UAV operation risk warning method based on enhanced SORA model and AIGC generated map

Abstract

The present invention discloses a UAV operation risk warning method based on an enhanced SORA model and an AIGC generated map. The real-time value of the evaluation index of the UAV operation risk assessment is calculated according to real-time data, and the real-time risk coefficient is calculated in combination with the corresponding fixed risk coefficient. The regional UAV operation dynamic risk value is calculated according to the artificial potential field method; the operation risk map of a single UAV based on the enhanced SORA model is generated according to the risk value, time and space coordinates; the operation risk maps of all UAVs in the region are superimposed, and the UAV operation risk map is generated based on AIGC. According to the status of the real-time UAV operation risk map, the UAV operation safety warning is performed. By integrating multi-dimensional inputs such as spatiotemporal data and real-time information of UAV operation, the regional UAV operation dynamic risk value is calculated, the dynamic response capability of the SORA assessment model to risks is enhanced, and the granularity and comprehensiveness of UAV risk assessment are improved.

Description

Technical Field

[0001] The present invention relates to unmanned aerial vehicle (UAV) low-altitude services, and in particular to an unmanned aerial vehicle (UAV) operation risk early warning method based on an enhanced SORA model and an AIGC generated map. Background Art

[0002] In terms of low-altitude UAV risk assessment technology, the Specific Operational Risk Assessment (SORA) method is a set of standardized methods proposed by the Civil Aviation Administration of China for assessing the operational risks of specific types of UAVs, mainly including ground risk and air risk assessment. Its main purpose is to ensure the safety of operations through systematic assessments in the trial operation management of UAVs. The SORA assessment process includes pre-application assessment, operational concept description, ground risk level assessment, air risk level assessment, and specific assurance and integrity level assessment. The SORA model can calculate the operational risks of UAVs and provide data support for the establishment of UAV risk maps.

[0003] However, although the SORA model calculates ground risks and air risks, due to the limitations of its review mechanism, its risk level is actually semi-static data and cannot reflect the changes in risks during the operation of drones. The indicators used in risk assessment are also relatively rough and lack consideration of more dynamic factors, resulting in relatively low model accuracy.

[0004] In terms of low-altitude drone risk map technology, mainstream low-altitude drone risk maps can help pilots avoid high-risk areas and reduce the probability of accidents; provide data support for the planning of drone flight missions, and help pilots or automatic systems make better flight decisions.

[0005] Although there have been some studies on the construction of drone risk maps, the relevant technologies still have room for improvement. For example, the current drone risk maps lack real-time performance, and the frequency of map updates of population density, airspace data and other data is low. Its evaluation mechanism does not include the dynamic operation data of the drone group, so the risk map can only reflect the relatively static drone flight risk status that is vaguely applicable to a limited number of drones. In addition, the existing drone system warning mechanism is also relatively simple, usually based on static rules, not flexible enough, and cannot dynamically adjust the warning information according to real-time data. Summary of the invention

[0006] Purpose of the invention: In view of the above shortcomings, the present invention provides a UAV operation risk warning method based on an enhanced SORA model and an AIGC generated map to improve UAV risk assessment.

[0007] Technical solution: To solve the above problems, the present invention adopts a UAV operation risk warning method based on an enhanced SORA model and an AIGC generated map, comprising the following steps:

[0008] (1) Construct an enhanced SORA risk assessment model for drone operations and calculate the dynamic risk value of regional drone operations; specifically,

[0009] Determine the evaluation indicators for drone operation risk assessment and the fixed risk factors corresponding to the evaluation indicators, obtain real-time data related to drone operation, calculate the real-time values ​​of the evaluation indicators based on the acquired real-time data, calculate the real-time risk factors of each evaluation indicator through the real-time values ​​of the evaluation indicators and the fixed risk factors corresponding to the evaluation indicators, and obtain the total real-time dynamic risk factor of the drone based on the sum of the real-time risk factors of each evaluation indicator. According to the principle of artificial potential field method, the total real-time dynamic risk factor of UAV is is the proportionality coefficient, and the dynamic risk value of regional drone operation is calculated;

