Unmanned aerial vehicle monitoring method and system for dam burst scene based on artificial intelligence

Abstract

The invention relates to a dam burst scene unmanned aerial vehicle monitoring method and system based on artificial intelligence. The method comprises the steps that hydrological data, meteorological data, satellite image data and topographic map data are acquired and input into a monitoring area identification artificial intelligence model, and target monitoring area risk level information and target monitoring area detection range information are generated; distributing the number of unmanned aerial vehicles and a preset route of each unmanned aerial vehicle based on the target monitoring area risk level information and the target monitoring area detection range information; controlling each unmanned aerial vehicle to fly along a corresponding preset route, and acquiring target monitoring data of each unmanned aerial vehicle; and inputting the target monitoring data of each unmanned aerial vehicle into the dam bursting and breaking scene analysis artificial intelligence model to generate a dam bursting and breaking scene analysis result. By adopting the method, high efficiency, accuracy and real-time performance of dam burst monitoring can be realized through multi-source data driving, dynamic task allocation and artificial intelligence model optimization.

Description

Technical Field

[0001] The present invention belongs to the field of flood monitoring, and particularly relates to a method and system for monitoring levee and dam breaches by drones based on artificial intelligence. Background Art

[0002] Levee and dam breach disasters are characterized by strong suddenness and great destructiveness, posing a serious threat to the safety of people's lives and property. Traditional manual inspection methods are inefficient and have limited coverage, making it difficult to meet the real-time monitoring requirements in complex scenarios. With the rapid development of drone technology, drones, with their high mobility and multi-angle observation capabilities, have become an important means for levee and dam breach monitoring, and intelligent monitoring technology based on drones has gradually been applied in the field of water conservancy project safety.

[0003] In traditional technologies, drone monitoring systems usually use fixed-route inspections and cannot dynamically adjust monitoring strategies according to real-time risks, resulting in problems such as missed inspections in high-risk areas and redundant monitoring in low-risk areas. Moreover, traditional drone monitoring methods usually require manual remote control operations, which have high requirements for the skills and experience of operators. This not only increases labor costs and operation risks, but also fails to meet the rapid response requirements in the face of emergencies.

[0004] In the prior art with the authorization announcement number CN110044338B, it is necessary to first detect by a fixed-wing drone flying along a first preset route before generating a specific monitoring plan for levee and dam breach scenarios by drones. Moreover, it is difficult to predict unoccurred levee and dam breaches in levee and dam breach scenarios, and there are defects such as slow response speed, weak end-cloud collaboration and anomaly recognition capabilities, and poor system robustness. Summary of the Invention

[0005] Based on this, it is necessary to provide a method and system for monitoring levee and dam breach scenarios by drones based on artificial intelligence that can respond quickly to the above technical problems.

[0006] In a first aspect, the present application provides a method for monitoring levee and dam breach scenarios by drones based on artificial intelligence, including: Obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, meteorological data, satellite image data, and topographic map data into a monitoring area recognition artificial intelligence model for target detection area recognition to generate target monitoring area risk level information and target monitoring area detection range information; Allocate the number of drones and the preset route of each drone based on the target monitoring area risk level information and the target monitoring area detection range information; Control each drone to fly along the corresponding preset route to obtain the target monitoring data of each drone; Input the target monitoring data of each drone into the artificial intelligence model for levee and dam breach scenario analysis to generate the analysis results of the levee and dam breach scenario, where the analysis results of the levee and dam breach scenario include levee and dam breach risk information and / or levee and dam breach disaster information.

[0007] In one embodiment, allocating the number of drones and the preset flight routes of each drone based on the target monitoring area risk level information and the target monitoring area detection range information includes: Set the maximum operation duration of the drone and the shooting range of the on-board monitoring equipment based on the target monitoring area risk level information; Calculate the maximum flight duration of the drone based on the maximum distance between the initial position of the drone and the target monitoring area detection range, and the maximum flight duration is used to represent the maximum time for the drone to reach the target monitoring area detection range from the initial position; Calculate the preset operation area of the drone based on the maximum operation duration, the shooting range of the on-board monitoring equipment, and the maximum flight duration; Allocate the number of drones based on the preset operation area and the target monitoring area detection range information; Allocate the preset flight routes of each drone based on the number of drones, the target monitoring area risk level information, and the target monitoring area detection range information.

[0008] In one embodiment, the calculation formula for the number of drones is: ; In the formula, is the number of drones, is the ceiling function, is the total area of the target monitoring area detection range, is the operation track overlap correction factor, is the maximum operation duration, is the maximum flight duration, is the environmental condition speed correction factor, is the average flight speed of the drone, is the width of the shooting range of the on-board monitoring equipment, is the fuzzy control membership function of the maximum operation duration, is the risk level of the target monitoring area, is the fuzzy control membership function of the shooting range of the on-board monitoring equipment.

[0009] In one embodiment, the analysis results of the levee and dam breach scenario include the second target monitoring area risk level information and the second target monitoring area detection range information. The method for monitoring drones in the levee and dam breach scenario based on artificial intelligence further includes: Calculate the risk level error between the risk level information of the target monitoring area and the risk level information of the second target monitoring area, and determine whether the risk level error exceeds the risk level deviation threshold; Calculate the detection range error between the detection range information of the target monitoring area and the detection range information of the second target monitoring area, and determine whether the detection range error exceeds the detection range deviation threshold; If the risk level error exceeds the risk level deviation threshold and / or the detection range error exceeds the detection range deviation threshold, adjust the number of drones based on the risk level information of the second target monitoring area and the detection range information of the second target monitoring area, and update the preset flight path of each drone based on the adjusted number of drones, the risk level information of the target monitoring area, the detection range information of the target monitoring area, and the monitored area information of the drones; among them, the monitored area information is generated based on the target monitoring data of the drones; Adjust and optimize the monitoring area identification artificial intelligence model based on the risk level information of the target monitoring area, the detection range information of the target monitoring area, the risk level information of the second target monitoring area, and the detection range information of the second target monitoring area.

[0010] In one embodiment, the artificial intelligence model for levee and dam breach scenario analysis includes a levee and dam breach risk analysis sub-model and a levee and dam breach disaster prediction sub-model. The levee and dam breach scenario analysis results include levee and dam breach risk information and levee and dam breach disaster information. Input the target monitoring data of each drone into the artificial intelligence model for levee and dam breach scenario analysis to perform levee and dam breach scenario analysis and generate levee and dam breach scenario analysis results, including: Input the target monitoring data into the levee and dam breach risk analysis sub-model to generate levee and dam breach risk area information and the corresponding levee and dam breach risk information for the levee and dam breach risk area information. The levee and dam breach risk level label includes the levee and dam breach disaster label; If the levee and dam breach risk information is the levee and dam breach disaster label, input the target monitoring data corresponding to the levee and dam breach disaster label and the topographic map data into the levee and dam breach disaster prediction sub-model to perform disaster prediction analysis and generate levee and dam breach disaster information.

[0011] In one embodiment, the levee and dam breach risk analysis sub-model is a deep learning neural network model constructed based on the improved YOLOV8 model; The levee and dam breach disaster analysis sub-model is a deep learning neural network model constructed based on the U-shaped network combined with the transformer module.

