition" is crucial in this step, requiring the system to assign a unified or correlated timestamp to all collected data from different sources (images, environment, operation), ensuring that all data are strictly corresponding in the time dimension during subsequent analysis, providing a foundation for multi-information fusion and judgment.

[0037] Optionally, in some embodiments, the fire monitoring system incorporates a main controller, which is connected to the multispectral imaging unit, the environmental data acquisition module, and the power plant operation data interface via a hardware trigger signal line. At the beginning of each monitoring cycle, the main controller synchronously sends out trigger pulses. Upon receiving the trigger pulses, each module immediately performs data acquisition operations, attaches the same trigger sequence number or timestamp to the acquired data packets, and then sends them back to the main controller for data aggregation.

[0038] S102. Extract the spectral feature data of the multispectral image data, and perform spectral complementarity verification based on the spectral feature data to obtain the spectral complementarity verification result. The spectral complementarity verification result is used to identify the image area that meets the preset fire feature conditions.

[0039] Among them, spectral feature data refers to the set of numerical values ​​or vectors extracted from the original multispectral image data that can quantitatively describe specific physical phenomena (such as flames, smoke, and high temperatures); spectral complementarity verification refers to a verification strategy that fuses multi-source information. Based on the principle that real fires usually exhibit multiple related features (such as high temperatures, flame textures, and smoke) simultaneously in different spectral bands, cross-validation is used to confirm whether a suspicious phenomenon meets all fire features, thereby improving the accuracy of identification; preset fire feature conditions refer to a set of predefined rules and thresholds used to determine whether an image region has fire features.

[0040] In some embodiments, the setting scheme of the preset fire characteristic conditions is usually based on statistical analysis and machine learning training of a large amount of image data of real fires and interference objects (such as lights and reflections). For example, the flame texture condition of the visible light image can be set to the energy ratio within a specific frequency range, the smoke brightness condition of the near-infrared image can be set to be 3 standard deviations higher than the average brightness of the surrounding environment, and the high temperature condition of the thermal infrared image can be set to a specific value (such as 150°C) higher than the upper limit of the normal operating temperature of the photovoltaic module (such as 85°C).

[0041] After acquiring synchronous multi-source data, the fire monitoring system filters areas that initially match fire characteristics. Specifically, the system first extracts features from the three types of synchronously acquired spectral image data. For example, for visible light images, wavelet transform or a deep learning-based texture analysis model is used to extract the flickering and irregular texture features unique to flames; for near-infrared images, high-pass filtering or direct brightness threshold segmentation is performed to highlight high-reflectivity smoke areas; for thermal infrared images, temperature calibration is performed to identify all pixels whose temperature values ​​exceed a preset abnormal high-temperature threshold. After feature extraction, the fire monitoring system performs spectral complementarity verification based on the extracted feature data, calculating the spatial intersection of these feature regions identified in different spectral images. This yields an image region that simultaneously exhibits abnormal high temperature in the thermal infrared image, flame texture in the visible light image, and is identified as a bright area (representing smoke) in the near-infrared image. This image region satisfies all preset fire characteristic conditions and is thus recorded in the spectral complementarity verification results.

[0042] S103. Based on the spectral complementarity verification results, the corresponding image region in the multispectral image data is determined as the first suspected fire area.

[0043] The first suspected fire area refers to the candidate fire area obtained after preliminary screening using multispectral features. It represents an area with a high probability of fire, but has not yet ruled out false alarms caused by specific environmental factors such as sunglass reflection.

[0044] The fire monitoring system performs this step after completing spectral complementarity verification. Specifically, the fire monitoring system receives the spectral complementarity verification result (e.g., a binary mask image). The fire monitoring system performs connected component analysis on this mask image. Each independent connected component (i.e., a continuous block of pixels that meets the verification conditions) is considered an independent potential fire event. The fire monitoring system generates a unique identifier (ID) for each connected component and calculates its key attributes, such as the bounding box, centroid coordinates, and pixel area. This image region, assigned an ID and attributes, is identified and named the "first suspected fire region" and added to a pending list, awaiting subsequent false alarm elimination and confirmation processes.

[0045] S104. Based on real-time solar position information and preset photovoltaic module geometric data, determine the fire interference area, which is the area of ​​specular reflection generated by sunlight on the photovoltaic module.

