Slope radar and rockfall radar cooperative emergency monitoring method and system

Abstract

The present application relates to the technical field of geological disaster emergency monitoring, and discloses a kind of slope radar and rockfall radar coordinated emergency monitoring method and system, to solve the problems of existing method, such as co-site deployment difficulty, low data fusion accuracy and poor early warning accuracy, the scheme mainly includes: slope radar data is obtained by interference solution to obtain deformation parameters, combined with DEM image feature matching to solve pose, generate deformation point cloud;Rockfall radar data is converted into dynamic trajectory point cloud by coordinate conversion;Two kinds of point clouds are unified reference and divided into voxel grid, and the average deformation rate and the cumulative collapse energy are determined respectively, and the coupling early warning index is calculated;According to rainfall dynamic adjustment threshold, issue early warning;According to rockfall trajectory identification blind area, adaptively reduce the coherence threshold of high-frequency rockfall area, and extract low SNR deformation information to update point cloud.The present application realizes accurate fusion of heterogeneous data without requiring two radars to be physically co-located, and improves early warning accuracy.

Description

Technical Field

[0001] This invention relates to the field of geological disaster emergency monitoring technology, specifically to an emergency monitoring method and system that combines slope radar and rockfall radar. Background Technology

[0002] In emergency rescue scenarios involving large landslides and steep rock slopes, non-contact, all-weather monitoring technologies play a crucial role. Because the disaster site is extremely unstable, personnel cannot enter to install contact sensors; therefore, remote sensing monitoring technologies, represented by radar, have become the mainstream. Among these, slope radar and rockfall radar are two core monitoring devices, with significant differences in their working principles and functional characteristics, and they are naturally complementary.

[0003] Slope radar typically employs stepped-frequency continuous wave combined with synthetic aperture technology, using interferometry to achieve high-precision deformation monitoring. Its core advantages lie in its wide field of view, enabling panoramic imaging, and sub-millimeter level monitoring accuracy. It can capture minute creep deformations before landslides and generate continuous deformation-displacement cloud maps. However, slope radar has low temporal resolution, long scanning cycles, and cannot capture instantaneous rockfall events; furthermore, its phase maps require specialized interpretation, resulting in a lag in response to sudden events.

[0004] Rockfall radar typically employs frequency-modulated continuous wave Doppler pulses combined with real-aperture scanning technology, offering extremely high real-time performance with refresh rates exceeding 20Hz. It can track high-speed falling rocks in real time, providing information on the rock's three-dimensional trajectory, speed, and impact point. However, rockfall radar cannot measure slow deformations of mountains; it can only detect rockfall events that have already occurred. Furthermore, its field of view is usually narrow, making it difficult to cover the entire mountainside. The ability to detect small rocks is inversely proportional to the fourth power of the distance, necessitating close-range deployment.

[0005] The two types of radars mentioned above are naturally complementary in function: slope radar detects trends, while rockfall radar detects rockfall events. However, how to organically combine these two heterogeneous devices to achieve collaborative monitoring remains a bottleneck in current technological development. Existing fusion solutions have the following limitations:

[0006] First, physical co-location schemes are difficult to implement. Existing technologies often require two radars to be installed on the same platform or at very close range, achieving spatial benchmark unification through mechanical axis alignment. However, in disaster relief sites with complex terrain, it is difficult to find a flat site that can satisfy both the long-range panoramic view of the slope radar and the close-range upward view of the rolling stone radar, resulting in extremely strong deployment constraints and poor adaptability to actual combat.

[0007] Second, the accuracy of heterogeneous data fusion is low. Existing systems simply display the two-dimensional deformation pseudo-color map from the slope radar and the two-dimensional trajectory map from the boulder radar side-by-side on the screen or simply overlay them. Because the two types of radars have completely different observation perspectives and inconsistent slant range projection planes, direct overlay will produce a positioning error of tens of meters, making it impossible for commanders to determine whether the boulder originated from the current deformation area, resulting in a break in the early warning logic.

[0008] Third, there is a lack of quantitative risk correlation analysis. Traditional rockfall radar reports any rockfall as it appears, unable to distinguish between precursory rocksfalls to large-scale landslides and random rocksfalls caused by wind, leading to frequent interruptions in rescue efforts. Slope radar, affected by shadows or loss of coherence due to surface fracturing, often loses data in the final moments of a disaster, failing to issue imminent landslide warnings. Existing systems cannot determine whether rockfalls are caused by accelerated deformation of the overlying rock mass, easily resulting in false alarms or missed alarms. Summary of the Invention

[0009] This invention aims to address the problems of existing slope emergency monitoring methods, such as difficulties in co-location deployment, low data fusion accuracy, and poor early warning accuracy. It proposes an emergency monitoring method and system that combines slope radar and rock-rolling radar.

