Landslide disaster identification and early warning method and device based on unmanned aerial vehicle autonomous inspection

Abstract

The application provides a landslide disaster identification and early warning method and device based on unmanned aerial vehicle autonomous inspection, relates to the technical field of geological disaster monitoring and early warning, and comprises the following steps: generating an unmanned aerial vehicle autonomous inspection route according to obtained target landslide monitoring area basic information; the unmanned aerial vehicle autonomously flies according to the generated inspection route and collects multi-time-phase monitoring data; the monitoring data is processed, including constructing a multi-time-phase surface model, multi-time-phase spatial registration and difference analysis, and extracting landslide disaster characteristic parameters; based on the landslide disaster characteristic parameters, the risk of the target landslide monitoring area is identified and graded, and the effectiveness of the current inspection result is determined in combination with the quality evaluation result, and when the preset condition is met, landslide disaster early warning information is output; the application realizes non-contact, automatic and quantitative inspection monitoring of the landslide area, improves the accuracy of landslide disaster identification, the consistency of inspection operation and the timeliness of early warning response.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, specifically to a method and device for landslide disaster identification and early warning based on unmanned aerial vehicle (UAV) autonomous inspection. Background Technology

[0002] Landslides are a common type of geological hazard in mountainous areas, highway slopes, railway lines, water conservancy project areas, and open-pit excavation areas. They are characterized by their suddenness, high degree of concealment, complex evolution process, and wide range of damage. Before a landslide occurs, the slope usually shows a series of abnormal signs such as crack expansion, local bulging, surface subsidence, boundary changes, and vegetation disturbance. If these abnormal signs are not detected and accurately identified in a timely manner, they can easily cause damage to engineering structures, traffic disruptions, and loss of life and property. Therefore, establishing an efficient, accurate, and highly automated landslide disaster monitoring and early warning technology system has important engineering application value and practical significance.

[0003] Existing landslide monitoring methods mainly include manual inspection, total station monitoring, GPS monitoring, fixed video monitoring, tilt monitoring, and ground-based radar monitoring. While manual inspection is simple to implement, it is greatly affected by the experience of the operators, weather conditions, and terrain, resulting in low monitoring efficiency, high subjectivity, poor continuity, and high risk. Contact or point-based monitoring methods such as total stations, GPS, and tilt sensors, while offering some accuracy, typically only acquire a small amount of discrete data, making it difficult to comprehensively reflect the overall surface changes of the landslide body. Furthermore, the equipment deployment and maintenance costs are high, and implementation in complex mountainous areas or steep slopes is challenging. Fixed video monitoring and ground-based radar monitoring can achieve continuous observation to some extent, but they still suffer from limited field of view, high requirements for equipment installation conditions, numerous monitoring blind spots, and insufficient system deployment flexibility.

[0004] With the development of low-altitude remote sensing and intelligent inspection technologies, drones are increasingly being applied to the field of geological disaster monitoring due to their advantages such as high mobility, flexible deployment, large monitoring range, non-contact operation, and high data acquisition efficiency. By equipping themselves with sensors such as visible light cameras, infrared thermal imaging devices, depth cameras, or lidar, drones can acquire image information, terrain information, and surface morphology information of landslide areas, providing rich data support for landslide disaster identification. However, existing UAV-based landslide monitoring technologies still have some shortcomings: First, most schemes rely mainly on manual remote control flight or simple preset routes, making it difficult to achieve truly autonomous inspection and dynamic re-flight based on historical key monitoring areas and complex terrain environments; second, the consistency of data acquisition positions, attitudes, and distances between different inspection phases is poor, resulting in insufficient registration accuracy between multi-temporal data, which in turn affects the quantitative identification of key features such as landslide crack expansion, surface displacement, and boundary changes; third, existing methods mostly focus on single image interpretation or local anomaly detection, lacking a unified fusion analysis mechanism for multiple disaster signs such as cracks, settlement, displacement, and boundary changes, making it difficult to achieve a comprehensive judgment of landslide risk levels; fourth, some technical solutions only stay at the level of data acquisition or change detection, lacking evaluation mechanisms for acquisition quality, flight stability, positioning deviation, and data integrity, resulting in insufficient reliability and consistency of identification results.

