Risk warning system based on dynamic video images
By using a risk warning system based on dynamic video images, fire areas are divided and mapped to physical space grids. The remaining amount of combustibles is calculated by combining duration and rate, which solves the problems of combustible quantification and fire front clustering in fires and improves the accuracy of fire assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHANGCHUN SHOUJIA TECH CO LTD
- Filing Date
- 2026-04-17
- Publication Date
- 2026-06-16
AI Technical Summary
Existing technologies are insufficient to quantify the remaining amount of combustibles in a fire area and cannot accurately cluster fire zones, leading to inaccurate assessments of fire intensity after multiple fires converge.
By using a risk warning system based on dynamic video images, the system divides the burned area, the fire head area, and the non-combustion area, and maps them to a physical space grid. It then calculates the amount of combustible material remaining by combining the duration of the fire and the combustion rate, performs fire head clustering and identification, and predicts the convergence of multiple fire heads to assess the fire intensity.
It achieves precise raster mapping of fire areas, accurately calculates the remaining amount of combustibles and fire head clustering, and improves the accuracy of fire intensity assessment after multiple fire heads merge.
Smart Images
Figure CN122050122B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image recognition technology, specifically a risk warning system based on dynamic video images. Background Technology
[0002] In real-world fire monitoring scenarios, fire zones exhibit dynamic spread characteristics, fire heads are spatially dispersed and their combustion states change in real time, and fire-related parameters lack a unified, gridded quantitative system. Existing technologies reveal the following shortcomings: First, existing technologies often estimate the remaining combustible material in the burned area as a whole, easily overlooking the actual differences between burned and unburned portions and failing to incorporate fire duration and combustion rate into the calculations. Second, existing solutions cannot accurately cluster and identify discretely distributed fire head grids to define independent fire heads, easily confusing the spatiotemporal characteristics of different fire heads. This leads to inaccurate predictions of multiple fire head convergence grids, resulting in assessments of fire intensity after multiple fire head convergence that do not match the actual combustion state.
[0003] Therefore, there is an urgent need for a risk warning system based on dynamic video images that can quantitatively calculate the remaining amount of each combustible material, cluster fire heads, predict the merging grid of multiple fire heads, and assess the intensity of the fire after merging, in order to solve the above-mentioned technical bottlenecks and improve the accuracy and practical application value of risk warning in dynamic fire monitoring. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a risk warning system based on dynamic video images, which solves the problems of existing technologies being unable to quantify the remaining amount of combustibles in the burned area and unable to predict the convergence of multiple fire heads, resulting in inaccurate assessment of the fire intensity after convergence.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a risk warning system based on dynamic video images, comprising: a region identification and rasterization module, used to divide the burned area, the fire head area and the non-burning area based on the gray value of each pixel in the video image, and to map the region to the corresponding physical space raster based on the actual physical space of the pixel.
[0006] The remaining combustible material calculation module is used to calculate the amount of remaining combustible material in each physical space grid based on the burned and non-combustion areas after the physical space grid is divided, combined with the burning time and combustion rate.
[0007] The fire convergence prediction module is used to cluster and identify fire head regions after physical space grid division to determine the fire head region grid set. Based on each fire head region grid set, the path and fire head movement speed of each fire head are determined, and the convergence grid of multiple fire heads is predicted.
[0008] The fire intensity assessment module is used to calculate the fire intensity characterization value after the convergence of multiple fire heads based on the amount of combustible material remaining in each physical space grid in the area surrounding the multi-fire head convergence grid.
[0009] Compared with existing technologies, the present invention has the following advantages: The present invention realizes the division of burned areas, fire heads, and non-combustible areas, and then performs gridded mapping on the fire area. The remaining combustible material calculation module can combine the burning duration and combustion rate of each physical space grid to calculate the remaining combustible material at the grid level. At the same time, the fire head convergence prediction module can cluster and identify discrete fire head area grids to define independent fire heads. The fire intensity assessment module calculates the fire intensity characterization value based on the accurate grid combustible material remaining amount and the clear multi-fire head convergence grid. This solves the problems of existing technologies that make it difficult to quantify the remaining combustible material in burned areas and predict the multi-fire head convergence grid, resulting in inaccurate fire intensity assessment after convergence. Attached Figure Description
[0010] Figure 1 This is a schematic diagram of the module connections of the risk warning system based on dynamic video images of the present invention;
[0011] Figure 2 This is a flowchart illustrating the division of burned areas, fire head areas, and non-combustion areas in the risk warning system based on dynamic video images of the present invention.