[0010] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0011] In a first aspect, the present invention provides an emergency monitoring method that combines slope radar and rockfall radar, the method comprising:

[0012] A slope radar is used to continuously scan the slope to be monitored, acquiring complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, the six-degree-of-freedom pose correction parameters of the radar system in geographic space are calculated through feature matching between simulation and real radar amplitude images. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are transformed to the geographic coordinate system, generating a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients.

[0013] The rolling stone radar is used to detect rolling stone targets on the slope and obtain the trajectory parameters of the rolling stones. Based on the observation geometric parameters of the rolling stone radar, the detected rolling stone targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, movement speed and normalized kinetic energy index.

[0014] The deformation point cloud and dynamic trajectory point cloud are unified to the same geospatial reference; the monitoring area is divided into a three-dimensional voxel grid under the geographic coordinate system; the deformation rate attribute in the deformation point cloud is mapped to the corresponding voxel, and the average deformation rate of each voxel is calculated; based on the rolling stone trajectory in the dynamic trajectory point cloud, the voxel where the collapse source of the rolling stone trajectory is located is determined by kinematic back-inference, and the corresponding normalized kinetic energy index is accumulated to the voxel to obtain the cumulative collapse energy of each voxel; based on the average deformation rate and cumulative collapse energy of each voxel, the deformation-collapse coupling early warning index of the voxel is calculated.

[0015] Set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; based on the comparison results, issue corresponding levels of early warning information to voxel areas that meet different combination conditions.

[0016] Based on the collapse source inverted from the dynamic trajectory point cloud, the data blind zone of the deformation point cloud is identified and marked; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, the coherence filtering threshold of the area in the slope radar data processing is adaptively reduced to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

[0017] Furthermore, feature matching is performed through simulation and real radar amplitude images, specifically including:

[0018] Simulated amplitude images are generated using digital elevation model (DEM) data. For each polygon in the DEM, the simulated amplitude value is... The calculation formula is:

[0019] ;

[0020] in, For the element index of the digital elevation model, The local incident angle between the radar beam and the surface element normal. The slant range from the surface element to the radar is given. It is a composite constant that includes transmit power, antenna gain, and average surface reflectivity. A binary function to characterize whether the surface element is occluded by the terrain;

[0021] Feature matching between simulated amplitude images and real radar amplitude images is achieved by maximizing mutual information, i.e., solving for the optimal radar pose correction parameters. This makes the true radar amplitude image The simulated amplitude image after transformation by the pose correction parameters The optimization objective is to maximize the mutual information between them, and the optimization objective is expressed as:

[0022] ;

[0023] in, For pose correction parameters that include 3D translation and 2D rotation, This represents a geometric transformation operation performed on a simulated image based on pose correction parameters. This is a mutual information function.

[0024] Furthermore, the detected rolling stone targets are converted to a geographic coordinate system, specifically based on the rolling stone radar's own location. Based on the measured distance, azimuth, and elevation angles, the three-dimensional spatial coordinates of the rolling stone point are calculated using the following coordinate transformation formula. :

[0025] ;

[0026] in, The slant range from the rolling stone target to the radar. It is the azimuth angle. It is the pitch angle.

[0027] Furthermore, the normalized kinetic energy index The calculation formula is:

[0028] ;

[0029] in, The radar cross section is used to characterize the size of the rolling stone. Let be the velocity vector of the rolling stone. This is the lithology-band calibration coefficient.

[0030] Furthermore, by using kinematic inverse deduction, the voxels at the collapse source of the rolling stone trajectory were determined, specifically including:

[0031] Extract the starting observation point of each rockfall trajectory, and search upwards for the nearest slope point using digital elevation model data to identify the collapse source.

[0032] Furthermore, the deformation-collapse coupling early warning index The calculation formula is:

[0033] ;

[0034] in, This corresponds to the average deformation rate within the voxel. This corresponds to the cumulative collapse energy within the voxel. and These are the historical maximum deformation rate and the historical maximum rockfall energy, respectively. and These are the weighting coefficients.

[0035] Furthermore, the deformation rate threshold is dynamically adjusted based on real-time rainfall data. Specifically, the adjusted deformation rate threshold is calculated using an exponential decay model with a lower bound constraint. The calculation formula is as follows:

[0036] ;

[0037] in, The initial deformation rate threshold, As the minimum safety threshold, For effective rainfall, This is the lithology sensitivity coefficient. It is a natural constant.