[0005] Therefore, existing technologies suffer from problems such as insufficient automation, poor consistency of multi-temporal data, incomplete identification of landslide disaster characteristics, and insufficient timeliness and reliability of early warning, which affect the accuracy, efficiency, and intelligence level of landslide disaster monitoring. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes a method and device for landslide disaster identification and early warning based on unmanned aerial vehicle (UAV) autonomous inspection. This method enables non-contact, automated, and quantitative inspection and monitoring of landslide areas, improving the accuracy of landslide disaster identification, the consistency of inspection operations, and the timeliness of early warning response.

[0007] According to some embodiments, the present invention adopts the following technical solution: Landslide disaster identification and early warning methods based on UAV autonomous inspection include: Based on the basic information of the target landslide monitoring area, generate an autonomous inspection route for the UAV. The drone flies autonomously along the generated inspection route and collects multi-temporal monitoring data; The monitoring data is processed, including constructing multi-temporal surface models, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters; Based on the landslide disaster characteristic parameters, risk identification and level assessment are performed on the target landslide monitoring area, and the effectiveness of the current inspection results is determined in conjunction with the quality evaluation results. When the preset conditions are met, landslide disaster early warning information is output.

[0008] According to some embodiments, the present invention adopts the following technical solution: A landslide disaster identification and early warning device based on drone autonomous inspection includes: The flight path generation module is configured to generate autonomous inspection flight paths for UAVs based on the acquired basic information of the target landslide monitoring area. The data acquisition module is configured to allow the UAV to fly autonomously along the generated inspection route and collect multi-temporal monitoring data. The data processing module is configured to process monitoring data, including constructing multi-temporal surface models, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters. The risk identification module is configured to: identify and assess the risk level of the target landslide monitoring area based on the landslide disaster characteristic parameters, and determine the effectiveness of the current inspection results in conjunction with the quality evaluation results; and output landslide disaster early warning information when preset conditions are met.

[0009] According to some embodiments, the present invention adopts the following technical solution: A computer program product includes a computer program that, when executed by a processor, implements the landslide disaster identification and early warning method based on unmanned aerial vehicle (UAV) autonomous inspection.

[0010] According to some embodiments, the present invention adopts the following technical solution: A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the landslide disaster identification and early warning method based on UAV autonomous inspection.

[0011] According to some embodiments, the present invention adopts the following technical solution: An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the landslide disaster identification and early warning method based on UAV autonomous inspection.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention forms a complete technical closed loop of "autonomous inspection - feature recognition - risk assessment - quality verification - graded early warning," which has the advantages of being non-contact, automated, quantitative, and intelligent. It can realize autonomous inspection of landslide areas, multi-temporal monitoring data acquisition, spatial registration, crack propagation identification, displacement and settlement calculation, boundary change analysis, quality evaluation, and risk early warning output. Compared with traditional manual inspection or fixed-point monitoring methods, it has the advantages of high automation, wide monitoring range, non-contact operation, good safety, and strong consistency of repeated observations, which can effectively improve the accuracy of landslide disaster identification and the timeliness of early warning response.

[0013] This invention generates autonomous inspection routes by combining topographic information, boundary information, historical key monitoring area distribution, and obstacle information of the target area. It can realize automatic inspection by drones in complex mountainous areas, steep slopes, and landslide environments with many obstacles. Compared with traditional manual inspection methods, it significantly improves monitoring efficiency and reduces inspection risks.

[0014] This invention enhances the spatial consistency and retest stability of multi-temporal inspection data by introducing priority inspection of key monitoring areas, densification of waypoints in key monitoring areas, and fixed-point re-flight mechanisms. This is beneficial for improving the quantitative identification accuracy of key disaster characteristics such as landslide crack expansion, surface displacement, surface settlement, and boundary changes.

[0015] This invention uses a combination of image data, depth data, and flight attitude data to construct a baseline temporal surface model and a current temporal surface model. It also achieves quantitative extraction of surface morphology changes in landslide areas through multi-temporal spatial registration and difference analysis. This solves the problems of low recognition accuracy and incomplete change information caused by relying solely on single image interpretation or local manual comparison in existing technologies.

[0016] This invention integrates multiple disaster characteristics, such as crack propagation, displacement, settlement, boundary changes, and local abnormal deformation, into a risk assessment model. By fusing multiple indicators, it achieves a comprehensive judgment on landslide risk levels, improving the comprehensiveness and scientific nature of risk identification results and enhancing the accuracy of early warning judgments.