[0012] Figure 3 This is a flowchart illustrating the calculation of fire intensity characterization values after multiple fireheads converge in the risk warning system based on dynamic video images according to the present invention. Detailed Implementation
[0013] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. Please refer to the accompanying drawings. Figure 1 This invention provides a technical solution: a risk warning system based on dynamic video images, including: a region identification and rasterization module, used to divide the burned area, the fire head area and the non-burning area based on the gray value of each pixel in the video image, and to map the region to the corresponding physical space raster based on the actual physical space of the pixel.
[0014] Considering the spatiotemporal differences in pixel grayscale values across different fire zones, executable judgment rules are constructed using quantization parameters to classify burned areas, fire front areas, and non-burning areas. Figure 2 As shown, the specific process is as follows: perform grayscale processing on the video image to determine the pixel grayscale value and pixel coordinates at the current instant.
[0015] Calculate the average grayscale value of a given pixel at the current instant and all pixels within its neighborhood, and calculate the grayscale change of this pixel between two adjacent instants. The neighborhood is an 8x8 area.
[0016] If the absolute value of the difference between the grayscale change of a pixel and 0 is less than or equal to a preset grayscale threshold, and the average grayscale value of the pixel is less than or equal to a fixed proportion of the global average grayscale value at the time of fire initiation, then the pixel coordinates of this pixel are classified as belonging to the burned area. The global average grayscale value at the time of fire initiation refers to the arithmetic mean of the pixel grayscale values collected at the time of fire initiation. The fixed proportion typically ranges from 0.6 to 0.8, and the preset grayscale threshold typically ranges from 0.01.
[0017] For pixels not belonging to the burned area, calculate the instantaneous change rate of pixel grayscale, and simultaneously calculate the pixel grayscale spatial neighborhood difference. The instantaneous change rate of pixel grayscale is the ratio of the grayscale change to the time interval between two instantaneous values.
[0018] If the instantaneous change rate of a pixel's grayscale is positive and shows a continuous increasing trend (e.g., increasing over three consecutive instantaneous values), and the neighborhood difference in the pixel's grayscale space is positive and the absolute value of the difference shows a continuous increasing trend (e.g., increasing over three consecutive instantaneous values), then the pixel coordinates of this pixel are assigned to the head region. The neighborhood difference in the pixel's grayscale space is the difference between the grayscale value of this pixel and the grayscale values of other pixels within its neighborhood.
[0019] The remaining pixel coordinates are assigned to the non-burning area.
[0020] The average grayscale value of a pixel is used to characterize the spatial correlation of a single pixel, avoiding misjudgment caused by abnormal grayscale values of a single pixel, and conforming to the spatial continuity characteristics of fire combustion; the grayscale change is used to reflect the change in the combustion state of a pixel in the time dimension. After the burning of the burned area stops, the grayscale tends to stabilize, so its grayscale change approaches zero.
[0021] For pixels not classified as burned areas, the instantaneous grayscale change rate quantifies the trend of combustion intensity, and the pixel grayscale spatial neighborhood difference quantifies the degree of fire concentration. Since the fire front area is in the stage of intensified combustion, both show a positive and continuously increasing trend.
[0022] This embodiment first divides the burned area by grayscale change and pixel grayscale mean, then filters the fire front area by grayscale instantaneous change rate and pixel grayscale spatial neighborhood difference, and finally classifies the remaining pixels as non-burning areas. The entire division process uses image grayscale features to quantify and simulate different stages of fire combustion, and the logic is consistent with the physical laws of actual fire.
[0023] Considering that pixel-level region division results are difficult to adapt to actual fire monitoring, and that the actual physical space range corresponding to a single pixel is extremely small and cannot reflect the overall distribution of the fire area, it is necessary to map the pixel-level regions to the physical space grid corresponding to the actual physical space. The specific process is as follows: extract the pixel coordinates of all pixels in the burned area, the fire head area, and the non-burning area, and convert them into the corresponding actual physical space coordinates.