[0012] In one embodiment, the method for monitoring levees and dams using drones based on artificial intelligence further includes: Input the target monitoring data into the levee / dam breach anomaly recognition edge artificial intelligence model for levee / dam breach anomaly recognition, and generate levee / dam breach recognition flag information, levee / dam breach anomaly area information, and levee / dam breach anomaly degree information; If the levee / dam breach recognition flag information indicates the existence of a levee / dam breach, control the drone to hover and monitor the area corresponding to the levee / dam breach anomaly area information; Based on the levee / dam breach anomaly area information, levee / dam breach anomaly degree information, target monitoring area risk level information, and target monitoring area detection range information, perform validity determination on the target monitoring area risk level information and validity determination on the target monitoring area detection range information; If the determination result of the validity determination of the target monitoring area risk level information is that the target monitoring area risk level information is invalid and / or the determination result of the validity determination of the target monitoring area detection range information is that the target monitoring area detection range information is invalid, generate levee / dam breach disaster emergency report information, which is used to indicate that the output result of the monitoring area recognition artificial intelligence model does not match the actual disaster situation.

[0013] In one embodiment, the monitoring area recognition artificial intelligence model includes a first feature extraction module, a second feature extraction module, a feature fusion module, and a feature decoding module. Input hydrological data, meteorological data, satellite image data, and topographic map data into the monitoring area recognition artificial intelligence model for target detection area recognition, and generate target monitoring area risk level information and target monitoring area detection range information, including: Input hydrological data, meteorological data, and topographic map data into the first feature extraction module to generate a meteorological and hydrological feature map and the corresponding meteorological and hydrological anomaly feature values of the generated meteorological and hydrological feature map; Input satellite image data and topographic map data into the second feature extraction module to generate a satellite image feature map and the corresponding satellite image anomaly feature values of the satellite image feature map; If the meteorological and hydrological anomaly feature values exceed the preset meteorological and hydrological anomaly threshold or the satellite image anomaly feature values exceed the preset satellite image anomaly threshold, input the meteorological and hydrological feature map and the satellite image feature map into the feature fusion module to generate a fusion feature map; Input the fusion feature map into the feature decoding module to generate target monitoring area risk level information and target monitoring area detection range information.

[0014] In one embodiment, control each drone to fly along the corresponding preset route to obtain the target monitoring data of each drone, including: Control each drone to fly along the corresponding preset route to obtain target monitoring sub-scene image data; Input the target monitoring sub-scenario image data into the image stitching edge artificial intelligence model for image stitching to generate target monitoring data.

[0015] In a second aspect, the present application also provides an unmanned aerial vehicle (UAV) monitoring system for levee and dam breach scenarios based on artificial intelligence, including: A monitoring area data analysis module, configured to obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, meteorological data, satellite image data, and topographic map data into a monitoring area recognition artificial intelligence model to identify the target detection area and generate target monitoring area risk level information and target monitoring area detection range information; A UAV working configuration module, configured to allocate the number of UAVs and the preset flight path of each UAV based on the target monitoring area risk level information and the target monitoring area detection range information; A UAV working control module, configured to control each UAV to fly along the corresponding preset flight path to obtain the target monitoring data of each UAV; A target monitoring data processing module, configured to input the target monitoring data of each UAV into a levee and dam breach scenario analysis artificial intelligence model for levee and dam breach scenario analysis to generate a levee and dam breach scenario analysis result, where the levee and dam breach scenario analysis result includes levee and dam breach risk information and / or levee and dam breach disaster information.

[0016] The above-mentioned UAV monitoring method and system for levee and dam breach scenarios based on artificial intelligence can improve the comprehensiveness of levee and dam breach scenario risk assessment and significantly increase the response speed of UAV monitoring for levee and dam breach scenarios by integrating hydrological, meteorological, satellite image, and topographic map data based on a monitoring area recognition artificial intelligence model and generating target monitoring area risk level information and target monitoring area detection range information; by intelligently allocating the number of UAVs and preset flight paths based on the risk level and range of the target monitoring area, it can ensure key monitoring of high-risk areas, improve the utilization efficiency of UAV resources, and ensure the rationality and pertinence of monitoring coverage; by controlling the UAVs to fly along the preset flight paths to obtain target monitoring data, it realizes automated and standardized data collection, reduces manual intervention, ensures the comprehensiveness and stability of data acquisition, and provides a high-quality data foundation for subsequent analysis; by using a levee and dam breach scenario analysis artificial intelligence model to process the target monitoring data and generate an analysis result including risk information and disaster information, it can help decision-makers quickly grasp the dynamic information of levee and dam breach disasters, provide accurate support for emergency response and rescue deployment, and improve the timeliness and scientificity of disaster response. Description of the Drawings

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the related art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 Schematic diagram of the application environment of an unmanned aerial vehicle monitoring method for levee and dam breach scenarios based on artificial intelligence provided by an embodiment of the present application; Figure 2 Schematic diagram of the flow of an unmanned aerial vehicle monitoring method for levee and dam breach scenarios based on artificial intelligence provided by an embodiment of the present application; Figure 3 Schematic diagram of the flow of allocating the number of unmanned aerial vehicles and flight routes provided by an embodiment of the present application; Figure 4 Schematic diagram of the flow of a feedback control method based on the analysis results of levee and dam breach scenarios provided by an embodiment of the present application; Figure 5 Schematic diagram of the flow of another unmanned aerial vehicle monitoring method for levee and dam breach scenarios based on artificial intelligence provided by an embodiment of the present application; Figure 6 Schematic diagram of the structure of the levee and dam breach risk analysis sub-model in the levee and dam breach scenario analysis artificial intelligence model provided by an embodiment of the present application; Figure 7 Schematic diagram of the structure of the levee and dam breach disaster situation analysis sub-model in the levee and dam breach scenario analysis artificial intelligence model provided by an embodiment of the present application; Figure 8 Schematic diagram of the structure of the deployment of an artificial intelligence model provided by an embodiment of the present application; Figure 9 Schematic diagram of the structure of an unmanned aerial vehicle monitoring system for levee and dam breach scenarios based on artificial intelligence provided by an embodiment of the present application. Detailed implementation manners

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the following further details the present application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0020] The unmanned aerial vehicle monitoring method for levee and dam breach scenarios based on artificial intelligence provided by the embodiments of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the sensing device 102 and the working device 103 can communicate with the computing device 101 through a communication channel. The data storage system can store the data that the computing device 101 needs to process. The data storage system can be integrated on the computing device 101, or placed in the cloud or other network servers. The computing device 101 can generate a levee breach and dam break disaster report and a working device control instruction based on the environmental data of the levee breach and dam break scenario obtained by the sensing device 102, and the computing device 101 can control the working device 103 to execute a work task based on the working device control instruction. Among them, the computing device 101 can be, but is not limited to, various servers, intelligent devices, and Geographic Information System Workstations (GIS Workstations). The server can be implemented by an independent server or a server cluster composed of multiple servers. The intelligent device can include, but is not limited to, various personal computers, laptops, smartphones, tablets, and Internet of Things devices. The Internet of Things devices can include edge computing devices. The sensing device 102 can include, but is not limited to, various water level gauges, piezometers, seepage sensors, sonar detection devices, remote sensing satellites, meteorological detection devices, cameras, infrared sensing devices, lidar, satellite navigation and positioning devices, and structural strain sensors. The working device 103 can include, but is not limited to, fixed-wing drones, multi-rotor drones, and rescue equipment.

