Method and system for unmanned aerial vehicle and satellite cooperative water conservancy facility inspection

Abstract

The application discloses a method and system for unmanned aerial vehicle and satellite cooperative water conservancy project facility inspection, relates to the technical field of water conservancy project inspection, and obtains satellite long-time sequence thermal infrared, short-wave infrared and SAR multi-band remote sensing data, reverses ground temperature anomaly, soil humidity change and micro-deformation gradient characteristics; a temperature-humidity-deformation coupling leakage sensitive feature fusion model is constructed, a leakage probability index graph is calculated, and a high-probability leakage area is automatically circled; an unmanned aerial vehicle target flight path and load strategy are dynamically planned, the unmanned aerial vehicle is controlled to approach and carry out multi-mode fine detection; satellite macro data and unmanned aerial vehicle high-precision local data are fused, a leakage comprehensive criterion model is established, leakage activity degree and risk grade are quantitatively evaluated, and a three-dimensional risk map and an early warning report are output; the problems of inaccurate positioning of traditional satellite inspection, low efficiency of blind unmanned aerial vehicle inspection and high missing rate are solved, early accurate identification of water conservancy project leakage hidden dangers is realized, and the intelligent level of inspection is improved.

Description

Technical Field

[0001] This invention relates to the field of water conservancy project inspection technology, specifically to a method and system for the collaborative inspection of water conservancy facilities using unmanned aerial vehicles (UAVs) and satellites. Background Technology

[0002] Large reservoirs, dikes, and other water conservancy projects are subject to long-term hydrological, geological, and environmental factors, making them prone to internal seepage, soil loosening, and micro-cracks. If these issues are not accurately identified early on, they can easily lead to major safety accidents such as piping, landslides, or even dam failures. Currently, water conservancy seepage inspections mainly rely on satellite remote sensing and independent drone operations. Satellite remote sensing, using technologies such as thermal infrared and InSAR, can achieve large-scale, wide-area continuous monitoring, capturing macroscopic features such as regional temperature anomalies and minor surface deformations. However, due to limitations in sensor resolution and satellite transit cycles, it suffers from poor positioning accuracy and lack of detail, making it difficult to accurately pinpoint the specific location of seepage and quantify its extent and development.

[0003] Drones possess advantages such as maneuverability, flexibility, and high-precision close-range observation. They can carry various sensors to acquire detailed surface data. However, current drone inspections mostly adopt a blind flight mode with a grid-like layout across the entire area, lacking guidance on macro-risk areas. This results in excessive redundancy in inspection paths, low operational efficiency, and a tendency to miss key areas with hidden leaks. Currently, the two monitoring methods operate independently with fragmented data, failing to form a collaborative mechanism of wide-area satellite screening combined with targeted drone inspection. Furthermore, the sensitive features related to leaks, such as temperature, humidity, and deformation, in multi-temporal satellite data are not fully explored, hindering the dynamic driving of intelligent drone task planning. Ultimately, this leads to industry pain points in water conservancy leak inspection, including inaccurate satellite coverage, incomplete drone flights, and delayed hazard identification, making it difficult to meet the actual needs of early warning and dynamic control of leak hazards in large-scale water conservancy projects. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for the collaborative inspection of water conservancy engineering facilities using unmanned aerial vehicles (UAVs) and satellites, in order to solve the aforementioned problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for inspecting water conservancy engineering facilities using a drone and satellite collaboration, comprising the following steps:

[0006] Step 1: Acquire long-term multi-band remote sensing data from satellites, including at least thermal infrared band data, shortwave infrared band data, and SAR band data;

[0007] Step 2: Based on the multi-band remote sensing data, the surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data are retrieved.

[0008] Step 3: The surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data are coupled and calculated using the leakage-sensitive feature fusion model to generate a leakage probability index map. Based on the leakage probability index map, the boundaries and centroid coordinates of high-probability leakage areas are automatically delineated. The leakage-sensitive feature fusion model is a temperature gradient-humidity-deformation coupling model.

[0009] Step 4: Based on the boundary and centroid coordinates of the high-probability leakage area, dynamically generate the UAV's multi-waypoint inspection path and payload triggering strategy; the payload triggering strategy includes the type of detection payload carried by the UAV and the triggering time, and the detection payload includes at least one of thermal imaging camera, ground penetrating radar, and hyperspectral imager;

[0010] Step 5: The UAV flies along the multi-waypoint inspection route. After reaching the designated high-probability leakage area, it automatically adjusts its flight altitude and attitude to collect high-resolution data from multiple angles and modes at suspected leakage points.

[0011] Step 6: Spatial registration and feature-level fusion of the local fine data collected by the UAV and the satellite multi-band remote sensing data are performed to construct a comprehensive leakage criterion model; the comprehensive leakage criterion model includes quantitative indicators of absolute temperature, humidity infiltration rate, and crack width;

[0012] Step 7: Based on the comprehensive leakage criterion model, determine the activity level and risk level of each leakage point, and output a leakage risk distribution map with three-dimensional location markers and an early warning report.

[0013] Furthermore, the collection period for the long-term satellite multi-band remote sensing data mentioned in step 1 is set according to the scale of the water conservancy project facilities and the degree of susceptibility to leakage hazards, with a collection duration of not less than 30 days.

[0014] Furthermore, the inversion of the surface temperature anomaly field in step 2 adopts the thermal infrared radiation transfer model, the inversion of the relative change data of soil moisture adopts the shortwave infrared spectral index method, and the inversion of the micro-deformation gradient data adopts the InSAR interferometric measurement technology.

[0015] Furthermore, in the calculation process of the leakage probability index map in step 3, the surface temperature anomaly field, the relative change data of soil moisture and the micro-deformation gradient data are normalized respectively, and then the leakage probability index is obtained by weighted summation. The weights are dynamically adjusted according to the geological structure characteristics of the water conservancy project.

