A three-dimensional terrain mapping method based on a UAV

Abstract

The application discloses a kind of three-dimensional terrain mapping methods based on unmanned aerial vehicle, it is related to unmanned aerial vehicle mapping field, the method is based on the multiple unmanned aerial vehicle three-dimensional terrain mapping of tailing dam, dam body is uniformly divided into regular dam surface monitoring unit, dam surface response parameter is extracted simultaneously on each monitoring unit scale, and after normalization processing, initial risk index R is calculated and output, so that in 2m×2m fine spatial scale, the comprehensive deviation degree of each monitoring unit in three dimensions of deformation, seepage and microstructure is quantitatively characterized.Compared with the way of traditional safety evaluation only relying on a small amount of monitoring section or single deformation index, the early signs such as local deformation acceleration, seepage anomaly and microtopographic disturbance can be strengthened in the initial risk index R through multi-parameter fusion and product type risk index construction, which improves the sensitivity and accuracy of tailing dam hidden danger identification.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) surveying technology, specifically to a UAV-based three-dimensional terrain surveying method. Background Technology

[0002] With the expansion of mining scale and the increase in the number of tailings dams, high-precision 3D topographic mapping of tailings dam surfaces has gradually shifted from traditional manual leveling and total station surveying to aerial photogrammetry using unmanned aerial vehicles (UAVs) equipped with optical cameras and inertial measurement units. This involves establishing high-precision, multi-phase 3D topographic models of the tailings dam area through aerial triangulation and bundle adjustment techniques. In this specific application scenario, how to utilize UAV 3D topographic mapping methods to monitor the deformation characteristics of tailings dam surfaces with high spatiotemporal resolution and provide reliable data support for the safe operation of tailings dams has become a key concern for the engineering and surveying technology communities.

[0003] The current common approach is to design relatively fixed UAV flight paths and fixed mapping cycles for the entire tailings dam, conducting full-coverage 3D topographic mapping of the dam body, and then analyzing deformation through a small number of feature sections or discrete monitoring points. This approach has several shortcomings: First, the flight path design and mapping cycle are mostly fixed in advance, resulting in weak coupling with the historical deformation evolution process of the dam body. It cannot adaptively adjust the spatial density and temporal frequency of the mapping according to the "accelerating trend" of local deformation of the dam body. Second, existing methods mostly take "uniform coverage" as the core objective. Even if the tailings dam is generally stable and only a few sections of the dam show concentrated deformation, the system still needs to repeatedly map the entire dam area at a uniform density, resulting in over-collection and data redundancy in stable areas. For dam sections that are truly in a state of accelerated deformation, it is difficult to increase the re-measurement frequency and 3D solution redundancy in a timely manner, resulting in coarse and untargeted resource allocation.

[0004] Under the aforementioned circumstances, once the tailings dam as a whole is in a generally stable state, but a local section of the dam experiences a significant increase in deformation vector field and deformation acceleration, traditional methods, relying solely on fixed routes and fixed-cycle full-dam mapping, often fail to capture the critical stage of the local dam section's evolution from "gradual deformation" to "accelerated deformation" in a timely manner. This can easily lead to delayed risk identification or even misjudgment of the overall stability of the dam. When the deformation of a local dam section develops rapidly between two fixed mapping cycles, the monitoring system may still perform tasks according to the conventional cycle, resulting in a lack of sufficient temporal density of three-dimensional topographic data to support the dangerous development stage. This hinders the early detection of signs of instability such as abnormal subsidence, bulging, and slippage of the dam. At the same time, because mapping resources are evenly distributed across the entire dam area, stable areas occupy a large amount of flight paths and processing capacity for a long time, while truly high-risk areas cannot obtain sufficient densified flight paths and shortened re-measurement cycles. This may amplify the risk of local instability of the tailings dam into an overall safety hazard, causing serious adverse consequences for the downstream environment and personnel safety. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs), which solves the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs), comprising the following steps:

[0007] S1. By using drones to conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route, a multi-phase three-dimensional topographic model under a unified coordinate system is obtained, and the tailings dam is divided into regular grids, with each regular grid defined as a monitoring unit.

[0008] S2. For each monitoring unit, extract the dam surface response parameters at different surveying cycles, and calculate the initial risk index R of the dam surface multidimensional response based on the dam surface response parameters;

[0009] S3. Perform time and spatial comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx; divide the tailings dam into several engineering dam sections Br along the dam axis, and statistically analyze the candidate risk monitoring unit set Hx within each engineering dam section Br to obtain the comprehensive risk index R of the dam section. Br And compare it with the preset trigger threshold Rth to determine the engineering dam section Br where the adaptive monitoring zone is located;

[0010] S4. Execute an adaptive monitoring zone generation strategy on the candidate risk monitoring unit set Hx, and adjust the retesting time based on the comprehensive risk index RBr to obtain the retesting interval T. Perform three-dimensional terrain mapping in the non-adaptive monitoring zone area according to the baseline route plan and the retesting baseline interval.

[0011] Preferably, S1 includes S11;

[0012] S11. Using a drone equipped with an optical camera and an inertial measurement unit, conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route;

[0013] The preset routes include an overall control route for the upper dam body covering the entire tailings dam area and a detailed route for the lower dam slope laid out along the dam slope; and forward-looking oblique photography routes and backward-looking oblique photography routes are set up near the water-facing slope of the tailings dam and the accumulation shoreline.

[0014] In each surveying mission, the drone is controlled to fly along the preset route in sequence, and three-dimensional terrain data is collected through the surveying module on the drone.

[0015] The mapping module includes an optical camera, an inertial measurement unit, and a flight control log.

[0016] The three-dimensional terrain data includes image data, acceleration data, and log data.

[0017] The 3D terrain data is transmitted to a cloud-based mapping server via a wireless network. The cloud-based mapping server performs aerial triangulation and bundle adjustment based on the 3D terrain data, and the 3D terrain data obtained from each phase of the survey are uniformly calculated into the same dam body reference coordinate system to form a multi-phase 3D terrain model under a unified coordinate system.

[0018] Preferably, S1 further includes S12;

[0019] S12. Project the upstream slope, dam crest, and downstream slope of the tailings dam in the multi-phase 3D terrain model under a unified coordinate system onto the dam reference coordinate system plane. Divide the dam axis direction and the slope perpendicular direction into regular grids. Define each regular grid cell as a dam surface monitoring unit. The regular grid is set to 2m×2m according to the ground resolution of UAV imagery and the accuracy of 3D calculation, so that each dam surface monitoring unit has complete point cloud and image coverage in each phase of the 3D terrain model. Extract the 3D coordinates of the corresponding dam surface monitoring unit and the image data associated with the current dam surface monitoring unit in each 3D terrain model.

[0020] Preferably, S2 includes S21;

[0021] S21. In each monitoring unit, based on the three-dimensional terrain model at different surveying cycles, the dam surface response parameters are extracted.

[0022] The dam surface response parameters include the vertical deformation rate parameter V of the j-th monitoring unit during the tk-th mapping period. j (tk), the deformation increment trend parameter A of the j-th monitoring unit in the tk-th survey period.j (tk), the seepage visual indication parameter Q of the j-th monitoring unit in the tk-th mapping cycle. j (tk) and the micro-topographic disturbance parameter M of the j-th monitoring unit in the tk mapping period. j (tk);

[0023] The dam surface response parameters were normalized using the maximum and minimum value normalization method to eliminate the unit dimensions of all parameters in the dam surface response parameters.

[0024] Preferably, S2 further includes S22;

[0025] S22. Based on the normalized dam surface response parameters, substitute them into the formula for the initial risk index of the multidimensional response of the dam surface, calculate and output the initial risk index R, and quantitatively analyze the comprehensive deviation of the dam body in the three dimensions of deformation, seepage and microstructure.

[0026] The initial risk index R is calculated and output using the following formula for the initial risk index of the multidimensional response of the dam surface;

[0027] ;

[0028] In the formula, R j (tk) represents the initial risk index of the j-th monitoring unit in the tk-th surveying period.