[0017] This invention further introduces quality evaluation indicators such as image clarity, point cloud density, data integrity, positioning deviation, and flight stability on the basis of disaster identification and risk assessment. It can determine the validity of the current inspection results and automatically trigger supplementary data collection or increased inspection when the quality does not meet the standards, thereby improving the reliability and stability of the monitoring results. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 This is a flowchart of the method in Example 1. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0023] Example 1 One embodiment of the present invention provides a method for landslide disaster identification and early warning based on unmanned aerial vehicle (UAV) autonomous inspection, including: Step S1: Generate an autonomous inspection route for the UAV based on the basic information of the target landslide monitoring area. Step S2: The UAV flies autonomously along the generated inspection route and collects multi-temporal monitoring data; Step S3: Process the monitoring data, including constructing a multi-temporal surface model, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters; Step S4: Based on the landslide disaster characteristic parameters, risk identification and level assessment are performed on the target landslide monitoring area, and the effectiveness of the current inspection results is determined in conjunction with the quality evaluation results. When the preset conditions are met, landslide disaster early warning information is output.

[0024] As one embodiment, the landslide disaster identification and early warning method based on UAV autonomous inspection of the present invention achieves non-contact, automated, and quantitative inspection and monitoring of landslide areas through a complete technical closed loop of "autonomous inspection—feature recognition—risk assessment—quality verification—tiered early warning," thereby improving the accuracy of landslide disaster identification, the consistency of inspection operations, and the timeliness of early warning response. The specific implementation process is described below, as follows: Figure 1 As shown, the steps are as follows: Step 1: Obtain basic information about the target landslide monitoring area.

[0025] First, basic information about the target landslide monitoring area is obtained. This basic information includes landslide area boundary information, slope elevation information, topographic relief information, historical inspection trajectory information, historical disaster characteristic information, and environmental obstacle distribution information. This basic information can be derived from one or more of the following: previous surveying data, historical inspection results, remote sensing image data, and manually labeled results.

[0026] In this step, the target landslide monitoring area can be divided into ordinary inspection areas and key monitoring areas based on the landslide area boundary, slope gradient, surface undulation changes, and historical key monitoring area distribution, so that different data collection densities and inspection priorities can be adopted for different areas in the future.

[0027] Step 2: Generate autonomous inspection routes.

[0028] Generate an autonomous inspection route for the drone based on the basic information obtained in step one.

[0029] Specifically, by combining the boundary range of the target landslide monitoring area, the distribution of obstacles, the distribution of key monitoring areas, and the location of historical key monitoring areas, an initial set of waypoints is generated within the landslide monitoring area. Based on obstacle distribution and flight constraints, unexecutable waypoints are eliminated to obtain a set of executable inspection waypoints. It can be represented as:

[0030] in, Indicates the first One waypoint, Indicates the area of ​​obstacles. Waypoints Area with obstacles The minimum distance between them This indicates the preset safe distance.

[0031] Furthermore, based on the aforementioned inspection waypoint set Based on inspection priority, spatial relationships between waypoints, and flight constraints, the waypoints are connected and sorted to generate several candidate inspection routes. The set of candidate inspection routes can be represented as:

[0032] in, This represents the set of candidate inspection routes. Indicates the first Candidate inspection routes, This indicates that the candidate inspection route meets the preset flight constraints.

[0033] Preferably, the flight constraints include distance constraints between adjacent waypoints, elevation change constraints, and turning constraints, which can be expressed as:

[0034] in, This indicates the maximum permissible flight distance between adjacent waypoints. This indicates the maximum permissible elevation change between adjacent waypoints. Indicates the first One steering angle, This indicates the maximum permissible steering angle.

[0035] Subsequently, by calculating and comparing the comprehensive value of each candidate inspection route, the candidate inspection route with the smallest comprehensive value is selected as the autonomous inspection route, and the UAV is controlled to fly along the autonomous inspection route.

[0036] To find the optimal autonomous inspection route that balances inspection efficiency and coverage of key areas, a route optimization objective function can be established:

[0037] in, To enhance the overall value of the route, For the inspection route length, For the cost of flight energy consumption, As a cost of obstacles and risks, The coverage benefits of key monitoring areas are represented by α, β, γ, and δ, which are the corresponding weighting coefficients.

[0038] In this embodiment, the candidate inspection route is assumed to pass through the following locations in sequence. The waypoint, the first The coordinates of each waypoint are as follows Then the inspection route length It can be obtained by summing the Euclidean distances between adjacent waypoints, that is:

[0039] in, Indicates the first The waypoint and the first The Euclidean distance between waypoints.

