Urban intelligent fire fighting system and method

Abstract

The invention provides an urban intelligent fire-fighting system and method. The method comprises the following steps: analyzing a thermal imaging image and a visible light image of a target floor from a fire-fighting monitoring video; constructing a potential risk area of fire spread when a fire occurs in the target floor based on a plurality of feature pixel points with sudden temperature change in the thermal imaging image; determining positioning loss when fire point positioning is carried out on the potential risk area through the omen characteristics of outward diffusion of the edge contour of the potential risk area in combination with the local temperature peak point in the thermal imaging image; performing dynamic area monitoring on the edge contour of the potential risk area at different moments based on the positioning loss, and when the area growth rate of the edge contour exceeds a preset threshold value, determining a confidence fire risk area of the target floor according to the edge contour at the corresponding moment and the visible light image; and the Internet of Things center sends a fire diffusion early warning signal based on the confidence fire risk area. According to the scheme, fire diffusion identification can be performed on the target floor in a complex urban environment based on the confidence fire risk area.

Description

Technical Field

[0001] This application relates to the technical field of fire warning, and more specifically, to an urban intelligent fire protection system and method. Background Art

[0002] With the acceleration of the urbanization process, the height and density of buildings are increasing continuously, and the fire risk is becoming increasingly prominent. Traditional fire protection relies on manual inspections and alarms, making it difficult to detect early fires in a timely manner. Modern fire warning makes use of cutting-edge technologies such as the Internet of Things, sensor technology, big data analysis, and artificial intelligence. The Internet of Things platform can effectively achieve the interconnection and interoperability between devices, and wireless communication technology can transmit the real-time collected data to the Internet of Things platform. The Internet of Things platform processes the data with the help of big data analysis and intelligent algorithms, and can quickly identify fire hazards and issue warnings. This kind of fire warning based on the Internet of Things effectively improves the real-time performance and accuracy of fire monitoring.

[0003] In the existing fire warning, the existing fire warning mainly conducts fire monitoring and warning based on a variety of sensors and intelligent algorithms. That is, by deploying sensors such as smoke, temperature, and combustible gas, environmental data is collected in real time. When the sensor detects an abnormal signal, such as the smoke concentration exceeding the standard or the temperature rising abnormally, the data processor uses the fire warning model, combines historical data and professional knowledge, and evaluates the current environmental conditions. If it is determined that there is a fire hazard, the alarm control system will be immediately activated and an alarm will be issued in various ways. However, in urban fire warning, it is difficult for traditional fire warning systems to accurately distinguish the early fire spread characteristics from other heat source interferences through single temperature monitoring in complex urban environments, resulting in identification deviations in the fire spread of the target floor. Therefore, how to identify the fire spread of the target floor in complex urban environments has become a difficult problem faced by the industry. Summary of the Invention

[0004] This application provides an urban intelligent fire protection system and method, which can identify the fire spread of the target floor in complex urban environments.

[0005] In the first aspect, this application provides a fire warning method for an urban intelligent fire protection system to conduct fire spread warning on the floor where a fire occurs. The method includes the following steps: Retrieve the fire monitoring video of the target floor from the Internet of Things platform, and parse the thermal imaging image and visible light image of the target floor from the fire monitoring video; Extract multiple characteristic pixel points with temperature mutations in the thermal imaging image, and then construct a potential risk area for the fire spread when a fire occurs on the target floor based on all the characteristic pixel points; Determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor. Based on the precursor characteristics and the local temperature peak points in the thermal imaging image, perform fire point location analysis on the potential risk area to obtain the location loss when locating the fire point in the potential risk area. Based on the location loss, perform dynamic area monitoring of the edge contour of the potential risk area at different times. When the area growth rate of the edge contour exceeds a preset threshold, perform mapping verification on the fire risk area of the target floor using the edge contour and the visible light image at the corresponding time to obtain the confidence fire risk area of the target floor. Upload the confidence fire risk area to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal.

[0006] In some embodiments, extracting multiple characteristic pixel points with temperature mutations in the thermal imaging image specifically includes: Obtain the thermal imaging images of the target floor at different time nodes; Determine the temperature distribution matrix corresponding to the thermal imaging image at each time node; For each time node, perform pixel-by-pixel differential calculation of the corresponding temperature values between the temperature distribution matrix of the current time node and the temperature distribution matrix of the previous time node to obtain the temperature difference matrix corresponding to each time node; Based on a preset temperature difference threshold, segment the temperature difference matrix corresponding to each time node to obtain multiple candidate pixel points corresponding to each time node; Extract the candidate pixel points that appear at each time node from all candidate pixel points as the characteristic pixel points with temperature mutations in the thermal imaging image, and thus obtain multiple characteristic pixel points with temperature mutations in the thermal imaging image.

[0007] In some embodiments, constructing the potential risk area of the fire spread when a fire occurs on the target floor based on all the characteristic pixel points specifically includes: Perform spatial density clustering analysis on all the characteristic pixel points to obtain an aggregated characteristic pixel point cluster and a discrete characteristic pixel point cluster; Perform morphological closing operation on all the aggregated characteristic pixel points in the aggregated characteristic pixel point cluster to obtain the initial heat source area of the target floor; Extract multiple correlation heat source points according to the correlation between the geometric center point of the initial heat source area and each discrete characteristic pixel point in the discrete characteristic pixel point cluster; Merge all the correlation heat source points with the initial heat source area to obtain the potential risk area of the fire spread when a fire occurs on the target floor.

[0008] In some embodiments, the precursor features of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor specifically include: Extract the center pixel point of the largest circumscribed circle of the potential risk area; Obtain all the edge pixel points on the edge contour of the potential risk area, and calculate the relative distances between the center pixel point and each edge pixel point; Determine the precursor features of the outward diffusion of the edge contour of the potential risk area based on all the relative distances.

[0009] In some embodiments, the dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss specifically includes: Compare the positioning loss with the positioning loss threshold. When the positioning loss is greater than the positioning loss threshold, perform dynamic area monitoring of the edge contour of the fire risk area at different times to obtain dynamic area monitoring data, and determine the area growth rate of the edge contour at different times according to the dynamic area monitoring data; When the positioning loss is less than or equal to the positioning loss threshold, do not perform dynamic monitoring on the edge contour of the fire risk area.

