Geological disaster modeling and dynamic risk assessment system and method based on unmanned aerial vehicle group

Abstract

The invention discloses a geological disaster modeling and dynamic risk assessment system and method based on an unmanned aerial vehicle group. The assessment method comprises the following steps: acquiring geological disaster body boundary data; partitioning according to the geological disaster body boundary data, and determining a plurality of to-be-detected areas; determining a B-spline curve path of the unmanned aerial vehicle group according to each to-be-detected area; the unmanned aerial vehicle group obtains acquisition data of each to-be-detected area according to the B spline curve path; fusing the collected data through a multi-modal feature pyramid fusion algorithm, and determining to-be-evaluated data; and performing entropy evaluation on the to-be-evaluated data to determine a risk level. The method comprises the steps of dividing a to-be-detected area and designing a path of an unmanned aerial vehicle group; and then a multi-modal feature pyramid fusion algorithm and entropy evaluation are matched, the efficiency, precision and dynamic response limitations of the traditional technology are broken through through systematic innovation, the disaster early warning capability of an emergency scene is remarkably improved, and rapid and accurate data collection and risk evaluation accuracy are effectively realized.

Description

Technical Field

[0001] This invention relates to the fields of geological disaster monitoring and artificial intelligence technology, and in particular to a geological disaster modeling and dynamic risk assessment system and method based on unmanned aerial vehicle (UAV) swarms. Background Technology

[0002] Geological disasters refer to geological processes or phenomena that, under the influence of natural or human factors, cause loss of human life and property and damage to the environment. The temporal and spatial distribution patterns of geological disasters are subject to both the natural environment and human activities, and are often the result of the interaction between humans and nature.

[0003] Geological disasters are natural disasters primarily caused by geological dynamic activities or abnormal changes in the geological environment. Under the influence of internal or external forces or anthropogenic geological forces, the Earth experiences abnormal energy release, material movement, deformation and displacement of rock and soil masses, and abnormal environmental changes, which endanger human life and property, livelihoods and economic activities, or damage the resources and environment upon which humanity depends for survival and development. Existing geological disaster monitoring systems generally use fixed monitoring equipment, which is easily damaged by geological disasters, affecting the accuracy of data collection and risk assessment. Summary of the Invention

[0004] The main objective of this invention is to provide a geological disaster modeling and dynamic risk assessment system and method based on unmanned aerial vehicle (UAV) swarms, aiming to solve the problem that existing monitoring equipment is easily damaged by geological disasters, affecting the accuracy of data collection and risk assessment.

[0005] To achieve the above objectives, the present invention proposes a method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms, comprising: Obtain boundary data of geological hazard bodies; divide the geological hazard bodies into zones based on the boundary data to determine several areas to be monitored; Based on the areas to be detected, determine the B-spline curve path of the drone swarm; The drone swarm acquires data from each of the areas to be detected based on the B-spline curve path. The collected data are fused using a multimodal feature pyramid fusion algorithm to determine the data to be evaluated. The entropy value of the data to be evaluated is assessed to determine the risk level.

[0006] Preferably, the step of acquiring geological hazard boundary data, dividing the geological hazard boundary data into zones, and determining several areas to be detected includes: Obtain the boundary data of the geological hazard body; Based on the Thiessen polygons, the boundary data of the geological hazard body is partitioned to determine several initial responsibility areas; Based on the Delaunay triangle, the initial responsibility regions are adjusted to determine the regions to be detected.

[0007] Preferably, after the step of partitioning the boundary data of the geological hazard body based on Thiessen polygons to determine several initial responsibility areas, the method includes: Obtain the slope data of each of the areas to be detected, and determine whether the slope data of each of the initial responsibility areas is greater than the preset slope; When the slope data of the initial responsibility area is greater than the preset slope, the initial responsibility area where the slope data is greater than the preset slope is determined as a steep slope area, and the boundary length of the steep slope area is adjusted to be less than or equal to a first preset length. When the slope data of the initial responsibility area is less than or equal to the preset slope, the initial responsibility area with the slope data less than or equal to the preset slope is determined as a gentle area, and the side length of the gentle area is adjusted to be less than or equal to the second preset length, and the side length of the gentle area is greater than the first preset length.

[0008] Preferably, the calculation formula for the B-spline curve path is as follows: ; in, For points on the B-spline curve, A function of parameter u; Let be the B-spline basis function, i be the index of the control point, representing the basis function corresponding to the i-th control point; k be the degree of the B-spline curve. Let i be a control point, and let i be the i-th control point.

[0009] Preferably, the drone swarm includes a navigator, at least one relay aircraft, and at least one mapping aircraft.

