Forest fire monitoring and early warning system based on big data

Abstract

The invention relates to the technical field of environment monitoring and early warning, in particular to a forest fire monitoring and early warning system based on big data, which performs multi-period data superposition calculation in combination with a temperature change difference value, remarkably improves the recognition precision of a suspected point of a fire source, enhances the capture capability of dynamic change of forest environment temperature, and improves the accuracy of forest fire monitoring and early warning. According to the method, the identification rate of potential fire behavior is improved, difference analysis is performed by using a time period highly overlapped with humidity in a temperature trend curve, a high-risk section is accurately marked in combination with the temperature rise rate and the wind direction continuity, the judgment capability of a fire risk range is improved, and the judgment is matched with the change rate of a risk point through path overlapping number judgment. Quantitative control of trigger conditions is realized, so that a risk control decision has multiple response capabilities of spatial dimension, time dimension and dynamic evolution characteristics, and timeliness, spatial accuracy and risk response perspectiveness of forest fire prediction are improved.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and early warning technology, and in particular to a forest fire monitoring and early warning system based on big data. Background Technology

[0002] Environmental monitoring and early warning aims to monitor, analyze, and warn of various natural phenomena in the environment in real time through various sensors, remote sensing technologies, data acquisition and processing technologies, and communication networks. By timely acquiring and analyzing environmental data, it can predict potential natural disasters, climate change, forest fires, and floods, thereby achieving effective protection of public life and property.

[0003] The purpose of forest fire monitoring and early warning based on big data is to provide an effective means of protecting forest resources through real-time monitoring and early warning of forest fires. By collecting and processing environmental data from different sources, such as temperature, humidity, wind speed, meteorological changes, and fire sensor data, through a big data platform, we can predict and warn of fire occurrences, quickly identify early signs of fires, issue timely warnings, and provide decision support so that relevant departments can quickly take effective emergency measures to reduce the damage of fires to the ecological environment and people's lives and property.

[0004] Existing technologies rely primarily on sensor data acquisition in practical operation, which is insufficient in responding to spatial anomalies and temporal trends. They lack an active mechanism for recognizing visual features in images, making the detection of initial fire signs heavily dependent on parameter threshold triggers. This results in low timeliness of early fire identification. Furthermore, current methods mainly rely on rule settings and individual indicator comparisons, failing to establish a fusion model of historical fire data and real-time data. This makes it difficult to form dynamic triggering condition judgments, leading to misjudgments and missed detections. The lack of path structure calculations and propagation area estimations hinders the modeling of fire evolution trends. Regarding early warning mechanisms, triggering conditions are mainly based on fixed threshold settings, without establishing numerical fluctuation judgment standards for the dynamic evolution process of high-risk areas. This makes it impossible to respond to rapid local changes and boundary propagation risks, impacting the effectiveness and proactivity of forest fire management. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a forest fire monitoring and early warning system based on big data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a forest fire monitoring and early warning system based on big data includes:

[0007] Fire data analysis module: It acquires meteorological data, remote sensing images and forest environment data through sensors, extracts infrared hotspot pixels from remote sensing images, and combines temperature change differences in meteorological data to overlay and calculate data from multiple time periods in the region to generate initial fire risk analysis results.

[0008] Fire risk assessment module: Based on the initial fire risk analysis results, combined with historical fire data and real-time meteorological data acquired by sensors, it determines whether the fire initiation conditions are met and generates a fire risk index.

[0009] Fire propagation simulation module: Based on the fire risk index, extract the coordinates of the fire source and load a forest area layer containing terrain boundary and forest distribution information. Perform spatial overlay analysis on the fire source coordinates and the forest area layer, construct a wind direction guidance path and divide the propagation section according to the wind speed and humidity distribution, calculate the coverage area of ​​the path at each time period, and generate fire spread path prediction results.

[0010] Dynamic risk prediction module: Based on the fire spread path prediction results, a path structure map is constructed using a graph neural network and meteorological features are embedded. High-risk sections are continuously marked by combining the temperature rise rate and wind direction to generate dynamic fire risk prediction results.

[0011] Fire warning triggering module: Based on the fire dynamic risk prediction results and the fire spread path prediction results, calculate the incremental number of high-risk path points within multiple preset time windows and obtain the rate of change per unit time, determine whether the current rate of change is greater than the past average rate, and generate and transmit fire warning information.

[0012] As a further aspect of the present invention, the fire data analysis module includes:

[0013] Infrared pixel extraction submodule: It acquires remote sensing images, meteorological data and forest environment data through sensors, extracts infrared band pixel values ​​and performs channel separation, calculates the mean and standard deviation of infrared pixel values ​​in the region and filters out abnormal pixels, extracts regions with abrupt changes in infrared values ​​in the filtered results and divides pixel block boundaries, processes wind speed and direction and corrects coordinate position errors, compares environmental humidity field and terrain slope data with the same spatial resolution based on the corrected coordinates, filters spatial areas that meet the requirements of low humidity and large slope, and generates hot spot pixel coordinate set;

[0014] The time-series temperature difference overlay submodule: Based on the hot spot pixel coordinate set, it extracts remote sensing images of multiple time periods by coordinate index and compares the pixel values ​​of the infrared band. It calculates the temperature rise difference using the pixel values ​​of adjacent time periods in the sequence, generates a temperature change numerical sequence, identifies temperature rise abrupt frames, marks areas with continuous temperature rise trends, compares with the wind direction sequence in meteorological data, adjusts the coordinate position, compares and adjusts the shape of hot spot areas and constructs a continuous trajectory, and generates a dynamic sequence map of temperature rise areas.

[0015] Risk level determination submodule: Based on the dynamic sequence map of the temperature rise area, extract the temperature, humidity and wind data of the corresponding area and discretize the indicators, screen the intersection of the low humidity area and the high temperature rise area, determine the coverage of the wind direction expansion range, identify overlapping areas and count the number of risk factors, match the risk classification zoning rules and generate multi-level classification results, delineate the boundaries of suspected fire source blocks and process the layer labels, and obtain the initial fire risk analysis results.

[0016] As a further aspect of the present invention, the fire risk assessment module includes:

[0017] Spatial overlap comparison submodule: Based on the initial fire risk analysis results, extract the location coordinates of the fire source points from the initial fire risk analysis results and process the partition numbers, batch standardize the location coordinates of the fire source points and uniformly perform projection transformation, record the alignment structure fields after importing coordinate data in historical fires, calculate the valley value of the distance between the two coordinate sets and filter the overlapping points, extract the outline of the overlapping area, calculate the area overlap ratio, and obtain the historical spatial overlap distribution set.