[0010] In some embodiments, mapping verification of the fire risk area of the target floor is performed by using the edge contour at the corresponding time and the visible light image to obtain the confidence fire risk area of the target floor, specifically including: Extract the contour pixel coordinate set of the edge contour at the corresponding time, and map the contour pixel coordinate set to the corresponding spatial position of the visible light image at the same time to obtain the mapped contour area of the target floor; Use a multi-scale filtering algorithm to perform light intensity detection on all the pixels in the mapped contour area to obtain a cluster of high-brightness pixel points when a fire occurs on the target floor; Construct the confidence fire risk area of the target floor according to all the high-brightness pixel points in the cluster of high-brightness pixel points.

[0011] In some embodiments, the fire monitoring video is obtained by a fire monitoring camera installed in the target floor.

[0012] In a second aspect, the present application provides an urban intelligent fire warning system, including a fire warning unit, and the fire warning unit includes: A retrieval module, configured to retrieve the fire monitoring video of the target floor from the Internet of Things platform, and parse the thermal imaging image and the visible light image of the target floor from the fire monitoring video; A processing module, configured to extract a plurality of feature pixel points with temperature mutations in the thermal imaging image, and further construct a potential risk area for the spread of the fire when a fire occurs on the target floor based on all the feature pixel points; The processing module is further configured to determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor, and perform fire point positioning analysis on the potential risk area based on the precursor characteristics in combination with the local temperature peak points in the thermal imaging image, so as to obtain the positioning loss when performing fire point positioning on the potential risk area. The processing module is further configured to perform dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss. When the area growth rate of the edge contour exceeds a preset threshold, perform mapping verification on the fire risk area of the target floor by using the edge contour at the corresponding time and the visible light image, so as to obtain the confidence fire risk area of the target floor. The execution module is configured to upload the confidence fire risk area to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal.

[0013] In a third aspect, the present application provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to obtain the code and execute the above-mentioned fire warning method.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned fire warning method is implemented.

[0015] The technical solution provided by the embodiments disclosed in the present application has the following beneficial effects: In the urban intelligent fire protection system and method provided by the present application, first, retrieve the fire protection monitoring video of the target floor from the Internet of Things platform, and parse the thermal imaging image and visible light image of the target floor from the fire protection monitoring video; secondly, extract multiple characteristic pixel points with temperature mutations in the thermal imaging image, and then construct a potential risk area for the spread of the fire when a fire occurs on the target floor based on all the characteristic pixel points; further, determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor, and perform fire point positioning analysis on the potential risk area based on the precursor characteristics in combination with the local temperature peak points in the thermal imaging image, so as to obtain the positioning loss when performing fire point positioning on the potential risk area; then, perform dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss. When the area growth rate of the edge contour exceeds a preset threshold, perform mapping verification on the fire risk area of the target floor by using the edge contour at the corresponding time and the visible light image, so as to obtain the confidence fire risk area of the target floor; finally, upload the confidence fire risk area to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal.

[0016] It can be seen that the present application can identify the spread of fire on the target floor in a complex urban environment. First, the thermal imaging image and visible light image of the target floor are parsed from the fire monitoring video, thereby providing data support for accurately evaluating the fire risk area and fire scale. Second, a potential risk area for the spread of fire when a fire occurs on the target floor is constructed based on multiple characteristic pixel points corresponding to the thermal imaging image, so as to provide early information on the development of the fire and distinguish other non-risky heat sources, and then effectively judge the possibility and influence range of the spread of the fire, thereby guiding subsequent emergency response and protection measures and avoiding the interference of the identification of the fire risk area caused by other heat sources. Further, the fire point positioning analysis of the potential risk area is carried out by combining the precursor characteristics spreading outward from the edge contour of the potential risk area when a fire occurs on the target floor with the local temperature peak points in the thermal imaging image, and the positioning loss during the fire point positioning of the potential risk area is obtained, so as to enhance the analysis ability of the fire morphological characteristics, improve the accuracy of fire spread prediction, and then optimize the trigger mechanism of the early warning system, thereby avoiding the inaccuracy of fire warning by single temperature monitoring in a complex urban environment. Then, based on the positioning loss, the dynamic area of the edge contour of the fire risk area is monitored at different times. When the area growth rate of the edge contour exceeds the preset threshold, the fire risk area of the target floor is mapped and verified by the edge contour and visible light image at the corresponding time, and the confidence fire risk area of the target floor is obtained, thereby identifying a more reliable fire risk area, reducing the false alarm rate of the fire warning, enabling the intelligent fire protection system to still play an effective role in a complex environment, and thus enhancing the intelligent level of fire prevention and control. Finally, the confidence fire risk area is uploaded to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal. In summary, the technical solution provided by the present application can identify the spread of fire on the target floor in a complex urban environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is an exemplary flowchart of a fire warning method according to some embodiments of the present application; Figure 2 is an exemplary flowchart of determining characteristic pixel points according to some embodiments of the present application; Figure 3 is an exemplary flowchart of determining precursor characteristics according to some embodiments of the present application; Figure 4 is a schematic structural diagram of a fire warning unit according to some embodiments of the present application; Figure 5 is a schematic structural diagram of a computer device for implementing the fire warning method according to some embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To better understand the technical solution of the present application, the technical solution of the present application will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments.

[0019] Referring to Figure 1 , this figure is an exemplary flowchart of a fire warning method shown according to some embodiments of the present application. The fire warning method 100 mainly includes the following steps: In step 101, retrieve the fire monitoring video of the target floor from the Internet of Things platform, and parse the thermal imaging image and visible light image of the target floor from the fire monitoring video.

[0020] In specific implementation, first, retrieve the fire monitoring video of the target floor from the Internet of Things platform. After obtaining the fire monitoring video, use OpenCV in video parsing technology to parse the fire monitoring video frame by frame, identify the thermal imaging image and visible light image. During the parsing process, decode video data in different formats, such as common video coding formats like H.264 and H.265, and extract the thermal imaging image and visible light image in each frame.

[0021] It should be noted that in this embodiment, the fire monitoring video is captured by fire monitoring cameras installed in the target floor, and these videos are mainly used to monitor and prevent fire safety accidents such as fires; in the present application, the thermal imaging image represents the monitoring image of the target floor generated by infrared thermal imaging technology; in the present application, the visible light image represents each frame image in the fire video image formed in the visible light band. The visible light image captures the visible light reflected or emitted by objects in the target floor environment to present the actual scene of the floor. Among them, the thermal imaging image and the visible light image are obtained after being aligned through a timestamp alignment mechanism. Through the thermal imaging image, the spread direction and speed of the fire can be monitored in real time, helping the fire command personnel accurately judge the development trend of the fire. Through the visible light image, the overall situation of the fire scene can be intuitively displayed, including information such as the color, shape, and size of the flame, the color, concentration, and diffusion range of the smoke, and the damage degree of the building structure. These information are of great reference value for fire fighters to accurately evaluate the scale and danger degree of the fire and formulate reasonable fire extinguishing and rescue plans.

