Low-altitude risk map generation method and device, medium and product
By constructing a low-altitude risk map using a spatiotemporal graph convolutional network and a dynamically coupled inference model, the problem of insufficient risk factor modeling in traditional methods is solved. This enables high-precision risk prediction and real-time updates in complex environments, improving the safety and efficiency of UAV flight missions.
Patent Information
- Application Number
- CN202511545584.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-27
- Publication Date
- 2026-02-06
AI Technical Summary
Traditional low-altitude risk map construction methods struggle to capture real-time changes in risk factors in complex and dynamic environments. Furthermore, risk factor modeling primarily relies on linear superposition, failing to consider the nonlinear relationships and dynamic coupling effects between risk factors, resulting in insufficient accuracy and reliability in risk prediction.
By employing a spatiotemporal graph convolutional network and a dynamic coupling inference model, a risk factor raster map is constructed by acquiring geographical, meteorological, biological, and social environmental data. The spatiotemporal graph convolutional network is used to extract spatial and temporal features, and the dynamic coupling inference model is combined to determine the overall coupling weight of the risk factors. Nonlinear fusion is then performed to generate a two-dimensional heat map of low-altitude risk, which is updated through real-time data sources.
It improves the accuracy and reliability of low-altitude risk prediction, can dynamically respond to sudden changes in complex environments, generate more accurate risk maps, meet the real-time planning needs of UAV flight missions, and reduce human intervention and subjective bias.
Smart Images
Figure CN121481218A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of risk map generation, in particular to a low-altitude risk map generation method, device, medium and product. BACKGROUND
[0002] In the field of low-altitude airspace management and unmanned aerial vehicle flight planning, traditional risk map construction methods generally rely on static rules, expert experience and linear superposition. Using the above methods for risk assessment, there are obvious deficiencies in the multi-source risk factor interaction modeling capability and map timeliness in complex dynamic environments. Specifically, there are two core problems: first, the risk perception mode has static defects and is difficult to adapt to changing environments. Existing methods mainly rely on static rules and prior experience to construct risk models, lack high-frequency perception ability for real-time dynamic data (such as weather, bird flocks, electromagnetic interference, etc.), and cannot capture the changes of risk factors in real time, resulting in lagging risk map updates and insufficient response. Second, the risk factor modeling method is mainly linear superposition, and the coupling relationship is not well described. Most current methods generally use linear models such as weighted summation to process multi-source risk factors, without considering the possible nonlinear relationships, dynamic coupling effects and linkage amplification mechanisms between risk factors, making it difficult to accurately identify potential high-order risk patterns in actual scenarios.
[0003] Therefore, based on the above problems, it is necessary to provide a low-altitude risk map generation method that can improve the accuracy and reliability of risk prediction in complex environments and provide a scientific basis for airspace dynamic scheduling. SUMMARY
[0004] The purpose of the present application is to provide a low-altitude risk map generation method, device, medium and product that can improve the accuracy and reliability of risk prediction in complex environments.
[0005] To achieve the above purpose, the present application provides the following solutions: In a first aspect, the present application provides a low-altitude risk map generation method, which comprises: obtaining low-altitude risk related data; the low-altitude risk related data includes geographic data, meteorological data, biological data and social environment data; preprocessing the low-altitude risk related data and calculating the corresponding risk values; determining a risk factor grid map according to the risk values; the preprocessing process includes spatial alignment and unit unification; the risk values include geographic risk values, meteorological risk values, biological risk values and social environment risk values; constructing a spatio-temporal graph dataset according to the risk factor grid map; the spatio-temporal graph dataset is a graph structure with grid nodes, spatial adjacency as graph edges, and time dimension as sequence; Based on the spatiotemporal graph structure dataset, spatial and temporal features of each node are extracted using a spatiotemporal graph convolutional network to determine the dynamic risk occurrence feature tensor of each node at multiple time steps. Based on the dynamic risk occurrence feature tensor, a dynamic coupling inference model is adopted to determine the overall coupling weight tensor of risk factors. The dynamic coupling inference model includes: a channel partitioning unit, a time state encoding unit, and a gated loop unit. The time state encoding unit is used to extract the time series features of risk factors at each node based on the dynamic risk occurrence feature tensor and a spatiotemporal attention mechanism, and to determine the coupling strength between risk factors. The gated loop unit is used to extract the time state representation of each risk factor at each node and to enhance the coupling strength based on the time state representation. The comprehensive risk intensity is determined by nonlinearly fusing the dynamic risk occurrence characteristic tensor and the overall coupling weight tensor. A two-dimensional heat map of low-altitude risk is determined based on the overall risk intensity.
