A forest fire early warning method, system and device based on satellite Internet of Things

Abstract

The present invention relates to the field of satellite Internet of Things technology, and in particular to a forest fire early warning method, system and equipment based on satellite Internet of Things, wherein the method specifically comprises: fusing the monitoring data of different monitoring devices in each target overlapping area according to a data fusion algorithm to generate a data fusion model; obtaining the monitoring data of the potential monitoring blind area by analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind area, and predicting the long-term change trend of the monitoring data of the potential monitoring blind area by a time series prediction algorithm; transmitting the monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area to a data center via the satellite Internet of Things. The present invention significantly improves the coverage and data accuracy of the forest monitoring system, and effectively reduces the problems caused by monitoring blind areas and data overlap.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite Internet of Things, and in particular to a forest fire early warning method, system and equipment based on satellite Internet of Things. Background Art

[0002] In the complex terrain environment of dense forests, due to factors such as uneven terrain and lush trees, the monitoring areas of different monitoring devices will inevitably have certain overlaps and blind spots, which will lead to redundancy in data collection and reduced recognition efficiency. Although each monitoring device has been calculated and processed by front-end AI and edge AI, there is still a problem that the overlapping areas and blind area data cannot be effectively identified and filtered when the data is transmitted to the center. For example, how to accurately determine the monitoring area of ​​the monitoring equipment, how to effectively identify and eliminate data overlap, how to accurately analyze the monitoring blind spots, how to judge and determine the overlapping areas between each monitoring device, and trace and analyze the data in these areas to find out the reasons for the overlap, etc. Summary of the invention

[0003] The purpose of the present invention is to provide a forest fire early warning method, system and equipment based on satellite Internet of Things, which significantly improves the coverage and data accuracy of the forest monitoring system, effectively reduces the problems caused by monitoring blind spots and data overlap, and solves at least one of the above-mentioned prior art problems.

[0004] In a first aspect, the present invention provides a forest fire early warning method based on satellite Internet of Things, the method specifically comprising:

[0005] A three-dimensional space model is established based on the position data of each monitoring device and the forest terrain data, and the monitoring area of ​​each monitoring device is simulated in the three-dimensional space model through the terrain map and the satellite image, and potential monitoring blind areas and potential overlapping areas are determined in the monitoring area;

[0006] Based on the three-dimensional space model, determining a plurality of first overlapping areas from the plurality of potential overlapping areas according to a spatial geometry algorithm, and determining a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data;

[0007] Acquire the monitoring data collected by each monitoring device in the target overlapping area, compare and analyze the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local anomaly factor algorithm, determine the consistency and difference of the monitoring data of each monitoring device, and determine the data interference factor through the consistency and the difference;

[0008] The monitoring data of different monitoring devices in each target overlapping area are fused according to a data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effect caused by the monitoring data of different monitoring devices in the target overlapping area;

[0009] Using image recognition technology to analyze the monitoring image in the potential monitoring blind area, extracting terrain feature data, vegetation density data and tree height data in the monitoring image, and determining the cause of the blind area based on the terrain feature data, the vegetation density data and the tree height data;

[0010] By analyzing the monitoring data of the adjacent areas of the potential monitoring blind area, the monitoring data of the potential monitoring blind area is deduced, and the long-term change trend of the monitoring data of the potential monitoring blind area is predicted by a time series prediction algorithm;

[0011] The monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area are transmitted to the data center via the satellite Internet of Things, and an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart are generated based on the data analysis.

[0012] In a second aspect, the present invention provides a forest fire early warning system based on satellite Internet of Things, the system specifically comprising:

[0013] The first early warning module is used to establish a three-dimensional space model based on the position data of each monitoring device and the forest terrain data, simulate the monitoring area of ​​each monitoring device in the three-dimensional space model through the terrain map and satellite image, and determine the potential monitoring blind area and potential overlapping area in the monitoring area;

[0014] A second early warning module is used to determine a plurality of first overlapping areas from the plurality of potential overlapping areas based on the three-dimensional space model according to a spatial geometry algorithm, and to determine a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data;

[0015] The third early warning module is used to obtain the monitoring data collected by each monitoring device in the target overlapping area, compare and analyze the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local abnormal factor algorithm, determine the consistency and difference of the monitoring data of each monitoring device, and determine the data interference factor through the consistency and difference;

[0016] A fourth early warning module is used to fuse the monitoring data of different monitoring devices in each target overlapping area according to a data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effect caused by the monitoring data of different monitoring devices in the target overlapping area;

[0017] a fifth warning module, configured to analyze the monitoring image in the potential monitoring blind area by using image recognition technology, extract terrain feature data, vegetation density data and tree height data in the monitoring image, and determine the cause of the blind area according to the terrain feature data, the vegetation density data and the tree height data;

[0018] A sixth early warning module is used to deduce the monitoring data of the potential monitoring blind area by analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind area, and predict the long-term change trend of the monitoring data of the potential monitoring blind area by a time series prediction algorithm;

[0019] The seventh early warning module is used to transmit the monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area to the data center via the satellite Internet of Things, and generate an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart based on data analysis.

[0020] In a third aspect, the present invention provides a computer device comprising: a memory and a processor and a computer program stored in the memory. When the computer program is executed on the processor, it implements a forest fire warning method based on satellite Internet of Things as described in any one of the above methods.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the forest fire warning method based on satellite Internet of Things as described in any one of the above methods is implemented.

[0022] Compared with the prior art, the present invention has at least one of the following technical effects:

[0023] 1. The present invention effectively improves the coverage efficiency and data accuracy of the forest monitoring system, reduces the problems of monitoring blind spots and data overlap, and enhances the real-time monitoring and fire warning capabilities of the forest environment.

[0024] 2. The present invention combines topographic maps, satellite images and monitoring equipment data to simulate and analyze the monitoring area of ​​the monitoring equipment in detail, effectively marking potential monitoring blind spots and overlapping areas. By using spatial geometry algorithms, the present invention can accurately identify and process the overlapping areas between monitoring devices, and combine the actual data of terrain undulations and vegetation occlusion to further optimize monitoring coverage and data accuracy.

[0025] 3. The present invention uses image recognition technology to extract terrain features, vegetation density and tree height data within the monitoring blind area, which helps to deeply analyze the causes of the blind area and determine the affected data points.

[0026] 4. The present invention uses a data fusion algorithm to integrate environmental data such as temperature, humidity and wind direction collected by different devices in the same area, and generates a more accurate and reliable data fusion model by eliminating the overlapping effects in the data.

[0027] 5. This method transmits the processed monitoring data to the data center in real time via the satellite Internet of Things, which not only ensures the timeliness and integrity of the data, but also provides high-quality input for the comprehensive evaluation of the data center and the generation of environmental monitoring reports.

[0028] 6. The present invention helps to timely discover and warn of potential fire risks through real-time and accurate environmental data analysis, and also provides strong technical support for the long-term monitoring of meteorological and ecological changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 It is a flowchart of a forest fire early warning method based on satellite Internet of Things provided by one embodiment of the present invention;

[0031] Figure 2 It is a structural schematic diagram of a forest fire early warning system based on satellite Internet of Things provided by an embodiment of the present invention;

[0032] Figure 3 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0034] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0035] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0036] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0037] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0038] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0039] In the embodiment of the present application, the execution subject of the process includes a terminal device, which includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the method disclosed in the present application. Figure 1 A schematic flow chart of a forest fire early warning method based on satellite Internet of Things disclosed in an embodiment of the present invention is shown, and the details are as follows:

[0040] S101, establishing a three-dimensional space model based on the posture data of each monitoring device and the forest terrain data, simulating the monitoring area of ​​each monitoring device in the three-dimensional space model through topographic maps and satellite images, and determining potential monitoring blind spots and potential overlapping areas in the monitoring area.

[0041] In this embodiment, according to the position coordinates and monitoring angle of each monitoring device, combined with the actual terrain data of the forest, a three-dimensional space model is established, and the monitoring area of ​​each monitoring device is simulated through the topographic map and satellite image, and potential monitoring blind areas and overlapping areas are marked. Specifically, according to the position coordinates and monitoring angle of each monitoring device, combined with the actual terrain data of the forest, a three-dimensional space model of the forest is established through a three-dimensional space modeling algorithm. The topographic map and the satellite image are aligned to obtain the position and orientation of the monitoring device in the three-dimensional space model. According to the field of view and monitoring distance of the monitoring device, the monitoring area of ​​each monitoring device is simulated in the three-dimensional space model to obtain the visible range of each monitoring device. The visible range of each monitoring device is superimposed and analyzed to identify the monitoring blind area and overlapping area. For the identified monitoring blind area, the Thiessen polygon algorithm is used for spatial interpolation to estimate the terrain features in the blind area. According to the estimated terrain features, the vegetation distribution in the blind area is simulated by the Monte Carlo method. The simulated vegetation distribution is compared with the actual vegetation distribution in the monitoring overlapping area to evaluate the accuracy of the simulation results. If the simulation results are highly consistent with the actual situation, the simulated vegetation distribution in the blind area is added to the three-dimensional spatial model; otherwise, the simulation parameters are adjusted and the simulation is repeated until the accuracy requirements are met. Finally, the monitoring blind areas and overlapping areas are marked in the three-dimensional spatial model to provide a basis for optimizing the layout of monitoring equipment.