[0022] In step 102, extract multiple feature pixel points with temperature mutations in the thermal imaging image, and then construct a potential risk area for the spread of the fire when a fire occurs on the target floor based on all the feature pixel points.

[0023] In some embodiments, referring to Figure 2 shown, this figure is an exemplary flowchart of determining feature pixel points shown according to some embodiments of the present application. In this embodiment, the extraction of multiple feature pixel points with temperature mutations in the thermal imaging image can be implemented by the following steps: First, in step 1021, obtain the thermal imaging images of the target floor at different time nodes; Second, in step 1022, determine the temperature distribution matrix corresponding to the thermal imaging image at each time node; Further, in step 1023, for each time node, perform a pixel-by-pixel differential calculation of the corresponding temperature values between the temperature distribution matrix of the current time node and the temperature distribution matrix of the previous time node, so as to obtain the temperature difference matrix corresponding to each time node; Then, in step 1024, based on a preset temperature difference threshold, segment the temperature difference matrix corresponding to each time node, so as to obtain multiple candidate pixel points corresponding to each time node; Finally, in step 1025, extract the candidate pixel points that appear at each time node from all the candidate pixel points as the characteristic pixel points with temperature mutations in the thermal imaging image, so as to obtain multiple characteristic pixel points with temperature mutations in the thermal imaging image.

[0024] When specifically implemented, obtain the thermal imaging images of the target floor at different time nodes from the Internet of Things platform, which will not be elaborated here. The thermal imaging images contain the temperature information of the target floor.

[0025] When specifically implemented, determine the temperature distribution matrix corresponding to the thermal imaging image at each time node, that is: for each time node, read the temperature values corresponding to each pixel point in the thermal imaging image at the time node, use the temperature values corresponding to each pixel point as matrix elements, and construct a temperature distribution matrix with the same resolution as the thermal imaging image at the time node, so as to obtain the temperature distribution matrix corresponding to the thermal imaging image at each time node; among them, the temperature values corresponding to each pixel point in the thermal imaging image at the time node can be read through a thermal imaging software development kit (SDK), which will not be elaborated here; in this embodiment, the temperature distribution matrix represents a two-dimensional matrix composed of the temperature values corresponding to the pixel points in the thermal imaging image, and is used to analyze the distribution state of the temperature.

[0026] In specific implementation, for each time node, a pixel-by-pixel difference calculation of the corresponding temperature values is performed between the temperature distribution matrix of the current time node and the temperature distribution matrix of the previous time node, so as to obtain a temperature difference matrix corresponding to each time node, that is: for each time node, a difference calculation of the corresponding temperature values is performed between the temperature distribution matrix of the current time node and the temperature distribution matrix of the previous time node, and the results of each difference calculation are arranged as elements of the temperature difference matrix according to the corresponding positions to obtain the temperature difference matrix corresponding to the time node, and then the temperature difference matrices corresponding to each time node are obtained; in this embodiment, the temperature difference matrix represents a two-dimensional matrix composed of multiple temperature difference values, and the temperature difference value represents the difference between the temperature values at the corresponding positions in the temperature distribution matrices.

[0027] In specific implementation, based on a preset temperature difference threshold, the temperature difference matrices corresponding to each time node are segmented, so as to obtain multiple candidate pixel points corresponding to each time node, that is: obtain the preset temperature difference threshold from the Internet of Things platform, for each time node, compare each temperature difference value in the temperature difference matrix corresponding to the time node with the temperature difference threshold, and use the pixel points with temperature difference values greater than the temperature difference threshold as the candidate pixel points corresponding to the time node, so as to obtain multiple candidate pixel points corresponding to each time node; wherein, the average temperature difference value when no fire occurs on the target floor can be used as the temperature difference threshold, and in other embodiments, it can also be set according to actual needs, which is not limited here; in this embodiment, the candidate pixel points represent pre-selected pixel points. Specifically, the candidate pixel points represent pixel points that meet the condition that the temperature change exceeds the preset temperature difference threshold. During the fire occurrence process, the surface temperature of the object will increase with time. Therefore, by analyzing the temperature difference between the same positions of adjacent time nodes, the positions with large temperature changes can be analyzed. Those positions with large temperature changes are usually the places with the highest fire occurrence probability. Then, the pixels at the positions with large temperature changes can be used as candidate pixel points through the set threshold.

[0028] In specific implementation, since each time point corresponds to multiple candidate pixel points, and the thermal imaging images collected at each time point are of the same object (i.e., the target floor), when a fire occurs on the target floor, if the temperature value at the pixel point in the thermal imaging image changes continuously, the position at the pixel point may have a fire risk. Therefore, the candidate pixel points that appear at each time node can be extracted from all candidate pixel points as the characteristic pixel points with temperature mutations in the thermal imaging image, so as to obtain multiple characteristic pixel points with temperature mutations in the thermal imaging image.

[0029] It should be noted that in this application, the characteristic pixel points represent the position points with stable and significant temperature changes at different time nodes. The characteristic pixel points characterize the position points where there may be a fire risk. By analyzing the temperature changes of the thermal imaging image at different time nodes and extracting the pixel points with continuous high-temperature changes as the characteristic pixel points, false alarms caused by short-term temperature fluctuations can be effectively excluded, ensuring the accurate tracking of the spread of the fire. These characteristic pixel points are not only used to construct the fire risk area, but also can be combined with smoke texture analysis and edge dynamic monitoring to improve the reliability of the early warning system.

[0030] In some embodiments, the potential risk area of the spread of the fire when a fire occurs on the target floor can be constructed based on all the characteristic pixel points by the following steps, that is: Perform spatial density clustering analysis on all the characteristic pixel points to obtain a cluster of aggregated characteristic pixel points and a cluster of discrete characteristic pixel points; Perform morphological closing operation on all the aggregated characteristic pixel points in the cluster of aggregated characteristic pixel points to obtain the initial heat source area of the target floor; Extract a plurality of correlation heat source points according to the correlation between the geometric center point of the initial heat source area and each discrete characteristic pixel point in the cluster of discrete characteristic pixel points; Merge all the correlation heat source points with the initial heat source area to obtain the potential risk area of the spread of the fire when a fire occurs on the target floor.