[0010] (2) Generate an operational risk map of a single UAV based on the enhanced SORA model according to the regional UAV operational dynamic risk value and the corresponding time and space coordinates of the UAV;

[0011] (3) Based on the spatiotemporal information, align the latitude, longitude and altitude information of the map, overlay the operational risk maps of all drones in the area, and obtain a fused airspace drone risk map;

[0012] (4) Automatically generate a drone operation risk map based on AIGC. The two-dimensional image of the fused airspace drone risk map combined with spatiotemporal information is used as input, and a multi-view diffusion model and ISOMER algorithm are used to generate a drone operation risk map.

[0013] (5) Provide drone operation safety warnings based on the status of the real-time drone operation risk map in the mission area.

[0014] Furthermore, the evaluation indicators of the UAV operation risk assessment in step (1) include UAV equipment characteristics, UAV operation characteristics, ground information, obstacle data, air traffic information, weather data and smoke events.

[0015] Furthermore, the operation characteristics and time data of the drone are obtained through the drone sensor equipment; the air traffic information is obtained through the drone management system; the real-time positioning, images and videos uploaded by the drone are used to perform geographic information retrieval and analysis, image recognition and analysis, and obtain ground information, obstacle data, weather data and smoke events.

[0016] Furthermore, the calculation formula for the dynamic risk value of drone operation in the area is:

[0017] ;

[0018] ;

[0019] in, Indicates the drone test point The sum of the repulsive forces in the area is the dynamic risk value of regional drone operations; Indicates the maximum value of the dynamic risk value of regional drone operations, Indicates that the drone test point is affected by the The repulsive force of a regional point is calculated as:

[0020] ;

[0021] in, Indicates the drone test point. Indicates the area Regional points, Indicates the drone test point and regional points The distance between represents the partial derivative vector with respect to distance, Indicates the radius of the area of ​​influence.

[0022] Furthermore, the layers of the operation risk map include a population density layer, an obstacle and shielding factor layer, and a regional high-precision map layer. The population density layer includes time data and ground information. The obstacle and shielding factor layer includes regional drone operation dynamic risk values, air traffic information, obstacle information, drone equipment information, and drone operation information. The regional high-precision map layer includes administrative area divisions and ground information.

[0023] Furthermore, the specific steps of automatically generating the UAV operation risk map based on AIGC in step (4) include:

[0024] (4.1) Using a multi-view diffusion model to generate four orthogonal view images and their corresponding normal images from a single input image;

[0025] (4.2) Use the ControlNet-Tile network for upsampling to improve the resolution of the images of the four orthogonal views and generate high-resolution images;

[0026] (4.3) The four high-resolution images and the normal map are input into the 3D mesh reconstruction algorithm ISOMER algorithm to reconstruct the UAV operation risk map.

[0027] Furthermore, the 3D mesh reconstruction algorithm in step (4.3) includes: initial mesh estimation, the four orthogonal views include a front view and a back view, the initial mesh structure is directly estimated from the front view and the back view, and the depth map is derived from the normal map in the front view:

[0028] ;

[0029] in, Indicates pixel position The depth value at represents the normal vector at pixel position x, Represents the component of the normal vector at pixel position t in the x-axis direction;

[0030] The loss function used for:

[0031] ;

[0032] in, is a mask-based loss used to adjust the mesh outline; is the normal-based loss.

[0033] Furthermore, a drone full-process management early warning mechanism based on the drone operation risk map is established. For each early warning judgment node, the real-time value is compared with the predefined early warning threshold, the alarm level is determined and the alarm information is automatically generated. The alarm information includes: alarm number, alarm name, alarm type, risk level, alarm description, drone information, ground information, mission information, route information, operator information, real-life photos or videos, operation instructions, and control measures.