[0040] Flight energy consumption cost The formula used to characterize the total energy consumed by a UAV flying along a candidate inspection route can comprehensively consider the horizontal flight distance, changes in climb or descent altitude, and the number of turns. It can be expressed as:

[0041] in, This represents the Euclidean distance between adjacent waypoints. This represents the elevation change between adjacent waypoints. This represents the number of turns in the candidate inspection route. , , This is the corresponding energy consumption weighting coefficient.

[0042] Obstacles, risks, and costs It is used to characterize the risk level between candidate inspection routes and obstacles, and can be calculated based on the minimum safe distance between each route and the obstacle. The calculation formula can be expressed as:

[0043] in, For the first The minimum distance between the route segment and the nearest obstacle. A very small positive number is set to avoid the denominator being zero; the closer the flight path gets to an obstacle, the more... The smaller the value, the lower the corresponding obstacle risk cost. The larger.

[0044] Key monitoring area coverage benefits This is used to characterize the coverage of key monitoring areas by candidate inspection routes. It can be calculated based on the coverage ratio of key monitoring areas, and the formula can be expressed as:

[0045] in, The number of key monitoring areas, For the first The area of ​​each key monitoring area For the candidate inspection route to the first Effective coverage area of ​​key monitoring areas For the first The priority weights correspond to the key monitoring areas. The higher the coverage rate of a key monitoring area, the greater the corresponding coverage benefit. The larger.

[0046] Preferably, it can be used for , , and After normalization, the data is substituted into the route optimization objective function to reduce the impact of different dimensions on the calculation results of the comprehensive cost of the route.

[0047] Based on the above , , and The calculation results are used to optimize and compare the objective function values ​​corresponding to each candidate inspection route, and the candidate inspection route with the smallest objective function value is selected as the autonomous inspection route. For historical key monitoring areas or currently identified key monitoring areas, denser waypoints or fixed-point bypass waypoints can be set to improve the data acquisition resolution and multi-temporal observation consistency in local areas.

[0048] Step 3: Perform autonomous inspection and collect current time-phase monitoring data.

[0049] The drone performs inspection flights according to the optimized autonomous inspection route. During the flight, it performs the following operations at each waypoint to collect monitoring data for the current time phase: 1) Continuously record the drone's flight attitude data. ,in, These are the current coordinates of the drone. These are roll angle, pitch angle, and yaw angle, respectively.

[0050] 2) The distance information between the UAV and the slope surface is acquired through laser ranging or depth sensing, and this distance information serves as the depth data for the current time phase. The depth data is used to characterize the acquisition distance between the UAV and the slope surface. Based on the depth data, the UAV is controlled to fly stably within a preset safe distance range, thereby improving the consistency of data acquisition between different inspection cycles.

[0051] 3) In this embodiment, subscripts are used to represent time phases. The monitoring data collected during the first inspection is recorded as the 0th time phase, corresponding to the reference time phase; the monitoring data collected during the subsequent t-th inspection is recorded as the t-th time phase, where t is a positive integer; two adjacent subsequent inspection time phases can be recorded as the t-th time phase and the (t+1)-th time phase, respectively. While maintaining a stable flight attitude, the UAV uses its onboard dual-light monitoring cameras—a visible light camera and an infrared thermal imaging camera—to continuously observe the landslide area, acquiring the visible light image data and infrared image data of the t-th time phase, denoted as:

[0052] in, For the visible light image data of the t-th time phase, This represents the infrared image data for the t-th time phase.

[0053] Step 4: Preprocess the monitoring data and construct a multi-temporal surface model. The current time-phase monitoring data obtained in step three is preprocessed, including image denoising, brightness equalization, distortion correction, shadow suppression, spatiotemporal synchronization, and coordinate unification. The data collected during the initial inspection is used as the 0th time-phase monitoring data, i.e., the baseline time-phase monitoring data; the data collected during subsequent inspections is used as the tth time-phase monitoring data and compared with the 0th time-phase monitoring data.

[0054] In this step, a baseline temporal surface model and a current temporal surface model are constructed based on the preprocessed image data. The surface model can be one or more of the following: an orthophoto model, a digital surface model, a digital elevation model, and a 3D point cloud model. For key monitoring areas, a more refined local surface model can be constructed to improve the accuracy of identifying cracks, settlement, and boundary changes.