[0031] In specific implementation, spatial density clustering analysis is performed on all feature pixel points to obtain an aggregated feature pixel point cluster and a discrete feature pixel point cluster, that is: set the neighborhood radius and the minimum number of cluster points for spatial density clustering, traverse the number of feature pixel points corresponding to each feature pixel point within the range of the neighborhood radius. If the number of feature pixel points is greater than or equal to the minimum number of cluster points, the feature pixel points corresponding to the number of feature pixel points are classified into the aggregated feature pixel point cluster. If the number of feature pixel points is less than the minimum number of cluster points, the feature pixel points corresponding to the number of feature pixel points are classified into the discrete feature pixel point cluster, thereby obtaining the aggregated feature pixel point cluster and the discrete feature pixel point cluster; among them, the neighborhood radius and the minimum number of cluster points for spatial density clustering can be set according to actual needs. For example, in this embodiment, the neighborhood radius of spatial clustering is set to 10 pixels, and the minimum number of cluster points is set to 5. In addition, in other embodiments, it can also be set according to actual needs, which is not limited here; in this embodiment, the aggregated feature pixel point cluster represents a clustering area composed of multiple feature pixel points with high adjacency. The aggregated feature pixel point cluster contains multiple aggregated feature pixel points, and the aggregated feature pixel point is the feature pixel point in the aggregated feature pixel point cluster; in this embodiment, the discrete feature pixel point cluster represents a set of feature pixel points that do not belong to the aggregated feature pixel point cluster. The discrete feature pixel point cluster contains multiple discrete feature pixel points, and the discrete feature pixel point is the feature pixel point in the discrete feature pixel point cluster.

[0032] In specific implementation, a morphological closing operation is performed on all the clustered feature pixel points in the clustered feature pixel point cluster to obtain the initial heat source area of the target floor, that is: all the clustered feature pixel points in the clustered feature pixel point cluster are converted into a binary mask image, a morphological closing operation is performed on the binary mask image through the image processing tool OpenCV, and the connected area contour after the morphological closing operation is extracted as the initial heat source area of the target floor; wherein, the binary mask image is a binary image with the pixels of the clustered feature pixel points set to 1 and the background pixels set to 0; wherein the cv2.morphologyEx function of the image processing tool OpenCV can be called to perform a morphological closing operation on the binary mask image, which will not be repeated here. ; In this embodiment, the binary mask image represents a binary image composed only of pixel values 0 and 1; in this embodiment, the initial heat source area represents the initial area where there is a fire risk on the target floor. By determining the initial heat source area, the probability of the existence of a fire risk on the target floor can be pre-evaluated, thereby providing an information basis for timely detection of fire risks on the target floor; during the occurrence of a fire, as the fire spreads, the local temperature area will show similar temperature change characteristics, that is, the temperature changes at multiple locations are similar. Therefore, the locations with more drastic temperature changes and close distances can be clustered to obtain a cluster of clustered characteristic pixel points, and then the area composed of the clustered characteristic pixel points in the clustered characteristic pixel point cluster can be used as the initial heat source area of the target floor.

[0033] In specific implementation, a plurality of correlation heat source points are extracted according to the correlation between the geometric center point of the initial heat source area and each discrete feature pixel point in the discrete feature pixel point cluster, that is: the geometric center point of the initial heat source area is extracted through the image processing tool OpenCV, and the correlation values between the geometric center point and each discrete feature pixel point in the discrete feature pixel point cluster are calculated respectively by the Euclidean distance. The discrete feature pixel points corresponding to the respective correlation values are compared with a correlation threshold, and the discrete feature pixel points with correlation values less than the correlation threshold are extracted as correlation heat source points, thereby obtaining a plurality of correlation heat source points; wherein, the geometric center point of the initial heat source area can be extracted through the cv2.moments function of the image processing tool OpenCV; wherein, the correlation value is determined by the calculation result of the Euclidean distance, which will not be elaborated here; wherein, the correlation threshold is a preset standard correlation value for discriminating the correlation heat source points, and can be specifically set according to actual requirements, which is not limited here; in this embodiment, the correlation value represents an index for measuring the correlation between two parameter points. Specifically, the correlation value is an index for the correlation between the geometric center point and the discrete feature pixel point; in this embodiment, the correlation heat source points represent discrete feature pixel points that can characterize the fire risk. In the identification of the fire risk area on the target floor, there may be discrete heat source points at the boundary of the fire risk area, that is, heat source points far from the main fire occurrence area but within a specific range. Therefore, the correlation between the aggregated feature pixel points and the discrete heat source points can be analyzed by the Euclidean distance, that is, the discrete heat source points within a specific distance are used as the correlation heat source points, and further, the distance between the aggregated feature pixel points and the discrete heat source points can be used as the correlation between the two.

[0034] In specific implementation, all the correlation heat source points are merged with the initial heat source area to obtain the potential risk area for the spread of the fire when a fire occurs on the target floor, that is: all the correlation heat source points are merged with the initial heat source area through the image processing tool OpenCV to obtain the potential risk area for the spread of the fire when a fire occurs on the target floor; wherein, all the correlation heat source points can be merged with the initial heat source area through the cv2.convexHull function of the image processing tool OpenCV to obtain the potential risk area for the spread of the fire when a fire occurs on the target floor, which will not be elaborated here.

[0035] It should be noted that in this application, the potential risk area represents the area where fire spread may occur. The potential risk area includes the fire source area and the heat diffusion path around it, reflecting the direction and scope of possible fire spread. In fire warning, the potential risk area is constructed based on the spatial distribution, temperature change pattern, and geometric relationship of characteristic pixel points, which can provide early warning information on the development of fire for the fire protection system and distinguish other non-risky heat sources. Furthermore, it is conducive to judging the possibility and influence scope of fire spread, thereby guiding subsequent emergency response and protection measures, and avoiding the interference of the identification of fire risk areas caused by other heat sources. When a fire occurs, the temperature in the fire area will increase rapidly over time. Therefore, the potential risk area of fire spread in the target floor during a fire can be constructed by multiple characteristic pixel points with temperature mutations.

[0036] In step 103, determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs in the target floor. Based on the precursor characteristics and the local temperature peak points in the thermal imaging image, perform fire point location analysis on the potential risk area to obtain the location loss during fire point location of the potential risk area.

[0037] In some embodiments, refer to Figure 3 As shown, this figure is an exemplary flowchart for determining precursor characteristics according to some embodiments of the present application. In this embodiment, the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs in the target floor can be implemented by the following steps: First, in step 1031, extract the center pixel point of the maximum circumscribed circle of the potential risk area; Then, in step 1032, obtain all the edge pixel points on the edge contour of the potential risk area, and calculate the relative distances between the center pixel point and each edge pixel point; Finally, in step 1033, determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area according to all the relative distances.