[0016] Furthermore, the multi-waypoint inspection path generation in step 4 adopts an improved A* algorithm, which uses the centroid coordinates of high-probability leakage areas as core waypoints, taking into account both flight efficiency and coverage integrity, thus shortening the inspection path length by more than 70% compared to conventional gridded UAV inspection.

[0017] Furthermore, in step 5, the UAV automatically adjusts its flight altitude and attitude, specifically by adjusting the flight altitude to 5-50 meters and the attitude to form a 30°-60° angle with the slope where the suspected leakage point is located, for different detection payloads, to ensure the clarity and completeness of multimodal data acquisition.

[0018] Furthermore, the spatial registration in step 6 adopts a registration algorithm based on feature point matching to unify the spatial coordinates of the local fine data collected by the UAV and the satellite multi-band remote sensing data into the same coordinate system; the feature-level fusion adopts a weighted fusion algorithm, and the fusion weight is dynamically allocated according to the data resolution and reliability.

[0019] Furthermore, the leakage risk level described in step 7 is divided into four levels: no risk, low risk, medium risk, and high risk. Each level of risk corresponds to a specific absolute temperature threshold, humidity penetration rate threshold, and crack width threshold, and the leakage identification accuracy rate is no less than 95%.

[0020] A drone and satellite collaborative water conservancy facility inspection system is used to implement the inspection method described in any one of claims 1-8. The system includes a satellite data acquisition module, a data inversion module, a leakage area delineation module, a drone mission planning module, a drone data acquisition module, a data fusion module, and a risk assessment module.

[0021] The satellite data acquisition module is used to acquire multi-band remote sensing data of satellites over a long period of time.

[0022] The data inversion module is used to invert surface temperature anomaly field, relative change data of soil moisture and micro-deformation gradient data based on multi-band remote sensing data;

[0023] The leakage area delineation module is used to calculate the leakage probability index map through the leakage sensitive feature fusion model, and delineate the boundary and centroid coordinates of high-probability leakage areas.

[0024] The UAV mission planning module is used to generate UAV multi-waypoint inspection paths and payload triggering strategies based on high-probability leakage area information.

[0025] The UAV data acquisition module is used to control the UAV to fly along the inspection path, adjust the flight attitude, and collect multimodal high-resolution data of suspected leakage points.

[0026] The data fusion module is used to achieve spatial registration and feature-level fusion of local fine data from UAVs and macroscopic background data from satellites, and to construct a comprehensive leakage judgment model.

[0027] The risk assessment module is used to determine the risk level of leakage points based on the comprehensive leakage criterion model, and output a leakage risk distribution map and early warning report.

[0028] Furthermore, the system also includes a data storage module for storing satellite remote sensing data, UAV-collected data, inversion data, leakage probability index maps, leakage risk distribution maps, and early warning reports. The data storage module supports real-time data updates and historical data retrieval. The UAV data acquisition module is connected to the UAV and supports automatic triggering of the payload and real-time data transmission, thereby reducing the leakage point location error to the centimeter level.

[0029] Compared with existing technologies, the UAV and satellite collaborative inspection method and system for water conservancy engineering facilities provided by this invention have the following advantages:

[0030] A new mechanism for collaborative inspection by satellites and drones has been established. Based on long-term multi-source remote sensing data from satellites, the initial screening of suspected leakage areas in a wide area is completed, driving drones to conduct targeted and refined inspections. This completely abandons the traditional blind grid-based flight mode of drones, reducing the length of inspection paths by more than 70%, significantly reducing flight energy consumption and operation time, and significantly improving the overall efficiency of water conservancy facility inspections.

[0031] By integrating multi-dimensional leakage-sensitive features such as temperature gradient, soil moisture, and micro-deformation, and quantifying leakage probability through a coupled model, combined with UAV thermal imaging, ground-penetrating radar, and hyperspectral multimodal fine detection, the leakage point location error is reduced from tens of meters at the satellite level to centimeters. The comprehensive identification accuracy of leakage hazards is improved to over 95%, effectively reducing the problems of missed detection and false detection.

[0032] It achieves spatial registration and feature fusion of multi-source heterogeneous data, establishes standardized comprehensive quantitative criteria for leakage, and can quantitatively analyze leakage rate, crack development degree and active status of hidden dangers. It can complete risk classification assessment and dynamic trend prediction, support quantitative estimation of daily leakage volume, and realize the upgrade of leakage hidden dangers from qualitative screening to quantitative assessment.

[0033] Automatically generating three-dimensional risk distribution maps and standardized early warning reports provides intuitive and accurate decision-making basis for water conservancy operation and maintenance. It can intervene in the early minor leakage hazards in advance, reduce the probability of major disasters such as dike piping and dam failure from the source, ensure the long-term safe and stable operation of water conservancy projects, and at the same time reduce the cost of manual inspection and operation and maintenance management pressure. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0035] Figure 1 This is a block diagram of the overall system structure of the present invention;

[0036] Figure 2 This is a schematic diagram of the satellite multi-band data inversion and leakage probability index generation process of the present invention. Detailed Implementation

[0037] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0038] Please see Figure 1 , 2 The method for inspecting water conservancy facilities using drones and satellites in a coordinated manner includes the following steps:

[0039] Step 1: Acquire long-term multi-band remote sensing data from satellites. The multi-band remote sensing data should include at least thermal infrared band data, shortwave infrared band data, and SAR band data.

[0040] Long-term satellite data refers to continuously and uninterrupted satellite data, with flexible acquisition cycles. Its purpose is to capture the dynamic changes in leakage risks from their initial stages to their development, avoiding the randomness of single data points. Thermal infrared data is primarily used to capture surface temperature differences, as leakage areas exhibit significant temperature anomalies compared to surrounding normal areas due to groundwater infiltration. This data is acquired through satellite-borne thermal infrared sensors. Short-wave infrared data is used to invert soil moisture. The principle is that soil moisture content affects short-wave infrared reflectivity; higher moisture content results in lower reflectivity. This data is acquired through satellite short-wave infrared sensors. SAR data is used to capture micro-deformations of the surface. Utilizing the interference effect of radar waves, it detects minute displacements on the surface, ranging from millimeters to centimeters, revealing micro-deformations such as soil loosening and settlement caused by leakage. This data is acquired through satellite-borne synthetic aperture radar. The coordinated acquisition of these three bands of data comprehensively covers the three core characteristics of leakage: temperature, humidity, and deformation. This provides a comprehensive data foundation for subsequent identification of suspected leakage areas, overcoming the limitations of single-band data and achieving all-round capture of leakage-sensitive features.