[0010] Preferably, the formula for calculating the entropy value is as follows: ; in, The entropy value is defined as the value between 0 and 1. The normalized weight of the i-th data to be evaluated includes displacement rate, crack propagation rate and rainfall intensity threshold; n is the total number of risk parameters.

[0011] Preferably, the normalized weights of the data to be evaluated are calculated using the following formula: ; in, The weights of each parameter in the data to be evaluated are given.

[0012] Preferably, the step of performing entropy value assessment on the data to be evaluated to determine the risk level includes: The entropy value is evaluated on the data to be evaluated, and the specific value of the entropy value is determined: When the specific value of the entropy assessment is less than the first set value, the risk level is determined to be a blue warning. When the specific value of the entropy assessment is greater than or equal to the first set value and less than the second set value, the risk level is determined to be a yellow warning. When the specific value of the entropy assessment is greater than or equal to the second set value and less than the third set value, the risk level is determined to be an orange alert. When the specific value of the entropy assessment is greater than or equal to the third set value, the risk level is determined to be a red alert.

[0013] To achieve the above objectives, the present invention proposes a geological disaster modeling and dynamic risk assessment system based on unmanned aerial vehicle (UAV) swarms. The assessment system applies any of the aforementioned geological disaster modeling and dynamic risk assessment methods based on UAV swarms. The system is characterized by comprising: a calculation module, a swarm module, and a decision module, wherein the calculation module is signal-connected to the swarm module and the decision module, respectively. The calculation module is used to acquire boundary data of geological hazards; partition the geological hazard boundary data to determine several areas to be detected; determine the B-spline curve path of the UAV swarm based on each area to be detected; and fuse the acquired data through a multimodal feature pyramid fusion algorithm to determine the data to be evaluated. The cluster module is used to acquire the collected data of each of the areas to be detected according to the B-spline curve path; The decision-making module is used to evaluate the entropy value of the data to be evaluated and determine the risk level.

[0014] Compared with the prior art, the present invention has at least the following beneficial effects: By dividing the detection area and designing the path of the drone swarm, the high accuracy of the collected data is ensured. Combined with the multimodal feature pyramid fusion algorithm and entropy assessment, the rapid quantification of geological disaster risk levels is achieved. Through systematic innovation, the limitations of traditional technologies in terms of efficiency, accuracy and dynamic response are broken through, significantly improving the three-dimensional modeling and disaster early warning capabilities in emergency scenarios. This effectively realizes rapid and accurate data collection and accurate risk assessment. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating an embodiment of the geological disaster modeling and dynamic risk assessment method based on unmanned aerial vehicle (UAV) swarms of the present invention. Figure 2 The logic diagram for entropy value evaluation.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0019] The following description, with reference to the accompanying drawings, illustrates an embodiment of the present invention: a geological disaster modeling and dynamic risk assessment system and method based on unmanned aerial vehicle (UAV) swarms.

[0020] Figure 1 This is a flowchart illustrating an embodiment of the geological disaster modeling and dynamic risk assessment method based on unmanned aerial vehicle (UAV) swarms of the present invention.

[0021] Please see Figure 1 and Figure 2 To achieve the above objectives, the first embodiment of the present invention provides a method for geological disaster modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms, comprising: Step S10: Obtain boundary data of geological hazard bodies; divide the geological hazard body into zones based on the boundary data to determine several areas to be detected; Step S20: Determine the B-spline curve path of the UAV swarm based on each area to be detected; Step S30: The drone swarm acquires the collected data of each area to be detected according to the B-spline curve path; Step S40: The collected data are fused using a multimodal feature pyramid fusion algorithm to determine the data to be evaluated; Step S50: Perform entropy value assessment on the data to be evaluated to determine the risk level.

[0022] By dividing the detection area and designing the path of the drone swarm, the high accuracy of the collected data is ensured. Combined with the multimodal feature pyramid fusion algorithm and entropy assessment, the rapid quantification of geological disaster risk levels is achieved. Through systematic innovation, the limitations of traditional technologies in terms of efficiency, accuracy and dynamic response are broken through, significantly improving the three-dimensional modeling and disaster early warning capabilities in emergency scenarios. This effectively realizes rapid and accurate data collection and accurate risk assessment.

[0023] Specifically, the boundary data of geological hazard bodies is GIS data (i.e., WGS84 coordinate system). GIS data contains detailed geographical location and boundary information of geological hazard bodies. By processing this data, we can more accurately understand the scope and characteristics of the hazard bodies.

[0024] Before acquiring and collecting data, the drone swarm will synchronize the time axis and spatial coordinates.