[0018] Triggering Factor Screening Module: Based on the historical spatial overlapping distribution set, it retrieves real-time meteorological data within the coverage area of ​​each fire source point obtained by the sensor and extracts temperature, humidity and wind force as meteorological factors. It loads meteorological data within the corresponding time period of historical fires and divides the parameter range of each meteorological data. It compares the current meteorological data with the parameter range corresponding to historical fires and marks the matching. It accumulates the meteorological factors that meet the conditions of each fire source point and calculates the factor coverage ratio. It compares the factor coverage ratio with the preset threshold and completes the determination of the triggering condition. It then obtains the triggering status table of the triggering factor.

[0019] Risk index calculation submodule: Based on the triggering state table of the ignition factor, the triggering state code of the fire source point is converted and an index matrix is ​​constructed. Each index in the index matrix is ​​weighted to generate a standardized score. Combining the overlapping area ratio of each fire source point in the historical spatial overlapping distribution set, the fusion value is calculated according to the proportional weight rule. The risk level corresponding to the fusion value is classified and the risk interval of the spatial point is marked to generate the fire risk index.

[0020] As a further aspect of the present invention, the fire propagation simulation module includes:

[0021] Spatial mapping construction submodule: Based on the fire risk index, extract the coordinates of the fire source point to generate two-dimensional coordinate points, import the layer clipping boundary grid of the forest area structure map, align the fire source point coordinates with the layers of the forest area structure map and perform layer space overlay processing, extract wind direction data from meteorological data and construct a direction vector, construct the main axis path of the direction vector and the center of the fire source point and extend continuous line segments to generate a set of fire source wind direction path line segments;

[0022] Path segment division submodule: Based on the set of fire source wind direction path segments, the cumulative path segment distance is divided into regular equal-length segments. The coordinates of the first segment of each segment are obtained and the wind speed data of the corresponding position in the meteorological data is retrieved for matching. The humidity data in the meteorological data of each segment area is projected and the corresponding value is extracted to generate a wind speed and humidity combination index. The segment attributes are assigned, the index sets of each segment of the path are classified and the level segment identification is completed to obtain a graded propagation path segment map.

[0023] Coverage calculation submodule: Based on the hierarchical propagation path segment map, the segment propagation duration is set, the radius range of the line segment endpoints is constructed, the radius range graphic is expanded and buffer zones on both sides of the path are created, the outline layers of each segment are merged to generate a time period envelope, the envelope polygons are continuously spliced, and boundary consistency processing is performed to generate the fire spread path prediction result.

[0024] As a further aspect of the present invention, the dynamic risk prediction module includes:

[0025] Temperature Change Extraction Submodule: Based on the fire spread path prediction results, a graph neural network is used to extract the endpoint coordinates of the path segments and perform grid number mapping. Historical meteorological databases and multi-time period temperature data corresponding to the grid number positions are retrieved. Time index alignment is obtained according to the predicted time label of the path segments. The time series of temperature data for each path segment is sorted. Continuous data interpolation is processed. The temperature rise rate in the interpolation sequence is extracted. Continuous inflection points are marked. The location of the temperature rise segment is associated and matched with the path coordinates and the segment is marked. A path temperature trend map is generated.

[0026] Factor difference calculation submodule: Based on the path temperature trend map, it synchronously extracts humidity and temperature data corresponding to the path segments, matches the corresponding time periods, compares the temperature change sequence and humidity observation sequence of each path segment within the synchronous time window, extracts the value range overlap interval, calculates the difference at the same time point to form a point-by-point difference sequence; uses a sliding window filtering algorithm to smooth the difference sequence and fill the difference matrix, extracts the edges of high difference blocks in the matrix, screens continuous height difference segments, matches and merges the height difference segment index and path coordinates, and then binds the spatial location to obtain the climate factor change set;

[0027] High-risk section identification submodule: Based on the set of climate factor changes, calculate the rate of temperature change within the path segment and perform numerical filtering, retrieve the wind direction time series that matches the path coordinate segment, analyze the magnitude of the dominant wind direction change in continuous time slices, filter continuous segments with strong directional stability, cross-over the warming rate segment and the wind direction continuous segment and align the position index, convert the cross-segment classification code, mark the nested path layers, and generate dynamic fire risk prediction results.

[0028] As a further aspect of the present invention, the fire early warning triggering module includes:

[0029] Regional overlap judgment submodule: Based on the fire dynamic risk prediction results and the fire spread path prediction results, extract the boundary vector data of the two layers and perform coordinate system one processing, perform vector operation on the overlapping area to extract the cross area number, filter the cross area area, merge the extracted cross area number according to spatial adjacency, generate an aggregated area set, count the number of aggregated areas and the total area, determine whether it exceeds the set high-risk spatial coverage threshold and complete the matching identification, and generate an overlapping area judgment value set;

[0030] The path fluctuation detection submodule extracts the path segment coordinates from the fire dynamic risk prediction results based on the overlapping area judgment value set and sorts them by time dimension. It counts high-risk points in each time period to generate a change curve sequence, calculates the numerical increase in continuous time intervals to construct a difference sequence, matches the difference sequence with the historical average value, extracts abnormal segments, and obtains a path point fluctuation result table.

[0031] The early warning information generation submodule: Based on the path point fluctuation result table, it filters out high-risk segment identifiers, extracts high-frequency trigger condition tags using fuzzy logic reasoning, converts the identified segment codes, maps the segments with the original path map, matches the segment mapping results with the spatial location field, establishes a graphic structure template, assigns values ​​to the fields in the graphic structure template and completes text output rendering, and generates fire early warning information.

[0032] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0033] 1. In this invention, by combining the temperature change difference to perform multi-time period data superposition calculation, the accuracy of identifying suspected fire sources is significantly improved, and the ability to capture dynamic changes in forest environmental temperature is enhanced. After identifying the fire source, by introducing the spatial overlap calculation of historical fire data and real-time meteorological data, a spatial matching logic is constructed to effectively determine whether the current conditions have the basis for fire initiation, thereby improving the identification rate of potential fires.

[0034] 2. In this invention, by performing difference analysis on the time period that highly overlaps with humidity in the temperature trend curve, and combining the heating rate and wind direction persistence, high-risk areas are accurately marked, thereby improving the ability to determine the range of fire risk.

[0035] 3. In this invention, quantitative control is achieved by judging the number of overlapping paths and matching the rate of change of risk points, thereby enabling risk control decisions to have multiple response capabilities in terms of spatial, temporal, and dynamic evolutionary characteristics, thus improving the timeliness, spatial accuracy, and forward-looking nature of forest fire prediction and risk response. Attached Figure Description

[0036] Figure 1 This is a system flowchart of the present invention;

[0037] Figure 2 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0039] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0040] Example 1

[0041] Please see Figure 1 This invention provides a technical solution: a forest fire monitoring and early warning system based on big data, comprising:

[0042] Fire data analysis module: It acquires meteorological data, remote sensing images and forest environment data through sensors, extracts infrared hotspot pixels from remote sensing images, and combines temperature change differences in meteorological data to overlay and calculate data from multiple time periods in the region to generate initial fire risk analysis results.