[0041] Step 2: Based on multi-band remote sensing data, the surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data are obtained by inversion.

[0042] Transforming the raw satellite remote sensing data acquired in Step 1 into quantitative characteristic data that can be directly used to determine leakage is a crucial step connecting the raw data with leakage identification. Through mature remote sensing inversion technology, the electromagnetic signals captured by satellite sensors are converted into surface parameters with practical physical meaning. Among these, the surface temperature anomaly field refers to the spatial distribution field formed by the deviation of the surface temperature from the surrounding normal area, used to visually represent temperature anomaly areas. Its inversion uses a thermal infrared radiation transfer model, with the specific formula as follows:

[0043]

[0044] In the formula, This represents the actual surface temperature. Radiance in the thermal infrared band The corresponding Planck function inversion value, Surface emissivity; relative soil moisture variation data refers to the difference between soil moisture in the detection area and historical baseline moisture, used to reflect the dynamic changes in soil moisture. Its inversion uses the shortwave infrared spectral index method, with the specific formula as follows:

[0045]

[0046] In the formula, The normalized moisture index, The surface reflectance is at the 1.24μm band. The surface reflectance in the 1.65μm band is used to calculate the NDWI difference at different time points, thus obtaining the relative change data of soil moisture. Microdeformation gradient data refers to the spatial rate of change of surface microdeformation, used to reflect the degree and trend of surface deformation caused by seepage. Its inversion uses InSAR interferometry technology. By interferometric processing of two SAR images acquired at different times in the same area, an interferometric phase map is obtained. Then, through steps such as phase unwrapping and geocoding, the surface microdeformation data is inverted, and the microdeformation gradient is calculated. The specific formula is as follows: In the formula, For micro-deformation gradient, For the slight variations between two adjacent detection points, This represents the distance between two adjacent detection points. The raw satellite data is transformed into quantifiable and analyzable leakage-sensitive characteristic data, providing a direct basis for subsequent leakage probability calculations. Simultaneously, it enables dynamic capture of leakage characteristics, avoiding the problem that static data cannot reflect the development trend of leakage.

[0047] Step 3: The leakage-sensitive feature fusion model is used to couple and calculate the surface temperature anomaly field, soil moisture relative change data and micro-deformation gradient data to generate a leakage probability index map. Based on the leakage probability index map, the boundary and centroid coordinates of high-probability leakage areas are automatically delineated. The leakage-sensitive feature fusion model is a temperature gradient-humidity-deformation coupling model.

[0048] This step addresses the problem in existing technologies that fail to utilize multi-temporal satellite data to correlate leakage-sensitive features, thus hindering the accurate identification of suspected leakage areas. It achieves precise quantification of leakage probability through multi-feature coupling, providing clear guidance for targeted UAV inspections. The principle of the leakage-sensitive feature fusion model is as follows: the occurrence of potential leakage simultaneously leads to abnormal surface temperature gradients, increased soil moisture, and intensified surface micro-deformation. These three factors are significantly coupled, and a single feature cannot accurately determine leakage. By coupling these three factors, interference factors can be eliminated, improving the accuracy of leakage identification. First, the surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data obtained in step 2 are normalized to unify data of different dimensions into the [0,1] interval, eliminating the influence of dimensional differences on the calculation results. The normalization formula is:

[0049]

[0050] In the formula, For the normalized data, This is the original data. The minimum value of this type of data. The maximum value of this type of data is used; then, a weighted summation is used to achieve coupled calculation to obtain the leakage probability index, the specific formula of which is:

[0051]

[0052] In the formula, This is the leakage probability index. This is the normalized surface temperature gradient data. This is the normalized data on relative changes in soil moisture. The normalized micro-deformation gradient data, , , The weights for the three features are dynamically adjusted based on the geological structure characteristics of the hydraulic engineering project. The leakage probability index map is a visual map drawn according to spatial coordinates, based on the calculated leakage probability index of each detection point. By setting a leakage probability threshold, areas with an index greater than the threshold are identified as high-probability leakage areas. Then, the boundary and centroid coordinates of the area are automatically delineated using an image segmentation algorithm. This achieves automatic and accurate identification and quantitative classification of suspected leakage areas, solving the problem of insufficient positioning accuracy of satellite remote sensing data and the inability to accurately delineate leakage areas. At the same time, through multi-feature coupling, the probability of misjudgment based on a single feature is significantly reduced, providing clear target guidance for subsequent UAV targeted inspection and laying the foundation for a collaborative mechanism of "satellite wide-area initial screening → UAV targeted fine inspection".

[0053] Step 4: Based on the boundary and centroid coordinates of the high-probability leakage area, dynamically generate the UAV's multi-waypoint inspection path and payload triggering strategy; the payload triggering strategy includes the type of detection payload carried by the UAV and the triggering time, and the detection payload includes at least one of the following: thermal imaging camera, ground penetrating radar, and hyperspectral imager;