[0025] A GPS-PPS pulse signal is used to synchronize the clocks of the entire drone swarm (error <1μs). The GPS-PPS pulse signal is a high-precision time signal; by receiving pulse signals transmitted by GPS satellites, clock synchronization across the entire drone swarm can be achieved. In cases of network latency, time deviations are fine-tuned using the NTP protocol (error <10ms). The NTP protocol is a network time synchronization protocol that fine-tunes the drone swarm's clocks to ensure accuracy in unstable network environments. For example, in a large geological disaster monitoring area, multiple drones simultaneously collect data. Through time synchronization, the data collected by each drone has the same timestamp, facilitating subsequent data processing and analysis.

[0026] Spatial registration was performed using the ICP algorithm. Coarse registration was based on LiDAR point clouds and SIFT feature points from the image (50 iterations, error <5cm). LiDAR point clouds provide 3D spatial information about geological hazards, while SIFT feature points from the image provide feature information. Preliminary spatial registration was achieved by matching LiDAR point clouds with SIFT feature points from the image. Fine registration used RANSAC to eliminate mismatched points, resulting in a final registration error ≤3cm. RANSAC is an algorithm used to estimate mathematical model parameters; it effectively eliminates mismatched points and improves registration accuracy. Synchronization of spatial coordinates aligned data from different sensors, generating a more accurate 3D model.

[0027] In the second embodiment of the present invention, based on the first embodiment, step S10 includes: Step S11: Obtain boundary data of the geological hazard body; Step S12: Based on the Thiessen polygons, the boundary data of the geological hazard body is partitioned to determine several initial responsibility areas; Step S13: Adjust each initial responsibility region based on the Delaunay triangle to determine each region to be detected.

[0028] First, a Voronoi diagram (i.e., Thiessen polygon) is generated based on the boundary vertices of the hazard body to delineate initial responsibility areas. The Voronoi diagram is a point-set-based geometric partitioning method that divides space into multiple regions, where each point within a region is closest to its corresponding generated point. In geological hazard monitoring, the Voronoi diagram can be used to divide the hazard body into multiple initial responsibility areas. Then, the Voronoi diagram is converted into Delaunay triangles. Delaunay triangles are a special triangulation method that ensures non-overlapping and blind-spot-free area coverage. This partitioning method ensures comprehensive and thorough monitoring of the entire hazard body by the UAV swarm. For example, in a large landslide hazard body, Voronoi-Delaunay mesh partitioning can divide it into multiple detection areas, allowing for separate monitoring of each area and avoiding overlapping and omissions in monitoring.

[0029] The output shows the area to be monitored by the drone swarm, annotating the coverage area and grid density of each drone swarm. This provides a more intuitive display of the monitoring range and grid density of the drone swarm, offering clear guidance for drone swarm flight and data collection.

[0030] In the third embodiment of the present invention, based on the second embodiment, after the step of partitioning the boundary data of the geological hazard body based on Thiessen polygons to determine several initial responsibility areas, the method includes: Acquire slope data for each area to be detected, and determine whether the slope data for each initial responsibility area is greater than the preset slope; When the slope data of the initial responsibility area is greater than the preset slope, the initial responsibility area with the slope data greater than the preset slope is determined as a steep slope area, and the boundary length of the steep slope area is adjusted to be less than or equal to the first preset length. When the slope data of the initial responsibility area is less than or equal to the preset slope, the initial responsibility area with slope data less than or equal to the preset slope is determined as a gentle area, and the side length of the gentle area is adjusted to be less than or equal to the second preset length, and the side length of the gentle area is greater than the first preset length.

[0031] Specifically, the preset slope is 30°; the first preset length is 20 meters; and the second preset length is 50 meters.

[0032] Different terrain conditions require different data collection methods. Steep slope areas (i.e., initial responsibility areas with a slope greater than 30°) have complex terrain and more frequent geological changes, requiring a higher data collection density. Therefore, in steep slope areas, the edge length of the grid is set to less than or equal to 20 meters. This allows for more detailed acquisition of geological information in steep slope areas and timely detection of potential geological hazards.

[0033] In flat areas (i.e., initial responsibility areas with a slope greater than or equal to 30°), the terrain is relatively stable, and the requirements for data collection are relatively lower. Setting the grid edge length to less than or equal to 50 meters in flat areas can optimize flight efficiency and reduce the flight time and energy consumption of the drone swarm while ensuring monitoring effectiveness. For example, in a geological hazard area with both steep slopes and flat areas, by dynamically adjusting the grid density, the drone swarm can rationally collect data according to the characteristics of different areas, improving monitoring efficiency and accuracy.

[0034] In the third embodiment of the present invention, based on the first embodiment, the calculation formula for the B-spline curve path is as follows: ; in, For points on the B-spline curve, A function of parameter u; Let be the B-spline basis function, i be the index of the control point, representing the basis function corresponding to the i-th control point; k be the degree of the B-spline curve. Let i be a control point, and let i be the i-th control point.