[0046] Among them, real-time solar position information refers to data describing the precise position of the sun in the sky at the current moment, usually expressed as azimuth and altitude angles; preset photovoltaic module geometric data represents static information pre-entered during the deployment phase of the fire monitoring system, describing the physical properties of all photovoltaic modules in the monitoring area, such as the precise three-dimensional spatial coordinates, size, tilt angle, orientation, and optical reflection characteristics of the surface material of each module; fire interference area refers to a specific area on the two-dimensional image plane that may be misjudged as a fire due to the specular reflection of sunlight; specular reflection area is used to represent the physical areas on the surface of the photovoltaic modules in the three-dimensional physical world that can directly reflect sunlight into the lens of the fire monitoring system imaging equipment according to the laws of optical reflection.

[0047] The fire monitoring system performs this step after acquiring environmental data. Specifically, the system uses three-dimensional spatial geometric calculations to accurately predict the occurrence of interference. First, a three-dimensional scene model is constructed, including the sun, photovoltaic modules, and imaging equipment. Based on the input real-time sun position information, the incident direction vector of sunlight is determined. Based on preset photovoltaic module geometric data, the position, orientation, and surface normal vector of each photovoltaic module in three-dimensional space can be determined. According to the specular reflection law in physics (i.e., the angle of reflection equals the angle of incidence, and the reflected ray, incident ray, and normal lie in the same plane), the fire monitoring system can calculate the outgoing direction of reflected light for each point on the photovoltaic module. By traversing the surface of all photovoltaic modules, the system selects all points whose reflected light direction points precisely to the position of the imaging equipment; these points physically constitute specular reflection areas. Finally, the system uses camera calibration parameters to project these three-dimensional specular reflection areas onto a two-dimensional image plane, thus obtaining the final fire interference area.

[0048] In this embodiment, the fire monitoring system not only calculates the static specular reflection area at the current moment, but also predicts the moving speed vector and moving trajectory of the specular reflection area on the image plane within a preset time period (such as the next 1-5 seconds) based on a solar trajectory algorithm (such as the SPA algorithm).

[0049] The fire monitoring system compares the centroid displacement characteristics of the first suspected fire area in the image with the predicted movement trajectory of the specular reflection area:

[0050] If the correlation between the motion vectors (direction and velocity) of the two is higher than the preset threshold (i.e., the hotspot moves with the angle of the sun), it is judged as strong interference and is directly eliminated.

[0051] If the first suspected fire area is stationary or exhibits non-directional spread (characteristics of fire spread) with the center of mass as the origin, which does not match the drift trajectory of solar reflection, it will be retained.

[0052] By introducing a spatiotemporal motion consistency test, a real fire source that is fixed / spreading is distinguished from a dynamic glare that drifts slowly with the sun's angle. In particular, it solves the risk of false rejection when a real fire happens to occur on the path of the glare.

[0053] S105. Remove the fire interference area from the first suspected fire area to obtain the second suspected fire area;

[0054] Among them, the second suspected fire area refers to the candidate fire area that is more suspected and needs further verification after the operation of eliminating fire interference areas.

[0055] The fire monitoring system performs this step after obtaining the first suspected fire area and the fire interference area. Specifically, the fire monitoring system performs spatial logical operations on the two areas in the image coordinate system. The first suspected fire area is considered as a pixel set A, and the fire interference area is considered as another pixel set B. The difference between set A and set B is calculated to obtain a new pixel set C, where all pixels in set C belong to A but not to B. The image area corresponding to this new pixel set C is defined as the second suspected fire area.

[0056] S106. When there is an anomaly in the power plant operation data corresponding to the second suspected fire area, it is determined that there is a fire in the second suspected fire area.

[0057] Among them, "abnormal" is used to indicate that the power plant's operating data has changed abruptly, which indicates physical damage or serious failure, such as the output power dropping to zero instantly, or abnormal and drastic changes in current or voltage.

[0058] The fire monitoring system executes this step after successfully identifying one or more non-empty second suspected fire areas. Specifically, the fire monitoring system first determines the device identifier of the photovoltaic modules covered by the second suspected fire area through coordinate mapping. Then, based on the device identifier, the fire monitoring system extracts the output power sequence of the photovoltaic modules within a preset time window before and after the start time of the second suspected fire area from the power plant operation database. The fire monitoring system analyzes this output power sequence. If it finds that the power value at a certain time point has decreased by more than a preset abrupt change amplitude compared to the previous time point (e.g., a decrease of more than 80%), and the deviation of this time point from the start timestamp is less than a preset time threshold, then it determines that a fire exists within the second suspected fire area.