[0042] It is understandable that a Beidou positioning module needs to be installed or integrated on each monitoring device. This module can receive Beidou satellite signals and calculate the precise location (latitude and longitude) of the device. The location data calculated by the Beidou positioning module needs to be collected and transmitted to the data processing center. In the data processing center, the collected location data will be cleaned, integrated, and matched and associated with other relevant data (such as the monitoring angle of the monitoring device, terrain data, etc.). In the process of three-dimensional space modeling, the integrated location data will be used to determine the precise location of the monitoring device in the three-dimensional model.

[0043] Satellite images are acquired through remote sensing satellites. The frequency of satellite image capture can be set based on the characteristics of the satellite, mission requirements, and cost-effectiveness. For example, for monitoring areas that require rapid response, low-resolution multispectral satellites can be used, and the capture frequency can be set to 1-2 days. For areas with lower requirements for real-time monitoring, high-resolution optical satellites, medium-resolution multispectral satellites, etc. can be used, and the capture frequency can be set to within several days to several weeks.

[0044] In some embodiments, the above step S101 specifically includes:

[0045] According to the position coordinate data of each monitoring device, the precise position of each monitoring device in the forest is obtained to obtain the three-dimensional spatial coordinate data of the monitoring device;

[0046] Based on the principle of triangulation, the monitoring range of each monitoring device is determined through the monitoring angle data of each monitoring device;

[0047] Acquire forest terrain data, and construct a three-dimensional forest terrain model using the forest terrain data according to digital elevation model and geographic information system technology;

[0048] Overlaying and analyzing the three-dimensional spatial coordinate data of the monitoring device, the monitoring range, and the three-dimensional terrain model of the forest, and generating a three-dimensional spatial model through spatial interpolation and gridding processing;

[0049] Simulating the monitoring area of ​​each monitoring device in the three-dimensional space model to obtain the visual range of each monitoring device;

[0050] Superimpose and analyze the visual range of each monitoring device to determine potential monitoring blind spots and potential overlapping areas;

[0051] Using the Thiessen polygon algorithm to perform spatial interpolation on the potential monitoring blind area, estimating the terrain features within the potential monitoring blind area, and using the Monte Carlo method to simulate the vegetation distribution within the potential monitoring blind area through the terrain features within the potential monitoring blind area to obtain simulated vegetation distribution;

[0052] The simulated vegetation distribution is compared with the actual vegetation distribution. If the degree of fit between the two is greater than a preset fit threshold, the simulated vegetation distribution is added to the three-dimensional space model. Otherwise, the simulation parameters are readjusted and re-simulated until the degree of fit between the new simulated vegetation distribution and the actual vegetation distribution is greater than the preset fit threshold.

[0053] In this embodiment, the precise position of each device in the forest is obtained according to the position coordinates of each monitoring device, and the three-dimensional spatial coordinate data of the device is obtained. The monitoring angle information of each monitoring device is used to determine the monitoring range of each device through the principle of triangulation, and the forest area covered by the device is obtained. The actual terrain data of the forest, including geographical information such as mountains, rivers, and vegetation, is obtained, and a three-dimensional terrain model of the forest is constructed through digital elevation model (DEM) and geographic information system (GIS) technology. The position coordinates and monitoring range of the monitoring device are superimposed and analyzed with the three-dimensional terrain model of the forest, and a three-dimensional spatial grid model covering the entire forest is generated through spatial interpolation and grid processing. According to the three-dimensional spatial grid model, three-dimensional spatial visualization technology, such as three-dimensional rendering, texture mapping, etc., is used to construct a realistic three-dimensional forest scene to obtain a three-dimensional display effect of the forest. By integrating the position and monitoring range information of the monitoring device in the three-dimensional forest model, all-round and three-dimensional monitoring of the forest is achieved, monitoring blind spots are determined, and equipment layout is optimized.

[0054] According to the spatial location information of the identified monitoring blind spot, the geographic coordinates and corresponding terrain feature data of the existing monitoring points around the blind spot are obtained. Through the Thiessen polygon algorithm, with the monitoring blind spot as the center, the existing monitoring points around the blind spot are used as the vertices of the polygon to generate a Thiessen polygon network containing the blind spot. Using the known terrain feature data of each polygon in the Thiessen polygon network, the terrain feature value inside each polygon is estimated through the spatial interpolation method to obtain the terrain feature estimation result inside the blind spot. According to the estimated terrain feature value inside the blind spot, combined with the area, shape and other attributes of the blind spot, the terrain complexity and monitoring difficulty in the blind spot are judged. By analyzing the difference between the estimated terrain features in the blind spot and the terrain features of the existing monitoring points around it, it is determined whether there are special terrain or abnormal areas in the blind spot, which provides a reference for the layout of subsequent monitoring points. The terrain feature data of the blind spot estimated by the Thiessen polygon algorithm is integrated with the original terrain data to obtain more complete and accurate regional terrain information, providing data support for the optimization of the monitoring network.

[0055] The terrain information of the study area, including altitude, slope, aspect and other data, is obtained through satellite Internet of Things as the input conditions for simulating vegetation distribution. The existing vegetation types and distribution in the study area are analyzed to obtain the distribution probability of different vegetation types under different terrain conditions. The Monte Carlo method is used to randomly select a large number of sampling points in the blind area, and each sampling point contains terrain information such as altitude, slope, and aspect. According to the distribution probability of different vegetation types, each sampling point is randomly simulated to determine the most likely vegetation type at the point. By statistically analyzing the simulation results of all sampling points, the distribution area and proportion of various vegetation types in the blind area are obtained. According to the simulated vegetation distribution, combined with the ecological characteristics and vegetation succession laws of the study area, the vegetation distribution in the blind area is further corrected and improved. Finally, a relatively accurate vegetation distribution map in the blind area is obtained, which provides an important reference for subsequent ecological protection and management.

[0056] For example, a three-dimensional spatial model of a forest with an area of ​​10 square kilometers was established through a three-dimensional spatial modeling algorithm. According to the field of view and monitoring distance of the monitoring equipment, the visual range of the monitoring equipment numbered A-001 was calculated to be 8 square kilometers. After superimposing and analyzing the visual range of each monitoring device, a monitoring blind area with a total area of ​​5 square kilometers was identified, accounting for 15% of the entire forest area. For the monitoring blind area numbered B-012, the Thiessen polygon algorithm was used for spatial interpolation, and its average altitude was estimated to be 850 meters and the slope was 23 degrees. According to the estimated terrain characteristics, the vegetation distribution in the blind area was simulated by the Monte Carlo method. The simulation results show that the coverage of the tree layer is 65%, the coverage of the shrub layer is 20%, and the coverage of the herb layer is 45%. The simulated vegetation distribution was compared with the actual vegetation distribution of the adjacent monitoring area numbered C-005, and the consistency between the two was 92%, meeting the accuracy requirements. Finally, 7 monitoring blind areas and 3 monitoring overlapping areas were marked in the three-dimensional spatial model.

[0057] S102, based on the three-dimensional space model, determine a plurality of first overlapping areas from the plurality of potential overlapping areas according to a spatial geometry algorithm, and determine a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data.

[0058] In this embodiment, based on the three-dimensional space model, a spatial geometry algorithm is used to identify the target overlapping area according to the marked potential overlapping area, and an optimization analysis is performed. Specifically, according to the three-dimensional space model data, a spatial geometry algorithm is used to analyze the potential overlapping area to obtain multiple first overlapping areas. By obtaining terrain undulation data and tree occlusion data, the first overlapping area is further screened to determine several target overlapping areas. Among them, the spatial geometry algorithm includes computational geometry algorithms and computer graphics algorithms, such as convex hull algorithms, scan line algorithms, area filling algorithms, etc. When determining the first overlapping area, by constructing a boundary representation of the three-dimensional space model, the convex hull algorithm is used to calculate the boundary polygon of the potential overlapping area, and then the scan line algorithm is used to determine the overlap of the polygon to obtain the first overlapping area. Then, according to the terrain undulation data, the digital elevation model is used to analyze the terrain features of the first overlapping area, and the area with large terrain undulation is eliminated. At the same time, using the tree occlusion data, the tree occlusion of the first overlapping area is calculated by the shadow analysis algorithm, and the area with severe tree occlusion is removed, and finally several target overlapping areas that meet the conditions are obtained.