[0021] In an exemplary embodiment, as Figure 2 shown, a method for monitoring drones in the levee breach and dam break scenario based on artificial intelligence is provided. Taking the computing device 101 in Figure 1 as an example, the method includes the following steps S201 to step S204. Among them: Step S201: Obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, meteorological data, satellite image data, and topographic map data into the monitoring area recognition artificial intelligence model to perform target detection area recognition, and generate target monitoring area risk level information and target monitoring area detection range information.

[0022] Specifically, the computing device 101 can input the hydrological data, meteorological data, satellite image data, and topographic map data obtained based on the sensing device into the monitoring area recognition artificial intelligence model carried on the computing device 101 to perform target detection area recognition, and generate target monitoring area risk level information and target monitoring area detection range information.

[0023] Optionally, the target monitoring area detection range information may include the detection range information of each target monitoring area corresponding to each target monitoring area risk level.

[0024] Optionally, when the difference between the target monitoring area risk level information and the target monitoring area detection range information is small, the target monitoring area risk level information may include the overall target monitoring area risk level information for measuring the overall risk level of the target monitoring area detection range information.

[0025] Optionally, when the difference between the target monitoring area risk level information and the target monitoring area detection range information is large, the target monitoring area risk level information may include the local target monitoring area risk level information corresponding to the local target monitoring areas with the same local target monitoring area risk level information.

[0026] Furthermore, the computing device 101 may construct a contour map of the target monitoring area risk level information based on the target monitoring area risk level information and the target monitoring area detection range information generated by the monitoring area recognition artificial intelligence model.

[0027] Illustratively, the monitoring area recognition artificial intelligence model carried on the computing device 101 may analyze the temporal dependence and / or spatial dependence of hydrological data, meteorological data, satellite image data, and topographic map data based on a sequence model. Among them, the sequence model may include, but is not limited to, a recurrent neural network (RNN), a long short-term memory network model (LSTM), a gated recurrent unit model (GRU), a temporal convolutional network (TCN), and a Transformer model.

[0028] Furthermore, the monitoring area recognition artificial intelligence model carried on the computing device 101 may perform image segmentation and object detection on the hydrological data, meteorological data, satellite image data, and topographic map data from which the temporal dependence and / or spatial dependence have been extracted based on a convolutional neural network model (CNN) and its variant models, and generate the target monitoring area risk level information and the target monitoring area detection range information.

[0029] Optionally, the variant models of the convolutional neural network model may include, but are not limited to, an AlexNet model, a Visual Geometry Group Network model (VGGNet), a Residual Network model (ResNet), Inception series models, a U-Net model, YOLO series models, and a SegNet model.

[0030] Step S202: Allocate the number of drones and the preset flight routes of each drone based on the target monitoring area risk level information and the target monitoring area detection range information.

[0031] Specifically, the computing device 101 can allocate the number of drones based on the target monitoring area risk level information and the target monitoring area detection range information generated by the artificial intelligence model for monitoring area recognition installed thereon, and allocate the preset flight routes of each drone in combination with the target monitoring area risk level information and the target monitoring area detection range information according to the allocation result of the number of drones.

[0032] Optionally, the computing device 101 can allocate the preset flight routes of each drone based on an intelligent optimization algorithm in combination with the target monitoring area risk level information and the target monitoring area detection range information according to the allocation result of the number of drones. Among them, the intelligent optimization algorithm can include but is not limited to evolutionary algorithms, swarm intelligence algorithms, simulated annealing algorithms, and model-based intelligent optimization algorithms. Evolutionary algorithms can include but are not limited to genetic algorithms (GA) and immune algorithms (IA). Swarm intelligence algorithms can include but are not limited to ant colony optimization (ACO), artificial bee colony (ABC), particle swarm optimization (PSO), and wolf pack algorithm (WCA).

[0033] Step S203: Control each drone to fly along the corresponding preset flight route and obtain the target monitoring data of each drone.

[0034] Specifically, the computing device 101 can control the preset flight routes of each drone based on the generated preset flight routes of each drone, obtain the detection data of each drone, and generate the target monitoring data of each drone after preprocessing the detection data of each drone. The preprocessing can include but is not limited to image enhancement, image filtering, image registration, and image stitching.

[0035] Optionally, the drones can include but are not limited to fixed-wing drones and rotary-wing drones. The sensors carried by the drones can include but are not limited to thermal infrared cameras, lidar, high-definition cameras, global positioning system and real-time kinematic measurement modules (GPS / RTK modules), and multispectral cameras. Among them, the thermal infrared camera can be used for detecting leakage temperature differences and searching for people, and the lidar can be used for precise ranging and terrain modeling.

[0036] Optionally, the drones can deploy lightweight artificial intelligence models to achieve real-time data processing.

[0037] Step S204: Input the target monitoring data of each drone into the artificial intelligence model for levee and dam breach scenario analysis to perform levee and dam breach scenario analysis and generate the levee and dam breach scenario analysis results. Specifically, the computing device 101 can input the target monitoring data of each drone into the artificial intelligence model for levee and dam breach scenario analysis carried by the computing device 101 to perform levee and dam breach scenario analysis and generate the levee and dam breach scenario analysis results. The levee and dam breach scenario analysis results may include levee and dam breach risk information and / or levee and dam breach disaster information.

[0038] Schematically, the main structure of the artificial intelligence model for levee and dam breach scenario analysis carried by the computing device 101 can be constructed based on a convolutional neural network model and a variant model of the convolutional neural network model.

[0039] Optionally, when the computing device 101 does not identify an existing levee and dam breach phenomenon, the levee and dam breach scenario analysis results may include levee and dam breach risk information; when the computing device 101 identifies an existing levee and dam breach phenomenon, the levee and dam breach scenario analysis results may include levee and dam breach disaster information; when the computing device 101 identifies an existing levee and dam breach phenomenon and an impending levee and dam breach phenomenon, the levee and dam breach scenario analysis results may include levee and dam breach risk information and levee and dam breach disaster information.

[0040] Furthermore, the computing device 101 can identify existing levee and dam breach phenomena and impending levee and dam breach phenomena based on the artificial intelligence model for levee and dam breach scenario analysis carried by the computing device 101.

[0041] Exemplarily, the levee and dam breach disaster information generated by the artificial intelligence model for levee and dam breach scenario analysis carried by the computing device 101 may include, but is not limited to, disaster area information and disaster severity information. The computing device 101 can generate early warning level information based on the disaster area information and the disaster severity information.

[0042] In the above method for monitoring levee and dam breaches by drones based on artificial intelligence, by integrating multi-source data, various factors affecting levee and dam breaches can be comprehensively considered; by constructing an artificial intelligence model for identifying monitoring areas, the target areas that need to be key monitored can be automatically identified, and the risk level information and detection range information of the target monitoring areas can be generated, thereby dynamically adjusting the monitoring strategy of the drones and improving the monitoring efficiency and emergency response speed; by inputting the collected target monitoring data into the artificial intelligence model for levee and dam breach scenario analysis, in-depth analysis of the levee and dam breach risk information and / or disaster information can be performed, thereby realizing the full-process intelligent and real-time processing from data collection to analysis result generation, greatly shortening the time interval from monitoring to decision-making, and enabling a rapid response in the face of sudden levee and dam breach events, thus winning precious time for emergency rescue.