[0059] In this embodiment, by simultaneously acquiring multispectral images, environmental data, and power plant operation data, identifying the first suspected fire area through spectral complementarity verification, accurately determining the fire interference area based on real-time solar position information and photovoltaic module geometric data, eliminating the fire interference area from the first suspected fire area to obtain the second suspected fire area, and finally confirming it by combining abnormal power plant operation data, a complete fire confirmation chain from multidimensional perception and physical model elimination to data correlation verification can be constructed. This effectively solves the problem in related technologies that rely solely on single thermal imaging and time-series judgment, which cannot distinguish false heat sources formed by the drift of sunglass reflections on the tracking bracket, resulting in low accuracy and high false alarm rate in fire monitoring. Thus, it achieves automated identification of photovoltaic power plant fires and reliable operation with an extremely low false alarm rate.

[0060] Based on the above embodiments, the method provided in this embodiment will be described in further detail below. Please refer to... Figure 2This is another flowchart illustrating the fire false alarm elimination method based on multispectral features and environmental parameters in the embodiments of this application.

[0061] S201. Simultaneously acquire multispectral image data, environmental data, and power plant operation data within the monitoring area. The multispectral image data shall include at least visible light images, near-infrared images, and thermal infrared images. The environmental data shall include at least real-time solar position information. The power plant operation data shall include the electrical parameters of each photovoltaic module.

[0062] Step S201 and Figure 1 The description of step S101 in the embodiment is similar and will not be repeated here.

[0063] S202. Extract the spectral feature data of the multispectral image data, and perform spectral complementarity verification based on the spectral feature data to obtain the spectral complementarity verification result. The spectral complementarity verification result is used to identify the image area that meets the preset fire feature conditions.

[0064] This step specifically includes:

[0065] Based on the spectral feature data of visible light images, a first candidate region with preset flame texture features is identified;

[0066] Based on the spectral feature data of near-infrared images, a second candidate region with a brightness higher than a preset smoke brightness threshold is identified;

[0067] Based on the spectral feature data of thermal infrared images, a third candidate region with a temperature higher than a preset abnormal high temperature threshold is identified.

[0068] Calculate the spatial intersection of the first candidate region, the second candidate region, and the third candidate region, and determine the image region corresponding to the spatial intersection as the spectral complementarity verification result.

[0069] The first candidate region refers to the set of all pixels in the visible light image that satisfy the preset flame texture features; the spectral feature data of the near-infrared image represents the pixel brightness information extracted from the near-infrared band image; the preset smoke brightness threshold is a brightness boundary value used to distinguish smoke from the background, its function is to filter out areas with abnormally high brightness in the image as potential smoke areas, and the setting of this threshold can be dynamically calculated according to the overall brightness distribution of the current image (for example, set to the global or local brightness mean of the image plus three times the standard deviation); the second candidate region is used to represent the region in the near-infrared image where all pixel brightness values ​​are higher than the preset smoke brightness threshold; the spectral feature data of the thermal infrared image... The information refers to the radiation intensity information directly related to the surface temperature of an object, extracted from thermal infrared images; the preset abnormal high temperature threshold is a critical temperature value used to determine whether an object is in an abnormally high temperature state. The setting scheme of this threshold should combine the upper limit of the normal operating temperature of photovoltaic modules (e.g., 85℃) and the starting temperature of a fire, and set a value significantly higher than the normal operating condition, such as 150℃, to ensure a high detection rate and a low false alarm rate; the third candidate region refers to the pixel region in the thermal infrared image where all temperature values ​​are higher than the preset abnormal high temperature threshold; spatial intersection represents the part that overlaps when multiple regions (pixel sets) are logically ANDed in the same coordinate system.

[0070] After acquiring synchronized multispectral image data, the fire monitoring system performs this step. By leveraging the differences in the physical characteristics of fires across different spectral bands, it constructs a multi-dimensional, complementary verification logic to filter out potential fires. Specifically, the fire monitoring system decomposes the entire process into three parallel sub-tasks and a final fusion step. First, image processing algorithms (such as color space analysis, edge detection, and Fourier transform) are applied to the visible light image to identify regions with preset flame texture features, marking them as first candidate regions. Simultaneously, the near-infrared image is processed, and adaptive thresholding is used to identify all pixels with brightness higher than a preset smoke brightness threshold, forming second candidate regions. Similarly, the temperature-calibrated thermal infrared image is thresholded, identifying all points with temperatures higher than a preset abnormal high temperature threshold, constituting third candidate regions. Finally, after all candidate regions have been identified, the fire monitoring system performs pixel-level spatial intersection calculations on these three independent candidate regions (usually three binary mask images) to obtain simultaneously overlapping image regions, which are then saved in the spectral complementarity verification results.