[0038] Specifically, the center pixel point of the maximum circumscribed circle of the potential risk area can be extracted by the image processing tool OpenCV, which will not be elaborated here. Among them, the center pixel point represents the center of the maximum circumscribed circle of the potential risk area. By determining the center pixel point, the outward expansion state of the potential risk area can be further analyzed.

[0039] Specifically, the relative distances between the center pixel point and each edge pixel point can be calculated by the Euclidean distance, which will not be elaborated here.

[0040] In specific implementation, the precursor features of the outward diffusion of the edge contour of the potential risk area are determined according to all relative distances, that is: the statistical features of all relative distances are used as the precursor features of the outward diffusion of the edge contour of the potential risk area. For example, the variance of all relative distances is used as the precursor features of the outward diffusion of the edge contour of the potential risk area. The variance measures the degree of fluctuation of all relative distances, that is, the degree of variation in the magnitudes of individual relative distances. The degree of fluctuation of all relative distances also measures the irregularity of the edge contour of the potential risk area. Usually, when a fire occurs, the fire area will present an irregular edge contour. The more irregular the edge contour is, the greater the risk of fire spread, and the greater the degree of outward diffusion of the fire risk area. Therefore, the precursor features of the outward diffusion of the edge contour of the potential risk area can be measured by the irregularity of the edge contour of the potential risk area.

[0041] It should be noted that in this application, the precursor features represent the dynamic trend presented in the initial stage of fire spread. Specifically, the precursor features are the dynamic trend indicating that the fire is about to spread outward from the potential risk area. By determining the precursor features, the physical and geometric dynamic signals in the early stage of fire spread can be effectively quantified and analyzed, realizing the accurate identification and prediction of fire risk from the potential state to the actual spread stage.

[0042] In some embodiments, based on the precursor features and the local temperature peak points in the thermal imaging image, the fire point location analysis of the potential risk area is carried out. The location loss when carrying out the fire point location of the potential risk area can be realized by the following steps, that is: Traverse the image pixels in the thermal imaging image through a preset sliding window, and extract the pixel points corresponding to the peak pixels in each sliding window as the local temperature peak points in the thermal imaging image; Compare the precursor features with the precursor feature threshold. When the precursor features are greater than the precursor feature threshold, dynamically correct the local temperature peak points in each sliding window, and then obtain the local temperature peak correction points corresponding to each sliding window; Extract the local temperature peak prediction points corresponding to each sliding window; Determine the location loss when carrying out the fire point location of the potential risk area according to the local temperature peak correction points and the local temperature peak prediction points corresponding to each sliding window.

[0043] When specifically implemented, the image pixels in the thermal imaging image are traversed through a preset sliding window, and the pixel points corresponding to the peak pixels in each sliding window are extracted as the local temperature peak points in the thermal imaging image, that is: the preset sliding window traverses the image pixels in the thermal imaging image from top to bottom and from left to right according to a preset sliding step length, and the pixel points corresponding to the peak pixels in each sliding window are extracted as the local temperature peak points in the thermal imaging image. The preset sliding window is set to 5×5 pixels, and the preset sliding step length is set to 3 pixels. In addition, in other embodiments, it can also be set according to actual needs, which is not limited here.

[0044] It should be noted that in this application, the local temperature peak point represents the pixel point corresponding to the maximum temperature in the sliding window. By determining the local temperature peak point, the fire points in the potential fire risk area can be further identified.

[0045] When specifically implemented, the precursor feature is compared with the precursor feature threshold. When the precursor feature is greater than the precursor feature threshold, it indicates that the irregularity of the fire spread edge at this time is stronger, and the mutability of the fire gradually increases. Therefore, when locating the fire point, it will be affected by the mutability of the fire, resulting in a deviation in the location of the fire point. Furthermore, it is necessary to dynamically correct the local temperature peak points in each sliding window, that is, to dynamically correct the local temperature peak points in each sliding window through the image processing tool OpenCV, and then obtain the local temperature peak correction points corresponding to each sliding window. Among them, the cv2.resize function of the image processing tool OpenCV can be used to dynamically correct the local temperature peak points in each sliding window, which will not be elaborated here; among them, the precursor feature threshold can be set through machine learning after learning a large number of historical precursor features, which will not be elaborated here.

[0046] When specifically implemented, the local temperature peak prediction points corresponding to each sliding window can be extracted from the floor fire monitoring database. The local temperature peak prediction points are the prediction points obtained by simulating the floor fire scene of the same scenario and making predictions. The specific prediction method can adopt the existing prediction method for prediction, which is not limited here; among them, the floor fire monitoring database in this embodiment is an intelligent database specifically used for storing and managing data related to floor fire monitoring; among them, the local temperature peak prediction point in this embodiment represents the local temperature peak point obtained after prediction.

[0047] In specific implementation, the positioning loss for fire point positioning in the potential risk area is determined based on the local temperature peak correction points and local temperature peak prediction points corresponding to each sliding window, that is: the pixel mean square error of the local temperature peak correction points and local temperature peak prediction points corresponding to all sliding windows is used as the positioning loss for fire point positioning in the potential risk area. In fire monitoring, by calculating the pixel mean square error of the local temperature peak correction points and local temperature peak prediction points corresponding to all sliding windows, the spatial deviation degree between the observed value and the physical model prediction value can be quantified. The greater the spatial deviation degree, the more inaccurate the fire point positioning, and the greater the positioning loss. The smaller the spatial deviation degree, the more accurate the fire point positioning, and the smaller the positioning loss. Therefore, the positioning loss for fire point positioning in the potential risk area can be measured by the spatial deviation degree.

[0048] It should be noted that in this application, the positioning loss represents the degree of deviation between the fire point positioning result and the prediction of the theoretical physical model. This loss value directly reflects the accuracy and reliability of the fire source position prediction. By determining the positioning loss, the deviation degree between the fire point positioning result and the theoretical prediction model can be quantified, providing a key quality assessment index for the intelligent fire protection system. When the positioning loss is low, it indicates that the system's recognition of the fire source highly coincides with the physical diffusion law, which can improve the early warning confidence level. When the loss value exceeds the threshold, it indicates that there may be sensor noise interference or model mismatch problems, triggering the system to further identify the fire risk area to enhance the analysis ability of the fire form characteristics, improve the accuracy of fire spread prediction, and then optimize the triggering mechanism of the early warning system to avoid the inaccuracy of fire early warning using single temperature monitoring in complex urban environments.