[0006] Optionally, the step of determining the two-dimensional heat map of low-altitude risk based on the comprehensive risk intensity further includes: The low-altitude risk two-dimensional heat map is periodically recalculated and updated using a real-time data source, resulting in multiple low-altitude risk two-dimensional heat maps. Each low-altitude risk two-dimensional heat map reflects the comprehensive airspace risk at the corresponding time step. The real-time data source is real-time low-altitude risk-related data.
[0007] Optionally, the preprocessing of low-altitude risk-related data and the calculation of corresponding risk values, followed by determining a risk factor raster based on these risk values, specifically includes: Geographic risk values are determined based on topography and no-fly zones; Using formula Determine meteorological risk value ; Using formula Determine biological risk value ; Using formula Determine the social and environmental risk value ; The risk factor raster map is determined by weighting and fusing geographical risk values, meteorological risk values, biological risk values, and social environmental risk values. in, This is the wind speed risk value. This represents the rainfall risk value. This represents the distance the bird travels from its current point to its takeoff or landing point. This represents the risk value for pedestrian injury or death. This represents the risk value for injury or death of occupants of the vehicle.
[0008] Optionally, the overall coupling weight tensor of the risk factors is determined according to the dynamic risk occurrence feature tensor and the dynamic coupling inference model, and specifically includes: The dynamic risk occurrence feature tensor is divided into a plurality of sub-channels by a channel division unit; each sub-channel corresponds to a risk factor; The coupling strength between the risk factors is determined by extracting the time sequence features of any two risk factors at each node by using a time state coding unit; The time state representation of each risk factor at each node is extracted by using a gated recurrent unit; The coupling strength between the risk factors at each node is enhanced by taking the time state representation as the input of the time state coding unit, and the enhanced coupling strength is obtained; The overall coupling weight tensor of the risk factors is determined according to the enhanced coupling strength at each node.
[0009] Optionally, the coupling strength between the risk factors is determined by extracting the time sequence features of any two risk factors at each node by using a time state coding unit, and specifically includes: The coupling strength between the risk factors is determined by using the formula ; Wherein, and are two different risk factors, is a normalization operation, is a nonlinear similarity function, is the time sequence feature of the risk factor.
[0010] Optionally, the time state representation of each risk factor at each node is extracted by using a gated recurrent unit, and specifically includes: The time state representation of each risk factor at each node is extracted by using the formula Wherein, is the final time state hidden vector of the risk factor at the node , the gated recurrent unit is , and the time sequence feature of the risk factor is .
[0011] Optionally, the comprehensive risk strength is determined by nonlinearly fusing the dynamic risk occurrence feature tensor and the overall coupling weight tensor, and specifically includes: The comprehensive risk strength is determined by using the formula Wherein, is the first The time step, the first An intermediate representation vector of the overall risk intensity of each node. and For two different risk factors, R Number of sub-channels Amplification factor that is set manually or obtained through model training. For the node i Above, risk factors right The intensity of action, It is a binary interaction function. For the first The time step, the first Risk factors of each node Time series characteristics, For the first The time step, the first Risk factors of each node The time series characteristics.
[0012] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for generating medium- and low-altitude risk maps.
[0013] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for generating medium- and low-altitude risk maps.
[0014] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for generating medium- and low-altitude risk maps.