[0071] S203. Based on the spectral complementarity verification results, the corresponding image region in the multispectral image data is determined as the first suspected fire area.

[0072] S204. Based on real-time solar position information and preset photovoltaic module geometric data, determine the fire interference area, which is the area of ​​specular reflection generated by sunlight on the photovoltaic module;

[0073] Steps S203 and S204 and Figure 1 Steps S103 and S104 in the embodiment are described similarly and will not be repeated here.

[0074] S205. Based on the spectral feature data, extract the interference spectral features corresponding to the fire interference area. The interference spectral features include the stability of brightness, texture and shape within the fire interference area.

[0075] The fire monitoring system performs this step after identifying the fire interference area. Because real flames are unstable during combustion, their brightness, texture, and shape exhibit dramatic and irregular dynamic changes. While glare points formed by sunglass reflections may drift due to the rotation of the solar tracking bracket, their optical properties remain relatively stable in the image for a short period (several seconds or tens of seconds). Specifically, the fire monitoring system locks onto the identified fire interference area and extracts the corresponding spectral features from the multispectral image data within a short time window (e.g., 10 frames). For each frame, the fire monitoring system calculates the average brightness value, texture feature vector, and shape parameters within the fire interference area.

[0076] S206. Determine whether there is a preset disturbance anomaly in the interference spectral features, and identify the image area with the disturbance anomaly as the third suspected fire area. The disturbance anomaly refers to the drastic decrease in brightness of the fire interference area in a short period of time, or the destruction of the stability of its texture and shape.

[0077] Among them, the preset disturbance anomaly represents a predefined rule or model used to determine whether the interference spectral characteristics exhibit a drastic change pattern that is inconsistent with stable glare. It indicates that the area may not be a simple specular reflection, but may be superimposed with a real fire event.

[0078] After extracting the interference spectral features, the fire monitoring system performs this step to avoid overlooking potential real fires in areas prone to interference. Specifically, the system compares the time series of interference spectral features with preset anomaly conditions. For example, it calculates the first-order difference of the brightness sequence; if the absolute value of the difference exceeds a preset brightness attenuation threshold at some point, it considers a brightness anomaly to exist. Simultaneously, it calculates the distance between texture feature vectors and the similarity of shape contours between adjacent frames; if these indicators exceed their respective stability thresholds, the stability of the texture or shape is considered compromised. For any fire interference area whose interference spectral features meet one or more of the above anomaly conditions, the fire monitoring system removes the image area from the interference area list and re-marks it as a new suspected area, i.e., the "third suspected fire area."

[0079] Specifically, determining whether the interference spectral features exhibit any abnormal disturbances also includes performing time-domain to frequency-domain transformation analysis:

[0080] For pixels falling into the fire interference area, collect their brightness change time series data within a preset short time window and perform Fast Fourier Transform (FFT).

[0081] Pure glare: Since the reflection is mainly dominated by the DC component, its spectral energy is mainly concentrated in the low frequency band (close to 0Hz), and the high frequency component is extremely low;

[0082] Flame glare: Real flames, due to turbulence and irregular flickering caused by incomplete combustion, exhibit significant energy peaks in brightness variations within a specific frequency range (typically 1-10Hz), which is the flame flicker frequency.

[0083] If the proportion of high-frequency energy in the area exceeds the preset turbulence threshold, it is determined that a real fire is superimposed in the interference area.

[0084] S207. When there is an anomaly in the power plant operation data corresponding to the third suspected fire area, it is determined that there is a fire in the third suspected fire area.

[0085] The fire monitoring system performs this step after identifying a third suspected fire area. Specifically, it is related to... Figure 1 The description of step S106 in the above embodiments is similar, both involving confirming the fire by associating abnormal electrical parameters, and will not be repeated here.