[0049] It should also be noted that in this application, the fire point positioning analysis refers to the process of positioning the fire points in the potential fire risk area. Among them, based on the precursor characteristics and the local temperature peak points in the thermal imaging image, the fire point positioning analysis of the potential risk area is carried out, that is: the image pixels in the thermal imaging image are traversed through a preset sliding window, and the pixel points corresponding to the peak pixels in each sliding window are extracted as the local temperature peak points in the thermal imaging image; the precursor characteristics are compared with the precursor characteristic threshold. When the precursor characteristics are greater than the precursor characteristic threshold, the local temperature peak points in each sliding window are dynamically corrected to obtain the local temperature peak correction points corresponding to each sliding window; the local temperature peak prediction points corresponding to each sliding window are extracted; the positioning loss for fire point positioning in the potential risk area is determined based on the local temperature peak correction points and local temperature peak prediction points corresponding to each sliding window, and then the positioning loss is used as the result of the fire point positioning analysis, thus completing the fire point positioning analysis of the potential risk area.

[0050] In step 104, based on the positioning loss, dynamic area monitoring of the edge contour of the potential risk area is performed at different times. When the area growth rate of the edge contour exceeds a preset threshold, mapping verification of the fire risk area of the target floor is performed using the edge contour and the visible light image at the corresponding time to obtain the confidence fire risk area of the target floor.

[0051] In some embodiments, the dynamic area monitoring of the edge contour of the potential risk area based on the positioning loss can be implemented by the following steps, namely: Compare the positioning loss with the positioning loss threshold. When the positioning loss is greater than the positioning loss threshold, perform dynamic area monitoring of the edge contour of the fire risk area at different times to obtain dynamic area monitoring data, and determine the area growth rate of the edge contour at different times according to the dynamic area monitoring data; When the positioning loss is less than or equal to the positioning loss threshold, do not perform dynamic monitoring of the edge contour of the fire risk area.

[0052] In specific implementation, when the positioning loss is greater than the positioning loss threshold, perform dynamic area monitoring of the edge contour of the fire risk area through an infrared camera, and combine the edge contour areas monitored at each time to obtain dynamic area monitoring data; the dynamic area monitoring data includes multiple edge contour areas, and the edge contour area represents the size of the edge contour in the fire risk area.

[0053] In specific implementation, determine the area growth rate of the edge contour at different times according to the dynamic area monitoring data, that is: extract the edge contour areas at each time in the dynamic area monitoring data. For each time, calculate the difference between the edge contour area at this time and the edge contour area at the previous time, and use the quotient of the difference calculation result and the time interval as the area growth rate of the edge contour at this time, so as to obtain the area growth rate of the edge contour at different times.

[0054] It should be noted that in this application, the area growth rate represents the growth speed of the area of the edge contour in the fire risk area with time. The area growth rate reflects the rate of fire spread. It is used to evaluate the development trend of the fire and is an important parameter for triggering early warnings in the intelligent fire protection system.

[0055] In specific implementation, when the area growth rate of the edge contour exceeds a preset threshold, the edge contour at the corresponding moment and the visible light image are used to perform mapping verification on the fire risk area of the target floor. The preset threshold can be set by machine learning based on the area growth rate data of the edge contour during historical fire occurrences, which is not limited here. The preset threshold is used to determine whether to perform spatial mapping verification between the edge contour at the corresponding moment and the visible light image. The area growth rate measures the fire spread state of the current potential risk area. The larger the area growth rate, the more obvious the fire spread state of the current potential risk area, and thus it is urgent to further monitor the fire spread area to determine the final fire risk area.

[0056] It should be noted that in this application, the edge contour at the corresponding moment and the visible light image are images that have been time-registered through image registration technology, which will not be elaborated here.

[0057] In some embodiments, the mapping verification of the fire risk area of the target floor by the edge contour at the corresponding moment and the visible light image to obtain the confidence fire risk area of the target floor can be achieved by the following steps, namely: Extract the contour pixel coordinate set of the edge contour at the corresponding moment, and map the contour pixel coordinate set to the corresponding spatial position of the visible light image at the same moment to obtain the mapped contour area of the target floor; Use a multi-scale filtering algorithm to detect the light intensity of all pixels in the mapped contour area to obtain a cluster of high-brightness pixel points when a fire occurs on the target floor; Construct the confidence fire risk area of the target floor according to all the high-brightness pixel points in the cluster of high-brightness pixel points.

[0058] In specific implementation, the contour pixel coordinate set of the edge contour at the corresponding moment is extracted, and the contour pixel coordinate set is mapped to the corresponding spatial position of the visible light image at the same moment to obtain the mapped contour area of the target floor, that is: the contour pixel coordinates of each edge pixel point in the edge contour at the corresponding moment are extracted through the image processing tool OpenCV, and all the contour pixel coordinates are combined to obtain the contour pixel coordinate set of the edge contour at the corresponding moment, and then each contour pixel coordinate in the contour pixel coordinate set is mapped to the corresponding spatial position of the visible light image at the same moment through OpenCV to obtain the mapped contour area of the target floor; among them, the contour pixel coordinates of each edge pixel point in the edge contour at the corresponding moment can be extracted through the cv2.findContours function of OpenCV, and each contour pixel coordinate in the contour pixel coordinate set can be mapped to the corresponding spatial position of the visible light image at the same moment through the cv2.warpPerspective function of OpenCV, which will not be elaborated here; in this embodiment, the mapped contour area refers to the area where the edge contour is mapped in the visible light image.

[0059] In specific implementation, a multi-scale filtering algorithm is adopted to detect the light intensity of all pixels in the mapped contour region, and a cluster of highlighted pixel points when a fire occurs on the target floor is obtained, that is: a Gaussian pyramid multi-scale space is constructed for the visible light image in the mapped contour region, the visible light image in the mapped contour region in each scale space layer is converted to the HSV color space, a converted contour region map corresponding to each scale space layer is obtained, highlighted pixel points are extracted from the converted contour region map corresponding to each scale space layer, a plurality of highlighted pixel points corresponding to each scale space layer are obtained, the plurality of highlighted pixel points corresponding to each scale space layer are upsampled to the original visible light image and marked, and the marked pixel points are combined to obtain a cluster of highlighted pixel points when a fire occurs on the target floor; wherein, the cv2.pyrDown function of the image processing tool OpenCV can be used to construct a Gaussian pyramid multi-scale space for the visible light image in the mapped contour region, and the visible light image in the mapped contour region is converted to the HSV color space in each scale space layer. In addition, when constructing a Gaussian pyramid multi-scale space for the visible light image in the mapped contour region, the scale series of the Gaussian pyramid is set to 3 to 5 layers, and the downsampling factor is set to 0.5. In addition, in other embodiments, it can also be set according to actual needs, which is not limited here; wherein, highlighted pixel points are extracted from the converted contour region map corresponding to each scale space layer, and a plurality of highlighted pixel points corresponding to each scale space layer are obtained, that is: for each scale space layer, the pixel value corresponding to each pixel point in the converted contour region map corresponding to the scale space layer is compared with a preset pixel value, and the pixel points with pixel values greater than the preset pixel value are used as highlighted pixel points, and then a plurality of highlighted pixel points corresponding to each scale space layer are obtained; wherein the plurality of highlighted pixel points corresponding to each scale space layer can be upsampled to the original visible light image and marked through the cv2.resize function of OpenCV, which is not elaborated here; wherein, the preset pixel value is a preset standard pixel value used to distinguish highlighted pixel points. The preset pixel value can be determined by analyzing a large amount of historical fire visible light image data and automatically learning the light intensity characteristics of fires in different scenarios using machine learning algorithms (such as clustering algorithms, neural networks, etc.), which is not limited here.