[0035] Path planning for UAV swarm data collection based on B-spline curves is a crucial step in geological disaster monitoring. Proper path planning can improve the flight efficiency of UAV swarms, reduce energy consumption, and ensure the comprehensiveness and accuracy of data collection.

[0036] Traditional path planning algorithms may generate discontinuous or uneven paths in complex terrain, leading to instability in drone swarm flight. The improved B-spline curve algorithm, however, generates smoother paths, reducing drone swarm jitter and improving flight safety. Furthermore, this algorithm has a dynamic obstacle avoidance response time of less than 50ms, enabling it to react quickly to obstacles, adjust its flight path, and avoid collisions.

[0037] Based on the generated B-spline curve path, the operational path of the UAV swarm is decomposed into a series of target points. The flight control system of the UAV swarm will use these target points as guides, adjusting the flight speed and direction to ensure that the UAV swarm flies along the B-spline curve.

[0038] In the fifth embodiment of the present invention, based on the first embodiment, the unmanned aerial vehicle swarm includes a navigator, at least one relay aircraft, and at least one mapping aircraft.

[0039] The navigator aircraft serves as the mission hub. In geological disaster monitoring missions in complex terrain, the navigator aircraft needs to possess powerful computing capabilities and precise detection equipment to achieve flight path planning and obstacle avoidance functions.

[0040] The surveying machine collects data using a spiral-lift scanning mode (vertical resolution 0.5m) and an intelligent overlap rate adjustment strategy (slope-sensitive). The spiral-lift scanning mode allows the machine to perform a comprehensive and detailed scan of the target area, improving the completeness of data acquisition. The intelligent overlap rate adjustment strategy dynamically adjusts the overlap rate based on the terrain slope, increasing it in areas with steeper slopes to improve the accuracy of modeling complex terrain. For example, on steep mountain slopes, increasing the overlap rate allows for the acquisition of more detailed information and reduces modeling errors.

[0041] To maintain communication links with the lead and mapping drones, the relay drone, flying along a B-spline curve path, adjusts its altitude and position based on communication range and signal strength requirements. For example, if the distance between the relay drone and the lead or mapping drone exceeds the communication range, the relay drone will increase its altitude to extend the communication range; when the signal strength is weak, the relay drone will adjust its flight direction to move closer to the signal source. The relay drone's flight speed is also coordinated with communication needs and the overall flight speed of the drone swarm, generally maintaining a moderate speed to ensure communication stability.

[0042] Specifically, the navigator is equipped with LiDAR for obstacle avoidance. The LiDAR scans the environment ahead in real time, acquiring distance and location information of obstacles. When an obstacle is detected, the navigator, based on the obstacle's position and size, and using a local adjustment algorithm based on B-spline curves, replans a path to bypass the obstacle. For example, an elastic waveform method can be used to adjust the path. Treating obstacles as elastic bodies, the drone's path is like a wave passing through an elastic body; obstacles affect the path, causing it to bend to avoid them. Simultaneously, the navigator must ensure the adjusted path retains a certain degree of smoothness to meet continuity requirements and reduce turbulence and vibration during flight. The presence of vegetation can affect data acquisition. Vegetation obscures the ground surface, preventing LiDAR point clouds from accurately acquiring surface information. A 77GHz millimeter-wave radar is used to penetrate the vegetation layer (thickness ≤30cm) to acquire surface deformation data. The 77GHz millimeter-wave radar has strong penetration capabilities and can penetrate a certain thickness of vegetation layer to acquire surface information beneath the vegetation.

[0043] The mapping drone employs a spiral ascent and descent scanning mode during flight. Based on the planned range of the B-spline curve path, it scans within a set pitch angle range (alternating between 30° and 60°) during ascent and descent. Pitch angle control is achieved by adjusting the drone's pitch rudder. For example, to increase the pitch angle, the drone's nose is raised, and vice versa. Simultaneously, the mapping drone's flight speed is adjusted according to terrain conditions (steep slopes and gentle slopes) on different sections of the B-spline curve path. In steep slopes, the speed is reduced to 2 m / s to ensure data acquisition accuracy; in gentle slopes, the speed is increased to 5 m / s to improve monitoring efficiency. Overlap rate control is achieved by adjusting the mapping drone's scanning frequency and flight speed. In steep slopes, the scanning frequency is increased to ensure a 40% overlap between each scanned area and the previous scanned area; in gentle slopes, the scanning frequency is reduced to maintain a 15% overlap rate to ensure data integrity and accuracy.