[0043] Schematically, in the above-mentioned UAV monitoring method for levee and dam breach scenarios based on artificial intelligence, through multi-source data fusion, application of artificial intelligence models, and UAV monitoring, accurate and efficient real-time monitoring and analysis of levee and dam breach scenarios can be achieved, providing strong technical support for flood control and disaster reduction work.

[0044] In one alternative embodiment, as Figure 3 shown, allocating the number of UAVs and the preset flight routes of each UAV based on the risk level information of the target monitoring area and the detection range information of the target monitoring area includes: Step S301, setting the maximum operation duration of the UAV and the shooting range of the on-board monitoring device based on the risk level information of the target monitoring area.

[0045] Specifically, the computing device can obtain the maximum operation duration of the UAV and the shooting range of the on-board monitoring device corresponding to the risk level information of the target monitoring area based on the preset UAV operation parameter setting rules. Among them, the maximum operation duration of the UAV can be used to represent the maximum monitoring duration of the UAV within the target monitoring area corresponding to the detection range information of the target monitoring area, and the maximum operation duration of the UAV can be used to measure the response speed of the UAV monitoring for levee and dam breach scenarios.

[0046] Optionally, the preset UAV operation parameter setting rules can be but are not limited to data mapping tables, calculation formulas, and BP neural networks.

[0047] Step S302, calculating the maximum flight duration of the UAV based on the maximum distance between the initial position of the UAV and the detection range of the target monitoring area.

[0048] Specifically, the computing device can calculate the maximum flight duration of the UAV based on the maximum distance between the initial position of the UAV and the detection range of the target monitoring area. The maximum flight duration of the UAV can be used to represent the maximum duration for the UAV to reach the detection range of the target monitoring area from the initial position. Among them, the initial position of the UAV can be set as the coordinate position of the UAV base station.

[0049] Step S303, calculating the preset operation area of the UAV based on the maximum operation duration, the shooting range of the on-board monitoring device, and the maximum flight duration.

[0050] Step S304, allocating the number of UAVs based on the preset operation area and the detection range information of the target monitoring area.

[0051] Step S305, allocating the preset flight routes of each UAV based on the number of UAVs, the risk level information of the target monitoring area, and the detection range information of the target monitoring area.

[0052] In the above-mentioned UAV monitoring method for levee and dam breach scenarios based on artificial intelligence, by dynamically adjusting the number and flight routes of UAVs according to real-time monitoring data, it is possible to cope with changes in the risk level and detection range of the monitoring area, and improve the flexibility and adaptability of levee and dam breach scenario monitoring.

[0053] Specifically, in the above-mentioned UAV monitoring method for levee and dam breach scenarios based on artificial intelligence, through means such as dynamic resource allocation, optimizing operation duration and shooting range, precise calculation of flight duration, and reasonable operation area planning, it is possible to achieve efficient and accurate monitoring of levee and dam breach scenarios, providing strong technical support for flood control and disaster reduction work.

[0054] In one alternative embodiment, the calculation formula for the number of UAVs is: ; In the formula, is the number of UAVs, is the ceiling function, is the total area of the detection range of the target monitoring area, is the operation track overlap correction factor, is the maximum operation duration, is the maximum flight duration, is the environmental condition speed correction factor, is the average flight speed of the UAV, is the width of the shooting range of the on-board monitoring equipment, is the fuzzy control membership function of the maximum operation duration, is the risk level of the target monitoring area, is the fuzzy control membership function of the shooting range of the on-board monitoring equipment.

[0055] Optionally, the risk level of the target monitoring area can be obtained by taking the weighted sum of the risk level information of the local target monitoring areas corresponding to the local target monitoring areas with the same risk level information for each part.

[0056] Furthermore, the calculation formula for the risk level of the target monitoring area can be: ; In the formula, is the risk level of the target monitoring area, is the number of types of risk level information of the local target monitoring area, is the weight of the th type of risk level information of the local target monitoring area, is the th type of risk level information of the local target monitoring area corresponding to the risk level, is the area of the local target monitoring area corresponding to the risk level information of the th kind of local target monitoring area, is the total area of the target monitoring area corresponding to the detection range information of the target monitoring area.

[0057] In one optional embodiment, as Figure 4 shown, the analysis result of the dike breach and dam break scenario includes the risk level information of the second target monitoring area and the detection range information of the second target monitoring area. The unmanned aerial vehicle monitoring method for the dike breach and dam break scenario based on artificial intelligence further includes: Step S401, calculate the risk level error between the risk level information of the target monitoring area and the risk level information of the second target monitoring area, and determine whether the risk level error exceeds the risk level deviation threshold.

[0058] Step S401, calculate the detection range error between the detection range information of the target monitoring area and the detection range information of the second target monitoring area, and determine whether the detection range error exceeds the detection range deviation threshold.

[0059] Step S401, if the risk level error exceeds the risk level deviation threshold and / or the detection range error exceeds the detection range deviation threshold, adjust the number of unmanned aerial vehicles based on the risk level information of the second target monitoring area and the detection range information of the second target monitoring area, and update the preset flight path of each unmanned aerial vehicle based on the adjusted number of unmanned aerial vehicles, the risk level information of the target monitoring area, the detection range information of the target monitoring area, and the monitored area information of the unmanned aerial vehicle.

[0060] Specifically, if the computing device determines that the risk level error exceeds the risk level deviation threshold and / or the computing device determines that the detection range error exceeds the detection range deviation threshold, the computing device can recalculate and adjust the number of unmanned aerial vehicles based on the risk level information of the second target monitoring area and the detection range information of the second target monitoring area, and update the preset flight path of each unmanned aerial vehicle based on the adjusted number of unmanned aerial vehicles, the risk level information of the target monitoring area, the detection range information of the target monitoring area, and the monitored area information of the unmanned aerial vehicle; wherein, the monitored area information can be generated based on the target monitoring data of the unmanned aerial vehicle.

[0061] Step S401, adjust and optimize the monitoring area recognition artificial intelligence model based on the risk level information of the target monitoring area, the detection range information of the target monitoring area, the risk level information of the second target monitoring area, and the detection range information of the second target monitoring area.

[0062] Optionally, the computing device can constrain the distribution consistency of the two predictions based on the KL divergence.

[0063] In the above method for monitoring dams and levees using drones based on artificial intelligence, by determining whether the risk level error and detection range error exceed the preset deviation thresholds, anomalies or deviations in the monitoring results can be detected in a timely manner, ensuring the accuracy and timeliness of the monitoring data. When the risk level error or detection range error exceeds the threshold, by dynamically adjusting the number of drones based on the risk level information and detection range information of the second target monitoring area, it is possible to ensure that the number of drones is increased in a timely manner in areas with high risks or large changes in the detection range, improving the coverage and accuracy of dam and levee breach monitoring. By combining the risk level information and detection range information of the target monitoring area and the second target monitoring area to adjust and optimize the artificial intelligence model for monitoring area identification, it helps the model continuously learn and adapt to new monitoring data and environmental changes, improving the accuracy and robustness of the artificial intelligence model for monitoring area identification.