[0029] Preferably, S3 includes S31;

[0030] S31. Perform time and space comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx;

[0031] The time comparison is performed by comparing the initial risk index R of the j-th monitoring unit corresponding to multiple mapping cycles in the tk-th mapping cycle at each monitoring unit. j (tk) forms a time series, and the difference ΔR between the initial risk index R of two adjacent periods is calculated. j (tk); When the difference ΔR between the initial risk indices R of two adjacent periods... j When (tk) is greater than the preset time change threshold ΔRs and appears continuously within the preset number of periods, the monitoring unit will be marked as a time risk monitoring unit.

[0032] The spatial comparison is performed by comparing the initial risk index R of the j-th monitoring unit within the same dam section during the tk-th mapping period, within the same mapping cycle. j (tk) Perform spatial statistics to calculate the local mean and local standard deviation; and set the initial risk index R of the j-th monitoring unit in the same dam section during the tk mapping period. jMonitoring units whose (tk) is greater than the spatial threshold formed by the weighted combination of the local mean and the local standard deviation are marked as spatial high value monitoring units;

[0033] Monitoring units that simultaneously satisfy both time risk markers and spatial high-value markers are aggregated through spatial connectivity analysis to obtain a high-value zone continuously distributed along the dam surface, and all monitoring units within the high-value zone constitute a candidate risk monitoring unit set Hx.

[0034] Preferably, S3 further includes S32;

[0035] S32. Extract the center line of the tailings dam crest in the dam reference coordinate system, and set multiple boundary points for the center line of the dam crest according to the engineering design station number. Define the spatial area corresponding to the center line of the dam crest between any two adjacent boundary points as an engineering dam segment Br, so that the engineering dam segments Br along the dam axis are connected end to end and do not overlap with each other.

[0036] Using the projection range of each dam segment Br in the dam body reference coordinate system, monitoring units whose centroids fall within this projection range are selected in the direction perpendicular to the slope along the dam axis, forming a set of monitoring units contained in the dam segment Br. Within each dam segment Br, monitoring units belonging to the current dam segment Br are selected from the candidate risk monitoring unit set Hx, and the initial risk index R of the j-th monitoring unit in the tk-th surveying period is extracted. j (tk), and the initial risk index R of these j-th monitoring units in the tk-th mapping period. j (tk) Sort by value from smallest to largest, and take the initial risk index R of the j-th monitoring unit in the upper quartile position during the tk-th surveying period. j (tk) serves as the comprehensive risk indicator R for the dam section Br during the tk survey period. Br (tk).

[0037] Preferably, S3 further includes S33;

[0038] S33. Based on the comprehensive risk index R of the engineering dam section Br during multiple historical surveying periods under normal operating conditions of the tailings dam. Br Distribution and statistics of the comprehensive risk index R of Br for each dam section in the tk survey period. Br The sample set of (tk) is used to calculate the product of the historical maximum value of the sample set and the safety margin coefficient to determine the trigger threshold Rth;

[0039] The comprehensive risk index R of the engineering dam section Br, which is acquired in real time, is used in the tk survey period. Br(tk) is compared with the trigger threshold Rth to determine the overall risk situation of the current dam section Br, and an adaptive monitoring zone generation request is triggered based on the comparison results; the specific comparison content is as follows:

[0040] When the comprehensive risk index R of the dam section Br during the tk survey period... Br When (tk)≥ trigger threshold Rth, it indicates that the overall risk of the current engineering dam section Br is abnormal. Then, an adaptive monitoring zone generation request is triggered for the engineering dam section Br, and the engineering dam section Br is determined as the dam section where the adaptive monitoring zone is located. Within the current engineering dam section Br, the adjacent monitoring units in the candidate risk monitoring unit set Hx are extended along the dam axis direction and the vertical direction of the slope to form the adaptive monitoring zone area Hd.

[0041] When the comprehensive risk index R of the dam section Br during the tk survey period... Br When (tk) ≥ trigger threshold Rth, it indicates that the overall risk of the current dam section Br is normal, and no operation is generated.

[0042] Preferably, S4 includes S41;

[0043] S41. After triggering the adaptive monitoring zone generation request for the engineering dam section Br, an adaptive monitoring zone region Hd is formed for the engineering dam section Br, and an adaptive monitoring generation strategy is executed; the adaptive monitoring zone generation strategy includes a bandwidth and location generation strategy, a monitoring zone internal point encryption strategy, and a monitoring zone external area sparsification strategy.

[0044] The bandwidth and location generation strategy determines the engineering dam section Br as the adaptive monitoring zone generation request, and takes the monitoring unit in the candidate risk monitoring unit set Hx within the engineering dam section Br as the center, and extends no less than 3 and no more than 5 adjacent monitoring units upstream and downstream along the dam axis in the dam body reference coordinate system.

[0045] Along the dam slope direction, the adaptive monitoring zone extends from the dam crest to the dam toe in the dam body reference coordinate system, ensuring that the adaptive monitoring zone covers at least 60%-100% of the dam height from the dam crest in the dam slope direction, and that the upper region of the water-facing slope and the toe region are both included within the adaptive monitoring zone region Hd. For dam sections that reach the design water level, the coverage height range is adjusted to 80%-100% of the dam height, thereby determining the spatial location and bandwidth of the adaptive monitoring zone region Hd.

[0046] The monitoring zone densification strategy involves adjusting the UAV flight path spacing from the baseline flight path spacing D0 to 0.5D0-0.8D0 within the adaptive monitoring zone area Hd, and adding at least one layer of flight paths with different flight altitudes and tilt angles on the basis of the original flight altitude. This is used to improve the density and solution reliability of the 3D point cloud within the adaptive monitoring zone area Hd.

[0047] The sparsification strategy for areas outside the monitoring zone maintains the baseline route scheme or adjusts the route spacing from the baseline route spacing D0 to 1.1D0-1.4D0 in engineering dam sections that are not identified as adaptive monitoring zones.

[0048] Preferably, S4 further includes S42;

[0049] S42. For the engineering dam section Br located within the adaptive monitoring zone Hd, based on the original re-measurement period Tjz and the comprehensive risk index R of the engineering dam section Br in the tk mapping period. Br (tk), calculate and output the retest interval T, and then introduce the retest interval T into the task degree layer of the UAV to retest the current engineering dam section Br according to the retest interval T;

[0050] The retest interval T is calculated and output using the following algorithm formula;

[0051] ;

[0052] In the formula, T Br (tk+1) represents the re-measurement interval of the engineering dam section Br in the next tk+1 surveying cycle, and g represents the cycle compression coefficient, with a value range of (0,1).

[0053] This invention provides a three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs). It has the following beneficial effects:

[0054] This method, based on multi-phase UAV 3D topographic mapping of tailings dams, uniformly divides the upstream slope, dam crest, and downstream slope of the dam body into regular dam surface monitoring units. At the scale of each monitoring unit, it simultaneously extracts dam surface response parameters such as vertical deformation rate parameter V, deformation increment trend parameter A, seepage visual indication parameter Q, and micro-topographic disturbance parameter M. After normalization processing, these parameters are input into the initial risk index formula for the multi-dimensional response of the dam surface to calculate and output the initial risk index R. This quantitatively characterizes the comprehensive deviation of each monitoring unit in the three dimensions of deformation, seepage, and microstructure at a fine spatial scale of 2m × 2m. Compared with traditional methods that rely solely on a few monitoring sections or a single deformation index for safety assessment, this invention, through multi-parameter fusion and the construction of a product-type risk index, enhances the early signs such as local "accelerated deformation," "abnormal seepage," and "micro-topographic disturbance" in the initial risk index R. This improves the sensitivity and accuracy of tailings dam hazard identification, reduces reliance on single deformation parameters and human experience judgment, and facilitates multi-dimensional comprehensive diagnosis of early anomalies.