[0059] In some embodiments, the above step S102 specifically includes:

[0060] Calculating the boundary polygon of the potential overlapping area by using a convex hull algorithm, determining the overlapping condition of the boundary polygon by using a scan line algorithm, and obtaining a first overlapping area;

[0061] Based on the terrain undulation data, a digital elevation model is used to analyze the terrain features of the plurality of first overlapping areas, and the areas in the first overlapping areas where the terrain undulation is greater than a preset terrain undulation threshold are eliminated to obtain a plurality of second overlapping areas;

[0062] Based on the tree occlusion data, a shadow analysis algorithm is used to analyze the tree occlusion situation of the second overlapping area, and areas in the second overlapping area where the tree occlusion degree is greater than a preset tree occlusion degree threshold are proposed to obtain several target overlapping areas.

[0063] In this embodiment, according to the boundary representation of the three-dimensional space model, the convex hull algorithm is used to calculate the boundary polygon of the potential overlapping area to obtain the geometric boundary of the potential overlapping area. The boundary polygon of the potential overlapping area is traversed by the scan line algorithm to obtain the pixel coordinate information inside the polygon to determine whether the pixel belongs to the overlapping area. According to the coordinates of the overlapping pixels obtained by the scan line algorithm, the specific position and range of the first overlapping area in the three-dimensional space model are determined. The boundary polygon and the overlapping pixel coordinate information are used to obtain an accurate geometric representation of the first overlapping area, providing a data basis for subsequent further analysis. The size, shape and spatial distribution characteristics of the overlapping area are determined by the geometric information of the first overlapping area, providing a basis for optimizing the three-dimensional space model. According to the analysis results of the first overlapping area, the redundancy and inconsistency existing in the three-dimensional space model are obtained to provide guidance for subsequent model optimization.

[0064] According to the terrain undulation data, the first overlapping area is modeled using a digital elevation model to obtain a three-dimensional terrain model of the first overlapping area. By performing a slope analysis on the three-dimensional terrain model of the first overlapping area, the slope values ​​of each position in the first overlapping area are obtained. According to the set slope threshold, it is determined whether each position in the first overlapping area belongs to an area with large terrain undulations, and the distribution of areas with large terrain undulations is obtained. The mask processing method is used to remove the area judged to have large terrain undulations from the first overlapping area, and the first overlapping area after removing the area with large terrain undulations is obtained. The first overlapping area after removing the area with large terrain undulations is visualized to intuitively present the terrain characteristics of the area, providing a reference for subsequent analysis.

[0065] According to the acquired tree occlusion data, the shadow analysis algorithm is used to calculate the tree occlusion of the first overlapping area, and the tree occlusion degree values ​​of each area are obtained. By comparing the tree occlusion degree values, the area with more serious tree occlusion is determined, and it is removed from the first overlapping area to obtain the remaining candidate target overlapping area. According to the area size and shape characteristics of the candidate target overlapping area, the region screening algorithm is used to further screen the candidate area to obtain the target overlapping area that meets the area and shape requirements. By analyzing the position coordinates of the target overlapping area, the spatial distribution of each area is obtained, and the relative position relationship between the areas is determined. According to the spatial distribution of the target overlapping area, the spatial clustering algorithm is used to cluster the target overlapping area to obtain several clusters, and each cluster contains multiple target overlapping areas with similar spatial positions. By performing attribute analysis on the target overlapping area in each cluster, the attribute information such as vegetation type and terrain characteristics of each area is obtained, and the ecological environment characteristics of each area are determined. According to the ecological environment characteristics of each cluster, the ecological suitability evaluation algorithm is used to evaluate each cluster to obtain the ecological suitability score of each cluster. By sorting and screening the ecological suitability scores of each cluster, several clusters with the highest ecological suitability are obtained, and the final target overlapping area set is determined. Based on the final target overlapping area set, the boundary extraction algorithm is used to extract the boundary of each area to obtain the vector boundary data of each area for subsequent analysis and application.

[0066] For example, when determining the first overlapping area, a three-dimensional space model containing 500 triangular facets was constructed. The convex hull algorithm was used to calculate that there were 20 boundary polygons of the potential overlapping area. Then, the scan line algorithm was used to determine that 8 of the 20 polygons overlapped, so 8 first overlapping areas were obtained. Next, based on the terrain undulation data, the digital elevation model was used to analyze the terrain characteristics of these 8 first overlapping areas. For example, areas with an average slope greater than 30 degrees were considered to have large terrain undulations. After analysis, 3 areas with large terrain undulations were eliminated, leaving 5 areas. At the same time, using tree occlusion data, the shadow analysis algorithm was used to calculate the tree occlusion of the first overlapping area. For example, areas with a canopy ratio of more than 50% were considered to have severe occlusion. It was found that there were 2 areas with severe tree occlusion, so they were removed. Finally, through the above analysis, evaluation and screening, 3 target overlapping areas that met various conditions were obtained from the initial 500 triangular facets after a series of calculations.

[0067] S103, acquiring the monitoring data collected by each monitoring device in the target overlapping area, comparing and analyzing the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local anomaly factor algorithm, determining the consistency and difference of the monitoring data of each monitoring device, and determining the data interference factor through the consistency and the difference.

[0068] In this embodiment, the temperature, humidity and wind direction data collected by each monitoring device in the overlapping area are obtained, and a comparative analysis is performed to determine the consistency and difference of the data, and to analyze the source and interference factors of the data. Specifically, the temperature, humidity and wind direction data collected by each monitoring device in the overlapping area are obtained, and the correlation between the data of each monitoring device is calculated using the Pearson correlation coefficient to obtain a correlation coefficient matrix. According to the correlation coefficient matrix, the monitoring devices are divided into different groups by cluster analysis, and the similarity of the monitoring device data in each group is high, and the similarity between groups is low. Outlier detection is performed on each group of monitoring device data, and the abnormality of each data point is calculated using the local abnormal factor algorithm, and the data points with abnormalities exceeding the threshold are marked as abnormal values. The consistency of the group of data is determined by abnormal value analysis. If there are more abnormal values, the consistency is poor, and vice versa. Variance analysis is performed on the monitoring device data of different groups, and the intra-group variance and inter-group variance are calculated. The difference of different groups of data is determined by the variance analysis results. If the inter-group variance is much larger than the intra-group variance, it indicates that the data of different groups are significantly different; if the inter-group variance is equivalent to the intra-group variance, it indicates that the data of different groups are not much different. According to the geographical distribution of each group of data, analyze the external environmental factors that cause data differences, such as terrain, building shielding, heat source, etc. Use spatial interpolation methods such as Kriging interpolation to estimate the environmental parameter values ​​of unknown locations and generate a continuous environmental parameter distribution map. Compare this distribution map with the distribution of each group of data to analyze the impact of environmental factors on the data. Perform frequency domain analysis on the data of each monitoring device, and use the fast Fourier transform algorithm to convert the time domain data into frequency domain data to obtain the amplitude and phase information of different frequency components. Use the frequency domain analysis results to determine the periodic interference components in the data, such as the vibration of the device itself, electromagnetic interference, etc. For data with obvious periodic interference, filtering processing is required to eliminate interference. Based on the above analysis results, the source of the data of each monitoring device is determined. If a group of data has good consistency, few outliers, little difference with other groups, little impact from environmental factors, and little periodic interference, it can be determined that the source of the data is reliable. Otherwise, it is necessary to further verify the source of the data and check the interference factors.

[0069] In some embodiments, in the above step S103, comparing and analyzing the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local anomaly factor algorithm to determine the consistency and difference of the monitoring data of each monitoring device specifically includes:

[0070] The Pearson correlation coefficient is used to calculate the correlation between the monitoring data of each monitoring device to obtain a correlation coefficient matrix;

[0071] Based on the correlation coefficient matrix, multiple monitoring devices are divided into different monitoring groups through cluster analysis;

[0072] Based on the monitoring data of each monitoring group, the abnormality of each data point is calculated using a local abnormality factor algorithm, and the data point whose abnormality exceeds a preset abnormality threshold is marked as an abnormal value, and the consistency of the monitoring data of each monitoring group is determined by analyzing the abnormal value;

[0073] The monitoring data of different monitoring groups are subjected to variance analysis, and the intra-group variance and inter-group variance of each monitoring group are calculated to obtain variance analysis results, and the degree of difference between the monitoring data of different monitoring groups is determined through the variance analysis results.

[0074] In this embodiment, according to the location information of each monitoring device in the overlapping area, the temperature, humidity and wind direction data collected by each monitoring device are obtained to obtain the data set of each monitoring device. By extracting the temperature data in the data set of each monitoring device, the temperature time series of each monitoring device is obtained, and the temperature data vector of each monitoring device is determined. The same method is used to extract the humidity data and wind direction data in the data set of each monitoring device respectively, and the humidity data vector and wind direction data vector of each monitoring device are obtained. According to the calculation formula of the Pearson correlation coefficient, the temperature data vector of each monitoring device is used to calculate the correlation coefficient of the temperature data between the devices, and the correlation coefficient matrix of the temperature data is obtained. Through the same calculation process, the humidity data vector and the wind direction data vector are used respectively to calculate the correlation coefficient of the humidity data and the wind direction data between the devices, and the humidity data correlation coefficient matrix and the wind direction data correlation coefficient matrix are obtained. The correlation coefficient matrix of temperature, humidity and wind direction is integrated to obtain the correlation matrix of the comprehensive environmental data between each monitoring device, and the correlation degree of the data collected by different monitoring devices is judged.