[0086] S208. Remove the fire interference area from the first suspected fire area to obtain the second suspected fire area;

[0087] This step specifically includes:

[0088] Based on the multispectral image data, obtain the first set of pixel coordinates corresponding to the first suspected fire area and the second set of pixel coordinates corresponding to the fire interference area;

[0089] Calculate the spatial difference between the first set of pixel coordinates and the second set of pixel coordinates, and determine the image region corresponding to the spatial difference as the second suspected fire area.

[0090] S209. Obtain the pixel position of the second suspected fire area in the multispectral image data and the preset photovoltaic module geometric data, the preset photovoltaic module geometric data including the physical boundary coordinates of the photovoltaic modules in the photovoltaic power station;

[0091] Among them, pixel position refers to the set of coordinates occupied by the second suspected fire area on the two-dimensional image plane, which can be represented as the (x, y) coordinates of a series of pixels, an outline, or an outer rectangle; physical boundary coordinates are used to represent the three-dimensional spatial coordinate points that constitute the core content of the preset photovoltaic module geometric data, such as the latitude, longitude and altitude coordinates of the four corner points of each photovoltaic module or the (X, Y, Z) coordinates in the local coordinate system of the power station. These coordinates together define the precise physical range of each module in the real world.

[0092] This step is performed by the fire monitoring system after identifying a second suspected fire area, but before determining the nature of the fire.

[0093] S210. When the pixel range of the second suspected fire area does not overlap with the preset photovoltaic module geometric data, the second suspected fire area is determined as a non-module suspected fire area.

[0094] If the fire is on the photovoltaic module, it can be verified using electrical data; if the fire is outside the module, the electrical data is invalid, and other visual features must be relied upon for judgment. Specifically, the fire monitoring system must first convert the three-dimensional preset geometric data of the photovoltaic modules into a two-dimensional image space. Using pre-calibrated camera intrinsic parameters (such as focal length and principal point) and extrinsic parameters (the camera's position and orientation in the world coordinate system), the physical boundary coordinates of each photovoltaic module are transformed through perspective projection to calculate their corresponding pixel coordinates in the current multispectral image data, thus obtaining a series of polygons representing the outline of the photovoltaic modules in the image. Then, the fire monitoring system performs a spatial intersection test on the pixel range of each polygon representing a module with the second suspected fire area. If the pixel range of the second suspected fire area has no overlap with the projected polygons of any of the modules, the fire monitoring system re-marks or classifies the second suspected fire area as a non-module suspected fire area.

[0095] Optionally, in some embodiments, considering the slight changes in camera pose caused by the outdoor environment, the fire monitoring system can also perform coordinate system adaptive calibration:

[0096] The fire monitoring system uses an edge detection algorithm to extract the straight edges and corner features of the photovoltaic module array in real time from visible light images. The extracted corner image coordinates are matched with the theoretical corner coordinates in the preset photovoltaic module geometric data, and the homography matrix is ​​calculated and updated in real time.

[0097] By using the updated matrix for coordinate transformation, it is ensured that even when the camera experiences slight vibrations due to wind load, the specific photovoltaic panel number covering the second suspected fire area can be accurately located. This ensures that the extracted electrical data (such as current and voltage) is indeed the data of the components in the suspected area, avoiding verification failures caused by misattribution.

[0098] S211. Within a preset time window, extract the temporal dynamic features of the suspected fire area in the non-component area based on the multispectral image data. The temporal dynamic features include at least the thermal radiation intensity value and the pixel contour area value.

[0099] The preset time window is used to provide sufficient time samples to distinguish the dynamic development process of a real fire from static or instantaneous thermal anomalies. Its setting scheme is usually based on experimental data on the early fire development speed and is generally set to several seconds to tens of seconds (e.g., 10 seconds). The temporal dynamic characteristics refer to a set of quantitative indicators that describe how the non-component suspected fire area changes over time within the preset time window. The thermal radiation intensity value is used to represent the pixel value extracted from the thermal infrared image that is positively correlated with the area temperature. The pixel contour area value represents the number of pixels occupied by the area in the image, and its change over time directly reflects the visual spread or extinguishing trend of the fire source.

[0100] The fire monitoring system executes this step after identifying the second suspected fire area as a non-component suspected fire area. Specifically, the fire monitoring system activates a tracker (e.g., a tracking algorithm based on kernel-based correlation filtering (KCF) or Siamese RPN) for this non-component suspected fire area to ensure continuous locking of the area in subsequent frames. The tracker updates the area's position with each frame arriving within the preset time window. The fire monitoring system then extracts the thermal radiation intensity value from the synchronized thermal infrared image data at the area's current location and calculates the number of pixels within its contour to obtain the pixel contour area value. When the preset time window ends, the fire monitoring system obtains two complete time-series data points, recording the continuous changes in the area's thermal radiation intensity and area over the past period.