[0055] Step 5: Perform multi-temporal spatial registration and difference analysis To ensure that monitoring data from different time phases are compared under a unified spatial reference, spatial registration is performed between the reference time phase surface model and the current time phase surface model.

[0056] First, based on historical inspection trajectory information, flight attitude data, and spatial location characteristics of key monitoring areas, coarse registration is performed between the reference time phase surface model and the current time phase surface model.

[0057] Then, fine registration is performed based on at least one of the following methods: feature point matching, contour matching, point cloud iterative nearest point registration, normal constraint registration, and raster elevation difference.

[0058] Finally, based on the registered model, a difference analysis was performed, specifically as follows: For the corresponding sampling point after registration, its three-dimensional displacement can be expressed as:

[0059] Among them, D i For the first i The three-dimensional displacement of each sampling point For the first i The reference phase coordinates of each sampling point For the first i The current phase coordinates of each sampling point.

[0060] Its elevation change is calculated according to the following formula:

[0061] in, For the first The sampling point at the th sampling point The elevation change of each time phase relative to the reference time phase For the first The sampling point at the th sampling point Elevation coordinates of each time phase For the first The elevation coordinates of each sampling point at the reference time phase.

[0062] Through the above spatial registration and difference analysis, the displacement change field, settlement change field, boundary morphology change area and local abnormal deformation area of ​​the landslide area surface can be obtained.

[0063] The displacement change field is composed of the three-dimensional displacement of the corresponding sampling points after registration. Spatial mapping is performed within the landslide monitoring area; the settlement change field is formed by the elevation changes at each sampling point. Spatial mapping is performed within the landslide monitoring area; the boundary morphology change area is obtained by extracting the slope boundary contours of the reference time phase and the current time phase, and comparing the overlay of the two time phase boundary contours; the local abnormal deformation area is obtained by analyzing the elevation change. and three-dimensional displacement Threshold discrimination is performed, and sampling points that meet the abnormal conditions are clustered or connected to obtain the results.

[0064] The displacement change field, settlement change field, boundary morphology change region, and local abnormal deformation region are used as the basis for the differential analysis results of subsequent landslide disaster characteristic parameter extraction. Among them, the surface displacement is obtained by statistical analysis of the displacement change field, the settlement is obtained by statistical analysis of the settlement change field, the slope boundary change is obtained by extraction of the boundary morphology change region, the local abnormal deformation is obtained by quantification of the local abnormal deformation region, and the crack propagation is obtained by identifying, matching, and quantifying the crack region in the reference time phase and the current time phase image.

[0065] Step Six: Extracting Landslide Disaster Characteristic Parameters After completing multi-temporal registration and difference analysis, based on the displacement change field, settlement change field, boundary morphology change region and local abnormal deformation region obtained in step five, the landslide area is subjected to disaster feature extraction to obtain landslide disaster feature parameters; the landslide disaster feature parameters include crack propagation amount, surface displacement amount, settlement amount, slope boundary change amount and local abnormal deformation amount.

[0066] Among them, the crack propagation index is used. The formula for characterizing crack propagation is:

[0067] in, For the first The crack propagation index of each time phase The number of cracks. and The first The cracks were at the reference time phase and the first The length of each phase and The first The cracks were at the reference time phase and the first Width under each time phase and These are the length weighting coefficient and the width weighting coefficient, respectively.

[0068] The surface displacement is obtained by statistically analyzing the three-dimensional displacement of each sampling point obtained in step five, and can be represented by the average value of the three-dimensional displacement of the sampling points within the key monitoring area, i.e.:

[0069] in, For the first Surface displacement at each time phase For the first The sampling point at the th sampling point The three-dimensional displacement of each time phase relative to the reference time phase. This represents the number of sampling points within the key monitoring area.

[0070] The average settlement within the key monitoring area can be expressed as:

[0071] in, For the first Average settlement at each time phase For the first The elevation values ​​of each sampling point at the reference time phase. For the first The sampling point at the th sampling point Elevation values ​​for each time phase This represents the number of sampling points.

[0072] Using the rate of change of slope boundary The formula for representing the change in slope boundary is:

[0073] in, For the first The rate of change of the slope boundary at each time phase and They are respectively the reference phase and the first The projected area of ​​the slope at each time phase.