[0038] Furthermore, the coherence filtering threshold is adaptively reduced, specifically including:

[0039] When the cumulative collapse energy corresponding to a certain voxel is greater than zero and continues to increase, and the coherence coefficient of the region corresponding to the voxel is less than 0.3, the coherence filtering threshold of the voxel in the slope radar data processing is automatically reduced from the normal value to 0.1 to force the extraction of deformation parameters under low signal-to-noise ratio conditions, and the extraction results are updated to the deformation point cloud.

[0040] Furthermore, corresponding levels of early warning information are issued for voxel regions that meet different combinations of conditions, specifically including:

[0041] A blue alert is issued when the average deformation rate of a voxel is greater than the adjusted deformation rate threshold and the cumulative collapse energy of the voxel is less than or equal to the rockfall energy threshold.

[0042] A yellow alert is issued when the average deformation rate of a voxel is less than or equal to the adjusted deformation rate threshold and the cumulative collapse energy of the voxel is greater than the rockfall energy threshold.

[0043] A red alert is issued when the average deformation rate of a voxel exceeds the adjusted deformation rate threshold and the cumulative collapse energy of the voxel exceeds the rockfall energy threshold.

[0044] Secondly, the present invention provides an emergency monitoring system that combines slope radar and rockfall radar to achieve the emergency monitoring method that combines slope radar and rockfall radar as described in the first aspect, the system comprising:

[0045] The deformation point cloud construction module is used to continuously scan the slope to be monitored using slope radar to acquire complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, feature matching between simulation and real radar amplitude images is used to calculate the six-degree-of-freedom pose correction parameters of the radar system in geographic space. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are converted to the geographic coordinate system to generate a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients.

[0046] The trajectory point cloud construction module is used to detect rolling rocks on slopes using rolling rock radar and obtain the motion trajectory parameters of the rolling rocks. Based on the observation geometric parameters of the rolling rock radar, the detected rolling rock targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, motion speed and normalized kinetic energy index.

[0047] The heterogeneous data fusion module is used to unify the deformation point cloud and dynamic trajectory point cloud to the same geospatial reference; divide the monitoring area into a three-dimensional voxel grid under the geographic coordinate system; map the deformation rate attribute in the deformation point cloud to the corresponding voxel, and calculate the average deformation rate of each voxel; based on the rolling stone trajectory in the dynamic trajectory point cloud, determine the voxel where the collapse source of the rolling stone trajectory is located through kinematic back-inference, and accumulate the corresponding normalized kinetic energy index to the voxel to obtain the cumulative collapse energy of each voxel; and calculate the deformation-collapse coupling early warning index of each voxel based on the average deformation rate and cumulative collapse energy of each voxel.

[0048] The collaborative early warning module is used to set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; and issue corresponding levels of early warning information to voxel areas that meet different combination conditions based on the comparison results.

[0049] The monitoring feedback optimization module is used to identify and mark the data blind spots of the deformation point cloud based on the collapse source inverted from the dynamic trajectory point cloud; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, it adaptively reduces the coherence filtering threshold of the area in the slope radar data processing to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

[0050] The beneficial effects of this invention are as follows: The emergency monitoring method and system for coordinated slope radar and rockfall radar provided by this invention achieves spatial benchmark unification between slope radar and rockfall radar by simulating and matching features of real radar amplitude images, without requiring physical co-location of the two radars or the deployment of ground control points; it maps deformation point clouds and dynamic trajectory point clouds to the same voxel grid, establishing a physical correlation between deformation rate and rockfall energy, effectively solving the technical problems of low accuracy of heterogeneous data fusion and broken early warning logic; it adaptively reduces the coherence filtering threshold of blind zone and high-frequency rockfall area based on the rockfall trajectory inversion results, filling the data gaps of slope radar in the pre-slide stage, significantly reducing the false alarm rate and missed alarm rate, and providing accurate and reliable coordinated early warning capabilities for emergency monitoring. Attached Figure Description

[0051] Figure 1 A flowchart illustrating the emergency monitoring method combining slope radar and rockfall radar provided in this embodiment;

[0052] Figure 2 A schematic diagram of the deployment of slope radar and rock-rolling radar provided for an embodiment;

[0053] Figure 3 A schematic diagram illustrating the spatial registration principle based on terrain texture feature matching is provided for this embodiment.