[0054] This step addresses the problems of blind flight, low efficiency, and easy omission of key areas in existing UAV inspections. Based on the high-probability leakage areas identified in step 3, it plans the optimal inspection path for the UAV and clarifies the detection scheme, ensuring that the UAV only performs precise detection on high-probability leakage points. Specifically, the generation of the multi-waypoint inspection path uses an improved A* algorithm. Its core principle is: using the centroid coordinates of the high-probability leakage areas as core waypoints and the area boundaries as constraints, the planned path can completely cover all high-probability leakage areas while avoiding obstacles, ensuring flight safety and efficiency. Specifically, the algorithm solves for the minimum cost path by setting a path cost function, ultimately generating an inspection path composed of multiple waypoints. This shortens the inspection path length by more than 70% compared to conventional gridded UAV inspections, significantly improving inspection efficiency and avoiding unnecessary flights. The payload triggering strategy refers to pre-setting the operating mode and triggering timing of the detection payloads carried by the UAV based on the characteristics of high-probability leakage areas. The principle is that different detection payloads have different advantages and need to be specifically matched to achieve multimodal, high-precision detection: Thermal imaging cameras are used to capture local temperature anomalies in leakage areas, detecting temperature differences at the centimeter level; the triggering timing is after the UAV reaches the airspace above the high-probability leakage area and adjusts to a suitable flight attitude. Ground-penetrating radar is used to detect the moisture content and crack distribution of underground soil, directly detecting underground leakage channels; the triggering timing is when the UAV flies directly above the suspected leakage point and adjusts its altitude to 10-30 meters. Hyperspectral imagers are used to capture the spectral characteristics of the soil; by analyzing differences in spectral curves, soil moisture and pollutant distribution are determined; the triggering timing is synchronized with the thermal imaging camera. This achieves precise and efficient UAV inspections, avoiding blind flights. Simultaneously, through targeted payload triggering strategies, it ensures that the collected local high-precision data can accurately compensate for the deficiencies of satellite data, providing high-quality local data support for subsequent leakage risk assessment.

[0055] Step 5: The UAV flies along a multi-waypoint inspection route. After reaching the designated high-probability leakage area, it automatically adjusts its flight altitude and attitude to collect high-resolution data from multiple angles and modes at suspected leakage points.

[0056] The purpose of this step is to perform targeted precision inspection using a drone, collecting detailed local data from high-probability leakage areas. This compensates for the low resolution and inability of satellite remote sensing data to capture local details. The principle is to use the drone's autonomous control function to precisely adjust its flight attitude and altitude, ensuring the clarity and completeness of multimodal data acquisition. Specifically, the drone's flight control is based on the GPS / BeiDou positioning system, allowing it to fly precisely along the multi-waypoint path planned in step 4. Upon reaching the designated high-probability leakage area, the drone's onboard attitude sensor detects the flight attitude in real time, and the flight control system automatically adjusts the altitude and attitude: for different detection payloads, the flight altitude is adjusted to 5-50 meters; the attitude is adjusted to an angle of 30°-60° with the slope where the suspected leakage point is located. This angle avoids interference from sunlight reflection and slope obstruction, ensuring multi-angle data acquisition and comprehensively capturing the detailed features of the suspected leakage point. Multimodal high-resolution data refers to temperature data, underground detection data, and spectral data collected by thermal imaging cameras, ground-penetrating radar, and hyperspectral imagers, respectively. Temperature data resolution can reach 0.1℃, underground detection data resolution can reach centimeter-level, and spectral data resolution can reach nanometer-level, significantly improving accuracy compared to satellite data. This enables refined, multi-angle, and multi-dimensional data collection of suspected leakage points, reducing the location accuracy of leakage points from tens of meters on satellites to centimeter-level. This provides high-precision local data support for subsequent data fusion and risk assessment. Furthermore, by autonomously adjusting flight attitude and altitude, it reduces human intervention and improves the automation level of inspections.

[0057] Step 6: Spatial registration and feature-level fusion of the local fine data collected by the UAV with the satellite multi-band remote sensing data are performed to construct a comprehensive leakage criterion model; the comprehensive leakage criterion model includes quantitative indicators of absolute temperature, humidity infiltration rate, and crack width;

[0058] This technology addresses the disconnect and inability to coordinate satellite and UAV data in existing technologies. By integrating macroscopic background data with local details, it constructs a scientific and comprehensive leakage assessment standard, providing a basis for judging leakage risk levels. Specifically, spatial registration unifies the spatial coordinates of detailed local data collected by UAVs and multi-band satellite remote sensing data into the same coordinate system, avoiding data misalignment caused by coordinate inconsistencies. It employs a feature point matching-based registration algorithm, extracting feature points from both satellite and UAV imagery. Through feature point matching, it calculates the coordinate transformation parameters between the two, transforming the UAV data coordinates to the satellite data coordinate system, ensuring precise spatial correspondence with registration errors controlled to the centimeter level. Feature-level fusion deeply integrates the macroscopic features of satellite data with the detailed local features of UAV data. A weighted fusion algorithm is used, with fusion weights dynamically allocated based on data resolution and reliability, ensuring that the fused data possesses both macroscopic background support and local detail accuracy. The comprehensive leakage judgment model is a quantitative judgment standard built based on fused data. It extracts key quantitative indicators of potential leakage hazards and sets clear thresholds to determine the activity level and risk grade of leakage. Absolute temperature refers to the actual temperature of suspected leakage points collected by the UAV thermal imaging camera, in °C; moisture infiltration rate refers to the rate at which water infiltrates the soil, reflecting the severity of leakage, and is calculated using UAV ground-penetrating radar data and soil moisture data. The specific formula is as follows: In the formula, For humidity penetration rate, Soil permeability coefficient, The difference in water level above and below the leakage point. The length of the seepage path is denoted as ; the crack width refers to the actual width of the surface cracks captured by the UAV hyperspectral imager, calculated using an image recognition algorithm. This approach achieves the synergistic fusion of macroscopic satellite data and detailed local UAV data, breaking down information barriers between the two. The constructed comprehensive seepage criterion model provides a scientific and quantitative standard for judging seepage risk levels, avoiding the limitations of relying on single-data judgments and further improving the accuracy of seepage identification.

[0059] Step 7: Based on the comprehensive leakage criterion model, determine the activity level and risk level of each leakage point, and output a leakage risk distribution map with three-dimensional location markers and an early warning report.