[0074] The amount of local abnormal deformation is obtained by quantifying the local abnormal deformation region obtained in step five. This will satisfy...

[0075] The sampling points were divided into local abnormal deformation regions, among which, The elevation anomaly threshold, This represents the displacement anomaly threshold. The amount of local abnormal deformation can be expressed as the average value of the abnormal deformation intensity within the local abnormal deformation region, i.e.:

[0076] in, For the first Local abnormal deformation amount in each time phase, This is the set of sampling points corresponding to the local abnormal deformation region. The number of sampling points in the set of sampling points. For the first The sampling point at the th sampling point The elevation change of each time phase relative to the reference time phase For the first The sampling point at the th sampling point The three-dimensional displacement of each time phase relative to the reference time phase. and These are the corresponding weighting coefficients.

[0077] Through the above calculations, the crack propagation amount, surface displacement amount, settlement amount, slope boundary change amount, and local abnormal deformation amount can be obtained respectively, and these can be used as input parameters for risk identification and level assessment in step eight.

[0078] Step 7: Conduct quality evaluation Because landslide areas may be affected by factors such as vegetation obstruction, changes in sunlight, wind disturbance, and terrain shading, a quality evaluation of the current phase of the inspection data is necessary to ensure the reliability of the identification results. The comprehensive quality evaluation value can be expressed as:

[0079] in, For the current number Quality evaluation value for each phase, For the current number Image sharpness index for each phase, For the current number Point cloud density or data integrity index for each time phase For the current number Positioning accuracy indicators for each time phase For the current number Flight stability indicators for each time phase, For the current number The trajectory deviation from the control index at each time phase , , , , These are the corresponding weighting coefficients.

[0080] In this embodiment, for ease of comprehensive evaluation, the... , , , and Normalization to The range is defined, and the larger the indicator value, the better the quality of the current inspection data.

[0081] Among them, image sharpness index Used to characterize the current number The clarity of the current phase image data. Several corresponding image frames acquired in the same key monitoring area or the same patrol segment as the previous time phase constitute the evaluation image frame set; preferably, the current image frame is used. The Laplacian variance of the corresponding image frame for each time period is characterized by the following formula:

[0082] in, For the current number The number of corresponding image frame pairs that participate in the evaluation, formed by each time phase and the previous time phase. For the first For the current image frame in the corresponding image frame A phase image, For the current number Phase 1 The Laplacian operator response of a frame image, For variance calculation, This is a preset normalized reference value. The clearer the image, the higher the corresponding normalized reference value. The larger the value.

[0083] Point cloud density or data integrity index Used to characterize the current number Spatial coverage quality of time-phase inspection data.

[0084] Current number When the temporal monitoring data includes point cloud data, It can be determined by the number of point clouds per unit area within the key monitoring area, and its calculation formula is:

[0085] in, The area of ​​the key monitoring area. For the current number The number of valid point cloud points in the region described in each time phase, This is a reference value for the normalization of point cloud density.

[0086] Current number When the phase monitoring data does not include point cloud data Data integrity can also be used as a metric, and its calculation formula is:

[0087] in, For the current number The amount of effective monitoring data within the preset evaluation range for each time phase For the current number The total amount of monitoring data that should be acquired in each time phase within the preset evaluation range; the preset evaluation range is the current time phase. The key monitoring areas or preset inspection task ranges corresponding to each time period.

[0088] Preferably, when the monitoring data is image data, it is determined to be valid monitoring data if the image data meets at least one or more of the following conditions: The data collection location is within the preset evaluation range; Image clarity meets the preset threshold requirements; The pose information and timestamp information are complete; It can also be used for subsequent registration analysis, surface model construction, or disaster feature extraction.

[0089] Data that is severely blurry, overexposed, occluded, has missing frames, lacks positioning information, or cannot be used for subsequent processing is deemed invalid monitoring data.

[0090] Positioning accuracy indicators This parameter, used to characterize the consistency between the actual location of the UAV and the preset inspection waypoints, can be obtained based on the average deviation between the UAV's positioning coordinates and the corresponding target waypoint coordinates. The calculation formula is as follows:

[0091] in, For the current number The number of waypoints participating in the evaluation at each time phase For the current number Phase 1 The actual flight position coordinates corresponding to each waypoint For the current number Phase 1 The preset reference coordinates corresponding to each waypoint This is the normalization threshold for positioning deviation; it can be set to 0 when the calculated result is less than 0. The smaller the positioning deviation, the better. The larger the value.