[0043] Fire risk assessment module: Based on the initial fire risk analysis results, combined with historical fire data and real-time meteorological data acquired by sensors, it determines whether the fire initiation conditions are met and generates a fire risk index.

[0044] Fire propagation simulation module: Based on the fire risk index, the coordinates of the fire source point are extracted and a forest area layer containing terrain boundary and forest distribution information is loaded. Spatial overlay analysis is performed on the fire source point coordinates and the forest area layer. The wind direction guidance path is constructed and the propagation section is divided according to the wind speed and humidity distribution. The coverage area of ​​the path at each time period is calculated and the fire spread path prediction results are generated.

[0045] Dynamic risk prediction module: Based on the fire spread path prediction results, a graph neural network is used to construct a path structure map and embed meteorological features. High-risk sections are marked by combining the temperature rise rate and wind direction to generate dynamic fire risk prediction results.

[0046] Fire warning triggering module: Based on the fire dynamic risk prediction results and the fire spread path prediction results, calculate the incremental number of high-risk path points within multiple preset time windows and obtain the rate of change per unit time, determine whether the current rate of change is greater than the past average rate, generate and transmit fire warning information.

[0047] Please see Figure 2 The fire data analysis module includes:

[0048] Infrared pixel extraction submodule: It acquires remote sensing images, meteorological data and forest environment data through sensors, extracts infrared band pixel values ​​and performs channel separation, calculates the mean and standard deviation of infrared pixel values ​​in the region and filters out abnormal pixels, extracts regions with abrupt changes in infrared values ​​in the filtered results and divides pixel block boundaries, processes wind speed and direction and corrects coordinate position errors, compares environmental humidity field and terrain slope data with the same spatial resolution based on the corrected coordinates, filters spatial areas that meet the requirements of low humidity and large slope, and generates hot spot pixel coordinate set;

[0049] The temporal temperature difference overlay submodule extracts remote sensing images from multiple time periods based on the hot spot pixel coordinate set and compares the infrared band pixel values. It calculates the temperature rise difference using the pixel values ​​of adjacent time periods in the sequence, generates a numerical sequence of temperature changes, identifies abrupt temperature rise frames, marks areas with continuous temperature rise trends, compares the wind direction sequence in meteorological data, adjusts the coordinate positions, compares the morphology of the hot spot areas after adjustment, constructs a continuous trajectory, and generates a dynamic sequence map of temperature rise areas.

[0050] Risk level determination submodule: Based on the dynamic sequence map of temperature rise area, extract the temperature, humidity and wind data of the corresponding area and discretize the indicators, screen the intersection of low humidity area and high temperature rise area, determine the coverage of wind direction expansion range, identify overlapping areas and count the number of risk factors, match the risk classification zoning rules and generate multi-level classification results, delineate the boundaries of suspected fire source blocks and process layer labels, and obtain the initial fire risk analysis results;

[0051] Infrared pixel extraction submodule: Based on remote sensing images acquired by sensors, the module uses the `cv2.split()` function from the OpenCV library to perform channel separation of image data, extracting the corresponding infrared band channels. It then uses NumPy array indexing operations to extract the region pixel matrix, calling `np.mean()` and `np.std()` functions to calculate the mean and standard deviation of the infrared channels within each region. Using anomaly pixel identification methods, it performs Z-score standardization on the pixel values ​​within the region, setting a threshold of Z>2.5 to filter out abnormal infrared pixels. Based on the set of abnormal pixel coordinates, it uses the OpenCV function `cv2.findContours()` to delineate pixel boundary blocks. The bounding box constraint rectangle function cv2.boundingRect() determines the coordinates of the boundary range. The GDAL library is used to process the wind speed and direction vectors in the meteorological data to calculate the spatial offset relationship between the wind direction angle vector and the center of the pixel block. The correction function shapely.affinity.translate() is used to offset and correct the coordinates of abnormal pixels according to the wind direction and distance. The correction error range is limited to no more than 2 pixel units. The slope layer and the environmental humidity raster layer in the DEM data are overlaid. The Rasterio library is used for raster overlay. Spatial hot spots are screened by setting a humidity threshold of less than 0.25 and a slope threshold of greater than 30 degrees, and a set of hot spot pixel coordinates is generated.

[0052] The temporal temperature difference overlay submodule: Based on a hotspot pixel coordinate set, it loads multi-time-segment infrared remote sensing images using a remote sensing image indexing method. The starting indexes on the horizontal and vertical axes are used as hotspot coordinate values, with a window size of 1×1 pixels. Grayscale values ​​are read frame by frame and organized into a two-dimensional matrix. Each column represents a hotspot location, and each row contains the grayscale values ​​of the same hotspot in different time frames, generating a pixel value time series matrix. A temporal difference calculation method is used to calculate the difference along the time dimension for each pixel sequence, specifying a calculation span of one frame and setting a temperature difference threshold of 3.5. Abrupt points exceeding the threshold are selected to establish an abrupt change time index table, generating a temperature rise abrupt change time index set. After converting temperature values ​​into a binary image, connected region boundaries are extracted, with an area filtering threshold of 15 pixels. The bounding box shapes are compared according to the time series, and similarity is assessed. The region below 0.3 is defined as a non-matching region. Hotspot regions with continuous occurrence characteristics are extracted and their location information is marked to generate a continuously warming region annotation map. Based on the wind direction sequence, a coordinate correction mapping method is used to extract the wind direction angle and convert it into coordinate offset instructions according to the time sequence. The maximum horizontal and vertical translation range is set to 3 pixels. The coordinates of all hotspot regions are updated to form a set of hotspot coordinates after wind direction correction. The coordinate points are arranged in time order using a trajectory construction algorithm. The time window is set to 5 frames and the path smoothing parameter is the average value within 3 frames. A hotspot region movement trajectory map is generated. Finally, an image synthesis method is used to superimpose each processed frame with a transparency weight of 0.5, maintain the original image resolution, and set the video frame rate to 2 frames per second. The result is exported as an image sequence video to generate a dynamic sequence map of the temperature rise region.