[0015] According to the specific embodiments provided in this application, this application has the following technical effects: The application provides a low-altitude risk map generation method and device, medium and product, collects low-altitude risk related data from multiple fields such as geography, meteorology, biology and society, constructs a multi-level risk factor grid, and improves the accuracy of airspace comprehensive risk determination; the spatio-temporal graph convolution network (ST-GCN) is used to extract the spatial and temporal features of each node, and the response capability to sudden change scenes is improved; the attention mechanism and gated recurrent unit (GRU) are introduced, the nonlinear interaction between risk factors is realized, and the coupling relationship between risk factors is automatically identified and the influence weight is quantified; the spatio-temporal graph convolution network and the dynamic coupling reasoning model are combined, the defects of subjective and fragmented risk coupling in the traditional method are solved, the determination accuracy and reliability of airspace comprehensive risk in complex environment are significantly improved, and a scientific basis is provided for airspace dynamic scheduling. BRIEF DESCRIPTION OF DRAWINGS
[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0017] Figure 1 The flowchart of a low-altitude risk map generation method in an embodiment of the present application. DETAILED DESCRIPTION
[0018] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0019] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.
[0020] In an exemplary embodiment, as shown in Figure 1 A low-altitude risk map generation method is provided, including the following S1 to S7. Wherein: S1: acquiring low-altitude risk related data.
[0021] Specifically, the low-altitude risk-related data includes geographic data, meteorological data, biological data, and social environment data. The geographic data, meteorological data, biological data, and social environment data each include a plurality of risk factors.
[0022] Specifically, the geographic data includes risk factors such as topographic maps, no-fly zones, population density centers, and green distribution; the meteorological data includes risk factors such as wind speed and rainfall; the biological data includes risk factors such as bird activity frequency and bird distribution; and the social environment data includes risk factors such as ground population density and key infrastructure locations.
[0023] S2: Preprocess the low-altitude risk-related data and calculate the corresponding risk values; determine the risk factor grid map according to the risk values.
[0024] As a specific example, the preprocessing process includes spatial alignment and unit unification; and the risk values include geographic risk values, meteorological risk values, biological risk values, and social environment risk values.
[0025] Specifically, when calculating the geographic risk values, according to the information of the terrain and the no-fly zone, the geographic risk values of the areas that are not allowed to pass through are directly set to 100 or higher, to ensure that the path planning will not pass through the areas that are not allowed to pass through, i.e., the geographic risk values are determined according to the terrain and the no-fly zone.
[0026] When calculating the meteorological risk values, the effects of wind speed and rainfall are considered. When the wind speed is greater than or equal to 10.7 m / s, the wind speed risk value is 2; when the wind speed is greater than or equal to 5.4 m / s but less than 10.7 m / s, the wind speed risk value is 1; and when the wind speed is less than 5.4 m / s, the wind speed risk value is 0. When the rainfall is greater than or equal to 25 mm, the rainfall risk value is 2; when the rainfall is greater than or equal to 10 mm but less than 25 mm, the rainfall risk value is 1; and when the rainfall is less than 10 mm, the rainfall risk value is 0. Specifically, the meteorological risk value determination formula is as follows: ; wherein, is the wind speed risk value, , is the rainfall risk value, , is the meteorological risk value.
[0027] Finally, the meteorological risk values are normalized to have a value range of [0, 1].
[0028] When calculating the biological risk value, a green level map is generated using the topographic map, and the green level map is divided into low, medium and high green levels according to the probability of the occurrence of birds, wherein the level corresponding to the bird occurrence probability of 0.2 is low, the level corresponding to the bird occurrence probability of 0.5 is medium, and the level corresponding to the bird occurrence probability of 0.9 is high. The position of the occurrence of birds is generated according to the green level and the bird occurrence probability. For each bird position, the biological risk value within a certain radius (for example, 2 grids) around it is calculated, and the biological risk value decays with the increase of the distance. The calculation formula of the biological risk value is as follows: is the distance from the current point to the take-off point or the landing point.
[0029] The biological risk value is normalized to make its value range between 0 and 1.