[0092] Flight stability indicators Used to characterize the current number The degree of flight attitude fluctuation of the UAV in each time phase can be determined based on the current time phase. The fluctuations in roll, pitch, and yaw angles during each phase of flight are obtained, and the attitude angle standard deviation is preferably used for characterization. The calculation formula is as follows:

[0093] in, , , Each is the current number The standard deviations of the roll, pitch, and yaw angles in each phase. This is the normalization threshold for attitude fluctuations; it can be set to 0 when the calculated result is less than 0. The more stable the flight, the better. The larger the value.

[0094] Trajectory Deviation Control Indicators Used to characterize the current number The degree of deviation of the actual flight trajectory of the UAV from the autonomous inspection route can be determined based on the current... The average deviation between the actual flight trajectory and the preset route at each time phase is obtained, and its calculation formula is:

[0095] in, For the current number The number of trajectory sampling points participating in the evaluation at each time phase. For the current number Phase 1 The minimum distance from each trajectory sampling point to the preset autonomous inspection route. This is the normalization threshold for trajectory deviation; it can be set to 0 when the calculated result is less than 0. The smaller the trajectory deviation, the better. The larger the value.

[0096] Based on the calculation results of the above indicators, the comprehensive quality evaluation value can be obtained. When the quality evaluation value Less than the preset quality threshold At the same time, the drone can be controlled to return to the corresponding key monitoring area to perform supplementary data collection and inspection, intensified inspection, or fixed-point re-flight inspection, so as to improve the accuracy and consistency of monitoring results.

[0097] Step 8: Conduct risk identification and early warning output The drone completes the current inspection data collection according to the current autonomous inspection route, i.e., the current time phase. After collecting monitoring data and completing multi-temporal spatial registration, difference analysis and disaster feature extraction, an overall risk assessment of the target landslide monitoring area is conducted based on the crack expansion, surface displacement, settlement, slope boundary change and local abnormal deformation corresponding to the current inspection.

[0098] Based on crack propagation index, surface displacement index, settlement index, slope boundary change index, and local abnormal deformation index, a risk assessment model is established, and its risk assessment value can be expressed as:

[0099] in, For the current number Risk assessment value for each time phase, For the current number The average value of the three-dimensional displacement of sampling points in the key monitoring area of ​​each time phase. For the current number The crack propagation index of each time phase For the current number Average settlement at each time phase For the current number The rate of change of the slope boundary at each time phase For the current number Local abnormal deformation indices for each time phase, , , , , These are the corresponding weighting coefficients.

[0100] Based on risk assessment value The relationship between the current and preset thresholds, for the current... Risk classification is performed on the corresponding target landslide monitoring areas at each time; when When it is judged as low risk, When it is determined to be medium risk, It was determined to be high risk at that time.

[0101] When the risk level reaches the preset warning conditions, a landslide disaster warning message is output, including the risk level, location of the abnormal area, type of abnormality, and corresponding monitoring data; when the risk level does not reach the preset warning conditions, the current warning message is stored. The monitoring results of each time phase are used as historical data for subsequent multi-time phase comparative analysis.

[0102] Example 2 One embodiment of the present invention provides a landslide disaster identification and early warning device based on unmanned aerial vehicle (UAV) autonomous inspection, comprising: The flight path generation module is configured to generate autonomous inspection flight paths for UAVs based on the acquired basic information of the target landslide monitoring area. The data acquisition module is configured to allow the UAV to fly autonomously along the generated inspection route and collect multi-temporal monitoring data. The data processing module is configured to process monitoring data, including constructing multi-temporal surface models, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters. The risk identification module is configured to: identify and assess the risk level of the target landslide monitoring area based on the landslide disaster characteristic parameters, and determine the effectiveness of the current inspection results in conjunction with the quality evaluation results; and output landslide disaster early warning information when preset conditions are met.

[0103] Example 3 One embodiment of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the landslide disaster identification and early warning method based on UAV autonomous inspection.

[0104] Example 4 In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided for storing computer instructions. When the computer instructions are executed by a processor, the landslide disaster identification and early warning method based on UAV autonomous inspection is implemented.

[0105] Example 5 One embodiment of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the landslide disaster identification and early warning method based on UAV autonomous inspection.