[0034] Furthermore, the early warning judgment nodes include: when the drone is put into storage management, when the drone receives a mission, before the drone takes off to perform a mission, during the drone’s mission, and during the drone’s return to storage management stage;

[0035] When the drone is put into storage for management, it is determined whether the drone has a stable heartbeat signal reported;

[0036] When the UAV receives a task, it determines whether the performance indicators of the UAV meet the task execution standards;

[0037] Before the UAV takes off to perform a mission, it is determined whether the UAV status, route environment, and airport status meet the take-off criteria;

[0038] During the execution of the mission, the drone is judged whether the take-off state is normal, whether the return state is normal, whether the operation risk value exceeds the threshold, whether it enters the fence, and whether the heartbeat is disconnected;

[0039] During the drone return and storage management phase, it is determined whether the drone has a stable reported heartbeat signal.

[0040] The present invention also adopts a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0041] The present invention also adopts a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0042] Beneficial effects: Compared with the prior art, the significant advantage of the present invention is that it integrates multi-dimensional inputs such as spatiotemporal data and real-time information on drone operations, calculates the dynamic risk value of regional drone operations, enhances the dynamic response capability of the SORA assessment model to risks, improves the granularity and comprehensiveness of drone risk assessment, provides data support for autonomous decision-making by drones in complex environments, lays the foundation for automatically generating drone operation risk maps, and greatly improves the safe operation capability of drones in complex low-altitude airspace, thereby further assisting drone-related regulators and operators such as the bureau, drone operators, and air traffic management to better monitor and manage low-altitude traffic. By integrating more detailed real-time data, such as ground information, obstacles, air traffic, and weather conditions, potential risk points can be effectively identified and predicted, thereby providing more accurate and dynamic risk assessments.

[0043] By adopting drone operation risk assessment technology and AIGC technology, an enhanced SORA risk assessment model indicator system for drone operation is constructed. A drone risk map algorithm based on fused airspace is proposed, and an AIGC-based risk map model is automatically generated. A drone full-process management and early warning mechanism based on risk maps is established to dynamically assess drone flight risks in real time and automatically generate warning information in a timely manner, helping drone pilots, operators, safety regulators and other stakeholders to understand the real-time risks of low-altitude flights through real-time risk maps and better manage flight missions.

[0044] By integrating multi-UAV flight data and air traffic planning information, a new three-dimensional simulation fusion risk map model is generated. The fusion map model can reflect the distribution of operational risks of all managed UAVs in real time, which is of great significance to the operation and management of UAV groups. Operators can monitor and adjust the flight paths of UAVs in real time according to the map to avoid high-risk areas and ensure the safety and efficiency of group flights, which also provides strong support for the application of UAVs in emergency management and disaster response.

[0045] The risk map-based low-altitude drone full-process management warning mechanism significantly improves the active safety and response speed of drones. Compared with the traditional fence warning method, this mechanism automatically judges and generates drone warning information in real time through dynamic risk map data, reducing the warning response time. This enables drones to identify risks and make adjustments more quickly during missions, greatly improving the drone's autonomous flight capabilities in complex environments. Through this warning mechanism, the safety of drones has been substantially enhanced, providing a strong security guarantee for the commercial application and widespread promotion of drone technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Shown is a schematic diagram of the flow chart of the UAV operation risk warning method in the present invention.

[0047] Figure 2 Shown is a schematic diagram of the enhanced SORA risk assessment framework for UAV operations in the present invention.

[0048] Figure 3 The figure is a flow chart of the full-process management and early warning mechanism of low-altitude UAV in the present invention. DETAILED DESCRIPTION

[0049] like Figure 1 As shown, a UAV operation risk warning method based on an enhanced SORA model and an AIGC generated map in this embodiment includes the following steps:

[0050] Step 1: Establish a risk assessment model for UAV operation enhanced SORA. The model consists of three parts: UAV operation enhanced SORA risk perception (hereinafter referred to as risk perception), UAV operation enhanced SORA risk assessment indicator system (hereinafter referred to as risk assessment indicator system) and UAV operation enhanced SORA risk assessment model (hereinafter referred to as risk assessment model). The specific model framework is as follows: Figure 2 As shown. The risk perception module extracts drone operation-related data, images and videos from the drone perception system, drone management system and high-precision map respectively, and selects the effective information required for risk assessment and passes it to the risk assessment index system module for use. The risk assessment index system module uses the data transmitted by risk perception to calculate the index value, and then calculates the real-time risk coefficient of each indicator by comparing the index value with the corresponding fixed risk coefficient. Among them, the module will first disassemble, recognize and analyze the image and video data that cannot be used directly, and then calculate the index value and compare it with the coefficient. The risk assessment model is mainly based on the principle of artificial potential field method, with risk assessment indicators and their risk coefficients as input to calculate the dynamic risk value of regional drone operations. For the formula, please refer to the corresponding steps.