[0044] Specifically, during drone swarm operations, the surveying drone is positioned approximately 30 to 50 meters below and behind the navigator drone, and it performs spiral ascent and descent scanning along a B-spline curve path. This positioning of the surveying drone and the navigator drone allows the surveying drone to conduct detailed data collection on the area to be inspected within a safe passage opened up by the navigator drone, while avoiding interference with the navigator drone and ensuring the stability and accuracy of the data collection.

[0045] The relay aircraft is positioned between the navigator and the mapping aircraft. The interval between the relay aircraft and the navigator, as well as between the relay aircraft and the mapping aircraft, is maintained at about 20 to 30 meters. This allows the relay aircraft to maintain a communication link with the navigator to receive path planning instructions and obstacle avoidance information, as well as maintain real-time communication with the mapping aircraft to transmit mapping data to the ground control center and receive feedback instructions.

[0046] Positioned above, the navigator aircraft can maximize the use of its lidar for obstacle detection, paving a safe path for the entire swarm. The mapping aircraft follows closely behind but below, making full use of the safe zone provided by the navigator for detailed mapping work. Its lower position also helps maintain stable ground monitoring in complex terrain. The relay aircraft, positioned in the middle, balances communication needs, ensuring efficient information transmission between the navigator, mapping aircraft, and ground control center, avoiding signal blockage and interference, and guaranteeing the smooth completion of collaborative operations and monitoring tasks for the entire UAV swarm.

[0047] In the sixth embodiment of the present invention, based on any one of the first to fifth embodiments, the calculation formula for entropy value evaluation is as follows: ; in, The entropy value is defined as the value between 0 and 1. The normalized weight of the i-th data point to be evaluated, which includes displacement rate, fracture propagation rate, and rainfall intensity threshold. n is the total number of risk parameters.

[0048] Displacement rate can reflect the movement of geological disaster bodies, fissure propagation rate can reflect the degree of damage to geological disaster bodies, and rainfall intensity threshold can reflect the impact of rainfall on geological disasters.

[0049] In the seventh embodiment of the present invention, based on the sixth embodiment, the normalized weight calculation formula for the data to be evaluated is as follows: ; in, The weights of each parameter in the data to be evaluated.

[0050] By adjusting the weights of each parameter in the data to be evaluated using a random forest model, the weights of each parameter can be dynamically adjusted based on historical and real-time data, making the entropy value assessment more accurately reflect the risk of geological disasters. The principle behind determining weights in a random forest is to determine the importance (i.e., weight) of a feature by evaluating its contribution to the splitting of a decision tree node. A commonly used method is the Gini impurity reduction factor.

[0051] Specifically, for each decision tree, the formula for calculating the sum of the Gini impurity reductions for each feature across all split nodes is as follows: ; ΔGini=Gini 父节点 -Gini 子节点 ; in, Gini represents the importance of feature i in tree t; ΔGini represents the reduction in Gini impurity when a node splits; Gini 父节点 Gini impurity of the parent node before splitting; 子节点 The weighted Gini impurity of the child nodes after splitting; The formula for averaging and normalizing the importance of feature i across all trees is as follows: ; The final weights satisfy .

[0052] See Figure 2 In the eighth embodiment of the present invention, based on any one of the first to fifth embodiments, step S50 includes: Step S51: Evaluate the entropy value of the data to be evaluated and determine the specific value of the entropy value: Step S52: When the specific value of the entropy assessment is less than the first set value, the risk level is determined to be a blue warning. Step S53: When the specific value of the entropy value assessment is greater than or equal to the first set value and less than the second set value, the risk level is determined to be a yellow warning. Step S54: When the specific value of the entropy value assessment is greater than or equal to the second set value and less than the third set value, the risk level is determined to be an orange warning. Step S55: When the specific value of the entropy assessment is greater than or equal to the third set value, the risk level is determined to be a red alert.

[0053] Specifically, the first setting is 0.3; the second setting is 0.6; and the third setting is 0.9.

[0054] Blue Alert: When the entropy value assessment reaches 0.3, routine inspections are conducted, and a monitoring report is generated. The monitoring report details the current monitoring data and risk assessment results, providing relevant personnel with comprehensive information to facilitate further analysis and decision-making.

[0055] Yellow Alert: When the specific value of the entropy assessment is greater than or equal to 0.3 and less than 0.6, combined with the monitoring of automated equipment, real-time tracking of rainfall and geological deformation data will be pushed and issued to residents through channels such as broadcasts and WeChat groups. It is recommended to increase the frequency of patrols to once every 2 hours, and technical experts will be stationed on-site to analyze the stability of potential hazards. Temporary control measures will be implemented for construction projects, scenic spots, etc., and grassroots rescue teams will be assembled and on standby to ensure that emergency resources are in place.