[0106] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0108] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A landslide disaster identification and early warning method based on UAV autonomous inspection, characterized in that, include: Based on the basic information of the target landslide monitoring area, generate an autonomous inspection route for the UAV. The drone flies autonomously along the generated inspection route and collects multi-temporal monitoring data; The monitoring data is processed, including constructing multi-temporal surface models, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters; Based on the landslide disaster characteristic parameters, risk identification and level assessment are performed on the target landslide monitoring area, and the effectiveness of the current inspection results is determined in conjunction with the quality evaluation results. When the preset conditions are met, landslide disaster early warning information is output.

2. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The generated autonomous inspection route for the UAV is specifically as follows: Acquire basic information, including terrain information, boundary information, slope elevation information, historical inspection trajectory information, historical disaster characteristic information, and environmental obstacle information; Based on the aforementioned basic information, the landslide monitoring scope and key monitoring areas were determined; Based on topographic information, slope elevation information, historical inspection trajectory information, historical disaster characteristics information, environmental obstacle information, and inspection priority of key monitoring areas, a set of inspection waypoints is generated within the landslide monitoring range; Based on the set of inspection waypoints, one or more candidate inspection routes are generated according to the spatial relationship between waypoints, flight constraints and inspection priorities. An objective function is constructed to minimize the comprehensive cost of the candidate inspection routes, and the objective function is optimized to obtain an autonomous inspection route that meets the requirements of inspection efficiency, flight safety, and coverage of key areas.

3. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The multi-temporal monitoring data consists of monitoring data collected during multiple temporal inspection processes; The monitoring data collected at each time phase includes image data, depth data, and flight attitude data. The image data includes at least one of visible light image data and infrared image data. The depth data is the acquisition distance between the UAV and the slope surface. The flight attitude data includes at least one of the UAV's position coordinates, attitude angle, flight altitude, flight speed, heading angle, and gimbal attitude angle.

4. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The construction of the multi-temporal surface model involves preprocessing, spatiotemporally synchronizing, and unifying the coordinates of the monitoring data for each temporal phase, and combining historical temporal phase data to construct a reference temporal phase surface model and a current temporal phase surface model. The surface model can be one or more of the following: orthophoto model, digital surface model, digital elevation model, and three-dimensional point cloud model.

5. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The extraction of landslide disaster characteristic parameters is based on the difference analysis of the registered reference time phase surface model and the current time phase surface model, and the extraction of disaster characteristics of the landslide area. The landslide disaster characteristic parameters are crack propagation amount, surface displacement amount, settlement amount, slope boundary change amount, and local abnormal deformation amount.

6. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The risk identification and level assessment of the target landslide monitoring area specifically includes: Using a risk assessment model that includes crack propagation index, surface displacement index, settlement index, slope boundary change index, and local abnormal deformation index, the risk assessment value for the current time phase is calculated. Based on the relationship between the risk assessment value and the preset threshold, the target landslide monitoring area is classified into risk levels.

7. The landslide disaster identification and early warning method based on UAV autonomous inspection as described in claim 1, characterized in that, The quality evaluation result is based on at least one quality indicator among image clarity, point cloud density, data integrity, positioning deviation, attitude stability, and trajectory deviation. A quality evaluation model is constructed to calculate the quality evaluation value of the current time-phase monitoring data. When the quality evaluation value is less than the preset quality threshold, the UAV is controlled to perform supplementary data collection and inspection, intensified inspection, or fixed-point re-flight inspection on the corresponding key monitoring area.

8. A landslide disaster identification and early warning device based on unmanned aerial vehicle (UAV) autonomous inspection, characterized in that, include: The flight path generation module is configured to generate autonomous inspection flight paths for UAVs based on the acquired basic information of the target landslide monitoring area. The data acquisition module is configured to allow the UAV to fly autonomously along the generated inspection route and collect multi-temporal monitoring data. The data processing module is configured to process monitoring data, including constructing multi-temporal surface models, multi-temporal spatial registration and difference analysis, and extracting landslide disaster characteristic parameters. The risk identification module is configured to: identify and assess the risk level of the target landslide monitoring area based on the landslide disaster characteristic parameters, and determine the effectiveness of the current inspection results in conjunction with the quality evaluation results; and output landslide disaster early warning information when preset conditions are met.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the landslide disaster identification and early warning method based on UAV autonomous inspection as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the landslide disaster identification and early warning method based on UAV autonomous inspection as described in any one of claims 1-7.

Images