[0075] The data of each group of monitoring equipment is obtained. Through data preprocessing, the data is converted into an input format suitable for the local anomaly factor algorithm to obtain the standardized monitoring equipment data. According to the standardized monitoring equipment data, the local anomaly factor value of each data point is calculated by the local anomaly factor algorithm. By setting appropriate k-nearest neighbor parameters, the k-nearest neighbors of each data point are obtained, and their local density is calculated to obtain the abnormality of each data point. By setting the abnormality threshold, the data points whose abnormality exceeds the threshold are marked as potential outliers, and a set of potential outliers is obtained. According to the attributes of the potential outliers such as the timestamp and device number, the potential outliers are clustered and analyzed. Through the density clustering algorithm, the outliers that are similar in time and space are classified into the same category to obtain a set of abnormal events. The statistical analysis method is used to evaluate the severity of each abnormal event. By calculating indicators such as the duration of the abnormal event, the number of outliers and the degree of deviation, the severity score of the abnormal event is obtained, and the abnormal events that need to be focused on are sorted from high to low according to the severity, and the abnormal events that need to be focused on are determined. According to the device number and timestamp of the abnormal event, the original data of the corresponding monitoring device is obtained. Through visualization technology, the time sequence diagram and heat map of the abnormal event are generated to intuitively show the occurrence process and impact range of the abnormal event, providing support for the subsequent abnormal cause analysis. The detected abnormal event is compared with the historical abnormal event, and through similarity calculation, it is determined whether the current abnormal event is a known type. If it is a known type, the processing plan for this type of abnormal event is obtained. If it is an unknown type, the alarm mechanism of the abnormal event is triggered and the relevant personnel are notified to handle it.

[0076] S104, fusing the monitoring data of different monitoring devices in each target overlapping area according to a data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effects caused by the monitoring data of different monitoring devices in the target overlapping area.

[0077] In this embodiment, a data fusion algorithm is used to comprehensively process the monitoring data of different devices in the overlapping area, eliminate the overlapping effects, generate a data fusion model, and integrate it into the front-end AI and edge AI.

[0078] In a possible implementation, the monitoring data of different monitoring devices in each target overlapping area are fused according to the data fusion algorithm, which specifically includes: judging the changing trend of temperature, humidity and wind direction according to the terrain and vegetation details of the area, and fusing the overlapping data of the overlapping area. Specifically, according to the terrain and vegetation details of the area, a digital elevation model (DEM) is used to obtain terrain factor data such as the altitude, slope and aspect of the area. Vegetation factor data such as vegetation coverage and vegetation type of the area are obtained through the vegetation index model (VIM). Based on the acquired terrain factor data and vegetation factor data, the temperature interpolation model (TIM) is used to determine the temperature distribution trend of the area. Based on the terrain factor data and vegetation factor data, the humidity interpolation model (HIM) is used to judge the humidity distribution trend of the area. Through the acquired terrain factor data, the wind direction interpolation model (WIM) is used to obtain the wind direction distribution trend of the area. If there is overlap in a certain area, the terrain factor data, vegetation factor data, temperature distribution data, humidity distribution data and wind direction distribution data of the overlapping area are fused, and the weighted average fusion algorithm (WAFA) is used to fuse the multi-source data of the overlapping area to obtain the fused terrain factor data, vegetation factor data, temperature distribution data, humidity distribution data and wind direction distribution data. According to the fused terrain factor data, vegetation factor data, temperature distribution data, humidity distribution data and wind direction distribution data, the spatial interpolation algorithm (SIA) is used to perform an overall trend analysis on the temperature, humidity and wind direction of the entire area, and the temperature change trend, humidity change trend and wind direction change trend of the area are obtained. Based on the trend analysis results, the spatial decision model (SDM) is used to predict the future change trend of the temperature, humidity and wind direction of the area, and the temperature change prediction results, humidity change prediction results and wind direction change prediction results of the area in the future period of time are obtained.

[0079] In some embodiments, in the above step S104, fusing the monitoring data of different monitoring devices in each target overlapping area according to the data fusion algorithm to generate a data fusion model specifically includes:

[0080] Using a Kalman filter algorithm to filter the monitoring data of different monitoring devices in the overlapping area of ​​the multiple templates to obtain filtered monitoring data;

[0081] Using a relevance vector machine algorithm to extract features from the filtered monitoring data to obtain a feature vector of the monitoring data;

[0082] Using the Leaf Bayesian network algorithm to perform probability reasoning on the feature vectors, determine the correlation between different feature vectors, and obtain the association probability between different feature vectors;

[0083] When the association probability between different feature vectors is greater than a preset association probability threshold, the monitoring data corresponding to each feature vector is fused to obtain fused monitoring data;

[0084] The fused monitoring data is modeled using a hidden Markov model algorithm to generate a data fusion model.

[0085] In this embodiment, based on the monitoring data of different monitoring devices in the overlapping area of ​​the multiple templates, the Kalman filter algorithm is used to filter the monitoring data to obtain filtered monitoring data. Through the filtered monitoring data, the position coordinates and monitoring values ​​of each monitoring device in the overlapping area are obtained, and the precise position coordinates of the target object in the monitoring area are calculated using the triangulation principle. According to the precise position coordinates of the target object, the specific monitoring device where it is located is obtained, and the monitoring value of the monitoring device is used as the final monitoring data of the target object. By comparing the monitoring data of multiple monitoring devices, the movement trajectory and state changes of the target object are judged to determine whether there is an abnormal situation. According to the judgment result, the real-time monitoring information of the target object is obtained, and the monitoring information is transmitted to the monitoring center so that corresponding measures can be taken in time.

[0086] According to the filtered monitoring data, the relevant vector machine algorithm is used to extract the features of the monitoring data to obtain the feature vector that reflects the characteristics of the monitoring data. Through the feature vector, the support vector machine algorithm is used to train the feature vector to obtain the support vector machine classification model. The new monitoring data to be predicted is obtained, and its feature vector is extracted according to the method in step 1. The feature vector of the new monitoring data is input into the support vector machine classification model, and the predicted classification result of the monitoring data is obtained through model prediction. According to the predicted classification result, the category to which the new monitoring data belongs is judged, and the corresponding equipment operation status is determined, such as normal, abnormal or faulty. According to the judged equipment operation status, corresponding control measures are taken, such as maintaining normal operation, alarm prompts, emergency shutdown, etc., to ensure the safe and stable operation of the equipment.

[0087] According to different feature vectors, the Leaf Bayesian network algorithm is used to perform probability reasoning calculation on the feature vectors to obtain the association probability between different feature vectors. The correlation between different feature vectors is judged by the calculated association probability, and the correlation measurement result of the feature vector is obtained. According to the correlation measurement result of the feature vector, the threshold filtering method is used to determine the highly correlated feature vector combination. The feature vector combination is used to obtain the fused high-dimensional feature vector through the feature fusion algorithm. The fused high-dimensional feature vector is input into the classification model, and the final business classification judgment result is obtained through model training and prediction.

[0088] S105, using image recognition technology to analyze the monitoring image in the potential monitoring blind spot, extracting terrain feature data, vegetation density data and tree height data in the monitoring image, and determining the cause of the blind spot based on the terrain feature data, the vegetation density data and the tree height data.

[0089] In this embodiment, image recognition technology is used to analyze monitoring images in the blind spot area, extract data on terrain features, vegetation density and tree height in the image, determine the specific cause of the potential blind spot, and identify data points affected by the blind spot.