[0053] Risk Level Determination Submodule: Based on the dynamic sequence map of temperature rise areas, meteorological data corresponding to the hot spots in the map is extracted. SciPy's `scipy.stats.binned_statistic()` function is used to bin the data for temperature, humidity, and wind speed, with a bin number set to 10. Discrete mapping is performed on each data interval, and a classification index table is constructed. The spatial intersection of areas with humidity values ​​less than 0.3 and areas with temperature increases exceeding 2.5 degrees Celsius is selected. The `intersection()` function of Shapely is called to extract the intersecting polygon regions, and the coverage of the boundaries with the main wind axis path is determined. The wind direction coverage area is determined by the criteria that the angle between the path angle and the regional center vector is less than 20 degrees and the overlap rate is greater than 40%. The number of high temperature, low humidity and high wind speed factors in each intersection area is counted. The number of risk factors is counted and compared with the preset risk level classification rules. The classification thresholds are set as low, medium and high levels respectively, with the number of risk factors being 1, 2 and 3. The classification results are matched and GeoPandas is used to assign layer attributes to the boundary of suspected high-risk areas. The risk_level field is set to 1, 2 and 3 respectively. The standard GeoTIFF layer is exported using gdf.to_file() to generate the initial fire risk layer.

[0054] Please see Figure 2 The fire risk assessment module includes:

[0055] Spatial overlap comparison submodule: Based on the initial fire risk analysis results, extract the location coordinates of the fire source points from the initial fire risk analysis results and process the zoning numbers, batch standardize the location coordinates of the fire source points and uniformly perform projection transformation, record the alignment structure fields after importing coordinate data in historical fires, calculate the valley value of the distance between the two coordinate sets and filter the overlapping points, extract the outline of the overlapping area, calculate the area overlap ratio, and obtain the historical spatial overlap distribution set.

[0056] Triggering Factor Screening Module: Based on the historical spatial overlapping distribution set, it retrieves real-time meteorological data within the coverage area of ​​each fire source point obtained by sensors and extracts temperature, humidity and wind force as meteorological factors. It loads meteorological data within the corresponding time period of historical fires and divides the parameter range of each meteorological data. It compares the current meteorological data with the parameter range corresponding to historical fires and marks the matching. It accumulates the meteorological factors that meet the conditions for each fire source point and calculates the factor coverage ratio. It compares the factor coverage ratio with the preset threshold and completes the determination of the triggering condition. It then obtains the triggering status table of the triggering factor.

[0057] Risk index calculation submodule: Based on the triggering state table of ignition factors, the triggering state code of fire source points is converted and an index matrix is ​​constructed. Each index in the index matrix is ​​weighted to generate a standardized score. Combining the overlapping area ratio of each fire source point in the historical spatial overlapping distribution set, the fusion value is calculated according to the proportional weight rule. The risk level corresponding to the fusion value is classified and the risk interval of spatial points is marked to generate the fire risk index.

[0058] Spatial Overlap Comparison Submodule: Based on the initial fire risk layer, the GeoPandas library is used to load the layer data and the `gdf.geometry.centroid` function is used to extract the center coordinates of the fire source. `gdf["zone_id"] = gdf.sindex.query_bulk()` is called to generate the zone number index. Coordinate normalization is employed, using the `pyproj.Transformer.from_crs()` function to perform a coordinate system-unified projection transformation, converting the original coordinates from EPSG: 4326 to EPSG: 3857. The historical fire record coordinate set is imported into the process, and the `merge()` function of Pandas is used to align the field structure of the historical records. The Euclidean distance matrix between the two coordinate sets is calculated, and the `scipy.spatial.distance.cdist()` function is called, setting a distance threshold of 200 meters for filtering. The minimum value position is identified, and the corresponding coincident point index is extracted. The outline region composed of the coincident points is extracted using the MultiPoint() and convex_hull methods of Shapely. The area intersection is calculated, and the intersection() function is called to overlay the current outline and the historical fire polygon. The area attribute is used to extract the intersection area and the union area respectively. The Python expression overlap_ratio = intersect_area / union_area is used to calculate the area overlap ratio, where intersect_area represents the intersection area between the currently extracted coincident point outline region and the historical fire polygon, union_area represents the area of ​​the union geometric region composed of the current outline region and the historical fire polygon, and overlap_ratio represents the ratio of the intersection area to the union area. This generates a historical spatial coincident distribution set.

[0059] Inducing factor screening sub-module: Based on the historical spatial coincidence distribution set, use the PostGIS spatial query statement ST_Intersects() to retrieve the area covered by the fire source points, use the SQL function ST_Within() to determine the meteorological data grid blocks into which each fire source point falls, extract the temperature, humidity, and wind force values in the meteorological factors, and assign them as temp_now, hum_now, and wind_now respectively. Load the meteorological records corresponding to the historical fire data for the corresponding time period, segment and aggregate the historical data at an hourly frequency. Use the np.percentile() function of Numpy to divide each index into 5 interval segments, namely 0-20, 20-40, 40-60, 60-80, and 80-100 percentiles. Compare the current temp_now, hum_now, and wind_now with each historical parameter interval, and call the Python boolean comparison operator (x >= lower_bound) & (x < upper_bound) for matching and marking. Count the number of successfully matched indicators. Set the maximum number of matching items for each fire source point to 3. Use the formula ratio = matched_items / 3 to calculate the factor coverage ratio. Call the np.where(ratio >= threshold, 1, 0) function of NumPy to compare the ratio with the preset threshold. The preset threshold is set to 0.67. Determine whether the trigger condition is established. Summarize the judgment result field and write it into a DataFrame with the field name trigger_state. The field value of 1 indicates triggering, and 0 indicates non-triggering. Generate the triggering state table of the excitation factor;

[0060] The risk index calculation submodule, based on the trigger state table of the ignition factor, calls the Pandas function `df["trigger_state"].apply()` to convert the binary states into codes. Trigger state 1 is converted to 100 points, and state 0 is converted to 0 points. A code column is constructed, and a two-dimensional array is built using NumPy, containing the scores of three indicators for each fire source. A standardized weighting method is used, setting the weight of temperature indicator to 0.4, humidity indicator to 0.3, and wind indicator to 0.3. The standardized score is generated using the expression `score = temp_score * 0.4 + hum_score * 0.3 + wind_score * 0.3`. The overlap ratio field from the aforementioned historical spatial overlap distribution set is called, and the fusion formula is set to `fused_value = ...`. The fusion value is calculated using `score * overlap_ratio`, where `score` represents the standardized risk score calculated based on the ignition factor, `overlap_ratio` represents the spatial overlap rate between the location of the fire source and historical fire records, and `fused_value` is the comprehensive risk index value generated after fusion. A rule-based grading method is used to set risk levels: 0-30 is low risk, 30-70 is medium risk, and above 70 is high risk. The `pd.cut()` function of Pandas is called to divide the risk level ranges, recording the spatial location and level label of each fire source. GeoPandas is used to generate a vector layer, with the `risk_index` field as the output field, exported as a Shapefile to generate the fire risk index.