[0024] Traverse the actual physical space coordinates corresponding to each pixel, compare the actual physical space coordinates corresponding to each pixel in the burned area, the fire head area and the non-burning area with the preset set of physical space coordinates of each physical space grid (i.e. the range of the actual physical space horizontal coordinate and the range of the vertical coordinate corresponding to each physical space grid), determine whether the actual physical space coordinates fall within the range of the physical space coordinates of a certain physical space grid, and thus determine the physical space grid that uniquely corresponds to each pixel.
[0025] Determine the physical space grid corresponding to the coordinates of each pixel. For each preset physical space grid, aggregate the region attributes of all pixels within its range. If more than half of the pixels in a certain physical space grid belong to a certain region type (burnt area, fire area, and non-combustible area), then the physical space grid is determined to be the corresponding region type, thus completing the mapping of burnt areas, fire areas, and non-combustible areas to the preset physical space grid.
[0026] The physical space grid is divided according to the actual physical space range. Each physical space grid has an equal area and the physical space grids are interconnected.
[0027] In this embodiment, the physical space grid has a fixed actual physical space range. After mapping, it can intuitively present the physical space distribution and range size of each area, making it easier for staff to quickly grasp the overall situation of the fire.
[0028] The remaining combustible material calculation module is used to calculate the amount of remaining combustible material in each physical space grid based on the burned and non-combustion areas after the physical space grid is divided, combined with the burning time and combustion rate.
[0029] The specific process is as follows: If the physical space grid corresponds to a non-combustion area, the remaining amount of combustible material is the initial combustible amount corresponding to the physical space grid. If the physical space grid corresponds to a burned area, the product of the initial combustible amount corresponding to the physical space grid, the grid combustion rate, and the grid burning time is taken as the burned amount, and the difference between the initial combustible amount and the burned amount is taken as the remaining amount of combustible material.
[0030] Initial combustible mass refers to the total mass of all combustible vegetation materials that can participate in combustion within the actual physical space corresponding to a single physical space grid before a fire occurs. The unit is kilograms, and it can be obtained directly through various existing means, such as on-site measurement.
[0031] The grid burning rate is the proportion of combustibles burning per unit time within a physical grid. It is a non-negative real number between 0 and 1 and can be obtained through various existing methods, such as indoor standard combustion experiments, on-site small sample burning measurements, or database lookup methods. The burning rate of each type of combustible is determined, and the burning rate of each type of combustible is standardized by using the maximum burning rate as the denominator to obtain the grid burning rate.
[0032] The grid fire duration refers to the cumulative duration of a single physical space grid from the start of combustion to the current instantaneous time. It is obtained by determining the target instantaneous time value at which the average instantaneous rate of change of pixel grayscale for each pixel corresponding to the physical space grid is 0. The time difference between the target instantaneous time value and the instantaneous time value at which grid combustion begins is recorded as the fire duration.
[0033] The remaining combustible material refers to the mass of combustible material that is not currently participating in combustion within a single physical space grid. It is measured in kilograms and is a non-negative real number, serving as a core indicator of the remaining combustion potential of a physical space grid. If the calculated remaining combustible material is negative, it is corrected to 0.
[0034] Additionally, it should be noted that some physical spatial grids may contain both burned and unburned areas. For such physical spatial grids, the remaining combustible material is calculated as follows: First, extract the combustion status (burned / unburned) of all pixels within the physical spatial grid, and calculate the proportion of burned pixels to the total number of pixels in the physical spatial grid, denoted as the grid combustion percentage. Then, recalculate the amount of combustible material, with the formula modified to equal the product of the initial combustible material, the grid combustion rate, the grid overheating time, and the grid combustion percentage. Finally, calculate the remaining combustible material based on the modified amount of combustible material, which is equal to the initial combustible material minus the amount of combustible material.
[0035] The fire convergence prediction module is used to cluster and identify fire head regions after physical space grid division to determine the fire head region grid set. Based on each fire head region grid set, the path and fire head movement speed of each fire head are determined, and the convergence grid of multiple fire heads is predicted.
[0036] Considering that the firehead regions exist as discrete grids after grid division and it is impossible to define independent fireheads, without clustering and labeling, the spatiotemporal characteristics of each firehead will be confused, making it impossible to carry out subsequent firehead path fitting and firehead movement speed calculation. In addition, considering that fireheads have similarities in the instantaneous change rate of pixel grayscale, the process of firehead clustering and labeling after physical space grid division is as follows: For each firehead region grid obtained after firehead region grid division, the corresponding central physical coordinates are extracted. At the same time, the mean instantaneous change rate of pixel grayscale at each instantaneous value under continuous time sampling sequence is extracted to form a change rate mean sequence, which characterizes the change characteristics of the combustion intensity of the firehead region grid over time.