[0090] In a possible implementation, the terrain features of hills and gullies in the image are identified, and it is determined whether the terrain undulation leads to a monitoring blind spot. The distribution of trees and the type of vegetation are analyzed to determine whether dense vegetation leads to a blind spot. The occlusion effect of the height of trees on the viewing angle is evaluated to determine whether the height of trees leads to a blind spot. Specifically, according to the input image data, an edge detection algorithm, such as the Canny algorithm, is used to detect the edge of the image to obtain edge information in the image. Based on the edge information, a regional growing algorithm, such as a seed filling algorithm, is used to perform regional segmentation on the edge to obtain different regions in the image. According to the regional segmentation results, a morphological analysis method, such as a skeleton extraction algorithm, is used to extract the skeleton information of the region to obtain the shape characteristics of the region. According to the shape characteristics of the region, a geometric shape analysis method, such as a convex hull analysis algorithm, is used to determine whether the region is a terrain feature such as a hill or a gully. If the region is a hill or a gully, a line of sight analysis algorithm, such as a ray tracing algorithm, is used according to the degree of terrain undulation to analyze the effect of the terrain undulation on the monitoring viewing angle to determine whether the terrain undulation leads to a monitoring blind spot. According to the regional segmentation results, a texture analysis method, such as a grayscale co-occurrence matrix algorithm, is used to analyze the texture characteristics in the region to obtain the texture information of the region. According to the texture information, a texture classification algorithm, such as a support vector machine algorithm, is used to classify the vegetation type of the area and determine the vegetation type in the area. According to the vegetation type classification results, a density estimation algorithm, such as a kernel density estimation algorithm, is used to analyze the vegetation density in the area and determine whether the vegetation density leads to a monitoring blind area. According to the regional segmentation results, a depth estimation algorithm, such as a binocular stereo vision algorithm, is used to estimate the tree height information in the area. According to the tree height information, a three-dimensional space analysis algorithm, such as an octree algorithm, is used to analyze the occlusion effect of the tree height on the monitoring perspective and evaluate whether the tree height leads to a monitoring blind area. By comprehensively analyzing factors such as terrain features, vegetation type, vegetation density, and tree height, a blind area recognition algorithm, such as a multi-factor weighted algorithm, is used to comprehensively determine the location and range of the monitoring blind area in the image to obtain the final monitoring blind area recognition result.

[0091] In some embodiments, the above step S105 specifically includes:

[0092] Using an image segmentation algorithm to segment the monitoring image in the potential monitoring blind area to obtain a plurality of image blocks;

[0093] Based on the decision tree classification algorithm, the image block category is classified according to the color feature, texture feature and shape feature of each image block to determine the environment category of each image block, wherein the environment category includes terrain, vegetation and trees;

[0094] Based on the environmental category of each image block, the region growing algorithm is used to merge image blocks of the same category to obtain terrain areas, vegetation areas, and tree areas;

[0095] In the terrain area, the grayscale value and gradient of each image block are extracted, and the elevation value of the terrain area is estimated by combining the support vector machine regression algorithm to obtain the terrain height information, and the terrain characteristic parameters are calculated according to the terrain height information;

[0096] In the vegetation area, a color histogram and a co-occurrence matrix of each image block are extracted, and vegetation density classification is performed in combination with a random forest classification algorithm to obtain vegetation density information, and vegetation density information is calculated based on the vegetation density information;

[0097] In the tree area, edge features of each image block are extracted, and the positions of tree trunks are detected by combining the Hough transform algorithm to obtain tree distribution position information, and tree height information is calculated based on the tree distribution position information;

[0098] Based on the terrain feature parameters, the vegetation density information and the tree height information, a blind spot discrimination model is established through a logistic regression algorithm, and the blind spot discrimination model is used to determine the blind spot cause of the potential monitoring blind spot.

[0099] In this embodiment, according to the monitoring image data, the image segmentation algorithm is used to segment the image to obtain image blocks of different regions. By extracting the color, texture, shape and other features of the image block, and combining the decision tree classification algorithm to classify the image block, different categories such as terrain, vegetation, and trees are determined. According to the classification results, the image blocks of the same category are merged by the regional growing algorithm to obtain complete terrain areas, vegetation areas, and tree areas. For the terrain area, by extracting the gray value, gradient and other features of the image block, and combining the support vector machine regression algorithm, the elevation value of the area is estimated to obtain the terrain height information. According to the terrain height information, the terrain characteristic parameters such as the slope and slope direction of the area are calculated. For the vegetation area, by extracting the color histogram, co-occurrence matrix and other texture features of the image block, and combining the random forest classification algorithm, the vegetation density is classified to obtain the vegetation density information. According to the vegetation density information, the vegetation area is processed by the morphological operation algorithm to extract the vegetation characteristic parameters such as vegetation coverage. For the tree area, by extracting the edge features of the image block, and combining the Hough transform algorithm, the trunk position is detected to obtain the tree distribution position information. According to the distribution of trees, the watershed segmentation algorithm is used to segment the crown to obtain a single crown area. By extracting the size of the circumscribed rectangle of the crown area and combining it with the least squares fitting algorithm, the tree height is estimated to obtain the tree height information. Based on the terrain characteristic parameters, vegetation characteristic parameters and tree height information, the logistic regression algorithm is used to establish a blind spot discrimination model to determine the cause of the potential blind spot. When the slope is greater than a certain threshold and the slope is facing away from the monitoring point, it is judged as a blind spot caused by terrain occlusion; when the vegetation coverage is greater than a certain threshold and the vegetation density is large, it is judged as a blind spot caused by vegetation occlusion; when the tree height is greater than a certain threshold and the tree distribution position is close to the monitoring point, it is judged as a blind spot caused by tree occlusion. According to the blind spot discrimination results, the ray casting algorithm is used to simulate the visible line from the monitoring point to the blind spot to obtain the pixel position affected by the blind spot. The position information is matched with the original monitoring image to determine the image data points affected by the blind spot and mark them as blind spot data.

[0100] In a possible implementation, the decision tree classification algorithm is used to classify the image blocks by color features, texture features and shape features of each image block, and the environmental category of each image block is determined, which specifically includes: according to the acquired color features, texture features and shape features of each image block, the decision tree classification algorithm is used to classify the environmental category of each image block, and the environmental category to which each image block belongs is obtained, including terrain, vegetation and trees. The color distribution information of the image block is obtained by extracting the color histogram features of each image block, and the color distribution information of the image block is used as one of the input features of the decision tree classification algorithm to determine the environmental category to which the image block belongs. The texture features of the image block are calculated by using the gray level co-occurrence matrix to obtain the texture information of the image block, such as contrast, correlation, energy, etc., and the texture features are used as another input feature of the decision tree classification algorithm to determine the environmental category of the image block. The shape features of the image block are extracted, such as area, perimeter, circularity, etc., to obtain the shape information of the image block, and the shape information is used as the third type of input feature of the decision tree classification algorithm to determine the environmental category to which the image block belongs. According to the color features, texture features and shape features of the acquired image blocks, the decision tree classification algorithm is used to build an environmental category classification model. The decision tree model parameters are optimized through training sample data to obtain the optimal decision tree classification model. The constructed decision tree classification model is used to classify and predict the features of each image block. Through the node division and path selection of the decision tree, the environmental category to which each image block belongs is determined to obtain the category label of the image block, i.e., terrain, vegetation or trees. According to the environmental category label of each image block predicted by the decision tree classification model, the environmental information of the entire image is integrated to obtain the distribution of environmental categories in different areas of the image, and determine the distribution characteristics of terrain, vegetation and trees in the image. By analyzing the environmental category classification results of the image blocks, the proportion and spatial distribution characteristics of different environmental categories in the image are determined, and the overall characteristics of the image environment are obtained, providing a basis for subsequent environmental recognition and analysis.

[0101] In a possible implementation, in the terrain area, the grayscale value and gradient of each image block are extracted, and the elevation value of the terrain area is estimated by combining the support vector machine regression algorithm to obtain the terrain height information, and the terrain feature parameters are calculated according to the terrain height information, which specifically includes: graying each image block, extracting the grayscale value as the grayscale feature of the image block, and obtaining the grayscale feature vector of each image block. Using the gradient operator to calculate the gradient of each image block, the gradient value of the image block in the horizontal and vertical directions is obtained as the gradient feature of the image block, and the gradient feature vector of each image block is obtained. The grayscale feature vector and the gradient feature vector of the image block are used as input, and the elevation value of each image block is estimated by the support vector machine regression algorithm to obtain the estimated elevation value corresponding to each image block. According to the position information and the estimated elevation value of each image block, the elevation value matrix of the terrain area is constructed to obtain the complete terrain height information. Based on the terrain height information, the terrain feature parameters such as the slope, slope direction, and curvature of the terrain area are calculated to obtain the terrain feature description of the terrain area. By combining terrain height information and terrain feature parameters, a three-dimensional model of the terrain area is constructed to achieve accurate expression and analysis of the terrain area and provide data support for subsequent terrain applications.