[0061] Please see Figure 2 The fire propagation simulation module includes:

[0062] Spatial mapping construction submodule: Based on the fire risk index, extract the coordinates of the fire source point to generate two-dimensional coordinate points, import the layer clipping boundary raster of the forest area structure map, align the fire source point coordinates with the layers of the forest area structure map and perform layer space overlay processing, extract wind direction data from meteorological data and construct a direction vector, construct the main axis path of the direction vector and the center of the fire source point and extend continuous line segments to generate a set of fire source wind direction path line segments;

[0063] The path segment division submodule is as follows: Based on the set of fire source wind direction path segments, the cumulative path segment distance is divided into regular equal-length segments. The coordinates of the first segment of each segment are obtained and the wind speed data of the corresponding position in the meteorological data is retrieved for matching. The humidity data in the meteorological data of each segment area is projected and the corresponding value is extracted to generate a wind speed and humidity combination index. The segment attributes are assigned, the index sets of each segment of the path are classified and the level segment identification is completed to obtain a graded propagation path segment map.

[0064] Coverage calculation submodule: Based on the hierarchical propagation path segment map, the segment propagation duration is set, the radius range of the line segment endpoints is constructed, the radius range graphic is expanded and buffer zones on both sides of the path are created, the outline layers of each segment are merged to generate a time period envelope, the envelope polygons are continuously spliced, and boundary consistency processing is performed to generate fire spread path prediction results.

[0065] The spatial mapping construction submodule: Based on the fire risk index, it uses `gdf.geometry.centroid` in GeoPandas to extract the coordinates of fire source points and generate a collection of Point objects. It then uses the `rasterio.mask.mask()` method in Rasterio to import the forest area structure map raster layer and constructs Shapely polygon clipping regions according to the fire source point boundaries. The `gdf.to_crs()` function is used to convert the coordinate system of the fire source point collection to the projected coordinate system of the forest area layer. Finally, the `gpd.overlay()` function is called to align the two layers. Spatial overlay is performed. The NetCDF4 library is used to read the wind direction angle field from the wind direction data and convert it into a unit vector. When constructing the direction vector, math.cos() and math.sin() are used to convert the angle into the x and y direction increments in Cartesian coordinates. Shapely's LineString() is used to construct the path axis between the direction vector and the coordinates of the fire source point. The length of the line segment is set to the wind direction prediction range multiplied by the unit step size of 100 meters, and node points are added every 50 meters. The Fiona library is called to convert all path line segments into GeoJSON structure format to generate a set of fire source wind direction path line segments.

[0066] The path segmentation submodule, based on the fire source wind direction path segment set, calls the `LineString.length` property of `Shapely` to calculate the total distance of each segment. It sets the segment spacing to 200 meters and uses the `LineString.interpolate()` method to obtain the coordinates of the first point of each segment at fixed intervals. It uses `xarray` to read the wind speed raster layer data and obtains the wind speed value corresponding to the starting point of each segment using `sel(x=lon, y=lat, method="nearest")`, where `x=lon` represents selecting the target position on the longitude axis of the wind speed dataset, `y=lat` represents selecting the target position on the latitude axis of the wind speed dataset, and `method="nearest"` automatically finds a position in the dataset that corresponds to the input coordinates. The system retrieves the wind speed value at the nearest grid cell to the punctuation mark. It then uses Rasterio's `sample()` function to extract the humidity from the grid cell and record the humidity value for each segment. A combined index of wind speed and humidity is constructed, with an upper limit of 15 m / s for wind speed and a lower limit of 0.3 for humidity as the classification criteria. NumPy's `np.where()` method is used to label and assign values ​​to each path segment, with the segment label field named `speed_hum_level`. Values ​​of 0 represent low risk, 1 represent medium risk, and 2 represent high risk. Matplotlib is used to construct the path visualization, and GeoPandas is called to group the segmented lines by label to generate layers. The output field is named `zone_level`, generating a hierarchical propagation path segment map.

[0067] The coverage calculation submodule, based on a hierarchical propagation path segment map, sets the default propagation time for each segment to 15 minutes and the average fire advance speed to 100 meters per minute. It uses the expression `radius = 100 * 15 = 1500` to construct the buffer radius of the line segment endpoints, where 100 * 15 represents the maximum linear advance distance of the fire within 15 minutes, and `radius = 1500` sets the radius parameter of the circular buffer. It calls Shapely's `Point.buffer()` method to expand the circular range centered on each path point, uses the Fiona library to convert it to polygonal data and records it in a `GeoDataFrame`, and creates buffers on both sides of the path segment using `GeoPand`. The `buffer(distance=radius, cap_style=2)` command of `as` constructs a semi-circular extended region. `gpd.overlay()` is called to merge the path buffer polygon layers. The `dissolve(by="path_id")` method of `GeoDataFrame` is executed to generate a continuous time splicing envelope. The connection order field is set to the path segment timestamp and sorted by time using the `sort_values()` method. `unary_union()` of `Shapely` is used to merge all time period region polygons. `buffer(0)` is called to handle boundary geometric distortion and topological consistency issues, and the fire spread path prediction results are generated.

[0068] Please see Figure 2 The dynamic risk prediction module includes:

[0069] Temperature Change Extraction Submodule: Based on the fire spread path prediction results, a graph neural network is used to extract the endpoint coordinates of the path segments and perform grid number mapping. Historical meteorological databases and multi-time period temperature data corresponding to the grid number positions are retrieved. Time index alignment is obtained according to the predicted time label of the path segments. The time series of temperature data for each path segment is sorted. Continuous data interpolation is processed. The temperature rise rate in the interpolation sequence is extracted. Continuous inflection points are marked. The location of the temperature rise segment is associated and matched with the path coordinates and the segment is marked. A path temperature trend map is generated.

[0070] Factor difference calculation submodule: Based on the path temperature trend map, it synchronously extracts humidity and temperature data corresponding to the path segments, matches the corresponding time periods, compares the temperature change sequence and humidity observation sequence of each path segment within the synchronous time window, extracts the overlapping interval of the value range, calculates the difference at the same time point to form a point-by-point difference sequence; it uses a sliding window filtering algorithm to smooth the difference sequence and fill the difference matrix, extracts the edges of high difference blocks in the matrix, screens continuous height difference segments, matches and merges the height difference segment index and path coordinates, and then binds the spatial location to obtain the climate factor change set;

[0071] High-risk section identification submodule: Based on the set of climate factor changes, calculate the rate of temperature change within the path segment and perform numerical filtering, retrieve the wind direction time series that matches the path coordinate segment, analyze the magnitude of the dominant wind direction change in continuous time slices, filter continuous segments with strong directional stability, cross-over the warming rate segment and the wind direction continuous segment and align the position index, convert the cross segment classification code, mark the nested path layer, and generate dynamic fire risk prediction results;