[0060] This step is the final stage of the entire inspection process. Its purpose is to accurately assess and visualize potential leakage risks, providing clear decision-making basis for water conservancy project operation and maintenance personnel, and solving the problems of existing inspections being unable to quantify leakage risks and provide accurate early warnings. The activity level of a leakage point refers to the development speed and intensity of potential leakage, dynamically judged through quantitative indicators in the comprehensive leakage judgment model. Leakage risk levels are divided into four levels: no risk, low risk, medium risk, and high risk, each corresponding to a specific quantitative threshold: no risk, low risk, medium risk, and high risk. By comparing the quantitative indicators of each leakage point with the above thresholds, its risk level can be determined. This method can improve the leakage identification accuracy rate to over 95%. The leakage risk distribution map with 3D location markers is a visualized map formed by overlaying the 3D coordinates and risk level of each leakage point onto a satellite image map. Operation and maintenance personnel can intuitively view the specific location, distribution range, and risk level of leakage points. The early warning report includes detailed information on each leakage point, leakage development trend predictions, and operation and maintenance suggestions. It enables quantitative assessment, dynamic monitoring, and visualization of leakage risks, accurately determines the activity level and risk level of leakage points, and achieves dynamic quantitative assessment of leakage trends. This provides accurate decision-making basis for the operation and maintenance of water conservancy projects and significantly reduces the risk of dam failure or piping in large-scale water conservancy facilities.

[0061] In step 1, the collection period for the long-term satellite multi-band remote sensing data is set according to the scale of the water conservancy project facilities and the degree of susceptibility to leakage, with a collection duration of no less than 30 days.

[0062] The data collection cycle is set based on the scale of the water conservancy project facilities and the likelihood of leakage risks. The core objective is to ensure timely capture of the entire process of leakage risks from their inception to development. The collection period is no less than 30 days because the development of leakage risks typically requires a certain amount of time. Insufficient collection time prevents the formation of effective long-term data series, making it impossible to deduce the dynamic trends of soil moisture, surface temperature, and micro-deformation, and consequently, accurately assess the development of leakage risks. Specific standards for satellite data collection have been defined to ensure the completeness and continuity of the collected data, providing reliable foundational data for subsequent data inversion and leakage probability calculations, further enhancing the stability and accuracy of the entire inspection method.

[0063] In step 2, the inversion of the surface temperature anomaly field uses the thermal infrared radiation transfer model, the inversion of the relative change data of soil moisture uses the shortwave infrared spectral index method, and the inversion of the micro-deformation gradient data uses InSAR interferometry.

[0064] Among them, the thermal infrared radiation transfer model is a mature temperature inversion method in the field of remote sensing. By simulating the transmission process of thermal infrared radiation in the atmosphere and eliminating atmospheric interference, it inverts the true surface temperature from the radiance captured by satellite sensors. Compared with other temperature inversion methods, this model has higher inversion accuracy and is suitable for inverting large-scale temperature anomaly fields in large-scale water conservancy projects. The shortwave infrared spectral index method is a mature soil moisture inversion method. It utilizes the influence of soil moisture on shortwave infrared reflectivity and converts reflectivity data into soil moisture data by constructing a spectral index. This method is simple to operate, fast inversion speed, and can effectively eliminate the interference of vegetation cover, making it suitable for inverting large-scale relative changes in soil moisture. InSAR interferometry is a mature micro-deformation inversion technology. Its principle is to use the interference effect of radar waves. By interferometric processing of SAR images of the same area at different times, it inverts the surface micro-deformation data. The advantages of this technology are high detection accuracy and no weather influence, making it suitable for long-term monitoring of surface micro-deformation in water conservancy projects. The inversion methods for various types of data were clarified, ensuring the standardization and repeatability of the inversion process, while improving the accuracy and reliability of the inversion data, providing high-quality feature data for subsequent leakage probability calculation.

[0065] In the calculation of the leakage probability index in step 3, the surface temperature anomaly field, the relative change data of soil moisture and the micro-deformation gradient data are normalized respectively, and then the leakage probability index is obtained by weighted summation. The weights are dynamically adjusted according to the geological structure characteristics of the water conservancy project.

[0066] The purpose of normalization is to eliminate the influence of data with different dimensions, enabling coupled calculations of the three, and avoiding calculation biases caused by differences in dimensions. The specific formula has been detailed above and will not be repeated here. Weighted summation assigns different weights based on the contribution of different features to seepage identification. These weights are dynamically adjusted according to the geological structure characteristics of the hydraulic engineering project. In areas with hard geological structures, such as rock embankments and reservoir dams, seepage mainly manifests as surface micro-deformation; therefore, the weight of micro-deformation gradient data can be set to 0.4-0.5, and the weights of temperature gradient and humidity can each be set to 0.25-0.3. In areas with loose geological structures, such as soil embankments and reservoir perimeters, seepage mainly manifests as increased soil moisture; therefore, the weight of relative soil moisture change data can be set to 0.4-0.5, and the weights of temperature gradient and micro-deformation gradient can each be set to 0.25-0.3. In areas with mixed geological structures, the weights of the three can be evenly distributed. This makes the calculation of the leakage probability index more accurate and flexible, adaptable to water conservancy projects with different geological structures, further reducing the probability of misjudgment of leakage and improving the accuracy of delineating high-probability leakage areas.

[0067] In step 4, the multi-waypoint inspection path is generated using an improved A* algorithm, which takes the centroid coordinates of high-probability leakage areas as the core waypoints, balancing flight efficiency and coverage integrity, thus shortening the inspection path length by more than 70% compared to conventional gridded UAV inspection.