[0051] Compared with the original SORA model, the enhanced SORA model improves the original SORA model by adding time data, ground information, obstacle data, air traffic information, weather data, smoke events and other airspace and drone real-time operation data as risk representation, calculating the dynamic risk value of drone operation, and incorporating artificial potential field method, etc., to improve the objectivity and accuracy of risk data. With the calculation of the dynamic risk value of drone operation based on the enhanced SORA model, the granularity of drone operation risk calculation can be effectively improved, making up for the lack of real-time and detailed considerations of the traditional SORA model.

[0052] Step 2: Based on the actual characteristics of low-altitude airspace and the actual needs of drone flights, a set of drone operation enhanced SORA risk assessment indicator system is constructed. The first-level indicators include drone equipment characteristics, drone operation characteristics, time data, ground information, obstacle data, air traffic information, weather data and smoke events. The specific content of the indicator system is shown in Table 1.

[0053] Table 1 UAV operation enhanced SORA risk assessment indicators:

[0054]

[0055] Step 3: Obtain / calculate the enhanced SORA risk assessment index values ​​of drone operations in all regions, and further refer to the principle of the artificial potential field method to calculate the dynamic risk value of regional drone operations.

[0056] Step 3-1: Calculate the value of the enhanced SORA risk assessment index for low-altitude drone operations and perform 0-1 scaling. Use drone sensor equipment to obtain drone operation characteristics and time data. Obtain air traffic information through the drone management system. Perform geographic information retrieval and analysis, image recognition and analysis through real-time positioning, images and videos uploaded by drones to obtain ground information, obstacle data, weather data and smoke events. UTC data and latitude and longitude data are not included in the risk index calculation and are reserved for the automatic generation of risk maps in step 4.

[0057] Step 3-2: Based on the real-time values ​​and fixed risk factors of the UAV operation enhanced SORA risk assessment indicators, the real-time risk factors of each assessment indicator are calculated. The total real-time dynamic risk factor of a single UAV is calculated based on the sum of the real-time risk factors of each assessment indicator. .in The higher the risk factor, the greater the risk threat to the safety of drone operations. The dynamic risk factor of drone operations is set at a minimum of 0 and a maximum of 10.

[0058] Table 2 Fixed hazard coefficients and impact ranges of UAV operation enhanced SORA risk assessment indicators:

[0059] Index (0-1) Fixed risk factor (0-10) Impact Maximum feature size 5 2 Typical kinetic energy expected 5 5 Takeoff weight 5 2 Viewing distance 10 9 Relative height 10 9 speed 5 9 Latitude and longitude 0 0 UTC time 0 0 Population concentration 10 9 Obstacle concentration 10 9 Geofence Airspace Type 10 9 Route congestion 10 9 Rainfall 10 9 Wind speed 5 9 Smog Index 5 9

[0060] Step 3-3: Refer to the principle of artificial potential field method and use the total real-time dynamic risk coefficient of the UAV operation risk assessment index As input, the dynamic risk of regional drone operations is calculated.

[0061] (1) Calculate the size of the repulsive field generated by the region on the surrounding area, that is, the potential energy of the regional repulsive field. The calculation formula is as follows:

[0062] ;

[0063] in, is the regional risk factor. is the total real-time dynamic risk factor of the UAV, which is determined by the total real-time risk factor of the UAV test point obtained in step 3-2. is the distance between the point to be measured and the regional point. Indicates the radius of the area of ​​influence, which is determined by the area of ​​influence. When , the influence of the region on the tested point is 0.