[0064] In one optional embodiment, please refer to Figure 5 , the artificial intelligence model for dam and levee breach scenario analysis includes a dam and levee breach risk analysis sub-model and a dam and levee breach disaster prediction sub-model. The results of dam and levee breach scenario analysis include dam and levee breach risk information and dam and levee breach disaster information. Inputting the target monitoring data of each drone into the artificial intelligence model for dam and levee breach scenario analysis to perform dam and levee breach scenario analysis, and generating the results of dam and levee breach scenario analysis, including: Step S507: Input the target monitoring data into the dam and levee breach risk analysis sub-model to generate dam and levee breach risk area information and the corresponding dam and levee breach risk information for the dam and levee breach risk area information.

[0065] Specifically, the computing device can input the target monitoring data into the dam and levee breach risk analysis sub-model to generate dam and levee breach risk area information and the corresponding dam and levee breach risk information for the dam and levee breach risk area information. Among them, the dam and levee breach risk level labels can include dam and levee breach disaster labels.

[0066] Step S508: If the dam and levee breach risk information is a dam and levee breach disaster label, input the target monitoring data corresponding to the dam and levee breach disaster label and topographic map data into the dam and levee breach disaster prediction sub-model to perform disaster prediction analysis and generate dam and levee breach disaster information.

[0067] Optionally, the dam and levee breach disaster information can include disaster area information and disaster level information.

[0068] In the above method for monitoring dams and levees using drones based on artificial intelligence, by dividing the dam and levee breach scenario analysis into two stages: risk analysis and disaster prediction, it is possible to achieve refined analysis of complex scenarios and improve the real-time performance and stability of dam and levee breach detection.

[0069] In one alternative embodiment, the embankment and dam breach risk analysis sub-model is a deep learning neural network model constructed based on the improved YOLOV8 model; The embankment and dam breach disaster analysis sub-model is a deep learning neural network model constructed based on the U-shaped network combined with the transformer module.

[0070] In one alternative embodiment, please refer to Figure 6 , the embankment and dam breach risk analysis sub-model in the embankment and dam breach scenario analysis artificial intelligence model can be a deep learning neural network model constructed based on the improved YOLOV8 model. The improved YOLOV8 model can replace the fast double convolutional partial direct connection module (C2F module) in the backbone network with a fast double convolutional partial direct connection dilated residual convolutional module based on the initial YOLOV8 model; replace the neck network with a bidirectional feature pyramid structure combined with the attention mechanism; and add a global attention mechanism in the backbone network and the neck network. Among them, the fast double convolutional partial direct connection dilated residual convolutional module can be obtained by replacing the convolutional module in the C2F module with a dilated residual convolutional module, and the bidirectional feature pyramid network combined with the attention mechanism can be obtained by adding a permutation attention module in the skip connection structure of the initial bidirectional feature pyramid network.

[0071] Schematically, the detection module in the head network of the embankment and dam breach risk analysis sub-model in the embankment and dam breach scenario analysis artificial intelligence model can include an embankment and dam breach risk area information detection head and an embankment and dam breach risk information detection head.

[0072] In one alternative embodiment, please refer to Figure 7 , the embankment and dam breach disaster analysis sub-model in the embankment and dam breach scenario analysis artificial intelligence model can be a deep learning neural network model constructed based on the U-shaped network combined with the transformer module. The deep learning neural network model constructed based on the U-shaped network combined with the transformer module can include a disaster situation feature fusion module and a disaster situation feature decoding module. The disaster situation feature fusion module can be used to fuse the topographic map data features and the target detection data features to generate a disaster situation fusion feature map. The disaster situation feature decoding module can be constructed based on the U-shaped network combined with the residual convolutional module and the transformer module, and the disaster situation feature decoding module can be used to analyze the disaster situation fusion feature map to generate embankment and dam breach disaster information.

[0073] Optionally, the downsampling layer of the disaster situation feature decoding module can include, but is not limited to, a max pooling layer.

[0074] In one alternative embodiment, please refer to Figure 5 and Figure 8 , the UAV monitoring method for embankment and dam breach scenarios based on artificial intelligence further includes: Step S509: Input the target monitoring data into the edge artificial intelligence model for levee / dam breach anomaly identification to generate levee / dam breach identification flag information, levee / dam breach anomaly area information, and levee / dam breach anomaly degree information.

[0075] Optionally, the edge artificial intelligence model for levee / dam breach anomaly identification can be deployed in edge computing devices.

[0076] Illustratively, the computing devices can include a central computing device and edge computing devices. Among them, the central computing device can be used to deploy the artificial intelligence model for monitoring area identification and the artificial intelligence model for levee / dam breach scenario analysis.

[0077] Illustratively, the edge device can be but is not limited to being deployed on a drone.

[0078] Step S510: If the levee / dam breach identification flag information indicates the existence of a levee / dam breach, control the drone to hover and monitor the area corresponding to the levee / dam breach anomaly area information.

[0079] Specifically, if the levee / dam breach identification flag information generated by the edge artificial intelligence model for levee / dam breach anomaly identification carried by the edge computing device indicates the existence of a levee / dam breach, the edge computing device can control the drone to hover and monitor the area corresponding to the levee / dam breach anomaly area information. At this time, the edge computing device can generate a drone hovering work report and send the drone hovering work report to the central computing device. After receiving the drone hovering work report, the central computing device can make a determination on the additional deployment of drones based on the drone hovering work report. If the determination result of the additional deployment of drones is that drones need to be additionally deployed, the central computing device can generate the number of drones to be additionally deployed and the preset flight route.

[0080] Step S511: Perform the validity determination of the target monitoring area risk level information and the validity determination of the target monitoring area detection range information based on the levee / dam breach anomaly area information, levee / dam breach anomaly degree information, target monitoring area risk level information, and target monitoring area detection range information.

[0081] Illustratively, the central computing device can perform the validity determination of the target monitoring area risk level information and the validity determination of the target monitoring area detection range information based on the levee / dam breach anomaly area information, levee / dam breach anomaly degree information, target monitoring area risk level information, and target monitoring area detection range information.

[0082] Step S512: If the determination result of the validity determination of the target monitoring area risk level information is that the target monitoring area risk level information is invalid and / or the determination result of the validity determination of the target monitoring area detection range information is that the target monitoring area detection range information is invalid, generate the levee / dam breach disaster emergency report information.

[0083] Specifically, if the determination result of the effectiveness determination of the risk level information of the target monitoring area is that the risk level information of the target monitoring area is invalid and / or the determination result of the effectiveness determination of the detection range information of the target monitoring area is that the detection range information of the target monitoring area is invalid, the central computing device can generate an emergency report information for dam break and levee breach disasters, and the emergency report information for dam break and levee breach disasters can be used to characterize that the result output by the artificial intelligence model for identifying the monitoring area does not match the actual disaster situation.

[0084] In the above UAV monitoring method for dam break and levee breach scenarios based on artificial intelligence, through automatic UAV monitoring, the frequency and workload of manual inspections can be reduced, and the labor cost and risk can be lowered; through real-time anomaly identification and monitoring, actions can be taken quickly after an anomaly is detected, shortening the emergency response time.

[0085] In one alternative embodiment, please refer to Figure 5 , the artificial intelligence model for identifying the monitoring area includes a first feature extraction module, a second feature extraction module, a feature fusion module, and a feature decoding module. Inputting hydrological data, meteorological data, satellite image data, and topographic map data into the artificial intelligence model for identifying the monitoring area for target detection area identification, generating the risk level information of the target monitoring area and the detection range information of the target monitoring area, including: Step S501: Input hydrological data, meteorological data, and topographic map data into the first feature extraction module to generate a meteorological and hydrological feature map and a meteorological and hydrological anomaly feature value corresponding to the generated meteorological and hydrological feature map.