[0054] Figure 4 A schematic diagram of heterogeneous data fusion based on spatiotemporal voxel grids is provided for an embodiment;

[0055] Figure 5 Another schematic diagram of heterogeneous data fusion based on spatiotemporal voxel grids is provided for the embodiment;

[0056] Figure 6 A schematic diagram of the closed-loop feedback mechanism provided in the embodiment;

[0057] Figure 7 This is a schematic diagram of the structure of an emergency monitoring system that combines slope radar and rockfall radar, provided as an example. Detailed Implementation

[0058] Because existing technologies cannot achieve precise spatial registration and physical correlation of heterogeneous data from slope radar and rockfall radar under non-co-located conditions, a quantifiable causal chain between deformation trends and rockfall events is lacking, making it difficult to effectively support collaborative early warning for geological disaster emergency monitoring. Therefore, there is an urgent need for a scheme that can achieve spatial benchmark unification, precise fusion of heterogeneous data, and collaborative early warning monitoring based on deformation-collapse causal chains between slope radar and rockfall radar without requiring the two radars to be physically co-located or ground control points to be deployed.

[0059] Based on this, the technical solution of this invention is proposed. In this invention, firstly, a simulated amplitude image is generated using digital elevation model data, and feature matching is performed with the actual radar amplitude image acquired by the slope radar to calculate the six-degree-of-freedom pose correction parameters of the radar system, thereby accurately converting the deformation point cloud acquired by the slope radar to a geographic coordinate system. Simultaneously, based on the observation geometric parameters of the bouldering radar, the detected boulder targets are converted to the same geographic coordinate system, generating a dynamic trajectory point cloud. Then, the two types of point clouds are unified to a geographic spatial reference and divided into a three-dimensional voxel mesh, and the deformation in the deformation point cloud is... The rate is mapped to the corresponding voxel to calculate the average deformation rate. The rockfall trajectory in the dynamic trajectory point cloud is back-inferred from the kinematics to determine the collapse source and the accumulated collapse energy is obtained by adding the normalized kinetic energy index. Then, the deformation-collapse coupling early warning index of each voxel is calculated. Finally, the data blind zone of the deformation point cloud is identified based on the rockfall trajectory inversion results, and the coherence filtering threshold is adaptively reduced for high-frequency rockfall areas to extract deformation parameters under low signal-to-noise ratio conditions and update them to the deformation point cloud. This achieves a complete closed loop of unified spatial benchmark, heterogeneous data fusion and monitoring feedback optimization.

[0060] The technical solutions in this embodiment 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.

[0061] Figure 1 A flowchart illustrating an emergency monitoring method that combines slope radar and rockfall radar is provided. Please refer to [link / reference]. Figure 1 The method includes the following steps:

[0062] Step 1: Deformation point cloud construction:

[0063] A slope radar is used to continuously scan the slope to be monitored, acquiring complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, the six-degree-of-freedom pose correction parameters of the radar system in geographic space are calculated through feature matching between simulation and real radar amplitude images. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are transformed to the geographic coordinate system, generating a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients.

[0064] This step aims to convert the raw radar data collected by the slope radar into three-dimensional point cloud data with deformation attributes in a geographic coordinate system, laying a spatial benchmark for subsequent fusion with the rolling stone radar data.

[0065] Specifically, please refer to Figure 2First, slope radar is used to continuously scan the slope to be monitored, acquiring complex image data. This complex image data is then subjected to interferometric processing to generate an interferometric phase map. Algorithms such as minimum cost flow are used for phase unwrapping to obtain the absolute phase difference. Combined with the radar wavelength, the phase difference is converted into line-of-sight deformation values, thereby calculating the cumulative deformation value, deformation rate, and deformation acceleration of each pixel within the monitoring period. Simultaneously, the coherence coefficient of each pixel is obtained through the interferometric processing to characterize the stability of the surface medium.

[0066] Since ground control points such as corner reflectors cannot be deployed at the emergency monitoring site, this step adopts a control-point-free geometric correction method based on terrain texture features. Figure 3 A schematic diagram of a spatial registration principle based on terrain texture feature matching is shown. Specifically, high-precision digital elevation model (DEM) data acquired by UAV is used to generate a simulated amplitude image by reverse simulation combined with a radar scattering physical model. The simulated amplitude image is then matched with the real radar amplitude image acquired by the slope radar. By maximizing the mutual information between the two, the six-degree-of-freedom pose correction parameters (including three-dimensional translation and two-dimensional rotation) of the radar system in geographic space are iteratively searched and calculated.