[0030] When calculating the social environmental risk value, the personnel casualty risk cost is defined as the social environmental risk value; the social environmental risk value includes the pedestrian casualty risk value and the in-vehicle personnel casualty risk value, and the calculation formula of the pedestrian casualty risk value is as follows: is the pedestrian casualty risk value, is the unmanned aerial vehicle system failure probability (per flight hour), is the number of pedestrians hit (related to population density), ， is the impact area of the unmanned aerial vehicle crash , is the population density gravity model, , is a natural constant, and its value is about 2.71828, is the distance from this point to the center point, which refers to a point with relatively high population density and is artificially set, is the average population density of the region, is the mortality rate of pedestrians in unmanned aerial vehicle accidents (mortality rate = total number of deaths / total number of events), ， is a shielding coefficient, reflecting the buffering effect of buildings / trees, , is the impact kinetic energy required to cause 50% of pedestrians to die when the shielding coefficient is 0.5 (unit: joule, J), is the minimum impact kinetic energy threshold (unit: joule, J) required to cause death when the shielding coefficient approaches 0 (no shielding), and in this application, , is the kinetic energy of the unmanned aerial vehicle, , is the mass of the unmanned aerial vehicle, is the flight speed of the unmanned aerial vehicle.
[0031] The calculation formula of the risk value of the in-vehicle personnel casualty is as follows: ; wherein, is the risk value of the in-vehicle personnel casualty, is the number of vehicles hit, , is the vehicle density gravity model, , is the regional average traffic density.
[0032] The calculation formula of the social environment risk value is as follows: ; wherein, is the social environment risk value.
[0033] The social environment risk value is normalized so that its value range is between [0, 1].
[0034] The geographical risk value, the meteorological risk value, the biological risk value and the social environment risk value are weighted and fused to determine a risk factor grid map, the risk factor grid map including 50 50 grids, each grid having a length of 100 m, and each grid being a total value of weighted fusion of four risk values, i.e. a comprehensive risk.
[0035] S3: constructing a spatio-temporal graph dataset according to the risk factor grid map.
[0036] The spatial structure of the risk factor grid map is 50 50 grids, each grid being 100 m 100 m; the time dimension of the risk factor grid map is T time steps (for example, 24 hours); the risk factor of the risk factor grid map has K-dimensional characteristics, at least including four core risks, i.e. geographical risk, meteorological risk, biological risk and social environment risk; the value range of the risk factor of the risk factor grid map is normalized to [0, 1] except that the geographical risk value is 100.
[0037] The spatio-temporal graph dataset constructed by the present application is a graph structure taking each grid as a node, spatial adjacency as a graph edge and time dimension as a sequence.
[0038] S4: Based on the spatio-temporal graph structure dataset, the spatial features and the temporal features of each node are extracted based on the spatio-temporal graph convolution network, and a dynamic risk occurrence feature tensor of each node at multiple time steps is determined.
[0039] Specifically, the spatio-temporal graph structure dataset is input into the spatio-temporal graph convolution network, the spatial features and the temporal features of each node in the spatio-temporal graph structure dataset are extracted by using the spatio-temporal graph convolution network, and a dynamic risk occurrence feature tensor of each node at multiple time steps is output, the dimension of the dynamic risk occurrence feature tensor is (T, N, F), wherein T is the time step, N is the number of nodes (i.e. the number of grids), and F is the feature dimension of the extracted dynamic risk occurrence feature tensor (which can be set to 64).
[0040] S5: Based on the dynamic risk occurrence feature tensor, a dynamic coupling inference model is used to determine an overall coupling weight tensor of the risk factors.
[0041] The dynamic coupling inference model includes a channel division unit, a time state coding unit and a gated recurrent unit. The time state coding unit is used to extract the time series features of the risk factors at each node based on the dynamic risk occurrence feature tensor and the spatio-temporal attention mechanism, and to determine the coupling strength between the risk factors; the gated recurrent unit is used to extract the time state representation of each risk factor at each node, and to enhance the coupling strength according to the time state representation.
[0042] S5 specifically includes: S51: The dynamic risk occurrence feature tensor is divided into multiple sub-channels by using the channel division unit; each sub-channel corresponds to a risk factor.