[0068] The improved A* algorithm, based on the traditional A* algorithm, adds path coverage integrity constraints and obstacle avoidance mechanisms. It uses the centroid coordinates of high-probability leakage areas as core waypoints and the area boundaries as constraints. By setting a path cost function, it solves for the minimum-cost path, ensuring that the generated path completely covers all high-probability leakage areas while avoiding obstacles and ineffective flights. Using the centroid coordinates of high-probability leakage areas as core waypoints is crucial because they represent the core location of the area, ensuring the path's focus and avoiding omissions of key areas. Balancing flight efficiency and coverage integrity means minimizing path length while ensuring coverage of all high-probability leakage areas, reducing UAV flight time and energy consumption. Conventional gridded UAV inspections do not differentiate area priorities and fly according to a fixed grid, covering a large number of areas without leakage risk. This invention, through the improved A* algorithm, plans paths that only cover high-probability leakage areas, thus significantly shortening path length and improving inspection efficiency. The specific algorithms and principles for path generation have been clarified to ensure the relevance, efficiency, and completeness of the paths, thereby further improving the efficiency of drone inspections and reducing operation and maintenance costs.

[0069] In step 5, the UAV automatically adjusts its flight altitude and attitude. Specifically, for different detection payloads, the flight altitude is adjusted to 5-50 meters, and the attitude is adjusted to form an angle of 30°-60° with the slope where the suspected leakage point is located, to ensure the clarity and completeness of multimodal data acquisition.

[0070] The flight altitude adjustment to 5-50 meters is set according to the working characteristics of different detection payloads: the optimal working altitude for thermal imaging cameras is 5-20 meters, which ensures temperature detection accuracy while avoiding insufficient acquisition range due to close proximity and blurred details due to excessive distance; the optimal working altitude for ground penetrating radar is 10-30 meters, which ensures radar wave penetration while avoiding detection errors caused by ground interference; and the optimal working altitude for hyperspectral imagers is 20-50 meters, which ensures spectral acquisition coverage while maintaining spectral resolution. The drone's attitude was adjusted to form an angle of 30°-60° with the slope where the suspected leakage point was located. If the angle was too small, the drone's flight direction would be parallel to the slope, easily causing slope obstruction and making it impossible to fully capture the details of the suspected leakage point. If the angle was too large, the drone's flight direction would be perpendicular to the slope, easily affected by sunlight reflection and slope glare, affecting the clarity of data acquisition. An angle of 30°-60° can effectively avoid the above interference, ensuring that the drone can collect multimodal data of the suspected leakage point from multiple angles and comprehensively capture detailed features. Specific standards for the drone's flight altitude and attitude were clarified to ensure the clarity and completeness of multimodal data acquisition, further improving the accuracy of local fine data and providing reliable support for subsequent data fusion and risk assessment.

[0071] In step 6, spatial registration adopts a registration algorithm based on feature point matching to unify the spatial coordinates of the local fine data collected by the UAV and the satellite multi-band remote sensing data into the same coordinate system; feature-level fusion adopts a weighted fusion algorithm, and the fusion weights are dynamically allocated according to the data resolution and reliability.

[0072] Feature point matching-based registration algorithms are mature in the field of remote sensing data fusion. First, feature points are extracted from satellite and UAV imagery. The SIFT algorithm is used to extract descriptors for these feature points. Then, a feature point matching algorithm is used to find corresponding feature points in the satellite and UAV imagery. Next, coordinate transformation parameters are calculated, and finally, the coordinates of the UAV data are transformed to the coordinate system of the satellite data, ensuring precise spatial correspondence with registration errors controlled to the centimeter level. The advantages of this algorithm are high registration accuracy, strong anti-interference ability, and adaptability to image registration under different lighting and weather conditions. Weighted fusion algorithms are commonly used for feature-level fusion. Fusion weights are dynamically assigned based on the resolution and reliability of the satellite and UAV data. UAV data has high resolution and reliability, so a fusion weight of 0.6-0.7 is suitable. Satellite data has lower resolution but wider coverage and provides a macroscopic background, so a fusion weight of 0.3-0.4 is suitable. The dynamic allocation of weights can be adjusted according to the actual quality of the data. The specific algorithms and implementation methods for data fusion were clarified to ensure that satellite data and UAV data can be accurately fused, give full play to the advantages of both, improve the quality of fused data, and provide reliable support for the construction of a comprehensive leakage judgment model.

[0073] In step 7, the leakage risk level is divided into four levels: no risk, low risk, medium risk, and high risk. Each level of risk corresponds to a specific absolute temperature threshold, humidity penetration rate threshold, and crack width threshold. The leakage identification accuracy rate is no less than 95%.

[0074] The leakage risk level is divided into four levels, based on the actual severity of leakage hazards in water conservancy projects: No risk indicates no obvious leakage signs and no maintenance measures are required; low risk indicates slight leakage signs with slow development, requiring regular monitoring; medium risk indicates obvious leakage signs with rapid development, requiring timely investigation; high risk indicates serious leakage signs with active leakage, which could easily lead to safety accidents such as dam failure and piping, requiring immediate emergency measures. The quantitative thresholds corresponding to each risk level are derived from a large number of water conservancy project leakage case data and can be appropriately adjusted according to the importance of the water conservancy project. The specific thresholds have been explained in detail in step 7 and will not be repeated here. The setting of these thresholds ensures the objectivity and quantification of risk assessment and avoids errors from subjective judgment. The leakage identification accuracy is high, which has been verified through a large number of experiments. Compared with existing technologies, the accuracy is significantly improved. This method effectively eliminates interference factors and improves the accuracy of leakage identification by using satellite and UAV collaboration, multi-feature fusion, and quantitative criteria. The classification standards and quantitative thresholds for leakage risk levels have been clarified, ensuring that risk assessment is standardized and repeatable, while achieving high-accuracy leakage identification, providing a precise risk assessment basis for the operation and maintenance of water conservancy projects.

[0075] A UAV and satellite collaborative water conservancy facility inspection system is used to implement the inspection method of any one of claims 1-8. The system includes a satellite data acquisition module, a data inversion module, a leakage area delineation module, a UAV mission planning module, a UAV data acquisition module, a data fusion module, and a risk assessment module.

[0076] The satellite data acquisition module is used to acquire long-term, multi-band remote sensing data from satellites.

[0077] The data inversion module is used to invert surface temperature anomaly fields, relative changes in soil moisture, and micro-deformation gradient data based on multi-band remote sensing data.