[0102] In a possible implementation, in the vegetation area, extracting a color histogram and a co-occurrence matrix of each image block, combining a random forest classification algorithm to classify vegetation density, and obtaining vegetation density information, and calculating the vegetation density information according to the vegetation density information specifically includes:

[0103] According to the acquired remote sensing image data, the vegetation index threshold segmentation method is adopted. By calculating the normalized vegetation index NDVI, the pixels with NDVI greater than the set threshold are divided into vegetation areas, and the vegetation area mask image is obtained. After obtaining the vegetation area mask image, the image block method is adopted. By setting a fixed-size sliding window, it slides on the vegetation area image with a certain step length to obtain several image blocks. According to the extracted image blocks, the color histogram calculation method is adopted. By counting the color distribution of pixels in each image block, the histogram vector representing the color characteristics of the image block is obtained. According to the extracted image blocks, the gray level co-occurrence matrix calculation method is adopted. By counting the co-occurrence of the gray value of the pixels in the image block, the co-occurrence matrix representing the texture characteristics of the image block is obtained. After obtaining the color histogram vector and the gray level co-occurrence matrix, the feature splicing method is adopted. By splicing the two features in a certain order, the fused image block feature vector is obtained. According to the actual vegetation density of the vegetation area, the manual annotation method is adopted. By selecting representative image blocks, professionals will annotate the vegetation density, and obtain the sample data set for training and testing. After obtaining the feature vectors of the spliced ​​image blocks and the corresponding vegetation density labels, the random forest classification algorithm is used to train and classify the image block features by constructing multiple decision trees to obtain the vegetation density category of each image block. According to the classified image block vegetation density category, a voting strategy is adopted to obtain the vegetation density distribution of the entire vegetation area by counting the number of image blocks at each vegetation density level. According to the vegetation density distribution, a weighted average method is used to assign corresponding weights to image blocks of different density levels to calculate the comprehensive vegetation density information of the vegetation area.

[0104] In a possible implementation, in the tree area, edge features of each image block are extracted, and the position of the tree trunk is detected in combination with the Hough transform algorithm to obtain tree distribution position information, and tree height information is calculated based on the tree distribution position information, specifically including: based on the acquired tree area image, an edge detection algorithm is used to extract the edge features of the tree trunk in each image block to obtain the edge contour information of the trunk; through the Hough transform algorithm, combined with the trunk edge contour information, the position coordinates of the tree trunk in the image are detected to obtain the distribution position information of each tree; based on the tree distribution position information, the triangulation principle is used to calculate the height difference between the top position and the bottom position of each tree to obtain the height information of each tree; based on the obtained height information of each tree, combined with the tree distribution position information, a tree distribution map containing the position and height of each tree is generated to provide data support for subsequent forest area management and monitoring.

[0105] S106, obtaining the monitoring data of the potential monitoring blind spot by analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind spot, and predicting the long-term change trend of the monitoring data of the potential monitoring blind spot by using a time series prediction algorithm.

[0106] In this embodiment, the K-means clustering algorithm is used to divide the monitoring area according to the terrain and vegetation conditions to obtain sub-areas with different characteristics. The Kriging spatial interpolation algorithm is used to predict and deduce the data in the blind area using the existing monitoring data in the sub-area. When dividing the sub-areas, the Voronoi diagram algorithm is used to optimize the sub-area boundaries to avoid overlapping monitoring coverage. According to the terrain and vegetation characteristics of different sub-areas, the decision tree algorithm is used to generate monitoring equipment configuration rules. The sensitivity and sampling frequency parameters of the equipment are dynamically adjusted according to the rules. The decision tree is optimized by the AdaBoost algorithm to improve the adaptability and robustness of the configuration rules. In the process of data prediction and deduction, the adaptive Kalman filtering algorithm is used to dynamically correct the prediction results to reduce the prediction error. Through the ARIMA time series model, combined with historical data and seasonal factors, the long-term trend prediction of blind area data is carried out. In the deployment and scheduling of monitoring equipment, the ant colony algorithm is used to solve the optimal layout location and path planning problems. Through the Dijkstra shortest path algorithm, the routes of equipment maintenance and data return are optimized to reduce operation and maintenance costs. Finally, the DBSCAN density clustering algorithm is used to detect anomalies in the monitoring data, and abnormal areas are discovered and located in a timely manner. According to the location and characteristics of the abnormal areas, the monitoring strategy is dynamically adjusted to initiate targeted monitoring tasks. The importance of abnormal areas is evaluated through the PageRank algorithm to optimize the allocation and scheduling of monitoring resources.

[0107] In some embodiments, the above step S106 specifically includes:

[0108] According to the terrain and vegetation conditions, the K-means clustering algorithm is used to divide the monitoring area to obtain sub-areas with different characteristics, and the regional boundaries of the sub-areas are optimized by the Voronoi diagram algorithm;

[0109] Determine several sub-areas adjacent to the potential monitoring blind area, and use the Kriging spatial interpolation algorithm to predict and deduce the monitoring data of the potential monitoring blind area based on the monitoring data in the several sub-areas adjacent to the potential monitoring blind area, and perform dynamic correction according to the adaptive Kalman filtering algorithm to obtain the blind area monitoring data;

[0110] According to the blind spot monitoring data and seasonal factors within a preset time period, the long-term change trend of the monitoring data of the potential monitoring blind spot is predicted by the ARIMA time series model to obtain the predicted monitoring data of the potential monitoring blind spot.

[0111] In this embodiment, according to the terrain and vegetation conditions of the monitoring area, the K-means clustering algorithm is used to perform cluster analysis on the terrain and vegetation characteristics of the monitoring area, and regions with similar terrain and vegetation characteristics are obtained, and sub-regions with different characteristics are obtained through cluster analysis. The Voronoi diagram algorithm is used to optimize the regional boundaries of the sub-regions with different characteristics obtained in the previous step, and the optimal boundaries of the sub-regions are obtained through the Voronoi diagram algorithm to obtain the optimized sub-region boundaries. According to the optimized sub-region boundaries, the terrain characteristics and vegetation characteristics of each sub-region are obtained, and the terrain type and vegetation type of each sub-region are determined by analyzing the terrain and vegetation characteristics of each sub-region. According to the terrain type and vegetation type of each sub-region, different monitoring strategies and monitoring methods are adopted to obtain the monitoring data of each sub-region, and the ecological environment status of each sub-region is judged by analyzing the monitoring data. According to the ecological environment status of each sub-region, different protection measures and management measures are adopted, and the ecological environment improvement of each sub-region is obtained by implementing the protection measures and management measures, and the overall ecological environment status of the monitoring area is determined.

[0112] According to the grids divided in the monitoring area, the location information of the potential monitoring blind area is obtained. By analyzing the grids around the potential monitoring blind area, several sub-areas adjacent to the potential monitoring blind area are determined. The monitoring data in several sub-areas adjacent to the potential monitoring blind area are obtained by using a wireless sensor network. Through data preprocessing, effective monitoring data that can be used for spatial interpolation are obtained. According to the principle of the Kriging spatial interpolation algorithm, the effective monitoring data of the adjacent sub-areas are used to construct the semivariogram function and covariance matrix to obtain the preliminary predicted monitoring data of the potential monitoring blind area. According to the adaptive Kalman filtering algorithm, the historical monitoring data and Preliminary prediction of monitoring data, through dynamic updating of state equations and observation equations, to obtain dynamic correction values ​​for monitoring data of potential blind spots; the final blind spot monitoring data of potential blind spots are obtained through weighted fusion based on preliminary prediction of monitoring data of potential blind spots obtained by Kriging spatial interpolation algorithm and dynamic correction values ​​obtained by adaptive Kalman filtering algorithm; the obtained blind spot monitoring data are used to judge the pollution status of potential blind spots through environmental monitoring visualization platform, and according to the excessive concentration of pollutants, determine whether it is necessary to add monitoring equipment in the potential blind spots, so as to improve the layout of monitoring network and improve the monitoring accuracy of the whole monitoring area.

[0113] By collecting and sorting the blind spot monitoring data within the preset time period, the historical monitoring data of the potential monitoring blind spot is obtained as the input data of the ARIMA model. According to the collected historical monitoring data, the data is preprocessed in combination with seasonal factors to remove outliers and noise data to obtain a standardized monitoring data series. The standardized monitoring data series is stabilized by the difference method to eliminate the non-stationarity of the data and obtain stable time series data. By performing autocorrelation and partial correlation analysis on the stabilized time series data, the order parameters of the ARIMA model are determined to obtain the optimal model structure. According to the determined ARIMA model structure, the model parameters are estimated by the maximum likelihood estimation method to obtain the coefficients and residual terms of the model. By performing diagnostic tests on the estimated ARIMA model, the fitting effect and prediction ability of the model are judged to ensure the validity and reliability of the model. According to the verified ARIMA model, combined with the historical monitoring data and seasonal factors of the potential monitoring blind spot, the monitoring data for a period of time in the future are predicted to obtain the predicted monitoring data of the potential monitoring blind spot. By analyzing and evaluating the predicted monitoring data, the long-term change trend of potential monitoring blind spots can be determined, providing a decision-making basis for the optimization of monitoring blind spots.

[0114] S107, transmitting the monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area to the data center via the satellite Internet of Things, and generating an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart based on the data analysis.