[0055] This method, based on obtaining the initial risk index R of each monitoring unit, first performs temporal and spatial comparisons of R to form a candidate risk monitoring unit set Hx. Then, it divides the tailings dam into several engineering dam segments B_r along the dam axis, and statistically analyzes the initial risk index R of the candidate risk monitoring units within each engineering dam segment to obtain the comprehensive risk index R of the dam segment. Br The adaptive monitoring zone is then compared with the trigger threshold Rth to determine the engineering dam section where the adaptive monitoring zone is located. For the triggered engineering dam section Br, an adaptive monitoring zone region Hd is further generated within this dam section. A bandwidth and location generation strategy, a monitoring zone point densification strategy, and an out-of-monitoring zone sparsification strategy are implemented. Within the adaptive monitoring zone region Hd, the flight path spacing is compressed to 0.5D0-0.8D0 of the baseline flight path spacing D0. Furthermore, multi-altitude and multi-inclination angle flight paths are combined to improve the redundancy of the three-dimensional solution. Simultaneously, at the engineering dam section scale, the retest interval T is utilized. Br The inverse scaling formula adaptively compresses the re-measurement time, enabling high-risk dam sections Br to achieve higher temporal resolution, while non-triggered dam sections are still surveyed according to the baseline flight path and re-measurement baseline interval. Through this mechanism, under the premise of unchanged hardware conditions, this invention transforms the limited UAV flight time and data processing capabilities from a coarse and uniform distribution across the entire dam to a spatial concentration towards the adaptive monitoring zone area Hd and a temporal concentration towards the high-risk dam section Br, significantly improving the efficiency of surveying resource utilization and reducing invalid repeated data collection in stable areas.

[0056] (3) This method combines the multidimensional initial risk index R constructed based on dam surface response parameters with the comprehensive risk index R of dam section. BrThe adaptive monitoring zone generation strategy and adaptive retest cycle control formula are organically combined to form a "three-dimensional mapping adaptive scheduling mechanism driven by dam deformation evolution": On the one hand, the initial risk index R integrates multi-source information such as vertical deformation rate, deformation increment trend, seepage visual indication, and micro-topographic disturbance into a comprehensive risk characterization at the monitoring unit scale, which can accurately identify local dam sections with significantly increased deformation acceleration; on the other hand, the comprehensive risk index R of the dam section Br High-risk information at the monitoring unit level is rolled up to the engineering dam section scale, and an adaptive monitoring zone region Hd is automatically constructed within the high-risk dam section Br using an adaptive monitoring zone generation strategy, with retest intervals T. Br The inverse scaling formula automatically shortens the repetitive surveying cycle for this dam section. Compared with the traditional static scheme that weakly couples the route design and surveying cycle with historical deformation and covers the entire area at a fixed cycle for a long time, this method transforms the three-dimensional topographic surveying mode from "static, uniform, and pre-set" to an adaptive monitoring mode that "dynamically adjusts the route density and re-survey interval as the initial risk index of the multi-dimensional response of the dam surface evolves." This achieves adaptive data encryption driven by the extrapolation of dam deformation trends, enabling dam sections with accelerated local deformation to be captured more frequently in time and to obtain higher density three-dimensional data support in space. This improves the early identification capability of local instability risks in tailings dams and the overall safety assurance level. Attached Figure Description

[0057] Figure 1 This is a schematic diagram illustrating the steps of a UAV-based three-dimensional terrain mapping method according to the present invention.

[0058] Figure 2 This is a schematic diagram of the regular grid and monitoring units on the dam surface. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1: This invention provides a method for three-dimensional terrain mapping based on unmanned aerial vehicles (UAVs). Please refer to [link / reference]. Figure 1 This includes the following steps:

[0061] S1. By using drones to conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route, a multi-phase three-dimensional topographic model under a unified coordinate system is obtained, and the tailings dam is divided into regular grids, with each regular grid defined as a monitoring unit.

[0062] S2. For each monitoring unit, extract the dam surface response parameters at different surveying cycles, and calculate the initial risk index R of the dam surface multidimensional response based on the dam surface response parameters;

[0063] S3. Perform time and spatial comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx; divide the tailings dam into several engineering dam sections Br along the dam axis, and statistically analyze the candidate risk monitoring unit set Hx within each engineering dam section Br to obtain the comprehensive risk index R of the dam section. Br And compare it with the preset trigger threshold Rth to determine the engineering dam section Br where the adaptive monitoring zone is located;

[0064] S4. Execute an adaptive monitoring zone generation strategy on the candidate risk monitoring unit set Hx, and adjust the retesting time based on the comprehensive risk index RBr to obtain the retesting interval T. Perform three-dimensional terrain mapping in the non-adaptive monitoring zone area according to the baseline route plan and the retesting baseline interval.

[0065] In this embodiment, the method first controls a drone equipped with an optical camera and an inertial measurement unit to conduct multi-stage three-dimensional topographic mapping of the tailings dam along a preset upper-level overall control route and a lower-level dam slope detail route via S1. After aerial triangulation and bundle adjustment, the data is unified to the same dam body reference coordinate system. The dam surface is divided into regular 2m×2m monitoring units along the dam axis and perpendicular to the slope, providing a stable and comparable spatial carrier for subsequent detailed analysis. In S2, using the dam surface monitoring units as basic units, dam surface response parameters such as vertical deformation rate parameter V, deformation increment trend parameter A, seepage visual indication parameter Q, and micro-topographic disturbance parameter M are extracted from the three-dimensional topographic model and images at different mapping cycles. After normalization, these parameters are input into the multi-dimensional response initial risk index formula to calculate the dimensionless initial risk index R. j (tk) converts multi-source information such as "elevation change, seepage marks, and small-scale bulging" into a single risk quantity that can be directly compared in time and space, so that even units with small deformation but obviously "accelerating" can be amplified in R; in S3, R is... j (tk) Perform time and spatial comparisons, retaining only monitoring units that both "continuously rise" and "have large areas of high values ​​on the dam surface" as the candidate risk monitoring unit set Hx, and divide the engineering dam segment Br along the dam axis according to the engineering design station number. Within each engineering dam segment, monitor the R in set Hx. j (tk) is statistically analyzed, and the upper quartile or other representative values ​​are taken as the comprehensive risk index R of the dam section. Br(tk), and then compared with the trigger threshold Rth obtained from historical statistics, high-risk engineering dam sections Br that need to be constructed with adaptive monitoring zones are selected to avoid the decision-making of the entire dam section being affected by a single noise point or isolated anomaly; in S4, only for the triggered engineering dam section Br, several monitoring units are extended along the dam axis and dam slope direction with the candidate risk monitoring unit set Hx as the center to form an adaptive monitoring zone area Hd. Within Hd, the spacing between UAV flight paths is compressed from D_0 to 0.5D0~0.8D0 and multi-altitude, tilted flight paths are superimposed to improve the redundancy of the point cloud. At the same time, according to R Br (tk) Calculate the adaptive retest interval T using the inverse scaling formula. Br (tk) allows high-risk dam sections to have a significantly shorter re-measurement cycle than the T_benchmark, while dam sections not included in the adaptive monitoring zone continue to be surveyed according to the benchmark route scheme and re-measurement benchmark interval; through the above process, this method, on the one hand, uses the initial risk index R on the monitoring unit U_j scale. j (tk) accurately characterizes the multidimensional response of the dam surface, and on the other hand, uses R on the Br scale of the engineering dam section. Br (tk), adaptive monitoring zone Hd and retest interval T Br (tk) directly transforms "risk level" into differential control of "route density and flight frequency", thereby adjusting the 3D mapping resources from a coarse and uniform distribution across the entire dam to spatial densification and temporal frequency enhancement for high-risk dam sections without changing the hardware. This reduces invalid repeated acquisition in stable areas, effectively lowers the probability of missed or late detection of local accelerated deformation, and significantly improves the early warning capability and resource utilization efficiency of 3D topographic monitoring of tailings dams.