[0072] Temperature Change Extraction Submodule: Based on fire spread path prediction results, a graph neural network node feature model is used. A graph structure model is constructed using the PyTorchGeometric library. The node feature tensor and edge relationships are defined using `torch_geometric.data.Data(x=node_features, edge_index=edge_index)`. Node features are the endpoint coordinates of path segments, and edges represent the connectivity between adjacent segments. The `GNNConv()` function is called to generate the aggregated position vector for each node. The k-means clustering algorithm is used to assign grid numbers to the endpoint positions, with a cluster size of 100 and each point assigned a unique grid number. The Pandas `merge()` function is called to merge the grid numbers with historical data. Real-time temperature data is correlated by timestamp, and the temperature time series of each path segment is sorted. The temperature data at discontinuous time points is linearly interpolated using scipy.interpolate.interp1d(), with the interpolation method set to kind='linear' and all missing points filled. The temperature change amplitude in each interpolated sequence is extracted, and the first-order difference is calculated using the np.gradient() function to identify rising segments with a slope greater than 0.2. The local inflection points of continuous warming segments are marked using scipy.signal.find_peaks(), with a minimum height of 1.0 and a distance of 3 as identification parameters. The inflection point positions are matched with the path coordinate set, and the path attribute fields of the warming segment are updated through coordinate indexing to generate a path temperature trend map.

[0073] Factor Difference Calculation Submodule: Based on the path temperature trend map, a data synchronization method is used to set the timestamps corresponding to the path segments as the main index. The synchronization window width is set to 5 minutes, and the temperature and humidity data are aligned and structured using a unified time step. The temperature and humidity values ​​at each time point are extracted to calculate the difference and construct a difference sequence. A sliding window filtering algorithm is used to smooth the difference sequence. The window length is set to 7 and the step size is 1. The center alignment method is used, and the mean value within the window is used to replace the original value to fill the structured matrix. A difference matrix is ​​constructed with the path number as the row and the time node as the column. Null values ​​are filled with the constant -999. A difference block extraction algorithm is used to traverse the matrix. The difference threshold is set to 2.0 and continuous elevation difference areas are identified using the eight-neighbor rule. The minimum size of continuous segments is set to 5 units. The row and column numbers of the effective elevation difference segments are extracted to construct an index set. The index set is bound to the path segment number using the path index matching method. The corresponding coordinate fields x and y are extracted to form a time-series coordinate point set, which is then sorted and merged to generate a climate factor change set.

[0074] The high-risk section identification submodule: Based on the climate factor change set, the time series temperature values ​​within the path segment are subjected to first-order difference using the np.diff() function to calculate the rate of temperature change at consecutive time points. A threshold of 0.25 is used for filtering, and segments with values ​​greater than the threshold are selected as rapidly warming areas. Wind direction data is extracted from the time axis of the path segment, and a continuous segment detection function is used to identify segments with the same continuous wind direction. The groupby() function in Pandas is used in conjunction with diff().ne(0).cumsum() to determine the continuity of wind direction over time. The merge() function is used to connect the warming rate segments and the continuous wind direction segments according to the path segment number, perform cross-merging operations, and generate an overlapping segment index. The np.select() function in NumPy is used to convert the overlapping segments into classification codes, and the coding rules are set as follows: 2 for both warming and wind direction are satisfied, 1 for only one is satisfied, and 0 for neither is satisfied. The overlay() function in GeoPandas is used to nest the coded value field into the path segment layer, update the risk_code attribute field, and export it as a Shapefile to generate dynamic fire risk prediction results.

[0075] Graph neural networks follow the formula:

[0076]

[0077] Where: p v Let p be the two-dimensional coordinate vector of node v, where p is the coordinate vector itself, the subscript v indicates the v-th node in the graph, M is a two-dimensional real matrix with a size of 2 times d, and · is the matrix-vector multiplication operator. Let be the embedding vector of node v calculated by the Kth layer of the graph neural network, h be the node feature representation vector, the superscript (K) indicate the output of the Kth layer of the graph neural network, η be the temperature-sensitive offset correction coefficient, and E be the embedding vector of node v. v Let v be the gradient vector of temperature change. This is the average vector of temperature change gradients for all nodes in the entire graph. (Parentheses) This is the difference vector between the current node and the average temperature gradient of the entire map;

[0078] Execution process: First, a node graph structure is constructed based on historical forest fire data and geographic raster data. Each node is initialized as an embedding vector containing grid number, temperature change rate, time label, and node geographic type code. Then, the vector is input into a graph neural network model for multi-layer propagation. After passing through the Kth layer, the embedding vector of each node is obtained. The mapping matrix M maps the embedding vector to a two-dimensional geographic coordinate space, resulting in the initial coordinate vector. Based on the temperature change gradient E at node v v Average temperature gradient across the entire map Calculate the correction term for the difference between them. To correct spatial coordinate offsets caused by climate anomalies, the correction term is added to the coordinate mapping value to obtain the coordinate vector p of the endpoint node. v .

[0079] Please see Figure 2 The fire early warning triggering module includes:

[0080] Regional overlap judgment submodule: Based on the fire dynamic risk prediction results and the fire spread path prediction results, extract the boundary vector data of the two layers and perform coordinate system one processing, perform vector operation on the overlapping area to extract the cross area number, filter the cross area area, merge the extracted cross area number according to spatial adjacency, generate an aggregated area set, count the number of aggregated areas and the total area, determine whether it exceeds the set high-risk spatial coverage threshold and complete the matching identification, and generate an overlapping area judgment value set;

[0081] The path fluctuation detection submodule extracts the path segment coordinates from the fire dynamic risk prediction results based on the overlapping area judgment value set and sorts them by time dimension. It counts high-risk points in each time period to generate a change curve sequence, calculates the numerical increase in continuous time intervals to construct a difference sequence, matches the difference sequence with the historical average value, extracts abnormal segments, and obtains a path point fluctuation result table.

[0082] The early warning information generation submodule: Based on the path location fluctuation result table, high-risk segment identifiers are screened out, high-frequency trigger condition labels are extracted using fuzzy logic reasoning, the identified segment codes are converted, and then the segments are mapped with the original path map. The segment mapping results are matched with the spatial location field, a graphic structure template is established, the fields in the graphic structure template are assigned values, and the text output rendering is completed to generate fire early warning information.