[0043] Specifically, each feature vector in the dynamic risk occurrence feature tensor is divided into R sub-channels, each sub-channel corresponds to a specific risk factor (such as wind speed, rainfall, bird activity frequency, bird distribution, ground population density, etc.). Assuming that the feature dimension F is equal to the sub-dimension allocated to each risk factor , then F = R d.
[0044] The time series features of the risk factors of the node can be expressed as: , wherein is the sub-channel number, which corresponds to the risk factor, , is the node number, and T is the time step.
[0045] The above process aims to retain the feature evolution information of each risk factor on the time axis, providing a basis for subsequent interaction fusion.
[0046] S52: Extract the time series features of any two risk factors at each node using the time state encoding unit, and determine the coupling strength between the risk factors.
[0047] At each node, the time series features and the time series features of any two risk factors and are extracted, respectively, and the coupling strength between the two risk factors is calculated by constructing a similarity function based on the attention mechanism. Common calculation methods include vector correlation, relative importance score, and normalized similarity measure. In this application, the coupling strength calculation formula is as follows: ; wherein, and are two different risk factors, is a normalization operation, is a nonlinear similarity function such as dot product or bilinear mapping, is the time series feature of the risk factor.
[0048] The coupling strength represents the degree of dependence or influence weight of the risk factor on the risk factor at the node . Repeat the above process, and each node gets an R×R coupling weight matrix, and the expression of the coupling weight matrix is: The coupling weight matrix includes multiple elements, and each element represents the coupling strength of the risk factor and the risk factor at the node i , and the calculation formula is: ; wherein, is the final time state hidden vector of the risk factor at the node , and is the final time state hidden vector of the risk factor at the node .
[0049] S53: Extract the time state representation of each risk factor at each node using the gated recurrent unit.
[0050] Specifically, to further enhance the expression ability of the dynamic evolution process of the risk factor, on the basis of the above, a gated recurrent unit is added to extract the time state representation of each risk factor at each node, and the time state representation of each risk factor r at the node is extracted : ; wherein, is the final time state hidden vector of the risk factor at the node , which contains the historical dynamic trend information of the risk factor, is the gated recurrent unit, and the time series feature of the risk factor.
[0051] S54: Taking the time state representation as the input of the time state encoding unit, the coupling strength between risk factors at each node is enhanced to obtain the enhanced coupling strength; Specifically, the time state representation is taken as the context input in the time state encoding unit, the coupling strength between risk factors at each node is enhanced, and the dynamic adaptability of the coupling strength is further improved, so that the dynamic coupling reasoning model can automatically adjust the influence weight between risk factors according to the historical changes.
[0052] S55: According to the enhanced coupling strength at each node, the overall coupling weight tensor of the risk factor is determined.
[0053] The above process is repeated for all nodes , and finally the overall coupling weight tensor is obtained, and its expression is: ; wherein, represents the risk factor i of the i-th node and the risk factor of the j-th node, and the risk factor coupling matrix of the risk factor , wherein the element represents the action strength of the risk factor i on at the node .
[0054] S6: Nonlinearly fusing the dynamic risk occurrence feature tensor and the overall coupling weight tensor to determine the comprehensive risk strength; For each graph node, the dynamic risk occurrence feature tensor and the overall coupling weight tensor are nonlinearly fused and calculated to output the node-level comprehensive risk strength, considering the interaction factors of the risk factors.
[0055] In the t Each time step, node i The feature vector can be decomposed into R time series features: ; in, For the first t Each time step, node i eigenvectors, For each risk factor r In the t Each time step, node i Time series features, For length is A real vector.
[0056] Define the intermediate representation vector of the overall risk intensity after interactive fusion as follows: ; in, For the first The time step, the first The intermediate representation vector of the overall risk intensity of each node, where R is the number of sub-channels. Amplification factor that is set manually or obtained through model training. For each element in the risk factor coupling matrix of node i, represents a risk factor. right The intensity of action, It is a binary interaction function. For the first The time step, the first Risk factors of each node Time series characteristics, For the first The time step, the first Risk factors of each node The time series characteristics.
[0057] S7: Determine the two-dimensional heat map of low-altitude risk based on the comprehensive risk intensity.