[0101] S212. When the time sequence dynamic characteristics are greater than the preset fire judgment threshold, it is determined that there is a fire in the non-component suspected fire area.

[0102] Among them, the preset fire judgment threshold represents a set of quantitative standards used to evaluate whether the temporal dynamic characteristics conform to the actual fire development pattern. The setting scheme is usually based on the analysis and statistics of a large number of real fire videos. For example, within the preset time window, the linear regression slope of the pixel contour area value must be greater than a certain positive number S1 (indicating that the area is increasing), and the final value of the thermal radiation intensity value must be higher than an absolute temperature T1 (indicating that a sufficiently high temperature has been reached).

[0103] The fire monitoring system performs this step after extracting the temporal dynamic features. Specifically, the system analyzes two time-series data points to verify whether they simultaneously meet the conditions defined by a preset fire judgment threshold. For example, the system performs linear fitting on the pixel contour area value sequence and calculates its slope to determine if the area is showing a continuous increasing trend; simultaneously, it checks the thermal radiation intensity value sequence to determine whether it is continuously maintained at a high temperature level or also shows an increasing trend. Only when the trends of these two temporal dynamic features meet the preset fire judgment threshold does the fire monitoring system determine that a fire exists in the non-component suspected fire area and initiate the corresponding alarm and emergency response procedures.

[0104] S213. When the pixel range of the second suspected fire area overlaps with the preset geometric data of the photovoltaic module, and the power station operation data corresponding to the second suspected fire area is abnormal, it is determined that there is a fire in the second suspected fire area.

[0105] This step specifically includes:

[0106] Obtain the start timestamp of the second suspected fire area, and obtain the device identifier of the photovoltaic modules within the coverage area of ​​the second suspected fire area;

[0107] Based on equipment identification and power plant operation data, extract the electrical parameters corresponding to the second suspected fire area;

[0108] Check whether there are any sudden changes in electrical parameters within a preset time window adjacent to the start timestamp;

[0109] When there is a parameter mutation, and the deviation between the occurrence time of the parameter mutation and the start timestamp is less than a preset time threshold, it is confirmed that the power plant operation data corresponding to the second suspected fire area is abnormal, and it is determined that there is a fire in the second suspected fire area.

[0110] Specifically, obtaining the device identifier of photovoltaic modules within the coverage area of ​​the second suspected fire zone includes: setting physical world coordinates based on preset photovoltaic module geometric data, and identifying the corresponding positioning pixel coordinates in multispectral image data. The physical world coordinates represent at least four reference points set based on the physical features on the photovoltaic modules. Based on the correspondence between the physical world coordinates and the positioning pixel coordinates, coordinate transformation parameters are calculated to convert the pixel coordinates into physical world coordinates. Based on the coordinate transformation parameters, the pixel coordinates corresponding to the second suspected fire zone are transformed and calculated to obtain the geographical location coordinates of the second suspected fire zone. The geographical location coordinates are matched with the preset photovoltaic module geometric data to determine the corresponding photovoltaic modules and obtain the device identifier of the photovoltaic modules.

[0111] Among them, the start timestamp represents the precise moment when the fire monitoring system first confirmed the second suspected fire area, serving as the benchmark for time correlation analysis; the equipment identifier refers to the unique identification code (such as serial number or asset number) assigned to each photovoltaic module, used to accurately locate and query its related electrical parameters in the power plant operation database; electrical parameters represent a set of physical quantities extracted from the power plant operation data that reflect the working status of the photovoltaic modules; parameter mutation refers to a drastic change in electrical parameters between one or several adjacent data sampling points, with an amplitude far exceeding normal fluctuations; the preset time window is a time interval set around the start timestamp, its function being to limit the scope of electrical parameter inspection to ensure the correlation and efficiency of the analysis. The setting of this window usually takes into account the delay in data acquisition and transmission, and is generally set to 5 to 10 seconds before and after the start timestamp; the preset time threshold is an upper limit of the time difference used to determine whether visual events and electrical events constitute a strong causal relationship, its function being to exclude accidental time coincidences. The setting scheme for this threshold is very strict, usually set to 1 to 3 seconds, because the damage to electrical performance caused by fire is almost instantaneous;