[0115] In this embodiment, the data center comprehensively evaluates the forest environment, including fire warning, meteorological changes and ecological changes, generates an environmental monitoring report through data analysis, and produces a fire risk map and a temperature and humidity change trend chart. The processed monitoring data is transmitted to the data center through the satellite Internet of Things, and the data of each device is summarized and integrated. According to the results of the overlapping area analysis, the empirical orthogonal decomposition algorithm is used to analyze the wind direction change, obtain humidity fluctuation data, and evaluate the fire risk through the fuzzy comprehensive evaluation method. The data center comprehensively evaluates the forest environment, including fire warning, meteorological changes and ecological changes. According to the fire warning results, the random forest algorithm is used for data analysis to obtain an environmental monitoring report. According to the meteorological change data, a fire risk map is produced by the support vector machine algorithm. According to the ecological change data, a temperature and humidity change trend chart is produced by the time series analysis method. If the fire risk is higher than the threshold, a fire warning is triggered, and the warning information is sent to the monitoring device through the satellite communication system. If the humidity is lower than the threshold and the temperature is higher than the threshold, it is judged as high fire risk weather, and the monitoring frequency is increased. According to the fire risk map, the ant colony algorithm is used to plan the patrol route and optimize the patrol efficiency. According to the temperature and humidity change trend chart, the meteorological changes in the next week are predicted by the hidden Markov model, and the monitoring strategy is adjusted. Obtain real-time data from each monitoring device, conduct comprehensive analysis through data fusion algorithms, and determine whether there are any abnormalities. If an abnormality is found, activate the emergency plan and dispatch rescue forces.

[0116] It can be understood that micro-meteorological instruments or meteorological sensors can be deployed at key locations of each monitoring device or nearby to collect real-time meteorological data such as temperature, humidity, atmospheric pressure, wind speed, wind direction, PM2.5, PM10, noise, illumination, rainfall, etc., and combined with satellite Internet of Things terminals and monitoring equipment for fireworks identification to build a multi-level monitoring system.

[0117] In some embodiments, the monitoring data includes wind speed and direction data, temperature monitoring data, and humidity monitoring data; the method further includes:

[0118] By determining the influence of the terrain feature data on the wind speed and wind direction data, analyzing the degree of difference between the wind speed and wind direction data of each monitoring device in the target overlapping area;

[0119] Analyzing the degree of difference between the temperature monitoring data of each monitoring device in the target overlapping area according to the temperature change trend in the forest environment;

[0120] The humidity characteristics of the forest environment are determined according to the spatial distribution map of vegetation types and the spatial distribution map of water sources in the forest environment, and the degree of difference between the humidity monitoring data of each monitoring device in the target overlapping area is analyzed through the humidity characteristics of the forest environment.

[0121] In this embodiment, a multivariate regression model is established based on the terrain feature data of the forest environment and the wind speed and direction data of the forest environment to obtain the influence of the terrain feature on the wind speed and direction; variance analysis is performed using the different terrain type data of the forest environment, the wind speed data and the wind direction data to obtain the difference in wind speed and direction between different terrain types; a temperature time series graph is drawn based on the temperature time series data of the forest environment to determine the stationarity of the temperature data, and the ARIMA model order is determined based on the autocorrelation and partial autocorrelation graphs to establish the ARIMA model; and the ARIMA A model is used to predict future temperatures, obtain temperature prediction results and confidence intervals, and compare the differences in temperature prediction results of different devices; a spatial distribution map of vegetation types and a spatial distribution map of water sources of the forest environment are obtained, and the spatial distribution map of vegetation types, the spatial distribution map of water sources and the humidity data of the forest environment are superimposed and analyzed to obtain the spatial correlation between the humidity and the vegetation types and water sources; a spatial interpolation method is used to perform spatial interpolation on the humidity data to generate a humidity spatial distribution map, and the spatial differentiation characteristics of the humidity are analyzed by combining the spatial distribution map of vegetation types and the spatial distribution map of water sources.

[0122] Reference Figure 2 An embodiment of the present invention provides a forest fire early warning system 2 based on satellite Internet of Things, and the system 2 specifically includes:

[0123] The first early warning module 201 is used to establish a three-dimensional space model based on the posture data of each monitoring device and the forest terrain data, simulate the monitoring area of ​​each monitoring device in the three-dimensional space model through the terrain map and satellite image, and determine the potential monitoring blind area and potential overlapping area in the monitoring area;

[0124] A second early warning module 202 is used to determine a plurality of first overlapping areas from the plurality of potential overlapping areas according to a spatial geometry algorithm based on the three-dimensional space model, and to determine a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data;

[0125] The third early warning module 203 is used to obtain the monitoring data collected by each monitoring device in the target overlapping area, compare and analyze the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local abnormal factor algorithm, determine the consistency and difference of the monitoring data of each monitoring device, and determine the data interference factor through the consistency and difference;

[0126] The fourth warning module 204 is used to fuse the monitoring data of different monitoring devices in each target overlapping area according to the data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effect caused by the monitoring data of different monitoring devices in the target overlapping area;

[0127] The fifth warning module 205 is used to analyze the monitoring image in the potential monitoring blind area by using image recognition technology, extract terrain feature data, vegetation density data and tree height data in the monitoring image, and determine the cause of the blind area according to the terrain feature data, the vegetation density data and the tree height data;

[0128] The sixth warning module 206 is used to deduce the monitoring data of the potential monitoring blind area by analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind area, and predict the long-term change trend of the monitoring data of the potential monitoring blind area by a time series prediction algorithm;

[0129] The seventh early warning module 207 is used to transmit the monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area to the data center via the satellite Internet of Things, and generate an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart based on data analysis.

[0130] It is understandable that if Figure 1 The contents of the forest fire early warning method embodiment based on satellite Internet of Things shown in the figure are applicable to the forest fire early warning system embodiment based on satellite Internet of Things. The functions specifically implemented by the forest fire early warning system embodiment based on satellite Internet of Things are similar to those in the embodiment shown in the figure. Figure 1 The forest fire early warning method embodiment based on satellite Internet of Things is the same as shown, and the beneficial effects achieved are the same as those of Figure 1 The beneficial effects achieved by the forest fire warning method embodiment based on satellite Internet of Things shown are also the same.

[0131] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0132] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0133] Reference Figure 3 The embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, a forest fire warning method based on satellite Internet of Things as described in any one of the above methods is implemented.

[0134] The computer device 3 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will appreciate that Figure 3 It is only an example of computer device 3 and does not constitute a limitation on computer device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, etc.

[0135] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0136] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 3. Further, the memory 302 may also include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0137] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the forest fire warning method based on satellite Internet of Things as described in any one of the above methods is implemented.

[0138] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0139] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0140] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0141] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0142] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Claims

1. A forest fire early warning method based on satellite Internet of Things, characterized in that: The method specifically comprises: A three-dimensional space model is established based on the position data of each monitoring device and the forest terrain data, and the monitoring area of ​​each monitoring device is simulated in the three-dimensional space model through the terrain map and the satellite image, and potential monitoring blind areas and potential overlapping areas are determined in the monitoring area; Based on the three-dimensional space model, determining a plurality of first overlapping areas from the plurality of potential overlapping areas according to a spatial geometry algorithm, and determining a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data; Acquire the monitoring data collected by each monitoring device in the target overlapping area, compare and analyze the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local anomaly factor algorithm, determine the consistency and difference of the monitoring data of each monitoring device, and determine the data interference factor through the consistency and the difference; The monitoring data of different monitoring devices in each target overlapping area are fused according to a data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effect caused by the monitoring data of different monitoring devices in the target overlapping area; Using image recognition technology to analyze the monitoring image in the potential monitoring blind area, extracting terrain feature data, vegetation density data and tree height data in the monitoring image, and determining the cause of the blind area based on the terrain feature data, the vegetation density data and the tree height data; By analyzing the monitoring data of the adjacent areas of the potential monitoring blind area, the monitoring data of the potential monitoring blind area is deduced, and the long-term change trend of the monitoring data of the potential monitoring blind area is predicted by a time series prediction algorithm; The monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area are transmitted to the data center via the satellite Internet of Things, and an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart are generated based on the data analysis.

2. The method according to claim 1, characterized in that The three-dimensional space model is established according to the posture data of each monitoring device and the forest terrain data, and the monitoring area of ​​each monitoring device is simulated in the three-dimensional space model through the terrain map and the satellite image, and the potential monitoring blind area and the potential overlapping area are determined in the monitoring area, specifically including: According to the position coordinate data of each monitoring device, the precise position of each monitoring device in the forest is obtained to obtain the three-dimensional spatial coordinate data of the monitoring device; Based on the principle of triangulation, the monitoring range of each monitoring device is determined through the monitoring angle data of each monitoring device; Acquire forest terrain data, and construct a three-dimensional forest terrain model using the forest terrain data according to digital elevation model and geographic information system technology; Overlaying and analyzing the three-dimensional spatial coordinate data of the monitoring device, the monitoring range, and the three-dimensional terrain model of the forest, and generating a three-dimensional spatial model through spatial interpolation and gridding processing; Simulating the monitoring area of ​​each monitoring device in the three-dimensional space model to obtain the visual range of each monitoring device; Superimpose and analyze the visual range of each monitoring device to determine potential monitoring blind spots and potential overlapping areas; Using the Thiessen polygon algorithm to perform spatial interpolation on the potential monitoring blind area, estimating the terrain features within the potential monitoring blind area, and using the Monte Carlo method to simulate the vegetation distribution within the potential monitoring blind area through the terrain features within the potential monitoring blind area to obtain simulated vegetation distribution; The simulated vegetation distribution is compared with the actual vegetation distribution. If the degree of fit between the two is greater than a preset fit threshold, the simulated vegetation distribution is added to the three-dimensional space model. Otherwise, the simulation parameters are readjusted and re-simulated until the degree of fit between the new simulated vegetation distribution and the actual vegetation distribution is greater than the preset fit threshold.