[0086] Specifically, the computing device can input hydrological data, meteorological data, and topographic map data into the first feature extraction module of the artificial intelligence model for identifying the monitoring area carried by the computing device to generate a meteorological and hydrological feature map and a meteorological and hydrological anomaly feature value corresponding to the generated meteorological and hydrological feature map. Among them, the first feature extraction module can be used to perform spatial pairing and feature fusion on hydrological data, meteorological data, and topographic map data.

[0087] Step S502: Input satellite image data and topographic map data into the second feature extraction module to generate a satellite image feature map and a satellite image anomaly feature value corresponding to the generated satellite image feature map.

[0088] Step S503: If the meteorological and hydrological anomaly feature value exceeds the preset meteorological and hydrological anomaly threshold or the satellite image anomaly feature value exceeds the preset satellite image anomaly threshold, input the meteorological and hydrological feature map and the satellite image feature map into the feature fusion module to generate a fusion feature map.

[0089] Step S504: Input the fusion feature map into the feature decoding module to generate the risk level information of the target monitoring area and the detection range information of the target monitoring area.

[0090] Schematically, the feature decoding module can be constructed based on, but not limited to, the YOLOV8 model, the U-shaped network model, and the transformer model.

[0091] In the above-mentioned UAV monitoring method for dam break and levee breach scenarios based on artificial intelligence, by setting the first feature extraction module and the second feature extraction module to process meteorological and hydrological data and satellite image data respectively, key features can be quickly extracted to generate corresponding abnormal feature values. When the abnormal feature values exceed the preset threshold, the subsequent feature fusion and decoding process can be triggered in a timely manner, which can not only improve the calculation efficiency but also respond to potential risks in real time, providing timely and accurate basis for emergency decision-making; through the feature fusion module, feature maps from different sources can be effectively integrated to generate a comprehensive fusion feature map, further enhancing the model's adaptability to complex environments; through the feature decoding module, the risk level and detection range information of the target monitoring area are generated based on the fusion feature map, which can effectively cope with the complex and changeable natural environment and improve the generalization ability and robustness of the artificial intelligence model for monitoring area recognition.

[0092] In one alternative embodiment, please refer to Figure 8 , control each UAV to fly along the corresponding preset route to obtain the target monitoring data of each UAV, including: Control each UAV to fly along the corresponding preset route to obtain the image data of the target monitoring sub-scene; Input the image data of the target monitoring sub-scene into the image stitching edge artificial intelligence model for image stitching to generate the target monitoring data.

[0093] Optionally, the image stitching edge artificial intelligence model can be carried on an edge computing device.

[0094] In the above-mentioned UAV monitoring method for dam break and levee breach scenarios based on artificial intelligence, through the image stitching edge artificial intelligence model, the image stitching accuracy can be improved, the robustness of the dam break and levee breach detection system can be enhanced, and the end-cloud collaboration efficiency of the dam break and levee breach detection system can be improved.

[0095] In an exemplary embodiment, as Figure 5 shown, another UAV monitoring method for dam break and levee breach scenarios based on artificial intelligence is provided, including the following steps S501 to step S512. Among them: Step S501, input the hydrological data, meteorological data, and topographic map data into the first feature extraction module to generate a meteorological and hydrological feature map and a meteorological and hydrological abnormal feature value corresponding to the generated meteorological and hydrological feature map.

[0096] Step S502: Input the satellite image data and topographic map data into the second feature extraction module to generate a satellite image feature map and the corresponding satellite image anomaly eigenvalue of the satellite image feature map.

[0097] Step S503: If the meteorological and hydrological anomaly eigenvalue exceeds the preset meteorological and hydrological anomaly threshold or the satellite image anomaly eigenvalue exceeds the preset satellite image anomaly threshold, input the meteorological and hydrological feature map and the satellite image feature map into the feature fusion module to generate a fusion feature map.

[0098] Step S504: Input the fusion feature map into the feature decoding module to generate the risk level information of the target monitoring area and the detection range information of the target monitoring area.

[0099] Step S505: Based on the risk level information of the target monitoring area and the detection range information of the target monitoring area, allocate the number of drones and the preset flight routes of each drone.

[0100] Step S506: Control each drone to fly along the corresponding preset flight route to obtain the target monitoring data of each drone.

[0101] Step S507: Input the target monitoring data into the dike and dam breach risk analysis sub-model to generate the dike and dam breach risk area information and the corresponding dike and dam breach risk information of the dike and dam breach risk area information.

[0102] Step S508: If the dike and dam breach risk information is the dike and dam breach disaster label, input the target monitoring data and topographic map data corresponding to the dike and dam breach disaster label into the dike and dam breach disaster prediction sub-model for disaster prediction analysis to generate the dike and dam breach disaster information.

[0103] Step S509: Input the target monitoring data into the dike and dam breach anomaly recognition edge artificial intelligence model for dike and dam breach anomaly recognition to generate the dike and dam breach recognition flag information, the dike and dam breach anomaly area information, and the dike and dam breach anomaly degree information.

[0104] Step S510: If the dike and dam breach recognition flag information indicates the existence of a dike and dam breach, control the drone to hover and monitor the area corresponding to the dike and dam breach anomaly area information.

[0105] Step S511: Based on the dike and dam breach anomaly area information, the dike and dam breach anomaly degree information, the risk level information of the target monitoring area, and the detection range information of the target monitoring area, conduct the effectiveness determination of the risk level information of the target monitoring area and the effectiveness determination of the detection range information of the target monitoring area.

[0106] Step S512: If the determination result of the effectiveness determination of the risk level information of the target monitoring area is that the risk level information of the target monitoring area is invalid and / or the determination result of the effectiveness determination of the detection range information of the target monitoring area is that the detection range information of the target monitoring area is invalid, an emergency report information on dike breach and dam break disasters is generated.

[0107] In the above-mentioned UAV monitoring method for dike breach and dam break scenarios based on artificial intelligence, through means such as multi-source data fusion, efficient feature extraction and fusion, intelligent risk assessment and disaster assessment, and real-time anomaly identification and monitoring, it is possible to achieve efficient and accurate monitoring of dike breach and dam break scenarios, thereby providing strong technical support for flood control and disaster reduction work.

[0108] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indication of the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same moment, but can be executed at different moments. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0109] Based on the same inventive concept, the embodiment of the present application also provides a UAV monitoring system for dike breach and dam break scenarios based on artificial intelligence for implementing the above-mentioned UAV monitoring method for dike breach and dam break scenarios based on artificial intelligence. The implementation solutions provided by this system to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the UAV monitoring system for dike breach and dam break scenarios based on artificial intelligence provided below can refer to the limitations on the UAV monitoring method for dike breach and dam break scenarios based on artificial intelligence in the above text, and will not be repeated here.

[0110] In an exemplary embodiment, as Figure 9 shown, a UAV monitoring system 900 for dike breach and dam break scenarios based on artificial intelligence is provided, including: A monitoring area data analysis module 901, which can be used to obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, meteorological data, satellite image data, and topographic map data into the monitoring area recognition artificial intelligence model to perform target detection area recognition, and generate target monitoring area risk level information and target monitoring area detection range information.