[0064] (2) Calculate the repulsive force exerted on the drone’s test point by the regional points. The calculation formula is as follows:

[0065] ;

[0066] in, Refers to the drone test point. Repulsion, is the distance between the drone’s test point and the regional point, Represents the partial derivative vector with respect to distance.

[0067] 3) Calculate the dynamic risk value of the drone, that is, calculate the sum of the repulsive forces on the drone's test point by addition and superposition. The calculation formula is as follows:

[0068] ;

[0069] in, Indicates the drone test point The combined repulsive force, is 10, indicating The risk range is 0-10. When it is equal to 10, the risk of drone operation in the area is the greatest, and drones are advised to avoid the area; when When When the drone is at the second highest risk level; when When the drone is at a general risk level; when When the drone is at a low risk level; when When it is equal to 0, there is no risk of drone operation in the area and drones can fly safely in the area.

[0070] Step 4: Use the drone risk map algorithm based on fused airspace to generate a multi-layer fused airspace drone risk map. The risk map layers are divided into population density layer, obstacle layer and shielding layer, which comprehensively describe the risk distribution in the airspace.

[0071] Step 4-1: Single UAV risk map. Based on the enhanced SORA risk assessment model in step 3, the equipment characteristics, operation characteristics, time data, obstacle information, air traffic information, weather data and smoke events of a single UAV are superimposed to calculate the risk value of a single UAV. According to the dynamic risk value and its corresponding time and space coordinates, a single UAV operation risk map based on enhanced SORA is generated. The regional population density distribution layer contains spatiotemporal information (time data, longitude and latitude, and relative altitude) and ground information, the obstacle and shielding factor layer contains UAV operation risks, air traffic information, obstacle information, UAV equipment information, operation information, etc. The regional high-precision map layer mainly depicts administrative divisions and ground information data.

[0072] Step 4-2: Fusion of airspace drone risk map. Based on the spatiotemporal information (time data, longitude and latitude, and relative altitude), align the longitude and latitude altitude information of the map, and superimpose the risk maps of all drones in the area to obtain a fused airspace drone risk map.

[0073] Step 5: Automatically generate a UAV operation risk map model based on AIGC. Figure 2 The bitmap image is combined with spatiotemporal information as input, and the multi-view diffusion model and ISOMER algorithm are used to quickly generate high-fidelity risk maps. Figure 3 dimensional model.

[0074] Step 5-1: Multi-view image generation and normal map prediction: Use a model based on the Stable Diffusion ImageVariation architecture and modify class embedding and training parameters.

[0075] Class Embedding: Enter an integer (0 to 3) indicating the view index.

[0076] Process four views simultaneously and achieve multi-view by concatenating them in a self-attention layer Figure 1 Consistency. Training parameters include learning rate Etc. Generate four orthogonal view images and their corresponding normal maps from a single input image.

[0077] Step 5-2: High-resolution multi-view image generation. Upsampling is performed using the ControlNet-Tile network, which is based on StableDiffusion 1.5 and enhanced by IP-Adapter. Training parameters include learning rate , batch size 128, etc. Upscaling the low-resolution multi-view image to high resolution. Using the four low-resolution orthogonal view images generated in step 5-1 as input, a high-resolution multi-view image is automatically generated.

[0078] Step 5-3: 3D mesh reconstruction. Using the 3D mesh reconstruction algorithm ISOMER, we integrate color and geometric prior information. Input the high-resolution multi-view RGB image and normal image obtained in the above steps to reconstruct a high-quality fused airspace UAV operation risk map. Figure 3 D grid. Reconstruction of drone operation risk areas Figure 3 The key contents of the dimensional grid are as follows.

[0079] (1) Initial grid estimation:

[0080] The multi-view image includes the front view and the back view, and the initial grid structure is directly estimated from the front view and the back view. The key here is to derive the depth map from the normal map in the front view:

[0081] ;

[0082] in Indicates pixel position The depth value at represents the normal vector at pixel position x, Represents the component of the normal vector at pixel position t in the x-axis direction.