[0056] Orange Alert: When the specific value of the entropy assessment is greater than or equal to 0.6 and less than 0.9, the drone will be used for re-measurement, the re-measurement path will be automatically planned and the monitoring frequency will be increased, "one-to-one" monitoring will be implemented in high-risk areas, and real-time early warning will be issued using equipment such as Beidou displacement gauges; the masses will be evacuated in an orderly manner, the hidden danger sections will be closed, and emergency broadcasts will be used to continuously broadcast evacuation instructions.

[0057] Red Alert: When the specific value of the entropy assessment is greater than or equal to 0.9, the system outputs evacuation coordinates and shortest path planning to help affected personnel quickly and safely evacuate from the danger zone. Those who refuse to evacuate will be forcibly relocated according to law, and hard isolation will be implemented in areas near slopes, cliffs, and ditches. Disaster relief supplies, medical personnel, and rescue teams will be stationed on the front line 24 hours a day, ensuring a comprehensive response to safeguard life and property.

[0058] By integrating UAV swarm collaborative control, multimodal data fusion, and entropy assessment, a comprehensive breakthrough has been achieved in geological disaster monitoring and emergency response capabilities. In terms of efficiency, the UAV swarm employs a dynamic partitioning algorithm and a spiral ascent / descending scanning mode, shortening regional modeling time, improving efficiency, and reducing packet loss rates in complex terrain. Regarding risk assessment, entropy assessment based on random forests improves early warning accuracy, reduces false alarm rates, and shortens response time. This solution maintains stable performance even under extreme weather conditions, providing end-to-end technical support for geological disaster prevention and control, encompassing "precise monitoring - intelligent early warning - rapid response."

[0059] To achieve the above objectives, the present invention also discloses a geological disaster modeling and dynamic risk assessment system based on unmanned aerial vehicle (UAV) swarms. The assessment system applies any of the above-mentioned geological disaster modeling and dynamic risk assessment methods based on UAV swarms and includes: a calculation module, a swarm module, and a decision module. The calculation module is signal-connected to the swarm module and the decision module, respectively. The calculation module is used to acquire boundary data of geological hazards; divide the geological hazard boundary data into zones to determine several areas to be detected; determine the B-spline curve path of the UAV swarm based on each area to be detected; and fuse the collected data through a multimodal feature pyramid fusion algorithm to determine the data to be evaluated. The cluster module is used to acquire the collected data of each area to be detected according to the B-spline curve path; The decision-making module is used to assess the entropy value of the data to be evaluated and determine the risk level.

[0060] The computing module utilizes the NVIDIA Jetson AGX Orin (200 TOPS AI computing power, 64GB memory) and supports high-speed PCIe 4.0 expansion. Its powerful AI computing capabilities enable rapid processing of large amounts of collected data, while the 64GB of memory can store and process complex algorithm models. The PCIe 4.0 high-speed expansion interface allows for easy connection to other devices, such as sensors and storage devices, improving system scalability.

[0061] The computation module employs statistical filtering for denoising (σ=1.5) and an improved PointNet++ algorithm. Statistical filtering effectively removes noisy points from point cloud data, improving data quality. The improved PointNet++ algorithm enhances the mean intersection-over-union (mIoU) ratio in semantic segmentation tasks, enabling more accurate identification of different objects and features in point cloud data, providing a more reliable foundation for subsequent analysis and modeling.

[0062] The decision-making module utilizes a Kubernetes cluster, with the 3D reconstruction and risk calculation modules deployed as microservices. The Kubernetes cluster features high availability, scalability, and automated management, dynamically allocating computing resources based on task requirements. Compute nodes with 16 CPU cores and 128GB of memory provide powerful computing capabilities, while Ceph distributed storage securely and efficiently stores large amounts of data. The microservice deployment breaks down the 3D reconstruction and risk calculation modules into multiple independent services, facilitating development, deployment, and maintenance, and improving system flexibility and maintainability.

[0063] The navigator in the swarm module is equipped with an Intel i7-1185G7 processor and an RTX 5000 graphics card, providing sufficient computing resources to meet the demands of complex algorithms. It integrates a laser rangefinder (accuracy ±2cm) and a 77GHz microwave radar (detection range 500m), enabling precise measurement of the distance between the drone and surrounding obstacles, providing accurate data for flight path planning and obstacle avoidance.