[0037] For a certain fire zone grid A, calculate the spatial straight-line distance between its central physical coordinates and the central physical coordinates of other fire zone grids. Specifically, this can be calculated using Euclidean distance.
[0038] If the spatial straight-line distance between fire area grid A and fire area grid B is less than or equal to a fixed multiple of the physical side length of the grid, then fire area grid B is determined to be a spatial neighbor grid of fire area grid A, and the two fire area grids are associated.
[0039] For all fire area grids associated with fire area grid A, calculate the numerical similarity of the mean instantaneous grayscale change rate of the corresponding pixels (that is, calculate the similarity of two mean change rate sequences). If the numerical similarity is less than or equal to the preset grayscale change rate similarity threshold, then the two fire area grids are determined to be spatiotemporally associated.
[0040] Firefield grids that have both spatial and spatiotemporal relationships with firefield grid A are included in the same firefield grid set, and each firefield is assigned a unique identifier.
[0041] Repeatedly perform spatial neighborhood association and spatiotemporal association verification on other firehead area grids to complete the firehead clustering identification.
[0042] The physical side length of a grid refers to the actual horizontal or vertical side length of a single physical space grid. A fixed multiplier is typically set at 2-3 times the grid length. This ensures coverage of the spatial range of grids in the discretely distributed fire area during the spread of the same fire, while avoiding the problem of misclassifying adjacent physical space grids of the same fire as unrelated grids due to an excessively small value.
[0043] Numerical similarity is a quantitative value of the consistency between the mean change rate sequences of two fire zone grids, characterizing the spatiotemporal consistency of the combustion intensity changes of grids in different physical spaces over time. The calculation process is as follows: calculate the sum of the absolute values of the differences in the instantaneous change rate of pixel grayscale of fire zone grid A and other fire zone grids at the same instantaneous value, and then divide by the total length of the time sampling sequence to obtain the numerical similarity.
[0044] In this embodiment, spatial neighborhood associations are first determined, and then the spatiotemporal associations of combustion intensity are determined through numerical similarity. This approach not only aligns with the spatial continuity of fire spread but also matches the consistency of combustion intensity changes within the same fire. Furthermore, the method for calculating numerical similarity is logically simple, highly operable, requires no complex algorithms, and can be rapidly executed in batches by computers, thus meeting the real-time requirements of fire monitoring.
[0045] In order to adapt to the overall continuity of the fire head movement, avoid the deviation caused by the position fluctuation of individual discrete fire head grids, and capture the overall spread direction of the fire head, the process of determining the path and fire head movement speed of each fire head based on the grid set of each fire head region is as follows: extract the center coordinates of each fire head region grid set at each instantaneous time value.
[0046] Based on the least squares method, a straight line is fitted to the trajectory of the center coordinates changing with the instantaneous value over time. The slope and intercept of the fitted straight line are calculated to obtain the path of each fire head. The width of the fire head is determined based on the grid set of each fire head region.
[0047] The ratio of the distance difference between the center coordinates at two instantaneous times to the time interval between the two instantaneous times is taken as the flame movement speed of the corresponding flame. The distance difference can be calculated using Euclidean distance.
[0048] The least squares method can fit a straight line that fits the trajectory of the fire head center coordinates as time changes, which can smooth out instantaneous position errors. The slope and intercept of the fitted straight line can quantify the spatial characteristics of the fire head path.
[0049] The process of determining the flame width based on the set of grid cells for each flame area is as follows: calculate the distance from the center physical coordinate to the center coordinate of each flame area grid cell in the set of grid cells for each flame area, and take twice the maximum distance as the flame width.
[0050] Furthermore, the process of predicting the convergence grid of multiple fire heads is as follows: determine the center physical coordinates at the latest instantaneous time value, then traverse all physical space grids along the spread direction of the fire head path, and determine whether each physical space grid is covered by the width of the fire head by the formula of the perpendicular distance from the point outside the plane straight line to the straight line.
[0051] For a physical space grid determined to be covered by the flame width, calculate the Euclidean distance from the center physical coordinates of the grid to the center physical coordinates of the flame at the latest instantaneous time value. Based on the flame's movement speed, determine the arrival time of the physical space grid corresponding to the flame width coverage along the flame's path.