[0066] Example 2: Please refer to Figure 1 and Figure 2 Specifically: S1 includes S11;

[0067] S11. Using a drone equipped with an optical camera and an inertial measurement unit, conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route;

[0068] The preset flight paths include an overall control flight path for the upper dam body covering the entire tailings dam area and detailed flight paths for the lower dam slope laid out along the dam slope; forward-looking oblique photography flight paths and backward-looking oblique photography flight paths are set up near the water-facing slope of the tailings dam and the accumulation shoreline to improve the three-dimensional solution quality of the dam toe and slope toe areas.

[0069] In each surveying mission, the drone is controlled to fly along the preset route in sequence, and three-dimensional terrain data is collected through the surveying module on the drone.

[0070] The mapping module includes an optical camera, an inertial measurement unit, and a flight control log.

[0071] 3D terrain data includes imagery data, acceleration data, and log data.

[0072] The 3D terrain data is transmitted to a cloud-based mapping server via a wireless network. Aerial triangulation and bundle adjustment are performed on the cloud-based mapping server based on the 3D terrain data. The 3D terrain data obtained from each phase of the survey are uniformly calculated into the same dam body reference coordinate system to form a multi-phase 3D terrain model under a unified coordinate system.

[0073] Triangulation is a measurement method that determines the location of unknown points by using known control points and directions. In photogrammetry, it is often used for aerial triangulation of regional networks to calculate the exterior orientation elements and ground point coordinates of images. Bundle adjustment, also known as bundle adjustment, is an advanced technique in triangulation. It uses bundles of images as units, establishes a mathematical model through collinearity equations, and optimizes the exterior orientation elements and ground point coordinates of all images using the least squares method to achieve high-precision 3D modeling. The core process of bundle adjustment includes: determining the initial values ​​of exterior orientation elements and ground point coordinates, constructing error equations, solving algorithm equations, and iteratively optimizing. This method directly uses the original image point coordinate observations, has a rigorous theoretical system, and can effectively compensate for systematic errors (such as lens distortion), thus it is widely used in high-precision mapping. In contrast, traditional triangulation methods (such as the flight strip method) may indirectly process data and have lower accuracy. In practical applications, bundle adjustment requires high-quality initial values ​​and powerful computing resources, and is suitable for scenarios such as aerial photogrammetry and topographic mapping.

[0074] S1 also includes S12;

[0075] S12. Project the upstream slope, dam crest, and downstream slope of the tailings dam in the multi-phase 3D terrain model under a unified coordinate system onto the dam reference coordinate system plane. Divide the dam axis direction and the slope perpendicular direction into regular grids. Define each regular grid cell as a dam surface monitoring unit. The regular grid is set to 2m×2m according to the ground resolution of UAV imagery and the accuracy of 3D calculation. This ensures that each dam surface monitoring unit has complete point cloud and image coverage in each phase of the 3D terrain model. Extract the 3D coordinates of the corresponding dam surface monitoring unit and the image data associated with the current dam surface monitoring unit in each 3D terrain model.

[0076] The dam slope is divided into regular grids along the dam axis and vertically along the slope. This means that several vertical grid lines are cut along the dam crest line (dam axis direction); and several horizontal grid lines are cut from the dam crest to the dam toe (vertical direction of the slope). When the two sets of grid lines intersect, the dam slope is divided into regular small grids, each of which is a dam surface monitoring unit.

[0077] In this embodiment, step S1 of the method, through the cooperation of S11 and S12, completes the construction of "unified, high-precision, and comparable" basic data of the dam surface. On the one hand, in S11, a UAV equipped with an optical camera and an inertial measurement unit is used to conduct multiple phases of aerial photography of the tailings dam according to the overall control flight path of the upper dam body, the detailed flight path of the lower dam slope, and the forward / backward oblique photography flight path of the water-facing slope. In the cloud mapping server, aerial triangulation and bundle adjustment are performed on the image data, acceleration data, and flight control logs, and the three-dimensional terrain data of each phase are uniformly calculated to the same dam body reference coordinate system. The purpose is to eliminate the coordinate deviation between different mapping batches, so that the elevation change in any subsequent mapping cycle is the actual dam body displacement rather than "coordinate system drift", fundamentally avoiding the problems of traditional multi-phase mapping. The problem of "misalignment at the same location" is addressed in the drawing. On the other hand, in S12, the upstream slope, dam crest, and downstream slope, which have been unified under the dam body reference coordinate system, are projected onto the plane. The dam surface monitoring units are divided into 2m×2m regular grids along the dam axis and the vertical direction of the slope. This ensures that each dam surface monitoring unit has sufficient density of point cloud and image coverage in the 3D terrain model of each period. Its true physical meaning is to break down the entire dam into stable and fixed "pixel grids", each representing a specific physical area of ​​the dam surface, such as a small element with a width of 2m and a height of 2m. When a certain upstream slope subsides by about 5cm between two periods, the 3D coordinate change of the corresponding dam surface monitoring unit can be clearly located and quantified, without being diluted by the smooth fitting of the adjacent area. By combining a unified coordinate system and a regular grid, this implementation method ensures both the temporal comparability of multi-period 3D terrain data and the uniform and detailed spatial coverage of key parts of the dam face (water-facing slope, dam crest, and downstream slope). This effectively avoids the problems of "invisibility between sections and local small-scale deformation being averaged out" caused by relying on a small number of cross sections or discrete monitoring points in the traditional method. It provides stable, reliable, and spatially reasonable basic data support for the subsequent extraction of dam face response parameters and calculation of the initial risk index R based on dam face monitoring units.

[0078] Example 3: Please refer to Figure 1 Specifically: S2 includes S21;

[0079] S21. In each monitoring unit, based on the three-dimensional terrain model at different surveying cycles, the dam surface response parameters are extracted.

[0080] The dam surface response parameters include the vertical deformation rate parameter V of the j-th monitoring unit during the tk-th survey period. j (tk), the deformation increment trend parameter A of the j-th monitoring unit in the tk-th survey period. j (tk), the seepage visual indication parameter Q of the j-th monitoring unit in the tk-th mapping cycle. j(tk) and the micro-topographic disturbance parameter M of the j-th monitoring unit in the tk mapping period. j (tk);

[0081] The dam surface response parameters are normalized using the maximum and minimum value normalization method to eliminate the unit dimensions of all parameters in the dam surface response parameters.

[0082] in:

[0083] The vertical deformation rate parameter V is obtained by extracting the three-dimensional point cloud of the monitoring unit and its neighborhood from the multi-period three-dimensional terrain model under a unified coordinate system at each monitoring unit, obtaining the local dam surface plane at the monitoring unit through least squares plane fitting, and calculating the normal vector as the dam surface normal direction; in each surveying cycle, the displacement vector of the three-dimensional coordinates of the representative point of the monitoring unit relative to the dam body reference point is vector-projected in the dam surface normal direction to obtain the dam surface normal direction elevation value. The dam surface normal direction elevation values ​​of multiple surveying cycles constitute the dam surface normal direction elevation value sequence. By performing least squares straight line fitting on the elevation value sequence of several recent periods, the fitting slope is used as the vertical deformation rate parameter V.

[0084] The deformation increment trend parameter A is obtained by constructing a deformation time series based on the elevation difference in the normal direction of the dam surface between adjacent monitoring periods, and by using difference or linear fitting methods.

[0085] The seepage visual indicator parameter Q is generated by projecting the position of the monitoring unit in the dam reference coordinate system onto the UAV image through the camera exterior orientation element. The image features such as brightness difference, color difference and texture direction consistency compared with the reference dry state are extracted in the corresponding image window and then standardized and nonlinearly mapped to form the seepage visual indicator parameter Q.

[0086] The micro-topographic disturbance parameter M is formed by statistically analyzing and normalizing the residuals of the point cloud relative to the fitted surface after stripping the large-scale dam slope structure from the point cloud of the monitoring unit and its neighboring area.