[0058] Specifically, the comprehensive risk intensity is mapped to the original geographic coordinate system to generate a two-dimensional heat map of low-altitude risk.
[0059] After generating a two-dimensional heatmap of low-altitude risk, the system connects to a real-time data source to periodically recalculate and update the heatmap, resulting in multiple two-dimensional heatmaps of low-altitude risk to meet the needs of UAV flight mission planning and dynamic airspace management. Each two-dimensional heatmap of low-altitude risk reflects the comprehensive airspace risk at the corresponding time step. The real-time data source is real-time low-altitude risk-related data.
[0060] In summary, by adopting the space-time graph convolution network, the application can simultaneously capture the spatial adjacency relationship and the time sequence evolution characteristics, effectively reflecting the complex spatial-time dependence between risk factors. At the same time, the dynamic coupling reasoning mechanism automatically learns the nonlinear interaction between risk factors by using attention and GRU nonlinear models, avoiding the subjective errors of traditional static linear superposition and greatly improving the authenticity and accuracy of risk assessment. The risk map of the traditional method has a low update frequency, and it is difficult to integrate real-time weather and airspace information to dynamically respond to operational needs. The application can automatically update the low-altitude risk two-dimensional thermal map according to real-time data sources, has stronger dynamic adaptability, and can effectively overcome the shortcomings of traditional methods.
[0061] In addition, the application accesses real-time data sources such as weather changes, temporary flight restrictions, etc., and through the establishment of an automatic update mechanism, the low-altitude risk two-dimensional thermal map is refreshed at a minute level. This real-time dynamic response capability ensures that the map can accurately reflect the latest comprehensive risk intensity of the airspace, meeting the actual needs of unmanned aerial vehicles and low-altitude flight missions. Further, the traditional method relies on manual determination of risk factors and setting of weights, which is a cumbersome process and requires high professional skills, and there is a subjective bias. The application reduces manual intervention and experience dependence through automatic data cleaning, spatial alignment, and standardization processing. The automatic weight adjustment of the nonlinear fusion calculation and the dynamic coupling reasoning model realizes the whole-process unmanned risk map generation, improves the work efficiency, and reduces the operation difficulty of professional personnel.
[0062] In an exemplary embodiment, a computer device, which can be a server or a terminal, is provided. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store low-altitude risk map generation data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement a low-altitude risk map generation method.
[0063] In an exemplary embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.
[0064] In an example embodiment, a computer readable storage medium storing a computer program is provided, the computer program, when executed by a processor, implements the steps of any of the above method embodiments.
[0065] In an example embodiment, a computer program product is provided, comprising a computer program, the computer program, when executed by a processor, implements the steps of any of the above method embodiments.
[0066] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.
[0067] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.
[0068] The database involved in each of the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each of the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.
[0069] The technical features of the above embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, but it should be considered that any combination of the technical features is within the scope of the present disclosure, as long as there is no contradiction.
[0070] The principles and implementation manners of the present application are described by using specific examples herein, and the above embodiments are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In summary, the content of the present description should not be understood as a limitation of the present application.
Claims
1. A method for generating a low-altitude risk map, characterized in that, The method for generating low-altitude risk maps includes: Acquire low-altitude risk-related data; low-altitude risk-related data includes geographic data, meteorological data, biological data, and social environmental data; Low-altitude risk-related data are preprocessed, and corresponding risk values are calculated; risk factor raster maps are determined based on the risk values; the preprocessing process includes spatial alignment and unit unification; risk values include: geographical risk values, meteorological risk values, biological risk values, and social and environmental risk values. A spatiotemporal graph dataset is constructed based on the risk factor raster graph; the spatiotemporal graph dataset is a graph structure with raster as nodes, spatial adjacency as graph edges, and time dimension as sequence; Based on the spatiotemporal graph structure dataset, spatial and temporal features of each node are extracted using a spatiotemporal graph convolutional network to determine the dynamic risk occurrence feature tensor of each node at multiple time steps. Based on the dynamic risk occurrence feature tensor, a dynamic coupling inference model is adopted to determine the overall coupling weight tensor of risk factors. The dynamic coupling inference model includes: a channel partitioning unit, a time state encoding unit, and a gated loop unit. The time state encoding unit is used to extract the time series features of risk factors at each node based on the dynamic risk occurrence feature tensor and a spatiotemporal attention mechanism, and to determine the coupling strength between risk factors. The gated loop unit is used to extract the time state representation of each risk factor at each node and to enhance the coupling strength based on the time state representation. The comprehensive risk intensity is determined by nonlinearly fusing the dynamic risk occurrence characteristic tensor and the overall coupling weight tensor. A two-dimensional heat map of low-altitude risk is determined based on the overall risk intensity.