[0067] Specifically, simulated amplitude images are generated using digital elevation model (DEM) data. For each surface element in the DEM, the simulated amplitude value is... The calculation formula is:

[0068] ;

[0069] in, For the element index of the digital elevation model, The local incident angle between the radar beam and the surface element normal. The slant range from the surface element to the radar. It is a composite constant that includes transmit power, antenna gain, and average surface reflectivity. A binary function to characterize whether the surface element is occluded by the terrain;

[0070] Feature matching between simulated amplitude images and real radar amplitude images is achieved by maximizing mutual information, i.e., solving for the optimal radar pose correction parameters. This makes the true radar amplitude image The simulated amplitude image after transformation by the pose correction parameters The optimization objective is to maximize the mutual information between them, and the optimization objective is expressed as:

[0071] ;

[0072] in, For pose correction parameters that include 3D translation and 2D rotation, This represents a geometric transformation operation performed on a simulated image based on pose correction parameters. This is a mutual information function.

[0073] Finally, based on the calculated pose correction parameters, a projection matrix from the radar coordinate system to the geographic coordinate system is established. The deformation parameters and coherence coefficients of each pixel in the radar coordinate system are then back-projected to the geographic coordinate system, generating a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficient. ,in, In three-dimensional space coordinates, For deformation rate, This is the deformation acceleration, used to determine whether the slip phase has begun. The coherence coefficient is used to characterize the stability of the surface medium and identify vegetation or exposed rock. This deformation point cloud enables precise geospatial encoding of slope radar monitoring data, providing a unified data foundation for subsequent heterogeneous data fusion.

[0074] Step 2, Trajectory Point Cloud Construction:

[0075] The rolling stone radar is used to detect rolling stone targets on the slope and obtain the trajectory parameters of the rolling stones. Based on the observation geometric parameters of the rolling stone radar, the detected rolling stone targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, movement speed and normalized kinetic energy index.

[0076] This step aims to convert moving targets detected by the Rolling Stone radar into dynamic point cloud data with kinematic and energy attributes in a geographic coordinate system, providing data support for subsequent spatiotemporal correlation analysis with deformed point clouds.

[0077] Specifically, please refer to Figure 2 First, a rolling stone radar was used to continuously detect the slope and obtain the raw echo signal. To eliminate environmental clutter interference such as rainfall and vegetation swaying, a short-time Fourier transform was performed on the echo signal to extract the micro-Doppler features (rigid body translation and spin) unique to the rolling stone target. A classifier was constructed to filter out vegetation clutter with reciprocating swaying characteristics, and a unit average constant false alarm rate detector was used to extract the range-Doppler trace of the real rolling stone target.

[0078] To address the coexistence of multiple motion modes during the rolling stone movement, including free fall, collision rebound, and rolling, an interactive multi-model algorithm is employed for trajectory tracking. During the tracking process, a high-precision digital elevation model is introduced as a state constraint: when the predicted trajectory penetrates the terrain model, the terrain normal vector is used for forced correction to prevent trajectory drift, thereby outputting a continuous and smooth rolling stone trajectory.

[0079] After obtaining a stable rolling stone trajectory, coordinate transformation is performed based on the observation geometric parameters of the rolling stone radar. This transformation is based on the rolling stone radar's own position. The measured distance, azimuth, and elevation angles are used to calculate the three-dimensional spatial coordinates of the rolling stone point using coordinate transformation formulas. This ensures that the rolling stone point cloud and the deformed point cloud generated in step 1 are on the same spatial reference. The coordinate transformation formula is as follows:

[0080] ;

[0081] in, The slant range from the rolling stone target to the radar. It is the azimuth angle. It is the pitch angle.

[0082] Meanwhile, to quantify the destructive potential of rolling stones, a normalized kinetic energy index was constructed. The normalized kinetic energy index is calculated based on the radar cross section (approximately representing the size of the boulder) and the boulder's velocity vector (synthesized from the Doppler radial velocity and angular rate of change), and corrected by introducing a lithology-band calibration coefficient. The calculation formula is as follows:

[0083] ;

[0084] in, The radar cross section is used to characterize the size of the rolling stone. Let be the velocity vector of the rolling stone. This is the lithology-band calibration coefficient.

[0085] The final generated dynamic trajectory point cloud includes timestamps. Three-dimensional spatial coordinates Speed ​​of movement and normalized kinetic energy index , recorded as This dynamic trajectory point cloud fully records the spatial location, motion state, and energy characteristics of each rolling stone event, providing a quantifiable data foundation for subsequent fusion analysis with deformation point clouds.