[0083] (2) Reconstruct the loss function of mask-based loss and normal-based loss:

[0084] These loss functions are used to optimize the generated fused airspace UAV operation risk map. Figure 3 D model to make it consistent with the input image. Mask-based loss Used to adjust mesh outlines. Normal-based loss , the final reconstruction loss function Combining these two loss terms:

[0085] ;

[0086] Step 6: Establish a low-altitude UAV full-process management warning mechanism based on the risk map. For each warning judgment node, compare the real-time value with the predefined warning threshold, determine the warning level and automatically generate warning information. During the UAV mission, monitor the risk map status of the UAV mission area. If red appears in the area, it means that the dynamic risk value of the UAV operation in the area is greater than the predefined warning threshold, and a UAV operation safety alarm is issued.

[0087] Drone alarm information includes but is not limited to: alarm number, alarm name, alarm type, risk level, alarm description, drone information (including drone CPN code, basic information, remaining battery power, etc.), ground information (including administrative district, street, specific location, longitude, latitude, relative altitude, interstellar altitude, risk radius, etc.), mission information, route information, operator information, real-life photos / videos, operating instructions, control measures, etc.

[0088] Figure 3 This is the workflow of the low-altitude UAV early warning mechanism, showing the early warning judgment nodes and corresponding treatment measures that may be encountered in the entire process of UAV receiving tasks, safety inspections, executing tasks, returning, and entering the warehouse. The early warning mechanism for the entire process of UAV management is briefly introduced as follows:

[0089] (1) When managing the drone in storage, if there is a stable heartbeat signal reported, the normal storage management phase will be entered; otherwise, a disconnection alarm message will be generated.

[0090] (2) When a drone receives a mission, its performance indicators such as endurance must meet the mission execution inspection standards before it can receive the mission; if it fails the inspection, an inspection abnormality alarm message is generated.

[0091] (3) Before a drone takes off to perform a mission, a takeoff safety check is required. The drone's status, route environment, airport status and other indicators are checked to see if they meet the takeoff standards. If not, an inspection abnormality warning message is generated; otherwise, takeoff is allowed.

[0092] (4) During the mission, it is necessary to take off and land on time and provide flight information reports. If there is a takeoff abnormality, return abnormality, operation safety risk value exceeding the threshold, entering the fence, heartbeat disconnection or other abnormal conditions, the execution abnormality alarm / operation safety alarm / fence alarm / disconnection alarm will be triggered.

[0093] (5) During the return and storage management phase, when all operations are completed, the drone is sent back to the warehouse for storage management. During the storage period, if there is a stable heartbeat signal reported, it will be stored and managed normally; otherwise, a disconnection alarm message will be generated.

[0094] During the whole process, problems may occur at each step, causing the drone to be unable to continue. Therefore, corresponding monitoring methods will be set up at each key node to ensure that the entire drone operation can be carried out smoothly, and corresponding configuration adjustments will be made according to the actual situation to ensure that the drone system operates stably and reliably.