[0060] It should be noted that in this embodiment, the converted contour region map refers to the map obtained after converting the visible light image in the mapped contour region to the HSV color space; in this embodiment, the cluster of highlighted pixel points refers to a set containing a plurality of highlighted pixel points, and the highlighted pixel points refer to the pixel points in the visible light image whose brightness is significantly higher than the surrounding environment. By determining the highlighted pixel points, the risk area when a fire occurs on the target floor can be identified more accurately.

[0061] In specific implementation, a confidence fire risk area of the target floor is constructed based on all the highlighted pixel clusters, that is: convex hull fitting is performed on all the highlighted pixels in the highlighted pixel clusters, and holes are filled by combining morphological closing operations to generate the confidence fire risk area of the target floor; among them, the convex hull fitting can be performed on all the highlighted pixels in the highlighted pixel clusters through the cv2.convexHull function of OpenCV, and holes are filled by combining morphological closing operations, and then the confidence fire risk area of the target floor is generated, which will not be elaborated here.

[0062] It should be noted that in this application, the confidence fire risk area represents a closed area where there is a high probability of a fire spreading threat. In floor fire early warning, traditional thermal imaging monitoring often relies on a fixed temperature threshold and is easily affected by environmental temperature fluctuations. Visible light flame recognition may be misreported due to factors such as light changes and reflections. However, the confidence fire risk area constructs a more reliable fire area assessment method in the spatial and temporal dimensions through multi-source information fusion, combining the dynamic changes of the edge contour, the disturbance of the smoke texture, and the light intensity distribution density of the highlighted pixel clusters. It can not only reduce the false alarm rate of fire early warning, but also provide real-time and evolvable fire early warning results by dynamically monitoring the fire spreading trend, enabling the intelligent fire protection system to still play an effective role in complex environments, thereby enhancing the intelligent level of fire prevention and control.

[0063] It should also be noted that in this application, spatial mapping verification means establishing a spatial correspondence relationship between the image data obtained by different sensors (such as thermal imaging and visible light cameras) and verifying the characteristics of the target area in different modalities; among them, the fire risk area of the target floor is mapped and verified by the edge contour and the visible light image at the corresponding moment, that is: the contour pixel coordinate set of the edge contour at the corresponding moment is extracted, and the contour pixel coordinate set is mapped to the corresponding spatial position of the visible light image at the same moment to obtain the mapped contour area of the target floor; a multi-scale filtering algorithm is used to detect the light intensity of all pixels in the mapped contour area to obtain the highlighted pixel clusters when a fire occurs on the target floor; a confidence fire risk area of the target floor is constructed based on all the highlighted pixels in the highlighted pixel clusters, that is, the confidence fire risk area is used as the result of spatial mapping verification, and then the mapping verification of the fire risk area of the target floor by the edge contour and the visible light image at the corresponding moment is completed.

[0064] In step 105, the confidence fire risk area is uploaded to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire early warning terminal.

[0065] In specific implementation, the confidence fire risk area is uploaded to the Internet of Things center in the form of data packets through Internet of Things communication protocols (such as NB-IoT, LoRa, etc.), and the Internet of Things center integrates the received confidence fire risk area to automatically generate a fire spread warning signal. The fire spread warning signal includes the spread area where the fire occurs. Subsequently, the Internet of Things center sends the fire spread warning signal to the fire warning terminal, such as a mobile phone APP, a display screen in the fire control room, or an on-site audible and visual alarm, etc., through a wireless communication network (such as 4G, 5G, or a dedicated Internet of Things communication link), so as to realize the rapid transmission and response of fire warning information and ensure that relevant personnel can take measures in time to cope with the fire risk.

[0066] In addition, on the other hand of the present application, in some embodiments, the present application provides an urban intelligent fire warning system, including a fire warning unit. Refer to Figure 4 , this figure is a schematic structural diagram of the fire warning unit shown according to some embodiments of the present application. The fire warning unit 200 includes: a retrieval module 201, a processing module 202, and an execution module 203, which are described as follows: Retrieval module 201. In the present application, the retrieval module 201 is mainly used to retrieve the fire monitoring video of the target floor from the Internet of Things platform and parse the thermal imaging image and visible light image of the target floor from the fire monitoring video; Processing module 202. In the present application, the processing module 202 is mainly used to extract multiple feature pixel points with temperature mutations in the thermal imaging image, and then construct a potential risk area for the spread of the fire when a fire occurs on the target floor based on all the feature pixel points; The processing module 202 is further used to determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor, and perform fire point positioning analysis on the potential risk area based on the precursor characteristics combined with the local temperature peak points in the thermal imaging image to obtain the positioning loss when performing fire point positioning on the potential risk area; In addition, the processing module 202 is further used to perform dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss. When the area growth rate of the edge contour exceeds a preset threshold, the fire risk area of the target floor is mapped and verified by the edge contour at the corresponding time and the visible light image to obtain the confidence fire risk area of the target floor; Execution module 203. In the present application, the execution module 203 is mainly used to upload the confidence fire risk area to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal.

[0067] In addition, the present application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to obtain the code and execute the above-mentioned fire warning method.

[0068] In some embodiments, referring to Figure 5 , this figure is a schematic structural diagram of a computer device for implementing the fire warning method according to some embodiments of the present application. The fire warning method in the above embodiments can be implemented by Figure 5 the computer device shown. The computer device 300 includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0069] The processor 301 can be a general-purpose central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more for controlling the execution of the fire warning method in the present application.

[0070] The communication bus 302 can be used to transfer information between the above components.