[0087] S2 also includes S22;

[0088] S22. Based on the normalized dam surface response parameters, substitute them into the formula for the initial risk index of the multidimensional response of the dam surface, calculate and output the initial risk index R, and quantitatively analyze the comprehensive deviation of the dam body in the three dimensions of deformation, seepage and microstructure.

[0089] The initial risk index R is calculated and output using the following formula for the initial risk index of the multidimensional response of the dam surface;

[0090] ;

[0091] In the formula, R j(tk) represents the initial risk index of the j-th monitoring unit in the tk-th surveying period;

[0092] The initial risk index R of the multidimensional response of the dam surface is essentially a multi-index risk assessment function, which combines the commonly used multi-index decision theory: weighted sum model (used to combine Q and M) and product model (used to reflect the amplification effect when multiple high-risk factors are superimposed).

[0093] The initial risk index R of the dam surface multidimensional response in this formula is an engineering combination and tailoring based on these two classical forms: the first term (1+V j (tk): Linear amplification from velocity-type indicators;

[0094] The second term (1+A) j (tk): Linear amplification from acceleration-type indicators

[0095] The third term (1+0.5·Qj(tk)+0.5·Mj(tk)) is a weighted sum of Q and M embedded in the product structure;

[0096] 0.5·Qj(tk) + 0.5·Mj(tk) introduces the "surface symptoms" and acts as an amplifier;

[0097] A weight of 0.5–0.5 will not allow the third item to easily “overtake” the first two items, but when Q and M are both high, it will significantly increase the overall initial risk index R. If the weight is set too extreme (e.g., 0.8–0.2), it is easy to cause either seepage or micro-topography to become overly dominant.

[0098] The overall approach uses a product model to create a "non-linear amplification effect" when multiple indicators are simultaneously high. This is a common model concept in the field of risk assessment, but the specific "combination of V, A, Q, M and 0.5+0.5" is an innovative design of this formula. At the same time, the product model with an overall baseline value of 1 is to prevent any parameter from approaching 0 and thus suppressing the overall R value, making it difficult to reflect the particularly dangerous situation of a certain item. With the addition of "1+", the overall R value is ≥1. When all parameters are 0 (completely normal), the output of R is 1, which naturally provides a baseline state of 1. The risk index of the normal monitoring unit is 1.

[0099] Explanation of dimensional consistency: The dam surface response parameters are normalized dimensionless parameters, therefore the initial risk index R obtained by multiplication is also a dimensionless value.

[0100] The physical significance of the mathematical structure: The reason for using a product structure instead of a simple summation is that if only addition is performed, when some parameters are large but others are small, the result will not reflect the high-risk scenario of "multiple factors being abnormal at the same time";

[0101] In the product structure, when multiple factors are simultaneously greater than 1, nonlinear amplification will occur, which is more in line with the engineering intuition that "large deformation, obvious seepage and severe micro-topography anomalies lead to a multiplied increase in risk."

[0102] The vertical deformation rate parameter and the deformation increment trend parameter are placed in two independent multipliers: ((1+V j (tk) reflects how fast the transformation is; (1+A) j (tk) indicates whether the deformation is accelerating;

[0103] Multiplying the two together is equivalent to emphasizing both absolute speed and whether the speed is increasing, avoiding looking at only one dimension.

[0104] The visual indicator parameters of seepage and the micro-topographic disturbance parameters are weighted and added to the third multiplier:

[0105] Q and M are both indicators of the "surface symptoms dimension" and belong to the same category (surface symptoms). The equal weighting of 0.5 + 0.5 reflects their equal importance. These two enter the third factor, which is multiplied by the first two deformation factors, indicating that "internal geometric deformation and surface seepage and structural anomalies" jointly determine the final risk.

[0106] In this embodiment, step S2 of the method quantifies "geometric changes, image features, and micro-topographic anomalies" into an initial risk index R for the monitoring unit through steps S21 and S22. j (tk): In S21, for each monitoring unit, firstly, the point cloud of the monitoring unit and its neighborhood is extracted based on a multi-period 3D terrain model. The local dam surface plane is fitted, and the dam surface normal direction is determined. The 3D coordinates of representative points from each period are projected along the dam surface normal direction to obtain the dam surface normal direction elevation value sequence. The least squares straight line slope of the elevation values ​​from recent periods is used as the vertical deformation rate parameter V. Then, the deformation time series is constructed using the dam surface normal direction elevation difference between adjacent monitoring periods, and the deformation increment trend parameter A is extracted. This allows both "how much has changed" and "whether it is accelerating" to be explicitly characterized. Simultaneously, the position of monitoring unit U_j is back-projected onto the UAV imagery using camera exterior orientation elements. Within the corresponding image window, brightness difference, chromaticity difference, and texture direction consistency are extracted to form the seepage visual indicator parameter Q. After stripping the large-scale dam slope structure, the point cloud residuals are statistically analyzed to obtain the micro-topographic disturbance parameter M. For example, when thin, elongated wet marks and slight bulging appear locally on the water-facing slope, even if the elevation change is not significant, Q and M will be significantly larger. Subsequently, the maximum-minimum normalization method is used to uniformly map V, A, Q, and M into dimensionless normalized parameters. These parameters are then substituted into the product-type dam surface multidimensional response initial risk index formula, and the initial risk index R is calculated and output in S22. This allows the four types of information—"deformation rate," "deformation acceleration," "seepage traces," and "micro-undulation anomalies"—to be processed through (1+V). j(tk), 1+A j (tk), 1+0.5Q j (tk)+0.5M j The structure (tk) works in tandem: on the one hand, when a single dimension is abnormal (e.g., deformation is accelerating but seepage is not obvious), the corresponding factor can increase R. j (tk), to avoid missing important trends; on the other hand, when multiple dimensions are simultaneously abnormal, the product structure produces nonlinear amplification, causing the R of the monitoring unit to be affected. j (tk) is much higher than that of the surrounding normal units, thus effectively distinguishing between "single noise or local disturbance" and "real danger caused by multiple factors". Through this design, this implementation method compresses dam surface response parameters with different dimensions and physical meanings into a unified dimensionless risk index R. j (tk) not only avoids the problem of "not being able to see early seepage + micro-bulging" caused by only looking at a single elevation difference in the traditional approach, but also provides a clear and calculable quantitative basis for the subsequent unified sorting and screening of monitoring units in time and space, which substantially improves the sensitivity and reliability of early anomaly identification of tailings dams.

[0107] Example 4: Please refer to Figure 1 Specifically: S3 includes S31;

[0108] S31. Perform time and space comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx;

[0109] Time comparison involves comparing the initial risk index R of the j-th monitoring unit corresponding to multiple mapping cycles in the tk-th mapping cycle at each monitoring unit. j (tk) forms a time series, and the difference ΔR between the initial risk index R of two adjacent periods is calculated. j (tk); The specific calculation formula is: ΔR j (tk)=R j (tk)-R j (tk-1), when the difference ΔR between the initial risk indices R of two adjacent periods j When (tk) is greater than the preset time change threshold ΔRs and appears continuously within the preset number of periods, the monitoring unit will be marked as a time risk monitoring unit.

[0110] Spatial comparison is achieved by comparing the initial risk index R of the j-th monitoring unit within the same dam section during the tk-th surveying period, within the same surveying cycle. j (tk) Perform spatial statistics to calculate the local mean and local standard deviation; and set the initial risk index R of the j-th monitoring unit in the same dam section during the tk mapping period. jMonitoring units whose (tk) is greater than the spatial threshold formed by the weighted combination of the local mean and the local standard deviation are marked as spatial high value monitoring units;

[0111] Monitoring units that simultaneously satisfy both time risk markers and spatial high-value markers are aggregated through spatial connectivity analysis to obtain a high-value zone continuously distributed along the dam surface, and all monitoring units within the high-value zone constitute a candidate risk monitoring unit set Hx.

[0112] S3 also includes S32;

[0113] S32. Extract the center line of the tailings dam crest in the dam reference coordinate system, and set multiple boundary points for the center line of the dam crest according to the engineering design station number. Define the spatial area corresponding to the center line of the dam crest between any two adjacent boundary points as an engineering dam segment Br, so that the engineering dam segments Br along the dam axis are connected end to end and do not overlap with each other.