[0112] Specifically, once a second suspected fire area is identified as "on a photovoltaic module," the fire monitoring system immediately records the start timestamp of the current time. Next, a location procedure is initiated to obtain the device identifier of the affected photovoltaic module. This includes: identifying the location pixel coordinates corresponding to preset physical world coordinates in the image, calculating precise coordinate transformation parameters, and using these parameters to project the pixel coordinates of the second suspected fire area back into the three-dimensional physical world to obtain its geographical location coordinates. These geographical location coordinates are then matched with a three-dimensional model of the preset photovoltaic module geometric data to pinpoint the unique device identifier. After obtaining the device identifier, the fire monitoring system requests the electrical parameter sequence of the module within a preset time window adjacent to the start timestamp from the photovoltaic power plant's data center. It analyzes whether there are any parameter abrupt changes in this sequence and records the time of these abrupt changes. Finally, the time difference between the time of the parameter abrupt change and the start timestamp is calculated. If this time difference is less than a preset time threshold, it is determined that there is an anomaly in the power plant's operating data that is highly synchronized with the visual event, and ultimately, a fire is determined to exist within the second suspected fire area.

[0113] In this embodiment, after verifying the spectral complementarity and eliminating fire interference areas, the system further determines whether the second suspected fire area overlaps with the preset geometric data of the photovoltaic module. Based on this, it intelligently selects two different paths for fire confirmation: "comparing whether there are parameter mutations in electrical parameters" or "analyzing whether the time-series dynamic characteristics are greater than the fire judgment threshold." Therefore, it can call the most relevant verification logic for differentiated analysis for the two types of potential fires, one occurring on the photovoltaic module and the other in the surrounding environment. This effectively solves the limitation that a single confirmation method (such as relying solely on electrical data) cannot cover fires in non-equipment areas, or that (such as relying solely on visual features) is not reliable enough when confirming fires on equipment. This achieves full-scene coverage and accurate identification of equipment fires and environmental fires in photovoltaic power plants, greatly improving the overall adaptability and reliability of the fire monitoring system.

[0114] The fire monitoring system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of a fire monitoring system in an embodiment of this application.

[0115] It should be noted that, Figure 3 The structure of the fire monitoring system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0116] like Figure 3As shown, the fire monitoring system includes a CPU 301, which can perform various appropriate actions and processes according to a program stored in the read-only memory ROM 302 or a program loaded from the storage section 308 into the random access memory RAM 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An I / O interface 305 is also connected to the bus 304.

[0117] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0118] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by CPU 301, it performs the various functions defined in the present invention.

[0119] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0121] Specifically, the fire monitoring system in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the fire false alarm elimination method based on multispectral features and environmental parameters provided in the above embodiment.

[0122] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the fire monitoring system described in the above embodiments; or it may exist independently and not assembled into the fire monitoring system. The storage medium carries one or more computer programs, which, when executed by a processor of the fire monitoring system, cause the fire monitoring system to implement the fire false alarm elimination method based on multispectral features and environmental parameters provided in the above embodiments.

[0123] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0124] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0125] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for eliminating false fire alarms based on multispectral features and environmental parameters, applied to a fire monitoring system, characterized in that, include: Simultaneously acquire multispectral image data, environmental data, and power plant operation data within the monitoring area. The multispectral image data includes at least visible light images, near-infrared images, and thermal infrared images. The environmental data includes at least real-time solar position information. The power plant operation data includes the electrical parameters of each photovoltaic module. Spectral feature data of the multispectral image data is extracted, and spectral complementarity verification is performed based on the spectral feature data to obtain spectral complementarity verification results. The spectral complementarity verification results are used to identify image regions that meet preset fire feature conditions. Based on the spectral complementarity verification results, the corresponding image region in the multispectral image data is identified as the first suspected fire area; Based on the real-time solar position information and the preset photovoltaic module geometric data, the fire interference area is determined, which is the specular reflection area generated by sunlight on the photovoltaic module. The first suspected fire area is removed from the fire interference area to obtain the second suspected fire area; When the power plant operation data corresponding to the second suspected fire area is abnormal, it is determined that there is a fire in the second suspected fire area.