[0111] The UAV working configuration module 902 can be used to allocate the number of UAVs and the preset flight path of each UAV based on the risk level information of the target monitoring area and the detection range information of the target monitoring area.

[0112] The UAV working control module 903 can be used to control each UAV to fly along the corresponding preset flight path and obtain the target monitoring data of each UAV.

[0113] The target monitoring data processing module 904 can be used to input the target monitoring data of each UAV into the artificial intelligence model for levee and dam break scenario analysis, perform levee and dam break scenario analysis, and generate the levee and dam break scenario analysis result, where the levee and dam break scenario analysis result includes levee and dam break risk information and / or levee and dam break disaster information.

[0114] In an optional embodiment, the UAV working configuration module 902 can also be used to: Set the maximum operation duration of the UAV and the shooting range of the on-board monitoring equipment based on the risk level information of the target monitoring area; Calculate the maximum flight duration of the UAV based on the maximum distance between the initial position of the UAV and the detection range of the target monitoring area, and the maximum flight duration is used to represent the maximum duration for the UAV to reach the detection range of the target monitoring area from the initial position; Calculate the preset operation area of the UAV based on the maximum operation duration, the shooting range of the on-board monitoring equipment, and the maximum flight duration; Allocate the number of UAVs based on the preset operation area and the detection range information of the target monitoring area; Allocate the preset flight path of each UAV based on the number of UAVs, the risk level information of the target monitoring area, and the detection range information of the target monitoring area.

[0115] In an optional embodiment, the UAV monitoring system 900 for levee and dam break scenarios based on artificial intelligence can also be used to: Calculate the risk level error between the risk level information of the target monitoring area and the risk level information of the second target monitoring area, and determine whether the risk level error exceeds the risk level deviation threshold; Calculate the detection range error between the detection range information of the target monitoring area and the detection range information of the second target monitoring area, and determine whether the detection range error exceeds the detection range deviation threshold; If the risk level error exceeds the risk level deviation threshold and / or the detection range error exceeds the detection range deviation threshold, adjust the number of drones based on the risk level information of the second target monitoring area and the detection range information of the second target monitoring area, and update the preset route of each drone based on the adjusted number of drones, the risk level information of the target monitoring area, the detection range information of the target monitoring area, and the monitored area information of the drones; wherein, the monitored area information is generated based on the target monitoring data of the drones. Adjust and optimize the monitoring area identification artificial intelligence model based on the risk level information of the target monitoring area, the detection range information of the target monitoring area, the risk level information of the second target monitoring area, and the detection range information of the second target monitoring area.

[0116] In an alternative embodiment, the target monitoring data processing module 904 may also be used for: Input the target monitoring data into the dam break and levee breach risk analysis sub-model to generate dam break and levee breach risk area information and the corresponding dam break and levee breach risk information. The dam break and levee breach risk level label includes the dam break and levee breach disaster-affected label. If the dam break and levee breach risk information is the dam break and levee breach disaster-affected label, input the target monitoring data corresponding to the dam break and levee breach disaster-affected label and the topographic map data into the dam break and levee breach disaster prediction sub-model for disaster prediction analysis to generate dam break and levee breach disaster information.

[0117] In an alternative embodiment, the dam break and levee breach scenario drone monitoring system 900 based on artificial intelligence may also be used for: Input the target monitoring data into the dam break and levee breach anomaly recognition edge artificial intelligence model for dam break and levee breach anomaly recognition to generate dam break and levee breach recognition flag information, dam break and levee breach anomaly area information, and dam break and levee breach anomaly degree information. If the dam break and levee breach recognition flag information indicates the existence of a dam break and levee breach, control the drone to hover and monitor the area corresponding to the dam break and levee breach anomaly area information. Perform the validity determination of the risk level information of the target monitoring area and the validity determination of the detection range information of the target monitoring area based on the dam break and levee breach anomaly area information, the dam break and levee breach anomaly degree information, the risk level information of the target monitoring area, and the detection range information of the target monitoring area. If the determination result of the validity determination of the risk level information of the target monitoring area is that the risk level information of the target monitoring area is invalid and / or the determination result of the validity determination of the detection range information of the target monitoring area is that the detection range information of the target monitoring area is invalid, generate dam break and levee breach disaster emergency report information, which is used to indicate that the result output by the monitoring area identification artificial intelligence model does not match the actual disaster situation.

[0118] In an alternative embodiment, the monitoring area data analysis module 901 may further be configured to: Input hydrological data, meteorological data, and topographic map data into the first feature extraction module to generate a meteorological and hydrological feature map and a corresponding meteorological and hydrological anomaly feature value for the meteorological and hydrological feature map; Input satellite image data and topographic map data into the second feature extraction module to generate a satellite image feature map and a corresponding satellite image anomaly feature value for the satellite image feature map; If the meteorological and hydrological anomaly feature value exceeds a preset meteorological and hydrological anomaly threshold or the satellite image anomaly feature value exceeds a preset satellite image anomaly threshold, input the meteorological and hydrological feature map and the satellite image feature map into the feature fusion module to generate a fused feature map; Input the fused feature map into the feature decoding module to generate target monitoring area risk level information and target monitoring area detection range information.

[0119] In an alternative embodiment, the UAV operation control module 903 may further be configured to: Control each UAV to fly along a corresponding preset route to obtain target monitoring sub-scene image data; Input the target monitoring sub-scene image data into the image stitching edge artificial intelligence model for image stitching to generate target monitoring data.

[0120] In an embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of an UAV monitoring method based on artificial intelligence as described above are implemented.

[0121] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0122] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial descriptions of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0123] The above-described embodiments merely represent several implementation manners of the embodiments of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the embodiments of the present application.

Claims

1. An unmanned aerial vehicle monitoring method for levee and dam breach scenarios based on artificial intelligence, characterized in that, The method includes: Obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, the meteorological data, the satellite image data, and the topographic map data into a monitoring area recognition artificial intelligence model to perform target detection area recognition, and generate target monitoring area risk level information and target monitoring area detection range information; Based on the target monitoring area risk level information and the target monitoring area detection range information, allocate the number of unmanned aerial vehicles and the preset flight routes of each unmanned aerial vehicle; Control each unmanned aerial vehicle to fly along the corresponding preset flight route to obtain the target monitoring data of each unmanned aerial vehicle; Input the target monitoring data of each unmanned aerial vehicle into a dam break and levee breach scenario analysis artificial intelligence model to perform dam break and levee breach scenario analysis, and generate a dam break and levee breach scenario analysis result, where the dam break and levee breach scenario analysis result includes dam break and levee breach risk information and / or dam break and levee breach disaster information.

2. The method according to claim 1, wherein The allocating the number of unmanned aerial vehicles and the preset flight routes of each unmanned aerial vehicle based on the target monitoring area risk level information and the target monitoring area detection range information includes: Set the maximum operation duration of the unmanned aerial vehicle and the shooting range of the on-board monitoring device based on the target monitoring area risk level information; Calculate the maximum flight duration of the unmanned aerial vehicle based on the maximum distance between the initial position of the unmanned aerial vehicle and the target monitoring area detection range, where the maximum flight duration is used to represent the maximum duration for the unmanned aerial vehicle to reach the target monitoring area detection range from the initial position; Calculate the preset operation area of the unmanned aerial vehicle based on the maximum operation duration, the shooting range of the on-board monitoring device, and the maximum flight duration; Allocate the number of the unmanned aerial vehicles based on the preset operation area and the target monitoring area detection range information; Allocate the preset flight routes of each unmanned aerial vehicle based on the number of the unmanned aerial vehicles, the target monitoring area risk level information, and the target monitoring area detection range information.