[0086] Step 3: Heterogeneous data fusion:

[0087] The deformation point cloud and dynamic trajectory point cloud are unified to the same geospatial reference; the monitoring area is divided into a three-dimensional voxel grid under the geographic coordinate system; the deformation rate attribute in the deformation point cloud is mapped to the corresponding voxel, and the average deformation rate of each voxel is calculated; based on the rolling stone trajectory in the dynamic trajectory point cloud, the voxel where the collapse source of the rolling stone trajectory is located is determined by kinematic back-inference, and the corresponding normalized kinetic energy index is accumulated to the voxel to obtain the cumulative collapse energy of each voxel; based on the average deformation rate and cumulative collapse energy of each voxel, the deformation-collapse coupling early warning index of the voxel is calculated.

[0088] This step aims to fuse and analyze the deformation point cloud generated in step 1 and the dynamic trajectory point cloud generated in step 2 within a unified spatial framework, establish a quantitative correlation between deformation trends and rolling stone events, and provide a basis for risk warning decisions.

[0089] Specifically, the first step is to unify the deformed point cloud and the dynamic trajectory point cloud to the same geospatial reference. Since steps 1 and 2 have already completed the unification of the spatial reference through terrain feature matching and coordinate transformation, this step directly loads the two sets of point cloud data into the same geographic coordinate system to ensure that they correspond precisely in spatial location.

[0090] Secondly, the three-dimensional space of the monitoring area is divided into a regular three-dimensional voxel grid, denoted as... The size of the voxel grid is set according to the monitoring resolution. Each voxel serves as a basic analysis unit, used to store the fusion attributes of that spatial location. Through voxelization, discrete point cloud data is converted into structured spatial grid data, facilitating subsequent statistical analysis.

[0091] Please see Figure 4 and Figure 5 Based on the voxel grid, bidirectional data mapping is performed:

[0092] Forward mapping: Traversing the deformed point cloud generated in step 1 It will fall into each voxel Deformation rate of all points within Statistical analysis was performed to calculate the average deformation rate of the voxel. Average deformation rate It characterizes the overall deformation intensity of the spatial region during the current monitoring period.

[0093] Reverse backtracking: For the dynamic trajectory point cloud generated in step 2 For each rolling stone trajectory, extract its starting observation point. Using digital elevation model data, the nearest slope point is searched upwards from the starting point and identified as the landslide source. The normalized kinetic energy index corresponding to this trajectory is then calculated. The accumulated collapse energy of the source voxel is obtained by adding it to the voxel containing the source voxel. Accumulated collapse energy It represents the total energy released by the region as a source of rolling stones over a period of time.

[0094] Finally, based on the average deformation rate of each voxel and accumulated collapse energy Calculate the deformation-collapse coupling early warning index :

[0095] ;

[0096] in, This corresponds to the average deformation rate within the voxel. This corresponds to the cumulative collapse energy within the voxel. and These are the historical maximum deformation rate and the historical maximum rockfall energy, used for normalization to unify the data to the [0, 1] interval. and This is a weighting coefficient, which is adjusted according to the monitoring focus. If the primary concern is preventing major landslides, it can be increased. If the primary concern is preventing injuries from falling rocks, the elevation can be increased. .

[0097] Deformation-collapse coupling early warning index The design philosophy is that the instability risk of a region depends both on its own deformation trend (potential energy accumulation) and on its actual collapse events (kinetic energy release). When both are significant, it indicates that the region is in an active collapse phase, with the highest risk level. This coupled early warning index quantitatively integrates the macroscopic deformation trend of slope radar with the microscopic collapse events of rockfall radar at the voxel level. It provides a unified quantitative index, which can be used to quantify the instability risk of each voxel. For example, if a region has a high deformation rate and the rolling stone radar continuously tracks rolling stones falling from that region, then... The surge indicates that the region is in a period of active disintegration.

[0098] Step 4, Collaborative Early Warning:

[0099] Set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; based on the comparison results, issue corresponding levels of early warning information to voxel areas that meet different combination conditions.

[0100] This step aims to determine dynamic thresholds based on the average deformation rate and cumulative collapse energy of each voxel calculated in step 3, combined with environmental factors, to achieve hierarchical collaborative early warning.