[0078] The leakage area delineation module is used to calculate the leakage probability index map through the leakage sensitive feature fusion model, and delineate the boundary and centroid coordinates of high-probability leakage areas.

[0079] The UAV mission planning module is used to generate multi-waypoint inspection paths and payload triggering strategies for UAVs based on information about high-probability leakage areas.

[0080] The UAV data acquisition module is used to control the UAV to fly along the inspection path, adjust the flight attitude, and collect multimodal high-resolution data of suspected leakage points.

[0081] The data fusion module is used to achieve spatial registration and feature-level fusion of local fine data from UAVs and macroscopic background data from satellites, and to build a comprehensive leakage judgment criterion model.

[0082] The risk assessment module is used to determine the risk level of leakage points based on a comprehensive leakage criterion model, and output a leakage risk distribution map and early warning report.

[0083] By working collaboratively through various modules, each step of the inspection method is transformed into a feasible system function, addressing the lack of existing systems capable of achieving collaborative inspection from "wide-area satellite screening to targeted UAV inspection." The satellite data acquisition module is the system's fundamental data input unit. Its function is to establish a communication connection with the satellite data service provider, automatically acquire long-term multi-band remote sensing data, and perform preliminary preprocessing to ensure data integrity and usability. Its hardware can utilize a satellite data receiving antenna and data processing terminal, while the software can employ satellite data receiving and preprocessing programs. The data inversion module is the core data processing unit. Its function is to invoke a preset inversion algorithm to process data from the satellite data acquisition module. The input raw data is inverted to output surface temperature anomaly fields, relative changes in soil moisture, and micro-deformation gradient data. The hardware for this module can utilize a high-performance computing server, and the software can employ a data inversion algorithm program. The seepage area delineation module runs a seepage-sensitive feature fusion model, performs coupled calculations on the feature data output by the data inversion module, generates a seepage probability index map, and automatically delineates the boundaries and centroid coordinates of high-probability seepage areas using an image segmentation algorithm. The hardware for this module can utilize a graphics processing server, and the software can employ a seepage area recognition program. The UAV mission planning module is the core unit of UAV control. Based on the area information output by the seepage area delineation module, it calls an improved A* algorithm to generate multi-waypoint navigation data for the UAV. The system detects the path and payload triggering strategy, and sends relevant instructions to the UAV data acquisition module. The hardware of this module can be an embedded controller, and the software can be a path planning and payload control program. The UAV data acquisition module is the core data acquisition unit. It establishes a communication connection with the UAV, receives instructions from the UAV mission planning module, controls the UAV to fly along the inspection path, automatically adjusts its flight altitude and attitude, controls the activation and deactivation of the detection payload, acquires multimodal high-resolution data, and transmits the acquired data back to the data fusion module in real time. Its hardware can be a UAV flight control interface and data transmission module, and the software can be a UAV control and data acquisition program. The data fusion module is the core data fusion unit, receiving instructions from the UAV... The local fine-grained data input from the machine data acquisition module and the satellite data input from the satellite data acquisition module are combined with a registration algorithm based on feature point matching and a weighted fusion algorithm to achieve spatial registration and feature-level fusion, and to construct a comprehensive leakage criterion model. The hardware for this model can be a data fusion server, and the software can be a data fusion and model building program. The risk assessment module is the output unit of the system. It calls the comprehensive leakage criterion model, analyzes the fused data output from the data fusion module, judges the activity level and risk level of each leakage point, generates a leakage risk distribution map with three-dimensional location markers and an early warning report, and outputs the report to the operation and maintenance terminal. The hardware for this model can be an output terminal, and the software can be a risk assessment and report generation program.The various modules work together to form a complete collaborative inspection system, realizing the automation and intelligent implementation of inspection methods. It can efficiently and accurately complete the inspection of potential leakage hazards in water conservancy facilities, solve the problems of "disconnection between satellite and drone, low efficiency, and inaccurate identification" in existing inspection systems, and at the same time reduce manual intervention and improve the automation level and reliability of inspection.

[0084] The system also includes a data storage module for storing satellite remote sensing data, UAV-collected data, inversion data, leakage probability index maps, leakage risk distribution maps, and early warning reports. The data storage module supports real-time data updates and historical data review. The UAV data acquisition module is connected to the UAV and supports automatic triggering of the payload and real-time data transmission, reducing the leakage point location error to the centimeter level.

[0085] The system enhances its practicality and reliability by adding data storage functionality and further clarifying the communication and functional characteristics of the UAV data acquisition module, while quantifying the technical effects. The data storage module, an auxiliary unit, stores all data generated during system operation, preventing data loss and supporting real-time updates and historical data review. Its hardware can utilize large-capacity hard drives or cloud storage servers, while the software employs data storage and management programs. This module achieves full lifecycle management of data, providing data support for tracing, analyzing, and optimizing potential leaks. Communication between the UAV data acquisition module and the UAV can be achieved via wireless or 5G communication. 5G communication offers advantages such as low latency and high bandwidth, enabling real-time transmission of payload trigger commands and real-time feedback of collected data, avoiding reduced inspection efficiency due to data delays. Automatic payload triggering is supported, meaning the UAV data acquisition module can automatically activate the corresponding detection payload after the UAV arrives at the designated area based on the payload triggering strategy sent by the UAV mission planning module, without manual intervention. Real-time data feedback means that the collected multimodal high-resolution data can be transmitted to the data fusion module in real time for real-time processing and analysis. The ability to reduce leak point location error to the centimeter level is achieved by the UAV data acquisition module using GPS / BeiDou positioning systems to obtain the UAV's precise location. Combined with detailed local data collected by the UAV, the three-dimensional coordinates of the leak point can be accurately determined, reducing the leak point location error from tens of meters via satellite to the centimeter level. This significantly improves the accuracy of leak point location and provides clear location guidance for subsequent maintenance and troubleshooting. The system's functionality has been improved, enhancing its practicality, reliability, and automation level, further increasing the accuracy of leak point location and providing more comprehensive support for the operation and maintenance of water conservancy projects.