Claims

1. A UAV operation risk warning method based on enhanced SORA model and AIGC generated map, characterized in that: The following steps are involved: (1) Construct an enhanced SORA risk assessment model for drone operations and calculate the dynamic risk value of regional drone operations; specifically, Determine the evaluation indicators for drone operation risk assessment and the fixed risk factors corresponding to the evaluation indicators, obtain real-time data related to drone operation, calculate the real-time values ​​of the evaluation indicators based on the acquired real-time data, calculate the real-time risk factors of each evaluation indicator through the real-time values ​​of the evaluation indicators and the fixed risk factors corresponding to the evaluation indicators, and obtain the total real-time dynamic risk factor of the drone based on the sum of the real-time risk factors of each evaluation indicator. According to the principle of artificial potential field method, the total real-time dynamic risk factor of UAV is is the proportionality coefficient, and the dynamic risk value of regional drone operation is calculated; The evaluation indicators for drone operation risk assessment include drone equipment characteristics, drone operation characteristics, ground information, obstacle data, air traffic information, weather data, and smoke events; (2) Generate an operational risk map of a single UAV based on the enhanced SORA model according to the regional UAV operational dynamic risk value and the corresponding time and space coordinates of the UAV; (3) Based on the spatiotemporal information, align the latitude, longitude and altitude information of the map, overlay the operational risk maps of all drones in the area, and obtain a fused airspace drone risk map; (4) Automatically generate a drone operation risk map based on AIGC. The two-dimensional image of the fused airspace drone risk map combined with spatiotemporal information is used as input, and a multi-view diffusion model and ISOMER algorithm are used to generate a drone operation risk map. (5) Provide drone operation safety warnings based on the status of the real-time drone operation risk map in the mission area; The calculation formula for the dynamic risk value of drone operation in the area is: ； ； in, Indicates the drone test point The sum of the repulsive forces in the area is the dynamic risk value of regional drone operations; Indicates the maximum value of the dynamic risk value of regional drone operations, Indicates that the drone test point is affected by the The repulsive force of a regional point is calculated as: ； in, Indicates the drone test point. Indicates the area Regional points, Indicates the drone test point and regional points The distance between represents the partial derivative vector with respect to distance, Indicates the radius of the area of ​​influence; The specific steps of step (4) for automatically generating a drone operation risk map based on AIGC include: (4.1) Using a multi-view diffusion model to generate four orthogonal view images and their corresponding normal images from a single input image; (4.2) Use the ControlNet-Tile network for upsampling to improve the resolution of the images of the four orthogonal views and generate high-resolution images; (4.3) Input the four high-resolution images and the normal map into the 3D mesh reconstruction algorithm ISOMER algorithm to reconstruct the UAV operation risk map; The 3D mesh reconstruction algorithm in step (4.3) includes: initial mesh estimation, four orthogonal views including a front view and a back view, directly estimating the initial mesh structure from the front view and the back view, and deriving the depth map from the normal map in the front view: ； in, Indicates pixel position The depth value at represents the normal vector at pixel position x, Represents the component of the normal vector at pixel position t in the x-axis direction; The loss function used for: ； in, is a mask-based loss used to adjust the mesh outline; is the normal-based loss.

2. The UAV operation risk warning method according to claim 1 is characterized in that: In step (1), the UAV operation characteristics and time data are obtained through the UAV sensor equipment; the air traffic information is obtained through the UAV management system; the real-time positioning, images and videos uploaded by the UAV are used to perform geographic information retrieval and analysis, image recognition and analysis, and obtain ground information, obstacle data, weather data and smoke events.

3. The UAV operation risk warning method according to claim 2 is characterized in that: The layers of the operation risk map include a population density layer, an obstacle and shielding factor layer, and a regional high-precision map layer. The population density layer includes time data and ground information. The obstacle and shielding factor layer includes regional drone operation dynamic risk values, air traffic information, obstacle information, drone equipment information, and drone operation information. The regional high-precision map layer includes administrative area divisions and ground information.

4. The UAV operation risk warning method according to claim 3 is characterized in that: Establish a drone full-process management early warning mechanism based on the drone operation risk map. For each early warning judgment node, compare the real-time value with the predefined early warning threshold, judge the alarm level and automatically generate alarm information. The alarm information includes: alarm number, alarm name, alarm type, risk level, alarm description, drone information, ground information, mission information, route information, operator information, real-life photos or videos, operation instructions, and control measures.

5. The UAV operation risk warning method according to claim 4 is characterized in that: The early warning judgment nodes include: when the drone is put into storage management, when the drone receives a mission, before the drone takes off to perform a mission, during the drone’s mission, and during the drone’s return to storage management stage; When the drone is put into storage for management, it is determined whether the drone has a stable heartbeat signal reported; When the UAV receives a task, it determines whether the performance indicators of the UAV meet the task execution standards; Before the UAV takes off to perform a mission, it is determined whether the UAV status, route environment, and airport status meet the take-off criteria; During the execution of the mission, the drone is judged whether the take-off state is normal, whether the return state is normal, whether the operation risk value exceeds the threshold, whether it enters the fence, and whether the heartbeat is disconnected; During the drone return and storage management phase, it is determined whether the drone has a stable reported heartbeat signal.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images