[0071] The memory 303 can be a read-only memory (ROM), or other types of static storage devices that can store static information and instructions, a random access memory (RAM), or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disks, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 303 can exist independently and be connected to the processor 301 through the communication bus 302. The memory 303 can also be integrated with the processor 301.

[0072] Among them, the memory 303 is used to store the program code for executing the solution of this application, and is controlled by the processor 301 for execution. The processor 301 is used to execute the program code stored in the memory 303. The program code may include one or more software modules. The determination of the fire warning method in the above embodiments can be implemented by one or more software modules in the processor 301 and the program code in the memory 303.

[0073] The communication interface 304, using any device such as a transceiver, is used to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0074] In a specific implementation, as an embodiment, the computer device may include multiple processors, and each of these processors may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0075] The above computer device may be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device may be a desktop computer, a laptop computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of this application do not limit the type of the computer device.

[0076] In addition, this application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above fire warning method is implemented.

[0077] Although the preferred embodiments of this application have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of this application.

[0078] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, this application is also intended to include these changes and modifications.

Claims

1. A fire warning method for a smart city fire protection system to give a fire spread warning for the floor where a fire occurs, characterized in that, The method includes the following steps: Retrieve the fire monitoring video of the target floor from the Internet of Things platform, and parse the thermal imaging image and visible light image of the target floor from the fire monitoring video; Extract multiple characteristic pixel points with temperature mutations in the thermal imaging image, and then construct a potential risk area for the spread of fire when a fire occurs on the target floor based on all the characteristic pixel points; Determine the precursor features of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor, and perform fire point positioning analysis on the potential risk area based on the precursor features combined with the local temperature peak points in the thermal imaging image to obtain the positioning loss when performing fire point positioning on the potential risk area; Perform dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss. When the area growth rate of the edge contour exceeds a preset threshold, perform mapping verification on the fire risk area of the target floor by the edge contour at the corresponding time and the visible light image to obtain the confidence fire risk area of the target floor; Upload the confidence fire risk area to the Internet of Things center, and the Internet of Things center sends a fire spread warning signal to the fire warning terminal.

2. The method according to claim 1, characterized in that Specifically, extracting multiple characteristic pixel points with temperature mutations in the thermal imaging image includes: Obtain the thermal imaging images of the target floor at different time nodes; Determine the temperature distribution matrix corresponding to the thermal imaging image at each time node; For each time node, perform pixel-by-pixel differential calculation of the corresponding temperature values between the temperature distribution matrix of the current time node and the temperature distribution matrix of the previous time node to obtain the temperature difference matrix corresponding to each time node; Segment the temperature difference matrix corresponding to each time node based on a preset temperature difference threshold to obtain multiple candidate pixel points corresponding to each time node; Extract the candidate pixel points that appear at each time node from all the candidate pixel points as the characteristic pixel points with temperature mutations in the thermal imaging image, and then obtain multiple characteristic pixel points with temperature mutations in the thermal imaging image; 3. The method according to claim 1, wherein Specifically, constructing a potential risk area for the spread of fire when a fire occurs on the target floor based on all the characteristic pixel points includes: Perform spatial density clustering analysis on all the characteristic pixel points to obtain an aggregated characteristic pixel point cluster and a discrete characteristic pixel point cluster; Perform morphological closing operation on all the aggregated characteristic pixel points in the aggregated characteristic pixel point cluster to obtain the initial heat source area of the target floor; Extract multiple correlation heat source points according to the correlation between the geometric center point of the initial heat source area and each discrete characteristic pixel point in the discrete characteristic pixel point cluster; Merge all the correlation heat source points with the initial heat source area to obtain a potential risk area for the spread of fire when a fire occurs on the target floor; 4. The method according to claim 1, wherein Specifically, determining the precursor features of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor includes: Extract the central pixel point of the largest circumscribed circle of the potential risk area; Obtain all the edge pixel points on the edge contour of the potential risk area, and calculate the relative distance between the central pixel point and each edge pixel point; Determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area based on all relative distances.

5. The method according to claim 1, characterized in that The dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss specifically includes: Compare the positioning loss with the positioning loss threshold. When the positioning loss is greater than the positioning loss threshold, perform dynamic area monitoring of the edge contour of the fire risk area at different times to obtain dynamic area monitoring data, and determine the area growth rate of the edge contour at different times according to the dynamic area monitoring data; When the positioning loss is less than or equal to the positioning loss threshold, do not perform dynamic monitoring of the edge contour of the fire risk area.

6. The method according to claim 1, wherein Perform mapping verification on the fire risk area of the target floor by the edge contour and the visible light image at the corresponding time to obtain the confidence fire risk area of the target floor, specifically including: Extract the contour pixel coordinate set of the edge contour at the corresponding time, and map the contour pixel coordinate set to the corresponding spatial position of the visible light image at the same time to obtain the mapped contour area of the target floor; Use a multi-scale filtering algorithm to detect the light intensity of all pixels in the mapped contour area to obtain a cluster of high-brightness pixel points when a fire occurs on the target floor; Construct the confidence fire risk area of the target floor according to all the high-brightness pixel points in the cluster of high-brightness pixel points.

7. The method according to claim 1, characterized in that, The fire monitoring video is obtained by a fire monitoring camera installed in the target floor.

8. An urban intelligent fire warning system, including a fire warning unit, characterized in that, The fire warning unit includes: A retrieval module for retrieving the fire monitoring video of the target floor from the Internet of Things platform and parsing the thermal imaging image and visible light image of the target floor from the fire monitoring video; A processing module for extracting a plurality of characteristic pixel points with temperature mutations in the thermal imaging image, and then constructing a potential risk area for the spread of the fire when a fire occurs on the target floor based on all the characteristic pixel points; The processing module is further configured to determine the precursor characteristics of the outward diffusion of the edge contour of the potential risk area when a fire occurs on the target floor, and perform fire point positioning analysis on the potential risk area based on the precursor characteristics combined with the local temperature peak points in the thermal imaging image to obtain the positioning loss when performing fire point positioning on the potential risk area; The processing module is further configured to perform dynamic area monitoring of the edge contour of the potential risk area at different times based on the positioning loss. When the area growth rate of the edge contour exceeds a preset threshold, perform mapping verification on the fire risk area of the target floor by the edge contour and the visible light image at the corresponding time to obtain the confidence fire risk area of the target floor; An execution module for uploading the confidence fire risk area to the Internet of Things center, and the Internet of Things center sending a fire spread warning signal to the fire warning terminal.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores code, and the processor is configured to obtain the code and execute the fire warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the fire warning method according to any one of claims 1 to 7.

Images