[0086] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles (UAVs) and satellites, characterized in that: Includes the following steps: Step 1: Acquire long-term multi-band remote sensing data from satellites, including at least thermal infrared band data, shortwave infrared band data, and SAR band data; Step 2: Based on the multi-band remote sensing data, the surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data are retrieved. Step 3: The surface temperature anomaly field, relative soil moisture change data, and micro-deformation gradient data are coupled and calculated using the leakage-sensitive feature fusion model to generate a leakage probability index map. Based on the leakage probability index map, the boundaries and centroid coordinates of high-probability leakage areas are automatically delineated. The leakage-sensitive feature fusion model is a temperature gradient-humidity-deformation coupling model. Step 4: Based on the boundary and centroid coordinates of the high-probability leakage area, dynamically generate the UAV's multi-waypoint inspection path and payload triggering strategy; the payload triggering strategy includes the type of detection payload carried by the UAV and the triggering time, and the detection payload includes at least one of thermal imaging camera, ground penetrating radar, and hyperspectral imager; Step 5: The UAV flies along the multi-waypoint inspection route. After reaching the designated high-probability leakage area, it automatically adjusts its flight altitude and attitude to collect high-resolution data from multiple angles and modes at suspected leakage points. Step 6: Spatial registration and feature-level fusion of the local fine data collected by the UAV and the satellite multi-band remote sensing data are performed to construct a comprehensive leakage criterion model; the comprehensive leakage criterion model includes quantitative indicators of absolute temperature, humidity infiltration rate, and crack width; Step 7: Based on the comprehensive leakage criterion model, determine the activity level and risk level of each leakage point, and output a leakage risk distribution map with three-dimensional location markers and an early warning report.

2. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles (UAVs) and satellites according to claim 1, characterized in that, The long-term satellite remote sensing data mentioned in step 1 has an acquisition cycle set according to the scale of the water conservancy project facilities and the susceptibility of leakage risks, with an acquisition duration of no less than 30 days.

3. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles and satellites according to claim 1, characterized in that, The inversion of the surface temperature anomaly field in step 2 uses the thermal infrared radiation transfer model, the inversion of soil moisture relative change data uses the shortwave infrared spectral index method, and the inversion of micro-deformation gradient data uses InSAR interferometry.

4. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles and satellites according to claim 1, characterized in that, In the calculation of the leakage probability index map in step 3, the surface temperature anomaly field, the relative change data of soil moisture and the micro-deformation gradient data are normalized respectively, and then the leakage probability index is obtained by weighted summation. The weights are dynamically adjusted according to the geological structure characteristics of the water conservancy project.

5. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles and satellites according to claim 1, characterized in that, The multi-waypoint inspection path generation in step 4 uses an improved A* algorithm, which takes the centroid coordinates of high-probability leakage areas as the core waypoints, taking into account both flight efficiency and coverage integrity, thus shortening the inspection path length by more than 70% compared to conventional gridded UAV inspection.

6. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles and satellites according to claim 1, characterized in that, In step 5, the UAV automatically adjusts its flight altitude and attitude. Specifically, for different detection payloads, the flight altitude is adjusted to 5-50 meters, and the attitude is adjusted to form an angle of 30°-60° with the slope where the suspected leakage point is located, to ensure the clarity and completeness of multimodal data acquisition.

7. The method for inspecting water conservancy engineering facilities using a combination of unmanned aerial vehicles and satellites according to claim 1, characterized in that, The spatial registration described in step 6 uses a registration algorithm based on feature point matching to unify the spatial coordinates of the local fine data collected by the UAV and the satellite multi-band remote sensing data into the same coordinate system; The feature-level fusion adopts a weighted fusion algorithm, and the fusion weights are dynamically allocated according to the data resolution and credibility.

8. The method for inspecting water conservancy facilities using a drone and satellite in collaboration according to claim 1, characterized in that, The leakage risk level described in step 7 is divided into four levels: no risk, low risk, medium risk, and high risk. Each level of risk corresponds to a specific absolute temperature threshold, humidity penetration rate threshold, and crack width threshold. The leakage identification accuracy rate is no less than 95%.

9. A drone-satellite collaborative inspection system for water conservancy engineering facilities, characterized in that: The system is used to implement the inspection method according to any one of claims 1-8, and includes a satellite data acquisition module, a data inversion module, a leakage area delineation module, a UAV mission planning module, a UAV data acquisition module, a data fusion module, and a risk assessment module. The satellite data acquisition module is used to acquire multi-band remote sensing data of satellites over a long period of time. The data inversion module is used to invert surface temperature anomaly field, relative change data of soil moisture and micro-deformation gradient data based on multi-band remote sensing data; The leakage area delineation module is used to calculate the leakage probability index map through the leakage sensitive feature fusion model, and delineate the boundary and centroid coordinates of high-probability leakage areas. The UAV mission planning module is used to generate UAV multi-waypoint inspection paths and payload triggering strategies based on high-probability leakage area information. The UAV data acquisition module is used to control the UAV to fly along the inspection path, adjust the flight attitude, and collect multimodal high-resolution data of suspected leakage points. The data fusion module is used to achieve spatial registration and feature-level fusion of local fine data from UAVs and macroscopic background data from satellites, and to construct a comprehensive leakage judgment model. The risk assessment module is used to determine the risk level of leakage points based on the comprehensive leakage criterion model, and output a leakage risk distribution map and early warning report.

10. The UAV and satellite collaborative water conservancy engineering facility inspection system according to claim 9, characterized in that, The system also includes a data storage module for storing satellite remote sensing data, UAV-collected data, inversion data, leakage probability index maps, leakage risk distribution maps, and early warning reports. The data storage module supports real-time data updates and historical data retrieval. The UAV data acquisition module is connected to the UAV and supports automatic payload triggering and real-time data transmission, reducing the leakage point location error to the centimeter level.

Images