3. The method according to claim 1, characterized in that The method of determining a plurality of first overlapping areas from the plurality of potential overlapping areas based on the three-dimensional space model according to a spatial geometry algorithm, and determining a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data, specifically includes: Calculating the boundary polygon of the potential overlapping area by using a convex hull algorithm, determining the overlapping condition of the boundary polygon by using a scan line algorithm, and obtaining a first overlapping area; Based on the terrain undulation data, a digital elevation model is used to analyze the terrain features of the plurality of first overlapping areas, and the areas in the first overlapping areas where the terrain undulation is greater than a preset terrain undulation threshold are eliminated to obtain a plurality of second overlapping areas; Based on the tree occlusion data, a shadow analysis algorithm is used to analyze the tree occlusion situation of the second overlapping area, and areas in the second overlapping area where the tree occlusion degree is greater than a preset tree occlusion degree threshold are proposed to obtain several target overlapping areas.

4. The method according to claim 1, characterized in that The monitoring data of each monitoring device is compared and analyzed according to the correlation coefficient algorithm and the local abnormal factor algorithm to determine the consistency and difference of the monitoring data of each monitoring device, specifically including: The Pearson correlation coefficient is used to calculate the correlation between the monitoring data of each monitoring device to obtain a correlation coefficient matrix; Based on the correlation coefficient matrix, multiple monitoring devices are divided into different monitoring groups through cluster analysis; Based on the monitoring data of each monitoring group, the abnormality of each data point is calculated using a local abnormality factor algorithm, and the data point whose abnormality exceeds a preset abnormality threshold is marked as an abnormal value, and the consistency of the monitoring data of each monitoring group is determined by analyzing the abnormal value; The monitoring data of different monitoring groups are subjected to variance analysis, and the intra-group variance and inter-group variance of each monitoring group are calculated to obtain variance analysis results, and the degree of difference between the monitoring data of different monitoring groups is determined through the variance analysis results.

5. The method according to claim 1, characterized in that The step of fusing the monitoring data of different monitoring devices in each target overlapping area according to the data fusion algorithm to generate a data fusion model specifically includes: Using a Kalman filter algorithm to filter the monitoring data of different monitoring devices in the overlapping area of ​​the multiple targets to obtain filtered monitoring data; Using a relevance vector machine algorithm to extract features from the filtered monitoring data to obtain a feature vector of the monitoring data; Using the Leaf Bayesian network algorithm to perform probability reasoning on the feature vectors, determine the correlation between different feature vectors, and obtain the association probability between different feature vectors; When the association probability between different feature vectors is greater than a preset association probability threshold, the monitoring data corresponding to each feature vector is fused to obtain fused monitoring data; The fused monitoring data is modeled using a hidden Markov model algorithm to generate a data fusion model.

6. The method according to claim 1, characterized in that The method of using image recognition technology to analyze the monitoring image in the potential monitoring blind spot, extracting terrain feature data, vegetation density data and tree height data in the monitoring image, and determining the cause of the blind spot according to the terrain feature data, the vegetation density data and the tree height data, specifically includes: Using an image segmentation algorithm to segment the monitoring image in the potential monitoring blind area to obtain a plurality of image blocks; Based on the decision tree classification algorithm, the image block category is classified according to the color feature, texture feature and shape feature of each image block to determine the environment category of each image block, wherein the environment category includes terrain, vegetation and trees; Based on the environmental category of each image block, the region growing algorithm is used to merge image blocks of the same category to obtain terrain areas, vegetation areas, and tree areas; In the terrain area, the grayscale value and gradient of each image block are extracted, and the elevation value of the terrain area is estimated by combining the support vector machine regression algorithm to obtain the terrain height information, and the terrain characteristic parameters are calculated according to the terrain height information; In the vegetation area, a color histogram and a co-occurrence matrix of each image block are extracted, and vegetation density classification is performed in combination with a random forest classification algorithm to obtain vegetation density information; In the tree area, edge features of each image block are extracted, and the positions of tree trunks are detected by combining the Hough transform algorithm to obtain tree distribution position information, and tree height information is calculated based on the tree distribution position information; Based on the terrain feature parameters, the vegetation density information and the tree height information, a blind spot discrimination model is established through a logistic regression algorithm, and the blind spot discrimination model is used to determine the blind spot cause of the potential monitoring blind spot.

7. The method according to claim 1, characterized in that The step of analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind area to deduce the monitoring data of the potential monitoring blind area, and predicting the long-term change trend of the monitoring data of the potential monitoring blind area by a time series prediction algorithm specifically includes: According to the terrain and vegetation conditions, the K-means clustering algorithm is used to divide the monitoring area to obtain sub-areas with different characteristics, and the regional boundaries of the sub-areas are optimized by the Voronoi diagram algorithm; Determine several sub-areas adjacent to the potential monitoring blind area, and use the Kriging spatial interpolation algorithm to predict and deduce the monitoring data of the potential monitoring blind area based on the monitoring data in the several sub-areas adjacent to the potential monitoring blind area, and perform dynamic correction according to the adaptive Kalman filtering algorithm to obtain the blind area monitoring data; According to the blind spot monitoring data and seasonal factors within a preset time period, the long-term change trend of the monitoring data of the potential monitoring blind spot is predicted by the ARIMA time series model to obtain the predicted monitoring data of the potential monitoring blind spot.

8. The method according to any one of claims 1 to 7, characterized in that The monitoring data includes wind speed and direction data, temperature monitoring data and humidity monitoring data; the method further includes: By determining the influence of the terrain feature data on the wind speed and wind direction data, analyzing the degree of difference between the wind speed and wind direction data of each monitoring device in the target overlapping area; Analyzing the degree of difference between the temperature monitoring data of each monitoring device in the target overlapping area according to the temperature change trend in the forest environment; The humidity characteristics of the forest environment are determined according to the spatial distribution map of vegetation types and the spatial distribution map of water sources in the forest environment, and the degree of difference between the humidity monitoring data of each monitoring device in the target overlapping area is analyzed through the humidity characteristics of the forest environment.

9. A forest fire early warning system based on satellite Internet of Things, characterized in that: The system specifically comprises: The first early warning module is used to establish a three-dimensional space model based on the position data of each monitoring device and the forest terrain data, simulate the monitoring area of ​​each monitoring device in the three-dimensional space model through the terrain map and satellite image, and determine the potential monitoring blind area and potential overlapping area in the monitoring area; A second early warning module is used to determine a plurality of first overlapping areas from the plurality of potential overlapping areas based on the three-dimensional space model according to a spatial geometry algorithm, and to determine a plurality of target overlapping areas from the plurality of first overlapping areas according to terrain undulation data and tree occlusion data; The third early warning module is used to obtain the monitoring data collected by each monitoring device in the target overlapping area, compare and analyze the monitoring data of each monitoring device according to the correlation coefficient algorithm and the local abnormal factor algorithm, determine the consistency and difference of the monitoring data of each monitoring device, and determine the data interference factor through the consistency and difference; A fourth early warning module is used to fuse the monitoring data of different monitoring devices in each target overlapping area according to a data fusion algorithm to generate a data fusion model, wherein the data fusion model is used to eliminate the overlapping effect caused by the monitoring data of different monitoring devices in the target overlapping area; a fifth warning module, configured to analyze the monitoring image in the potential monitoring blind area by using image recognition technology, extract terrain feature data, vegetation density data and tree height data in the monitoring image, and determine the cause of the blind area according to the terrain feature data, the vegetation density data and the tree height data; A sixth early warning module is used to deduce the monitoring data of the potential monitoring blind area by analyzing the monitoring data of the adjacent area of ​​the potential monitoring blind area, and predict the long-term change trend of the monitoring data of the potential monitoring blind area by a time series prediction algorithm; The seventh early warning module is used to transmit the monitoring data of each monitoring device, the analysis results of the target overlapping area and the prediction results of the potential monitoring blind area to the data center via the satellite Internet of Things, and generate an environmental monitoring report, a fire risk map and a temperature and humidity change trend chart based on data analysis.

10. A computer device, characterized in that: include: A memory and a processor and a computer program stored in the memory, when the computer program is executed on the processor, implements the forest fire early warning method based on satellite Internet of Things as described in any one of claims 1 to 8.

Images