[0101] Specifically, firstly, a preset initial deformation rate threshold is established. and rockfall energy threshold The initial deformation rate threshold is set based on the slope rock mass type and historical monitoring data, and is used to determine whether the deformation has entered the acceleration stage; the rockfall energy threshold is determined based on the rockfall radar detection capability and on-site rockfall risk assessment, and is used to determine whether the rockfall is hazardous.

[0102] Considering that rainfall can soften soil and rock and reduce slope stability, this step introduces a rainfall correction mechanism to dynamically adjust the initial deformation rate threshold. The effective rainfall is calculated based on real-time monitored rainfall data. (Using a weighted cumulative method, such as...) The adjusted deformation rate threshold is calculated using an exponential decay model with a lower bound constraint. :

[0103] ;

[0104] in, The minimum safety threshold for the system is determined by the measurement accuracy of the radar equipment, ensuring that the adjusted threshold is not lower than the noise floor. The lithology sensitivity coefficient reflects the sensitivity of different soil and rock masses to rainfall. This model causes the deformation rate threshold to decrease exponentially with increasing rainfall, while ensuring it does not fall below the physical limits of the equipment. This responds to the impact of rainfall on slope stability while avoiding false alarms caused by excessively low thresholds.

[0105] Then, the average deformation rate of each voxel. With the adjusted deformation rate threshold Compare the cumulative collapse energy of the voxel. Energy threshold of falling rocks A comparison is made. Based on the comparison results, corresponding warning messages are issued for voxel regions that meet different combinations of conditions:

[0106] Blue Alert (Deformation Concern Level): When and The data was released in a timely manner. Under these conditions, the slope radar detected significant deformation, but the rockfall radar did not detect any effective falling rocks, indicating that although the soil and rock mass was deforming, its overall integrity was still good and it was in the creep stage, requiring increased monitoring frequency.

[0107] Yellow Alert (Rolling Stone Warning Level): When and Released in a timely manner. In this state, deformation is not obvious but rockfalls occur frequently. It is usually a point-like disaster caused by surface weathering and peeling or loosening of isolated rocks. It is necessary to delineate rockfall danger zones and warn people to avoid them.

[0108] Red Alert (Precipitated Landslide / Landslide Level): When and The alarm was issued immediately. Under these conditions, accelerated deformation and high-frequency rockfall occur simultaneously, creating a positive feedback loop between deformation and collapse. This indicates that the rock and soil structure is fractured, and the slope is about to experience large-scale instability, necessitating an immediate evacuation alert.

[0109] Through the above three-level collaborative early warning mechanism, this step organically combines the macroscopic deformation trend of slope radar with the microscopic landslide events of rockfall radar, effectively distinguishing between tectonic landslide precursors (high deformation and high frequency rockfall) and random environmental rockfalls (no deformation and occasional rockfall), significantly reducing the false alarm rate and missed alarm rate, and providing a clear and reliable decision-making basis for emergency command.

[0110] Step 5: Monitoring and Feedback Optimization:

[0111] Based on the collapse source inverted from the dynamic trajectory point cloud, the data blind zone of the deformation point cloud is identified and marked; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, the coherence filtering threshold of the area in the slope radar data processing is adaptively reduced to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

[0112] Please see Figure 6 This step aims to use the dynamic monitoring results of the rockfall radar to optimize the data processing strategy of the slope radar, forming a closed-loop feedback mechanism of "deformation accumulation-collapse response" to achieve complementary and enhanced monitoring capabilities.

[0113] Specifically, firstly, based on the location of the landslide source determined by kinematic inversion of the rockfall trajectory in step 3, the data blind spots of the slope radar deformation point cloud are identified and marked. Since slope radars are usually deployed at long distances, there may be shadowed areas (areas that cannot be illuminated by the radar beam) due to terrain undulations. These areas appear as missing data in the deformation point cloud. When the landslide source determined by the rockfall trajectory detected by the rockfall radar is located in these shadowed areas, the area is automatically marked as a high-risk blind spot in the 3D model, indicating that supplementary reconnaissance by drones or adjustment of the radar deployment position is needed to avoid missed detections due to limited viewing angle.

[0114] Secondly, based on the high-frequency rockfall areas indicated by the dynamic trajectory point cloud, the data processing parameters of the slope radar in these areas are adaptively adjusted. In the lead-up to a landslide, severe surface fracturing often causes a sharp drop in the coherence coefficient of the slope radar (typically less than 0.3). Deformation signals in these areas are easily filtered out as noise in conventional processing, leading to the loss of crucial data during the pre-landslide phase. In this embodiment, when the cumulative landslide energy corresponding to a certain voxel... When the coherence coefficient of the voxel is greater than zero and continues to increase, and the coherence coefficient of the region corresponding to the voxel is less than 0.3, the coherence filtering threshold of the voxel in the slope radar data processing is automatically reduced from the normal value (such as 0.3-0.5) to below 0.1, so as to force the extraction of deformation parameters under low signal-to-noise ratio conditions.