[0114] Using the projection range of each dam segment Br in the dam body reference coordinate system, monitoring units whose centroids fall within this projection range are selected in the direction perpendicular to the slope along the dam axis, forming a set of monitoring units contained in the dam segment Br. Within each dam segment Br, monitoring units belonging to the current dam segment Br are selected from the candidate risk monitoring unit set Hx, and the initial risk index R of the j-th monitoring unit in the tk-th surveying period is extracted. j (tk), and the initial risk index R of these j-th monitoring units in the tk-th mapping period. j (tk) Sort by value from smallest to largest, and take the initial risk index R of the j-th monitoring unit in the upper quartile position during the tk-th surveying period. j (tk) serves as the comprehensive risk indicator R for the dam section Br during the tk survey period. Br (tk);

[0115] It should be noted that the initial risk index R of the j-th monitoring unit in the tk-th mapping period is... j (tk) represents a very fine grid level, with a scale of 2m×2m, representing the comprehensive risk in three dimensions: deformation, seepage, and micro-topographic anomalies on this small block;

[0116] The comprehensive risk index R of the dam section Br during the tk survey period. Br (tk) represents the level of an entire dam section. A single engineering dam section Br contains many monitoring units. Because the dam is very long and there are many monitoring units, it is difficult to make direct decisions based on the initial risk index R of only one monitoring unit. Therefore, the comprehensive risk index R of the engineering dam section Br during the tk survey period is crucial. BrAfter (tk), it can be seen that the overall risk of a section is significantly higher. Risk comparison and ranking are carried out at the Br level of the engineering dam section to determine which section to focus on first and which to focus on later.

[0117] S3 also includes S33;

[0118] S33. Based on the comprehensive risk index R of the engineering dam section Br during multiple historical surveying periods under normal operating conditions of the tailings dam. Br Distribution and statistics of the comprehensive risk index R of Br for each dam section in the tk survey period. Br The sample set of (tk) is used to calculate the product of the historical maximum value of the sample set and the safety margin coefficient to determine the trigger threshold Rth;

[0119] The comprehensive risk index R of the engineering dam section Br, which is acquired in real time, is used in the tk survey period. Br (tk) is compared with the trigger threshold Rth to determine the overall risk situation of the current dam section Br, and an adaptive monitoring zone generation request is triggered based on the comparison results; the specific comparison content is as follows:

[0120] When the comprehensive risk index R of the dam section Br during the tk survey period... Br When (tk)≥ trigger threshold Rth, it indicates that the overall risk of the current engineering dam section Br is abnormal. In this case, an adaptive monitoring zone generation request is triggered for the engineering dam section Br, and the engineering dam section Br is determined as the dam section where the adaptive monitoring zone is located. Within the current engineering dam section Br, with the monitoring unit in the candidate risk monitoring unit set Hx as the center, adjacent monitoring units are extended along the dam axis direction and the vertical direction of the slope to form an adaptive monitoring zone area Hd, which is used for subsequent steps to perform encrypted three-dimensional terrain mapping tasks within the adaptive monitoring zone area Hd.

[0121] When the comprehensive risk index R of the dam section Br during the tk survey period... Br When (tk) ≥ trigger threshold Rth, it indicates that the overall risk of the current dam section Br is normal, and no operation is generated.

[0122] In this embodiment, step S3 of the method uses S31, S32, and S33 to process "many grids of R". j (tk) converges into a decision chain of "whether a few key dam sections need to activate the adaptive monitoring zone": In S31, the initial risk index R of each monitoring unit is first... j (tk) performs time and space comparisons, using ΔR for the time comparison. j(tk)Judge whether it is "climbing up for several consecutive periods" to avoid misjudging as dangerous due to an accidental high value in a certain period; calculate the local average value and standard deviation within the same dam section in space, and mark the monitoring units that are significantly higher than the surrounding average level and exceed the spatial threshold as spatially high-value monitoring units, so as to filter out "scattered isolated points", only retain the monitoring units that are continuously increasing and成片偏高 (I'm not sure what this exactly means in English, maybe "significantly higher in patches") on the dam surface, aggregate them into a high-value zone continuously distributed along the dam surface through spatial connectivity analysis, and form a set Hx of candidate risk monitoring units. The real meaning is to select the truly "continuously deteriorating and already成片异常 (again, not sure about this exact English translation)" dangerous areas from a large number of 2m×2m small grids; in S32, divide the center line of the dam crest into multiple consecutive and non-overlapping engineering dam sections Br according to the engineering design stake number, and then only count the monitoring units R that belong to the set Hx of candidate risk monitoring units within each engineering dam section Br j (tk), take the upper quartile and other representative values as the comprehensive risk index R of the dam section Br (tk), so that "there is a series of grids with continuously high R j (tk)" will be concentratedly reflected in R Br (tk), and for those dam sections with only occasional fluctuations in individual grids, their R Br (tk) will not be pulled too high by a few abnormal points, solving the problem that the dam body is very long, there are many monitoring units, but engineering decisions must be made in units of dam sections; in S33, then use the distribution of R Br (tk) in multiple historical periods during the normal operation stage of the tailings dam, count the historical maximum value and multiply it by the safety margin coefficient to determine the trigger threshold Rth, compare the real-time R Br (tk) with Rth, and when R Br (t_k)≥Rth, trigger the request for generating an adaptive monitoring band, and expand to form an adaptive monitoring band area Hd centered on the set Hx of candidate risk monitoring units within the corresponding engineering dam section Br. If R Br (tk)<Rth, maintain regular monitoring and do not perform additional operations. For example, a certain dam section Br1 has always been stable historically, and R Br1 (tk) has been close to 1 for a long time, while a certain section Br2 has R Br2 (tk) quickly approaching and exceeding Rth in recent periods, and it will be automatically identified as a dam section that needs to be monitored with increased focus. Through such a design, the implementation of S3 achieves three things in a very straightforward manner: First, using the dual constraints of "time + space", filter out accidental noise and isolated anomalies from a large number of monitoring units, and only include the truly continuous and成片异常 (again, unsure about this English translation) areas into the set Hx of candidate risk monitoring units; Second, using the engineering dam section Br and the comprehensive risk index R of the dam section Br(tk) Completes risk roll-up from grid level to management unit level, providing a clear and quantifiable basis for prioritizing "which section of the dam to monitor first"; third, uses a trigger threshold Rth based on historical statistics to distinguish between "normal fluctuations" and "real anomalies", avoiding two extremes: either monitoring all dams as key targets and severely dispersing resources, or missing early dangerous dam sections due to excessively high thresholds. This lays a reliable decision-making foundation for subsequent implementation of route densification and retesting cycle compression only within the adaptive monitoring zone Hd.

[0123] Example 5: Please refer to Figure 1 Specifically: S4 includes S41;

[0124] S41. After triggering the adaptive monitoring zone generation request for the engineering dam section Br, an adaptive monitoring zone area Hd is formed for the engineering dam section Br, and the adaptive monitoring generation strategy is executed; the adaptive monitoring zone generation strategy includes bandwidth and location generation strategy, monitoring zone collection point encryption strategy and monitoring zone sparsification strategy.

[0125] The bandwidth and location generation strategy determines the engineering dam segment Br as the source of the adaptive monitoring zone generation request. Taking the monitoring unit in the candidate risk monitoring unit set Hx within the engineering dam segment Br as the center, it extends upstream and downstream along the dam axis direction in the dam body reference coordinate system by no less than 3 and no more than 5 adjacent monitoring units, so that the total bandwidth of the adaptive monitoring zone in the dam axis direction is the length of the candidate risk monitoring unit interval plus the extension width of 6-10 monitoring units.