[0064] The surveying drone in the fleet module is based on the DJI M300RTK platform, which boasts excellent flight stability and reliability, making it suitable for surveying operations in complex terrain. Equipped with a gimbal with image stabilization and dual payload interfaces, the drone can carry various surveying equipment. It features a Riegl VUX-240 LiDAR (300 points / m² at 100m altitude, achieving ±3cm absolute accuracy with RTK-GNSS) and a Sony ILME-FR7 multispectral camera (8K resolution, covering RGB, near-infrared, and thermal infrared bands), enabling the acquisition of high-precision point cloud data and multispectral imagery, providing rich information for geological disaster monitoring.

[0065] The relay units in the drone swarm module are equipped with 5G millimeter-wave repeaters (28GHz band, bandwidth ≥200Mbps) and a dual-link redundancy design (primary 5G + backup LoRa). The 5G millimeter-wave repeater provides high-speed data transmission and meets the real-time data transmission needs between the drone swarm and the ground control center. The dual-link redundancy design improves communication reliability; when the primary link fails, the backup LoRa link can continue to operate, ensuring uninterrupted communication. Combined with a self-heating system, it achieves stable operation in environments as low as -20℃. In cold environments, the self-heating system prevents equipment damage due to low temperatures, ensuring the normal operation of the repeaters in harsh conditions.

[0066] The following is an example: The location is Mayang County, Hunan Province, at 27.74°N, 109.74°E, with a landslide risk area of ​​approximately 4.8 square kilometers.

[0067] The terrain features an average slope of 35°, a maximum slope of 65°, a vegetation coverage of 75%, and a loose deposit layer on the surface, with a thickness of 2 to 5 meters.

[0068] The triggering factor was three consecutive days of heavy rainfall (cumulative rainfall of 280 mm), which caused a sudden increase in seepage pressure on the slope.

[0069] I. Specific Configuration of the Cluster Module II. Configuration of the Calculation Module Hardware: NVIDIA Jetson AGX Orin ×2, 64GB memory, 200 TOPS computing power.

[0070] Real-time processing tasks: point cloud denoising (DBSCAN algorithm), image super-resolution reconstruction (ESRGAN network).

[0071] III. Configuration of the Decision Module Cloud platform: Alibaba Cloud ECS cluster (10 nodes), configuration: Intel Xeon 8-core CPU + NVIDIA T4 GPU×2.

[0072] Software architecture: Based on Kubernetes microservices, supporting 3D reconstruction (NeRF+Transformer) and risk entropy calculation.

[0073] The implementation process is as follows: (1) The detection area is divided by the improved Voronoi-Delaunay algorithm: the boundary length of the steep slope area (slope > 30°) is 18m, and the boundary length of the gentle area is 45m.

[0074] (2) Rapid networking of UAV swarm (5 minutes); the relay establishes a 5G communication link, with a measured bandwidth of 215Mbps and a delay of 9ms.

[0075] (3) The drone swarm follows the B-spline curve path (1.5 hours) LiDAR data: Point cloud density 312 points / m², effective point cloud coverage of 87% in vegetation-covered areas (≤40% using traditional methods). Absolute accuracy ±3cm (RTK-GNSS correction).

[0076] Multispectral image: original resolution 0.28cm, super-resolution reconstruction reaches 0.09cm (PSNR=29.5dB).

[0077] Millimeter-wave radar compensation: penetrates the vegetation layer (maximum thickness 30cm) and identifies 3 potential slip surfaces (length > 15m).

[0078] (4) The collected data are fused using a multimodal feature pyramid fusion algorithm (0.5 hours) Crack identification: Two main fractures were detected (23.5m and 18.7m in length), with a width of 8-15cm and a strike of N35°E.

[0079] (5) Entropy assessment (18 minutes) Parameter input: Entropy calculation: H(t)=-1 / ln(3)×(0.40×ln(0.40)+ 0.36×ln(0.36)+ 0.24×ln(0.24))=0.98 Warning output: Red alert, the system automatically pushes: danger zone, outputs evacuation coordinates and shortest path planning to help affected personnel quickly and safely evacuate the danger zone.

[0080] (6) Effect comparison The technological innovations are as follows: Unmanned aerial vehicle (UAV) swarm collaboration: Four surveying machines operate simultaneously, increasing efficiency by 4 times.

[0081] Multimodal fusion: LiDAR + imagery + radar penetration compensation, reducing vegetation area modeling error by 65%.

[0082] Dynamic weight adjustment: The random forest model automatically increases the rainfall weight to 0.26 (originally 0.15) based on the rainfall triggering conditions.

[0083] In the description of this specification, references are made to the terms "one embodiment", "another embodiment", "other embodiments" Descriptions such as "example" or "first embodiment to Xth embodiment" refer to descriptions in conjunction with that embodiment or example. Specific features, structures, materials, or characteristics are included in at least one embodiment or example of the present invention.

[0084] In this specification, the illustrative expressions of the terms used above do not necessarily refer to the same embodiments or examples.