2. The method according to claim 1, characterized in that, Spectral complementarity verification is performed based on the spectral feature data to obtain spectral complementarity verification results, specifically including: Based on the spectral feature data of the visible light image, a first candidate region with preset flame texture features is identified; Based on the spectral feature data of the near-infrared image, a second candidate region with a brightness higher than a preset smoke brightness threshold is identified; Based on the spectral feature data of the thermal infrared image, a third candidate region with a temperature higher than a preset abnormal high temperature threshold is identified. Calculate the spatial intersection of the first candidate region, the second candidate region, and the third candidate region, and determine the image region corresponding to the spatial intersection as the spectral complementarity verification result.

3. The method according to claim 1, characterized in that, After determining the fire disturbance area based on the real-time solar position information and preset photovoltaic module geometric data, the method further includes: Based on the spectral feature data, the interference spectral features corresponding to the fire interference area are extracted. The interference spectral features include the stability of brightness, texture and shape within the fire interference area. Determine whether the interference spectral features have a preset disturbance anomaly, and identify the image area with the disturbance anomaly as the third suspected fire area. The disturbance anomaly refers to the brightness of the fire interference area decreasing sharply in a short period of time, or the stability of its texture and shape being destroyed. When the power plant operation data corresponding to the third suspected fire area is abnormal, it is determined that there is a fire in the third suspected fire area.

4. The method according to claim 1, characterized in that, The first suspected fire area is removed from the fire interference area to obtain the second suspected fire area, which specifically includes: Based on the multispectral image data, obtain the first set of pixel coordinates corresponding to the first suspected fire area and the second set of pixel coordinates corresponding to the fire interference area; Calculate the spatial difference between the first set of pixel coordinates and the second set of pixel coordinates, and determine the image region corresponding to the spatial difference as the second suspected fire area.

5. The method according to claim 4, characterized in that, When the power plant operation data corresponding to the second suspected fire area is abnormal, it is determined that there is a fire in the second suspected fire area, specifically including: Obtain the start timestamp of the second suspected fire area, and obtain the device identifier of the photovoltaic module within the coverage area of ​​the second suspected fire area; Based on the equipment identification and the power plant operation data, extract the electrical parameters corresponding to the second suspected fire area; Check whether there are any sudden changes in the electrical parameters within a preset time window adjacent to the start timestamp; When there is a sudden change in the parameter, and the deviation between the occurrence time of the parameter change and the starting timestamp is less than a preset time threshold, it is confirmed that the power plant operation data corresponding to the second suspected fire area is abnormal, and it is determined that there is a fire in the second suspected fire area.

6. The method according to claim 5, characterized in that, Before obtaining the device identifiers of the photovoltaic modules within the coverage area of ​​the second suspected fire area, the method further includes: The pixel positions of the second suspected fire area in the multispectral image data and the preset geometric data of the photovoltaic module are obtained. The preset geometric data of the photovoltaic module includes the physical boundary coordinates of the photovoltaic module in the photovoltaic power station. When the pixel range of the second suspected fire area does not overlap with the preset photovoltaic module geometric data, the second suspected fire area is determined as a non-module suspected fire area. Within a preset time window, the temporal dynamic features of the non-component suspected fire area are extracted based on the multispectral image data. The temporal dynamic features include at least thermal radiation intensity values ​​and pixel contour area values. When the time-series dynamic characteristics exceed the preset fire judgment threshold, it is determined that there is a fire in the non-component suspected fire area.

7. The method according to claim 5, characterized in that, Obtaining the device identifier of the photovoltaic modules within the coverage area of ​​the second suspected fire area specifically includes: Physical world coordinates are set according to the preset photovoltaic module geometric data, and the positioning pixel coordinates corresponding to the physical world coordinates are identified in the multispectral image data. The physical world coordinates represent at least four reference points set according to the physical features on the photovoltaic module. Based on the correspondence between the physical world coordinates and the location pixel coordinates, coordinate transformation parameters are calculated, which are used to convert pixel coordinates into physical world coordinates; Based on the coordinate transformation parameters, the pixel coordinates corresponding to the second suspected fire area are transformed and calculated to obtain the geographical location coordinates of the second suspected fire area; The geographical coordinates are matched with the preset photovoltaic module geometric data to determine the corresponding photovoltaic module and obtain the device identifier of the photovoltaic module.

8. A fire monitoring system, characterized in that, The fire monitoring system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the fire monitoring system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the fire monitoring system, the fire monitoring system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the fire monitoring system, it causes the fire monitoring system to perform the method as described in any one of claims 1-7.

Images