3. The method according to claim 2, wherein The calculation formula for the number of the unmanned aerial vehicles is: ； Wherein, is the quantity of the UAVs, is the ceiling function, is the total area of the detection range of the target monitoring area, is the overlap correction factor of the operation track, is the maximum operation duration, is the maximum navigation duration, is the speed correction factor of the environmental conditions, is the average flight speed of the UAVs, is the width of the shooting range of the on-board monitoring device, is the fuzzy control membership function of the maximum operation duration, is the risk level of the target monitoring area, is the fuzzy control membership function of the shooting range of the on-board monitoring device.

4. The method according to claim 1, characterized in that The dam break and levee breach scenario analysis result includes second target monitoring area risk level information and second target monitoring area detection range information, and the method further includes: Calculate the risk level error between the target monitoring area risk level information and the second target monitoring area risk level information, and determine whether the risk level error exceeds the risk level deviation threshold; Calculate the detection range error between the target monitoring area detection range information and the second target monitoring area detection range information, and determine whether the detection range error exceeds the detection range deviation threshold; If the risk level error exceeds the risk level deviation threshold and / or the detection range error exceeds the detection range deviation threshold, adjust the number of the drones based on the risk level information of the second target monitoring area and the detection range information of the second target monitoring area, and update the preset flight path of each drone based on the adjusted number of the drones, the risk level information of the target monitoring area, the detection range information of the target monitoring area, and the monitored area information of the drones; wherein, the monitored area information is generated based on the target monitoring data of the drones. Adjust and optimize the monitoring area identification artificial intelligence model based on the risk level information of the target monitoring area, the detection range information of the target monitoring area, the risk level information of the second target monitoring area, and the detection range information of the second target monitoring area.

5. The method according to claim 1, wherein The artificial intelligence model for levee and dam break scenario analysis includes a levee and dam break risk analysis sub-model and a levee and dam break disaster prediction sub-model. The levee and dam break scenario analysis result includes the levee and dam break risk information and the levee and dam break disaster information. Inputting the target monitoring data of each drone into the artificial intelligence model for levee and dam break scenario analysis to perform levee and dam break scenario analysis and generate a levee and dam break scenario analysis result, including: Input the target monitoring data into the levee and dam break risk analysis sub-model to generate levee and dam break risk area information and the levee and dam break risk information corresponding to the levee and dam break risk area information. The levee and dam break risk information includes levee and dam break disaster labels. If the levee and dam break risk information is the levee and dam break disaster label, input the target monitoring data corresponding to the levee and dam break disaster label and the topographic map data into the levee and dam break disaster prediction sub-model to perform disaster prediction analysis and generate the levee and dam break disaster information.

6. The method according to claim 5, wherein: The levee and dam break risk analysis sub-model is a deep learning neural network model constructed based on the improved YOLOV8 model. The levee and dam break disaster analysis sub-model is a deep learning neural network model constructed based on the U-shaped network combined with a transformer module.

7. The method according to claim 1, wherein The method further includes: Input the target monitoring data into the edge artificial intelligence model for levee and dam break anomaly identification to perform levee and dam break anomaly identification and generate levee and dam break identification flag information, levee and dam break anomaly area information, and levee and dam break anomaly degree information. If the levee and dam break identification flag information indicates the existence of a levee and dam break, control the drone to hover and monitor the area corresponding to the levee and dam break anomaly area information. Perform validity determination of the risk level information of the target monitoring area and validity determination of the detection range information of the target monitoring area based on the levee and dam break anomaly area information, the levee and dam break anomaly degree information, the risk level information of the target monitoring area, and the detection range information of the target monitoring area. If the determination result of the validity determination of the risk level information of the target monitoring area is that the risk level information of the target monitoring area is invalid and / or the determination result of the validity determination of the detection range information of the target monitoring area is that the detection range information of the target monitoring area is invalid, an emergency report information on dike breach and dam break disasters is generated, and the emergency report information on dike breach and dam break disasters is used to characterize that the result output by the artificial intelligence model for identifying the monitoring area does not match the actual disaster situation.

8. The method according to claim 1, characterized in that, The artificial intelligence model for identifying the monitoring area includes a first feature extraction module, a second feature extraction module, a feature fusion module, and a feature decoding module. Inputting the hydrological data, the meteorological data, the satellite image data, and the topographic map data into the artificial intelligence model for identifying the monitoring area to perform target detection area identification and generate the risk level information of the target monitoring area and the detection range information of the target monitoring area includes: Inputting the hydrological data, the meteorological data, and the topographic map data into the first feature extraction module to generate a meteorological and hydrological feature map and a meteorological and hydrological anomaly feature value corresponding to the generated meteorological and hydrological feature map; Inputting the satellite image data and the topographic map data into the second feature extraction module to generate a satellite image feature map and a satellite image anomaly feature value corresponding to the satellite image feature map; If the meteorological and hydrological anomaly feature value exceeds the preset meteorological and hydrological anomaly threshold or the satellite image anomaly feature value exceeds the preset satellite image anomaly threshold, inputting the meteorological and hydrological feature map and the satellite image feature map into the feature fusion module to generate a fusion feature map; Inputting the fusion feature map into the feature decoding module to generate the risk level information of the target monitoring area and the detection range information of the target monitoring area.

9. The method according to any one of claims 1 to 8, characterized in that, Controlling each of the unmanned aerial vehicles to fly along the corresponding preset flight path to obtain the target monitoring data of each of the unmanned aerial vehicles includes: Controlling each of the unmanned aerial vehicles to fly along the corresponding preset flight path to obtain target monitoring sub-scene image data; Inputting the target monitoring sub-scene image data into the artificial intelligence model for image stitching at the edge to perform image stitching and generate the target monitoring data.

10. An unmanned aerial vehicle monitoring system for dam break scenarios based on artificial intelligence, characterized in that, The system includes: A monitoring area data analysis module, configured to obtain hydrological data, meteorological data, satellite image data, and topographic map data, and input the hydrological data, the meteorological data, the satellite image data, and the topographic map data into the artificial intelligence model for identifying the monitoring area to perform target detection area identification and generate the risk level information of the target monitoring area and the detection range information of the target monitoring area; An unmanned aerial vehicle work configuration module, configured to allocate the number of unmanned aerial vehicles and the preset flight path of each unmanned aerial vehicle based on the risk level information of the target monitoring area and the detection range information of the target monitoring area; An unmanned aerial vehicle work control module, configured to control each of the unmanned aerial vehicles to fly along the corresponding preset flight path to obtain the target monitoring data of each of the unmanned aerial vehicles; A target monitoring data processing module is configured to input the target monitoring data of each of the drones into an artificial intelligence model for analyzing the levee breach and dam break scenarios, perform the analysis of the levee breach and dam break scenarios, and generate an analysis result of the levee breach and dam break scenarios. The analysis result of the levee breach and dam break scenarios includes levee breach and dam break risk information and / or levee breach and dam break disaster information.

Images