[0083] The regional overlap judgment submodule, based on the fire dynamic risk prediction results and fire spread path prediction results, uses a coordinate unified projection method to adjust the spatial reference of the boundary vector data of the two layers, sets the target coordinate system as EPSG:3857 and completes the coordinate transformation of all layers, uses a spatial overlay analysis method to extract the area numbers of the layers with spatial intersection in the intersection mode and constructs a set of intersection area numbers, uses a spatial attribute filtering algorithm to extract area numbers that meet the conditions with a lower limit of 500 square meters, uses a spatial adjacency merging algorithm to merge adjacent numbers according to the boundary contact relationship with a maximum merging distance of 10 meters to generate an aggregated area set, uses an interval matching judgment algorithm to set the number of reference intervals to 10 and set the tolerance to ±2, matches and compares the number of aggregated areas with the interval values ​​and assigns an identifier value to generate an overlapping area judgment value set;

[0084] The path fluctuation detection submodule, based on the overlapping area judgment value set, uses GeoPandas to read the dynamic risk prediction map layer and calls gpd.sjoin() to extract the path segment coordinate set by spatial range. Based on the time field in the segment attributes, it calls Pandas' sort_values(by="timestamp") function to complete the time dimension sorting. It then counts the high-risk points within each time period, using groupby("timestamp")["risk_flag"].sum() to generate a sequence of high-risk point counts.

[0085] Numpy's `np.diff()` function calculates the numerical increase between adjacent time points, extracts the set of difference points and records the difference field as `delta_count`, sets the threshold for judging abnormal fluctuations to twice the historical average difference, calls Pandas' `rolling(window=5).mean()` to generate the historical moving average, compares the current difference with the moving average to form an anomaly judgment boolean sequence, uses boolean filtering to extract the paragraph index that meets the anomaly judgment, calls GeoPandas' `merge()` function to concatenate the abnormal paragraph identifier with the path paragraph spatial data field, and generates a path point fluctuation result table;

[0086] The early warning information generation submodule, based on the path point fluctuation result table, calls `df[df["risk_flag"]==1]` to filter out high-risk segment identifiers. It employs a fuzzy logic reasoning system, using the `fuzz.interp_membership()` method from scikit-fuzzy to define the input variable range and construct membership functions. The temperature input range is set to [20, 50] degrees Celsius, humidity to [0.1, 0.6], and wind speed to [0, 20] meters per second. The rule base is set to include three variable condition combinations for judgment, calling `ctrl.Rule()` to establish rule expressions. For each segment's input variables, high-frequency trigger condition labels are calculated, and the trigger label numbers are converted to `trigger_code`. The process involves using the GeoPandas spatial join function `sjoin()` to perform segment mapping matching between the segment code and the original path map layer fields. The `merge()` function is then called to link the segment code with the spatial fields in the layer. The spatial location field is named `segment_location`. A structured text template is constructed using the Jinja2 template engine. Template fields include path segment number, coordinates, trigger label, risk level, and spatial location description. The template filling function `template.render()` is executed to generate structured text content. Finally, Matplotlib and PIL libraries are used to call layer plotting and rendering functions to draw the path layer as an image and embed the structured template content, generating fire warning information.

[0087] Fuzzy logic reasoning systems follow the formula:

[0088]

[0089] Where: y is the output of the fuzzy inference system, representing the label of the high-frequency triggering condition. μ is the summation symbol. i Let w be the membership value of the i-th fuzzy rule. i Let ρ be the output weight value of the i-th fuzzy rule. i Let exp(-λ·σ) be the spatial consistency weight of the i-th fuzzy rule. i ) is an exponential function, λ is the volatility suppression exponential weighting coefficient, and σ is the exponential function. i Let represent the variance of the temperature change sequence of the path segment corresponding to the i-th rule, and · represent the multiplication sign. is the normalization factor for the weights of fuzzy rules, and n is the total number of fuzzy rules;

[0090] Execution process: First, extract the temperature change amplitude, the number of temperature inflection points, and the inflection point interval as input features, and calculate the membership value μ for each current paragraph. i Then, the output weight value w is set according to the corresponding trigger frequency in the historical fire records. iThe spatial consistency weight ρ of the query rule in the current segment region i The variance of the temperature change sequence of the current path segment is calculated as σ. i Through the exponential function exp(-λ·σ i The values ​​of all activated rules are then added together in product form and normalized by a normalization factor. Constraint normalization is performed to calculate the fuzzy inference output value y. When the value of y is greater than the set threshold, the segment is judged as a high-risk segment, and the trigger condition label data is extracted.

[0091] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A forest fire monitoring and early warning system based on big data, characterized in that, The system includes: Fire data analysis module: It acquires meteorological data, remote sensing images and forest environment data through sensors, extracts infrared hotspot pixels from remote sensing images, and combines temperature change differences in meteorological data to overlay and calculate data from multiple time periods in the region to generate initial fire risk analysis results. Fire risk assessment module: Based on the initial fire risk analysis results, combined with historical fire data and real-time meteorological data acquired by sensors, it determines whether the fire initiation conditions are met and generates a fire risk index. Fire propagation simulation module: Based on the fire risk index, extract the coordinates of the fire source and load a forest area layer containing terrain boundary and forest distribution information. Perform spatial overlay analysis on the fire source coordinates and the forest area layer, construct a wind direction guidance path and divide the propagation section according to the wind speed and humidity distribution, calculate the coverage area of ​​the path at each time period, and generate fire spread path prediction results. Dynamic risk prediction module: Based on the fire spread path prediction results, a path structure map is constructed using a graph neural network and meteorological features are embedded. High-risk sections are continuously marked by combining the temperature rise rate and wind direction to generate dynamic fire risk prediction results. Fire warning triggering module: Based on the fire dynamic risk prediction results and the fire spread path prediction results, calculate the incremental number of high-risk path points within multiple preset time windows and obtain the rate of change per unit time, determine whether the current rate of change is greater than the past average rate, and generate and transmit fire warning information.

2. The forest fire monitoring and early warning system based on big data according to claim 1, characterized in that, The fire data analysis module includes: Infrared pixel extraction submodule: It acquires remote sensing images, meteorological data and forest environment data through sensors, extracts infrared band pixel values ​​and performs channel separation, calculates the mean and standard deviation of infrared pixel values ​​in the region and filters out abnormal pixels, extracts regions with abrupt changes in infrared values ​​in the filtered results and divides pixel block boundaries, processes wind speed and direction and corrects coordinate position errors, compares environmental humidity field and terrain slope data with the same spatial resolution based on the corrected coordinates, filters spatial areas that meet the requirements of low humidity and large slope, and generates hot spot pixel coordinate set; The time-series temperature difference overlay submodule: Based on the hot spot pixel coordinate set, it extracts remote sensing images of multiple time periods by coordinate index and compares the pixel values ​​of the infrared band. It calculates the temperature rise difference using the pixel values ​​of adjacent time periods in the sequence, generates a temperature change numerical sequence, identifies temperature rise abrupt frames, marks areas with continuous temperature rise trends, compares with the wind direction sequence in meteorological data, adjusts the coordinate position, compares and adjusts the shape of hot spot areas and constructs a continuous trajectory, and generates a dynamic sequence map of temperature rise areas. Risk level determination submodule: Based on the dynamic sequence map of the temperature rise area, extract the temperature, humidity and wind data of the corresponding area and discretize the indicators, screen the intersection of the low humidity area and the high temperature rise area, determine the coverage of the wind direction expansion range, identify overlapping areas and count the number of risk factors, match the risk classification zoning rules and generate multi-level classification results, delineate the boundaries of suspected fire source blocks and process the layer labels, and obtain the initial fire risk analysis results.