[0085] Furthermore, the specific features, structures, materials, method steps, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0086] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to... This encompasses non-exclusivity inclusion, thereby allowing a process, method, article, or device to include a range of elements. The setting includes not only those elements, but also other elements not explicitly listed, or may also include... Elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0087] The sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0088] The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited thereto. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art, guided by the teachings of this invention, will apply the principles and claims of this invention without departing from its spirit and scope. Within the scope of protection, many other forms can be made, all of which fall within the protection scope of this invention.

Claims

1. A method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms, characterized in that, include: Obtain boundary data of geological hazard bodies; Based on the boundary data of the geological hazard body, several areas to be monitored are identified by dividing the area into zones. Based on the areas to be detected, determine the B-spline curve path of the drone swarm; The drone swarm acquires data from each of the areas to be detected based on the B-spline curve path. The collected data are fused using a multimodal feature pyramid fusion algorithm to determine the data to be evaluated. The entropy value of the data to be evaluated is assessed to determine the risk level.

2. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that, The steps of acquiring geological hazard boundary data, dividing the geological hazard boundary data into zones, and determining several areas to be detected include: Obtain the boundary data of the geological hazard body; Based on the Thiessen polygons, the boundary data of the geological hazard body is partitioned to determine several initial responsibility areas; Based on the Delaunay triangle, the initial responsibility regions are adjusted to determine the regions to be detected.

3. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in claim 2, characterized in that, After the step of partitioning the boundary data of the geological hazard body based on Thiessen polygons to determine several initial responsibility areas, the following steps are included: Obtain the slope data of each of the areas to be detected, and determine whether the slope data of each of the initial responsibility areas is greater than the preset slope; When the slope data of the initial responsibility area is greater than the preset slope, the initial responsibility area where the slope data is greater than the preset slope is determined as a steep slope area, and the boundary length of the steep slope area is adjusted to be less than or equal to a first preset length. When the slope data of the initial responsibility area is less than or equal to the preset slope, the initial responsibility area with the slope data less than or equal to the preset slope is determined as a gentle area, and the side length of the gentle area is adjusted to be less than or equal to the second preset length, and the side length of the gentle area is greater than the first preset length.

4. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that, The formula for calculating the B-spline curve path is as follows: ； in, For points on the B-spline curve, A function of parameter u; Let be the B-spline basis function, i be the index of the control point, representing the basis function corresponding to the i-th control point; k be the degree of the B-spline curve. Let i be a control point, and let i be the i-th control point.

5. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that, The drone swarm includes a navigator, at least one relay drone, and at least one mapping drone.

6. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in any one of claims 1-5, characterized in that, The formula for calculating the entropy value is as follows: ； in, The entropy value is a number that ranges from 0 to 1. The normalized weight of the i-th data to be evaluated includes displacement rate, crack propagation rate and rainfall intensity threshold; n is the total number of risk parameters.

7. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in claim 6, characterized in that, The formula for calculating the normalized weights of the data to be evaluated is as follows: ； in, The weights of each parameter in the data to be evaluated are given.

8. The method for geological hazard modeling and dynamic risk assessment based on unmanned aerial vehicle (UAV) swarms as described in any one of claims 1-5, characterized in that, The step of performing entropy value assessment on the data to be evaluated to determine the risk level includes: The entropy value is evaluated on the data to be evaluated, and the specific value of the entropy value is determined: When the specific value of the entropy assessment is less than the first set value, the risk level is determined to be a blue warning. When the specific value of the entropy assessment is greater than or equal to the first set value and less than the second set value, the risk level is determined to be a yellow warning. When the specific value of the entropy assessment is greater than or equal to the second set value and less than the third set value, the risk level is determined to be an orange alert. When the specific value of the entropy assessment is greater than or equal to the third set value, the risk level is determined to be a red alert.

9. A geological hazard modeling and dynamic risk assessment system based on unmanned aerial vehicle (UAV) swarms, wherein the assessment system applies the geological hazard modeling and dynamic risk assessment method based on UAV swarms as described in any one of claims 1-8, characterized in that, include: The system comprises a computing module, a cluster module, and a decision module, wherein the computing module is signal-connected to the cluster module and the decision module, respectively. The calculation module is used to acquire boundary data of geological hazards; partition the geological hazard boundary data to determine several areas to be detected; determine the B-spline curve path of the UAV swarm based on each area to be detected; and fuse the acquired data through a multimodal feature pyramid fusion algorithm to determine the data to be evaluated. The cluster module is used to acquire the collected data of each of the areas to be detected according to the B-spline curve path; The decision-making module is used to evaluate the entropy value of the data to be evaluated and determine the risk level.

Images