[0052] If the time difference between at least two different fireheads arriving at the same physical space grid is less than or equal to a preset time threshold, then it is determined that multiple fireheads have converged, and the corresponding physical space grid is designated as the multiple firehead convergence grid. The preset time threshold is set to 1-5 time sampling intervals.
[0053] In this embodiment, the center physical coordinates of the latest instantaneous value of the flame are used as the reference. The coverage of the flame width on the physical space grid is determined along the path using the formula of the vertical distance from the point outside the plane to the line, which fits the characteristic of the flame spreading in a strip. At the same time, the time for the flame to reach each covered physical space grid is accurately calculated by combining Euclidean distance with the flame's moving speed, ensuring the objectivity and numerical accuracy of the arrival time determination.
[0054] The fire intensity assessment module is used to calculate the fire intensity characterization value after the convergence of multiple fire heads based on the amount of combustible material remaining in each physical space grid in the area surrounding the multi-fire head convergence grid.
[0055] Considering that fire intensity is related to both the material basis of combustion (the amount of remaining combustible material) and the spatial morphology of the merged fire heads (length, width, and coverage area), the specific calculation of the fire intensity characterization value after the convergence of multiple fire heads is as follows: Figure 3 As shown: using the grid where multiple fire heads converge as the connection point, connect each fire head and calculate the length of the fire head after connection.
[0056] Using the width direction of each connected flame as a baseline, a coverage area of a set length is taken along the direction perpendicular to the width of the flame, and the total amount of combustible material remaining in all physical spatial grids within the coverage area is calculated. The set length is 1-3 times the width of the flame, and can be adjusted according to the early warning accuracy requirements of fire monitoring. The higher the early warning accuracy requirement, the smaller the set length should be.
[0057] The ratio of the total amount of remaining combustible material to the length of the connected fireheads is used as a characterization value of the fire intensity after multiple fireheads merge.
[0058] The total amount of remaining combustible material is the material basis for the continued burning and spread of the fire within the covered area, while the length of the connected fire head is the spatial scale of the ribbon-like fire head formed by the convergence of multiple fire heads. The ratio of the two can combine the material basis of combustion with the spatial shape of the fire head. Its physical meaning is the amount of remaining combustible material that a unit length of fire head can rely on in the ribbon-like fire head formed by the convergence of multiple fire heads. The value directly reflects the intensity and potential of the fire after the convergence of multiple fire heads. The larger the value, the more combustible material a unit length of ribbon-like fire head has after the convergence, and the higher the risk of causing a larger fire later.
[0059] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.
[0060] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0061] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.
[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0063] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A risk warning system based on dynamic video images, characterized in that, include: The region recognition and rasterization module is used to divide the burned area, the fire head area and the non-burning area based on the gray value of each pixel in the video image, and to map the region to the corresponding physical space raster based on the actual physical space of the pixel. The remaining combustible material calculation module is used to calculate the amount of remaining combustible material in each physical space grid based on the burned and non-combustion areas after the physical space grid is divided, combined with the burning time and the combustion rate. The fire convergence prediction module is used to cluster and identify fire head regions after physical space grid division to determine the fire head region grid set, determine the path and fire head movement speed of each fire head based on each fire head region grid set, and predict the multi-fire head convergence grid. The fire intensity assessment module is used to calculate the fire intensity characterization value after the convergence of multiple fire heads based on the amount of combustible material remaining in each physical space grid in the area surrounding the multi-fire head convergence grid.
2. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of dividing the burned area, the flame area, and the non-burning area based on the grayscale values of each pixel in the video image is as follows: Perform grayscale processing on the video image to determine the pixel grayscale value and pixel coordinates at the current instant. Calculate the average grayscale value of a pixel at the current instant and all pixels in its neighborhood, and calculate the grayscale change of this pixel between two adjacent instants. If the absolute value of the difference between the grayscale change of a certain pixel and 0 is less than or equal to the preset grayscale threshold, and the average grayscale value of the pixel is less than or equal to a fixed proportion of the global average grayscale value at the time of fire onset, the pixel coordinates of this pixel will be classified as the burned area. For pixels that do not belong to the over-fire area, calculate the instantaneous change rate of pixel grayscale and the difference between pixel grayscale neighborhoods. If the instantaneous change rate of the pixel grayscale of a certain pixel is positive and shows a continuous increasing trend, and the difference between the pixel grayscale spatial neighborhoods is positive and the absolute value of the difference shows a continuous increasing trend, then the pixel coordinates of this pixel are assigned to the fire head region. The remaining pixel coordinates are assigned to the non-burning area.
3. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of mapping the region to the corresponding physical space grid based on the actual physical space of the pixels is as follows: The actual physical space coordinates corresponding to the pixel coordinates of each pixel in the burned area, the fire head area, and the non-combustion area are compared with the physical space coordinate set of each preset physical space grid to determine the physical space grid corresponding to each pixel coordinate, and the burned area, the fire head area, and the non-combustion area are mapped to the corresponding physical space grid.
4. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of calculating the amount of combustible material remaining in each physical space grid is as follows: If the physical space grid corresponds to a non-combustion area, then the remaining amount of combustible material is the initial amount of combustible material corresponding to the physical space grid. If a physical space grid corresponds to a burned area, the product of the initial combustible amount, the grid burning rate, and the grid burning time is taken as the burned amount, and the difference between the initial combustible amount and the burned amount is taken as the remaining combustible amount.
5. The risk warning system based on dynamic video images according to claim 4, characterized in that, The process for determining the duration of fire exposure is as follows: Determine the target instantaneous time value at which the average instantaneous rate of change of pixel grayscale for each pixel corresponding to the physical space raster is 0; The time difference between the instantaneous value of the target time and the instantaneous value of the grid combustion start time is recorded as the overfire duration.
6. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of clustering and labeling firehead regions after physical space grid division is as follows: For each fire zone grid obtained after division, the corresponding central physical coordinates are extracted, and the mean instantaneous change rate of pixel grayscale at each instantaneous value in the continuous time sampling sequence is also extracted. For a given fire zone grid, calculate the spatial straight-line distance between its center physical coordinates and the center physical coordinates of other fire zone grids; If the spatial straight-line distance is less than or equal to a fixed multiple of the physical side length of the grid, then the two fire head area grids will be associated. For all fire area grids associated with a certain fire area grid, calculate the numerical similarity of the average instantaneous grayscale change rate of the corresponding pixels. If the numerical similarity is less than or equal to the preset grayscale change rate similarity threshold, then the two fire area grids are determined to be spatiotemporally associated. Assign a unique firehead identifier to a firehead area grid and other firehead area grids that have spatiotemporal relationships, and merge all firehead area grids into a firehead area grid set. Repeatedly perform spatial neighborhood association and spatiotemporal association verification on other firehead area grids to complete the firehead clustering identification.
7. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of determining the path and movement speed of each fire head based on the grid set of each fire head area is as follows: Extract the center coordinates of the grid set of each fire area at each instantaneous time value; Based on the least squares method, the trajectory of the center coordinates changing with the instantaneous value of time is fitted with a straight line, the slope and intercept of the fitted straight line are calculated, the path of each fire head is obtained, and the width of the fire head is determined based on the grid set of each fire head area. The ratio of the distance difference between the center coordinates at two instantaneous times to the time interval between the two instantaneous times is taken as the flame movement speed of the corresponding flame.
8. The risk warning system based on dynamic video images according to claim 7, characterized in that, The process of determining the width of a firehead based on the grid set of each firehead region is as follows: Calculate the distance from the center physical coordinates of each fire area grid in the fire area grid set to the center coordinates, and take twice the maximum distance as the fire width.
9. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of predicting the convergence grid of multiple fireheads is as follows: Based on the fire head's movement speed, determine the arrival time of the physical space grid corresponding to the fire head's width coverage along the fire head's path. If the time difference between at least two different fireheads arriving at the same physical space grid is less than or equal to a preset time threshold, it is determined that multiple fireheads have converged, and the corresponding physical space grid is designated as the multiple firehead convergence grid.
10. The risk warning system based on dynamic video images according to claim 1, characterized in that, The process of calculating the intensity characterization value of a fire after multiple fires converge is as follows: Using the grid where multiple fireheads converge as the connection point, connect each firehead and calculate the length of the firehead after connection; Using the width direction of each connected flame as the baseline, a coverage area of a set length is taken along the direction perpendicular to the width of the flame, and the total amount of combustible material remaining in all physical space grids within the coverage area is calculated. The ratio of the total amount of remaining combustible material to the length of the connected fireheads is used as a characterization value of the fire intensity after multiple fireheads merge.