3. The forest fire monitoring and early warning system based on big data according to claim 1, characterized in that, The fire risk assessment module includes: Spatial overlap comparison submodule: Based on the initial fire risk analysis results, extract the location coordinates of the fire source points from the initial fire risk analysis results and process the partition numbers, batch standardize the location coordinates of the fire source points and uniformly perform projection transformation, record the alignment structure fields after importing coordinate data in historical fires, calculate the valley value of the distance between the two coordinate sets and filter the overlapping points, extract the outline of the overlapping area, calculate the area overlap ratio, and obtain the historical spatial overlap distribution set. Triggering Factor Screening Module: Based on the historical spatial overlapping distribution set, it retrieves real-time meteorological data within the coverage area of ​​each fire source point obtained by the sensor and extracts temperature, humidity and wind force as meteorological factors. It loads meteorological data within the corresponding time period of historical fires and divides the parameter range of each meteorological data. It compares the current meteorological data with the parameter range corresponding to historical fires and marks the matching. It accumulates the meteorological factors that meet the conditions of each fire source point and calculates the factor coverage ratio. It compares the factor coverage ratio with the preset threshold and completes the determination of the triggering condition. It then obtains the triggering status table of the triggering factor. Risk index calculation submodule: Based on the triggering state table of the ignition factor, the triggering state code of the fire source point is converted and an index matrix is ​​constructed. Each index in the index matrix is ​​weighted to generate a standardized score. Combining the overlapping area ratio of each fire source point in the historical spatial overlapping distribution set, the fusion value is calculated according to the proportional weight rule. The risk level corresponding to the fusion value is classified and the risk interval of the spatial point is marked to generate the fire risk index.

4. The forest fire monitoring and early warning system based on big data according to claim 1, characterized in that, The fire propagation simulation module includes: Spatial mapping construction submodule: Based on the fire risk index, extract the coordinates of the fire source point to generate two-dimensional coordinate points, import the layer clipping boundary grid of the forest area structure map, align the fire source point coordinates with the layers of the forest area structure map and perform layer space overlay processing, extract wind direction data from meteorological data and construct a direction vector, construct the main axis path of the direction vector and the center of the fire source point and extend continuous line segments to generate a set of fire source wind direction path line segments; Path segment division submodule: Based on the set of fire source wind direction path segments, the cumulative path segment distance is divided into regular equal-length segments. The coordinates of the first segment of each segment are obtained and the wind speed data of the corresponding position in the meteorological data is retrieved for matching. The humidity data in the meteorological data of each segment area is projected and the corresponding value is extracted to generate a wind speed and humidity combination index. The segment attributes are assigned, the index sets of each segment of the path are classified and the level segment identification is completed to obtain a graded propagation path segment map. Coverage calculation submodule: Based on the hierarchical propagation path segment map, the segment propagation duration is set, the radius range of the line segment endpoints is constructed, the radius range graphic is expanded and buffer zones on both sides of the path are created, the outline layers of each segment are merged to generate a time period envelope, the envelope polygons are continuously spliced, and boundary consistency processing is performed to generate the fire spread path prediction result.

5. The forest fire monitoring and early warning system based on big data according to claim 1, characterized in that, The dynamic risk prediction module includes: Temperature Change Extraction Submodule: Based on the fire spread path prediction results, a graph neural network is used to extract the endpoint coordinates of the path segments and perform grid number mapping. Historical meteorological databases and multi-time period temperature data corresponding to the grid number positions are retrieved. Time index alignment is obtained according to the predicted time label of the path segments. The time series of temperature data for each path segment is sorted. Continuous data interpolation is processed. The temperature rise rate in the interpolation sequence is extracted. Continuous inflection points are marked. The location of the temperature rise segment is associated and matched with the path coordinates and the segment is marked. A path temperature trend map is generated. Factor difference calculation submodule: Based on the path temperature trend map, it synchronously extracts humidity and temperature data corresponding to the path segments, matches the corresponding time periods, compares the temperature change sequence and humidity observation sequence of each path segment within the synchronous time window, extracts the value range overlap interval, calculates the difference at the same time point to form a point-by-point difference sequence; uses a sliding window filtering algorithm to smooth the difference sequence and fill the difference matrix, extracts the edges of high difference blocks in the matrix, screens continuous height difference segments, matches and merges the height difference segment index and path coordinates, and then binds the spatial location to obtain the climate factor change set; High-risk section identification submodule: Based on the set of climate factor changes, calculate the rate of temperature change within the path segment and perform numerical filtering, retrieve the wind direction time series that matches the path coordinate segment, analyze the magnitude of the dominant wind direction change in continuous time slices, filter continuous segments with strong directional stability, cross-over the warming rate segment and the wind direction continuous segment and align the position index, convert the cross-segment classification code, mark the nested path layers, and generate dynamic fire risk prediction results.

6. The forest fire monitoring and early warning system based on big data according to claim 1, characterized in that, The fire early warning triggering module includes: Regional overlap judgment submodule: Based on the fire dynamic risk prediction results and the fire spread path prediction results, extract the boundary vector data of the two layers and perform coordinate system one processing, perform vector operation on the overlapping area to extract the cross area number, filter the cross area area, merge the extracted cross area number according to spatial adjacency, generate an aggregated area set, count the number of aggregated areas and the total area, determine whether it exceeds the set high-risk spatial coverage threshold and complete the matching identification, and generate an overlapping area judgment value set; The path fluctuation detection submodule extracts the path segment coordinates from the fire dynamic risk prediction results based on the overlapping area judgment value set and sorts them by time dimension. It counts high-risk points in each time period to generate a change curve sequence, calculates the numerical increase in continuous time intervals to construct a difference sequence, matches the difference sequence with the historical average value, extracts abnormal segments, and obtains a path point fluctuation result table. The early warning information generation submodule: Based on the path point fluctuation result table, it filters out high-risk segment identifiers, extracts high-frequency trigger condition tags using fuzzy logic reasoning, converts the identified segment codes, maps the segments with the original path map, matches the segment mapping results with the spatial location field, establishes a graphic structure template, assigns values ​​to the fields in the graphic structure template and completes text output rendering, and generates fire early warning information.

Images