[0126] Along the dam slope direction, the adaptive monitoring zone extends from the dam crest to the dam toe in the dam body reference coordinate system, ensuring that the adaptive monitoring zone covers at least 60%-100% of the dam height from the dam crest in the dam slope direction, and that the upper region of the water-facing slope and the toe region are both included within the adaptive monitoring zone region Hd. For dam sections that reach the design water level, the coverage height range is adjusted to 80%-100% of the dam height, thereby determining the spatial location and bandwidth of the adaptive monitoring zone region Hd.

[0127] The monitoring zone densification strategy involves adjusting the UAV flight path spacing from the baseline flight path spacing D0 to 0.5D0-0.8D0 within the adaptive monitoring zone area Hd, and adding at least one layer of flight paths with different flight altitudes and tilt angles on the basis of the original flight altitude. This is used to improve the density and solution reliability of the 3D point cloud within the adaptive monitoring zone area Hd.

[0128] The sparsification strategy for areas outside the monitoring zone reduces track density by maintaining the baseline flight path or adjusting the flight path spacing from the baseline spacing D0 to 1.1D0-1.4D0 in engineering dam sections that are not identified as adaptive monitoring zones. This reserves flight time and data processing resources for flight path densification and high-frequency retesting in the adaptive monitoring zone area Hd.

[0129] S4 also includes S42;

[0130] S42. For the engineering dam section Br located within the adaptive monitoring zone Hd, based on the original re-measurement period Tjz and the comprehensive risk index R of the engineering dam section Br in the tk mapping period. Br (tk), calculate and output the retest interval T, and then introduce the retest interval T into the task degree layer of the UAV to retest the current engineering dam section Br according to the retest interval T, so that the high-risk dam section can obtain more time resolution, while the low-risk dam section does not occupy too many resources.

[0131] The retest interval T is calculated and output using the following algorithm formula;

[0132] ;

[0133] In the formula, T Br (tk+1) represents the re-measurement interval of the dam section Br in the next tk+1 surveying cycle, and g represents the cycle compression coefficient, which takes values ​​in the range of (0,1] and is used to control the comprehensive risk index R of the dam section Br in the tk surveying cycle. Br The effect of (tk) on the degree of compression of the retest interval;

[0134] This formula is essentially a function in which the sampling period scales inversely with the risk. It can be inspired by two classic ideas: the "higher risk means higher sampling frequency" principle in control theory / sampling theory, which is common in control systems: the larger the error, the more frequent the control actions; corresponding to monitoring: the greater the risk, the shorter the retesting period (the higher the sampling frequency).

[0135] The denominator is a dimensionless parameter, and the numerator is a time dimension. Therefore, the output result of the retest interval T is in the time dimension.

[0136] In this embodiment, method S4 transforms "detecting high-risk dam sections" directly into an executable control strategy of "where to fly more densely and how often": In S41, once the corresponding engineering dam section Br is determined by S3 to trigger an adaptive monitoring band generation request, the density of monitoring is no longer averaged across the entire dam section. Instead, it expands 3-5 monitoring units upstream and downstream along the dam axis, centered on the monitoring units in the candidate risk monitoring unit set Hx within the engineering dam section Br, covering 60%-100% of the dam height in the dam slope direction (increasing to 80%-100% near the design water level). This is equivalent to creating a safety buffer around the "highest-risk grid," preventing high-risk areas from being diluted by sampling just outside the flight path, while also ensuring thorough coverage of the dam. The top and toe of the dam are the two most critical parts most prone to leakage and slippage. Simultaneously, within the adaptive monitoring zone Hd, the spacing between UAV flight paths is compressed from D0 to 0.5D0–0.8D0, and a layer of flight paths at different heights or tilt angles is superimposed. This is equivalent to "flying multiple paths and viewing from multiple angles" on a high-risk strip tens of meters wide, directly increasing the 3D point cloud density and solution redundancy. On the engineering dam section not within the adaptive monitoring zone, D0 is maintained or widened to 1.1D0–1.4D0. By flying fewer "long-term stable sections," battery time and computing resources are freed up, thus avoiding the problem of "one-size-fits-all densification across the entire dam" in the traditional model, which is both wasteful of aircraft and results in invalid data. In S42, the comprehensive risk index R of the engineering dam section Br in the tk mapping cycle is further calculated. Br (tk), using the inverse scaling formula, automatically calculates and outputs the next retest interval T from the original retest period Tjz. Br (tk+1), for example, for the dam section Br with a large RBr(tk), T Br (tk+1) will be significantly smaller than Tjz, which means that this dam section can acquire a set of high-precision three-dimensional data every 15 days or even less during the accelerated deformation phase, while R Br (t_k) The dam section that is close to normal remains at or slightly longer than Tjz, without encroaching on the limited flight window. In summary, this implementation process is achieved spatially through adaptive monitoring of the Hd area with "fixed-point densification," and temporally through retesting intervals T. Br (tk+1) "Frequency increase based on risk" means in reality that UAVs and 3D mapping systems should focus more lines of sight, more images, and more temporal resolution on dam sections that have shown an accelerating trend of deformation. This will significantly increase the probability of detecting local instability risks without changing the hardware, and reduce the risk of missed or late detections caused by relying solely on fixed routes and fixed cycles.

[0137] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended technical solutions and their equivalents.

Claims

1. A method for three-dimensional terrain mapping based on unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: S1. By using drones to conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route, a multi-phase three-dimensional topographic model under a unified coordinate system is obtained, and the tailings dam is divided into regular grids, with each regular grid defined as a monitoring unit. S2. For each monitoring unit, extract the dam surface response parameters at different surveying cycles, and calculate the initial risk index R of the dam surface multidimensional response based on the dam surface response parameters; S3. Perform time and spatial comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx; divide the tailings dam into several engineering dam sections Br along the dam axis, and statistically analyze the candidate risk monitoring unit set Hx within each engineering dam section Br to obtain the comprehensive risk index R of the dam section. Br And compare it with the preset trigger threshold Rth to determine the engineering dam section Br where the adaptive monitoring zone is located; S3 includes S31; S31. Perform time and space comparisons on the initial risk index R of the monitoring units to obtain the candidate risk monitoring unit set Hx; The time comparison is performed by comparing the initial risk index R of the j-th monitoring unit corresponding to multiple mapping cycles in the tk-th mapping cycle at each monitoring unit. j (tk) forms a time series, and the difference ΔR between the initial risk index R of two adjacent periods is calculated. j (tk); When the difference ΔR between the initial risk indices R of two adjacent periods... j When (tk) is greater than the preset time change threshold ΔRs and appears continuously within the preset number of periods, the monitoring unit will be marked as a time risk monitoring unit. The spatial comparison is performed by comparing the initial risk index R of the j-th monitoring unit within the same dam section during the tk-th mapping period, within the same mapping cycle. j (tk) Perform spatial statistics to calculate the local mean and local standard deviation; and set the initial risk index R of the j-th monitoring unit in the same dam section during the tk mapping period. j Monitoring units whose (tk) is greater than the spatial threshold formed by the weighted combination of the local mean and the local standard deviation are marked as spatial high value monitoring units; Monitoring units that simultaneously satisfy both time risk markers and spatial high value markers are aggregated through spatial connectivity analysis to obtain a high value zone continuously distributed along the dam surface, and all monitoring units within the high value zone constitute a candidate risk monitoring unit set Hx; S3 further includes S32; S32. Extract the center line of the tailings dam crest in the dam reference coordinate system, and set multiple boundary points for the center line of the dam crest according to the engineering design station number. Define the spatial area corresponding to the center line of the dam crest between any two adjacent boundary points as an engineering dam section Br. Using the projection range of each dam segment Br in the dam body reference coordinate system, monitoring units whose centroids fall within this projection range are selected in the direction perpendicular to the slope along the dam axis, forming a set of monitoring units contained in the dam segment Br. Within each dam segment Br, monitoring units belonging to the current dam segment Br are selected from the candidate risk monitoring unit set Hx, and the initial risk index R of the j-th monitoring unit in the tk-th surveying period is extracted. j (tk), and the initial risk index R of these j-th monitoring units in the tk-th mapping period. j (tk) Sort by value from smallest to largest, and take the initial risk index R of the j-th monitoring unit in the upper quartile position during the tk-th surveying period. j (tk) serves as the comprehensive risk indicator R for the dam section Br during the tk survey period. Br (tk); S4. Execute an adaptive monitoring zone generation strategy on the candidate risk monitoring unit set Hx, and adjust the retesting time based on the comprehensive risk index RBr to obtain the retesting interval T. Perform three-dimensional terrain mapping in the non-adaptive monitoring zone area according to the baseline route plan and the retesting baseline interval.