2. The low-altitude risk map generation method according to claim 1, characterized in that, The process of determining a two-dimensional heat map of low-altitude risks based on comprehensive risk intensity also includes: The low-altitude risk two-dimensional heat map is periodically recalculated and updated using a real-time data source, resulting in multiple low-altitude risk two-dimensional heat maps. Each low-altitude risk two-dimensional heat map reflects the comprehensive airspace risk at the corresponding time step. The real-time data source is real-time low-altitude risk-related data.
3. The low-altitude risk map generation method according to claim 1, characterized in that, The data related to low-altitude risks are preprocessed, and the corresponding risk values are calculated. The risk factor raster is determined based on the risk value, specifically including: Geographic risk values are determined based on topography and no-fly zones; Using formula Determine meteorological risk value ; Using formula Determine biological risk value ; Using formula Determine the social and environmental risk value ; The risk factor raster map is determined by weighting and fusing geographical risk values, meteorological risk values, biological risk values, and social environmental risk values. in, This is the wind speed risk value. This represents the rainfall risk value. This represents the distance the bird travels from its current point to its takeoff or landing point. This represents the risk value for pedestrian injury or death. This represents the risk value for injury or death of occupants of the vehicle.
4. The method for generating low-altitude risk maps according to claim 1, characterized in that, The step of determining the overall coupling weight tensor of risk factors based on the dynamic risk occurrence characteristic tensor and using a dynamic coupling inference model specifically includes: The dynamic risk occurrence feature tensor is divided into multiple sub-channels using channel partitioning units; each sub-channel corresponds to a risk factor. The temporal state coding unit is used to extract the time series features of any two risk factors at each node to determine the coupling strength between the risk factors. The time state representation of each risk factor at each node is extracted using a gated loop unit; Using the time state representation as the input to the time state encoding unit, the coupling strength between risk factors at each node is enhanced to obtain the enhanced coupling strength. The overall coupling weight tensor of the risk factor is determined based on the enhanced coupling strength at each node.
5. The low-altitude risk map generation method according to claim 4, characterized in that, The step of using a time-state coding unit to extract the time-series features of any two risk factors at each node to determine the coupling strength between the risk factors specifically includes: Using formula Determine the coupling strength between risk factors ; in, and For two different risk factors, For normalization operations, It is a nonlinear similarity function. The time series characteristics of risk factors.
6. The low-altitude risk map generation method according to claim 4, characterized in that, The extraction of the temporal state representation of each risk factor at each node using a gated loop unit specifically includes: Using formula Extract the temporal state representation of each risk factor at each node; in, Risk factors At the node The final time-state latent vector, For gated loop unit, The time series characteristics of risk factors.
7. The method for generating low-altitude risk maps according to claim 1, characterized in that, The nonlinear fusion of the dynamic risk occurrence feature tensor and the overall coupling weight tensor to determine the comprehensive risk intensity specifically includes: Using formula Determine the overall risk level; in, For the first The time step, the first An intermediate representation vector of the overall risk intensity of each node. and For two different risk factors, R Number of sub-channels Amplification factor that is set manually or obtained through model training. For the node i Above, risk factors right The intensity of action, It is a binary interaction function. For the first The time step, the first Risk factors of each node Time series characteristics, For the first The time step, the first Risk factors of each node The time series characteristics.
8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the low-altitude risk map generation method according to any one of claims 1-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the low-altitude risk map generation method according to any one of claims 1-7.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the low-altitude risk map generation method according to any one of claims 1-7.