2. The method for three-dimensional terrain mapping based on unmanned aerial vehicles according to claim 1, characterized in that: S1 includes S11; S11. Using a drone equipped with an optical camera and an inertial measurement unit, conduct multi-phase three-dimensional topographic mapping of the tailings dam along a preset route; The preset routes include an overall control route for the upper dam body covering the entire tailings dam area and a detailed route for the lower dam slope laid out along the dam slope; and forward-looking oblique photography routes and backward-looking oblique photography routes are set up near the water-facing slope of the tailings dam and the accumulation shoreline. In each surveying mission, the drone is controlled to fly along the preset route in sequence, and three-dimensional terrain data is collected through the surveying module on the drone. The mapping module includes an optical camera, an inertial measurement unit, and a flight control log. The three-dimensional terrain data includes image data, acceleration data, and log data. The 3D terrain data is transmitted to a cloud-based mapping server via a wireless network. The cloud-based mapping server performs aerial triangulation and bundle adjustment based on the 3D terrain data, and the 3D terrain data obtained from each phase of the survey are uniformly calculated into the same dam body reference coordinate system to form a multi-phase 3D terrain model under a unified coordinate system.

3. The method for three-dimensional terrain mapping based on unmanned aerial vehicles according to claim 2, characterized in that: S1 further includes S12; S12. Project the upstream slope, dam crest, and downstream slope of the tailings dam in the multi-phase 3D terrain model under a unified coordinate system onto the dam reference coordinate system plane. Divide the dam axis direction and the slope perpendicular direction into regular grids. Define each regular grid cell as a dam surface monitoring unit. The regular grid is set to 2m×2m according to the ground resolution of UAV imagery and the accuracy of 3D calculation, so that each dam surface monitoring unit has complete point cloud and image coverage in each phase of the 3D terrain model. Extract the 3D coordinates of the corresponding dam surface monitoring unit and the image data associated with the current dam surface monitoring unit in each 3D terrain model.

4. The method for three-dimensional terrain mapping based on unmanned aerial vehicles according to claim 3, characterized in that: S2 includes S21; S21. In each monitoring unit, based on the three-dimensional terrain model at different surveying cycles, the dam surface response parameters are extracted. The dam surface response parameters include the vertical deformation rate parameter V of the j-th monitoring unit during the tk-th mapping period. j (tk), the deformation increment trend parameter A of the j-th monitoring unit in the tk-th survey period. j (tk), the seepage visual indication parameter Q of the j-th monitoring unit in the tk-th mapping cycle. j (tk) and the micro-topographic disturbance parameter M of the j-th monitoring unit in the tk mapping period. j (tk); The dam surface response parameters were normalized using the maximum and minimum value normalization method to eliminate the unit dimensions of all parameters in the dam surface response parameters.

5. A three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs) according to claim 4, characterized in that: S2 further includes S22; S22. Based on the normalized dam surface response parameters, substitute them into the formula for the initial risk index of the multidimensional response of the dam surface, calculate and output the initial risk index R, and quantitatively analyze the comprehensive deviation of the dam body in the three dimensions of deformation, seepage and microstructure. The initial risk index R is calculated and output using the following formula for the initial risk index of the multidimensional response of the dam surface; ； In the formula, R j (tk) represents the initial risk index of the j-th monitoring unit in the tk-th surveying period.

6. The method for three-dimensional terrain mapping based on unmanned aerial vehicles according to claim 1, characterized in that: S3 also includes S33; S33. Based on the comprehensive risk index R of the engineering dam section Br during multiple historical surveying periods under normal operating conditions of the tailings dam. Br Distribution and statistics of the comprehensive risk index R of Br for each dam section in the tk survey period. Br The sample set of (tk) is used to calculate the product of the historical maximum value of the sample set and the safety margin coefficient to determine the trigger threshold Rth; The comprehensive risk index R of the engineering dam section Br, which is acquired in real time, is used in the tk survey period. Br (tk) is compared with the trigger threshold Rth to determine the overall risk situation of the current dam section Br, and an adaptive monitoring zone generation request is triggered based on the comparison results; the specific comparison content is as follows: When the comprehensive risk index R of the dam section Br during the tk survey period is... Br When (tk)≥ trigger threshold Rth, it indicates that the overall risk of the current engineering dam section Br is abnormal. Then, an adaptive monitoring zone generation request is triggered for the engineering dam section Br, and the engineering dam section Br is determined as the dam section where the adaptive monitoring zone is located. Within the current engineering dam section Br, the adjacent monitoring units in the candidate risk monitoring unit set Hx are extended along the dam axis direction and the vertical direction of the slope to form the adaptive monitoring zone area Hd. When the comprehensive risk index R of the dam section Br during the tk survey period... Br When (tk) ≥ trigger threshold Rth, it indicates that the overall risk of the current dam section Br is normal, and no operation is generated.

7. A three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs) according to claim 6, characterized in that: S4 includes S41; S41. After triggering the adaptive monitoring zone generation request for the engineering dam section Br, an adaptive monitoring zone region Hd is formed for the engineering dam section Br, and an adaptive monitoring generation strategy is executed; the adaptive monitoring zone generation strategy includes a bandwidth and location generation strategy, a monitoring zone internal point encryption strategy, and a monitoring zone external area sparsification strategy. The bandwidth and location generation strategy determines the engineering dam section Br as the adaptive monitoring zone generation request, and takes the monitoring unit in the candidate risk monitoring unit set Hx within the engineering dam section Br as the center, and extends no less than 3 and no more than 5 adjacent monitoring units upstream and downstream along the dam axis in the dam body reference coordinate system. Along the dam slope direction, the adaptive monitoring zone extends from the dam crest to the dam toe in the dam body reference coordinate system, ensuring that the adaptive monitoring zone covers at least 60%-100% of the dam height from the dam crest in the dam slope direction, and that the upper region of the water-facing slope and the toe region are both included within the adaptive monitoring zone region Hd. For dam sections that reach the design water level, the coverage height range is adjusted to 80%-100% of the dam height, thereby determining the spatial location and bandwidth of the adaptive monitoring zone region Hd. The monitoring zone collection point encryption strategy adjusts the UAV flight path spacing from the baseline flight path spacing D0 to 0.5D0-0.8D0 within the adaptive monitoring zone area Hd, and adds at least one layer of flight paths with different flight altitudes and different tilt angles on the basis of the original flight altitude. The sparsification strategy for areas outside the monitoring zone maintains the baseline route scheme or adjusts the route spacing from the baseline route spacing D0 to 1.1D0-1.4D0 in engineering dam sections that are not identified as adaptive monitoring zones.

8. A three-dimensional terrain mapping method based on unmanned aerial vehicles (UAVs) according to claim 7, characterized in that: S4 further includes S42; S42. For the engineering dam section Br located within the adaptive monitoring zone Hd, based on the original re-measurement period Tjz and the comprehensive risk index R of the engineering dam section Br in the tk mapping period. Br (tk), calculate and output the retest interval T, and then introduce the retest interval T into the task degree layer of the UAV to retest the current engineering dam section Br according to the retest interval T; The retest interval T is calculated and output using the following algorithm formula; ； In the formula, T Br (tk+1) represents the re-measurement interval of the engineering dam section Br in the next tk+1 surveying cycle, and g represents the cycle compression coefficient, with a value range of (0,1).

Images