[0115] Specifically, in the signal processing flow of slope radar, the coherence coefficient is typically used to screen reliable observation points: pixels with high coherence are considered stable surface observation points and included in deformation calculation; pixels with low coherence are usually filtered out. In this embodiment, once the rockfall radar identifies a region with frequent rockfalls, it automatically locks the voxel corresponding to that region, forcibly lowering the data filtering standard for that region, allowing signals from low-coherence regions to participate in deformation calculation. In this way, even under conditions of extremely low coherence, it can forcibly extract severe deformation signals from a noisy background, ensuring that monitoring targets are not lost in the final stages of a disaster.

[0116] Finally, the deformation parameters obtained by forced extraction are updated into the deformation point cloud, enabling the deformation point cloud to continuously reflect the true state of the slope during the pre-slip stage. This feedback optimization mechanism utilizes the real-time detection results of the rock-rolling radar to effectively compensate for the data loss caused by the loss of coherence in the slope radar during the pre-slip stage, achieving comprehensive monitoring of the entire process of the slope from initial creep to final collapse.

[0117] In summary, this embodiment achieves high-precision spatial benchmark unification of slope radar deformation point clouds and rockfall radar dynamic trajectory point clouds by simulating and matching features with real radar amplitude images, without requiring physical co-location of two radars or the deployment of ground control points. Furthermore, by mapping the two types of point clouds to the same voxel grid and calculating the average deformation rate and cumulative collapse energy of each voxel, a physical correlation between deformation trends and rockfall events is established, realizing quantitative fusion of heterogeneous data and accurate assessment of the deformation-collapse coupling early warning index. Simultaneously, based on the rockfall trajectory inversion results, data blind spots in the deformation point cloud are identified, and the coherence filtering threshold is adaptively reduced for high-frequency rockfall areas. Deformation parameters under low signal-to-noise ratio conditions are forcibly extracted and updated to the deformation point cloud, forming a closed-loop feedback mechanism. This effectively fills the data gaps in the slope radar during the pre-slide stage, significantly reducing false alarm and missed alarm rates, and providing accurate, reliable, and full-process collaborative early warning capabilities for geological disaster emergency monitoring.

[0118] Based on the above technical solution, this embodiment also proposes an emergency monitoring system that combines slope radar and rockfall radar to achieve the emergency monitoring method described in the embodiment. Please refer to [link to relevant documentation]. Figure 7 The system includes:

[0119] The deformation point cloud construction module is used to continuously scan the slope to be monitored using slope radar to acquire complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, feature matching between simulation and real radar amplitude images is used to calculate the six-degree-of-freedom pose correction parameters of the radar system in geographic space. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are converted to the geographic coordinate system to generate a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients.

[0120] The trajectory point cloud construction module is used to detect rolling rocks on slopes using rolling rock radar and obtain the motion trajectory parameters of the rolling rocks. Based on the observation geometric parameters of the rolling rock radar, the detected rolling rock targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, motion speed and normalized kinetic energy index.

[0121] The heterogeneous data fusion module is used to unify the deformation point cloud and dynamic trajectory point cloud to the same geospatial reference; divide the monitoring area into a three-dimensional voxel grid under the geographic coordinate system; map the deformation rate attribute in the deformation point cloud to the corresponding voxel, and calculate the average deformation rate of each voxel; based on the rolling stone trajectory in the dynamic trajectory point cloud, determine the voxel where the collapse source of the rolling stone trajectory is located through kinematic back-inference, and accumulate the corresponding normalized kinetic energy index to the voxel to obtain the cumulative collapse energy of each voxel; and calculate the deformation-collapse coupling early warning index of each voxel based on the average deformation rate and cumulative collapse energy of each voxel.

[0122] The collaborative early warning module is used to set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; and issue corresponding levels of early warning information to voxel areas that meet different combination conditions based on the comparison results.

[0123] The monitoring feedback optimization module is used to identify and mark the data blind spots of the deformation point cloud based on the collapse source inverted from the dynamic trajectory point cloud; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, it adaptively reduces the coherence filtering threshold of the area in the slope radar data processing to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

[0124] It is understood that the emergency monitoring system for the coordinated use of slope radar and rock-rolling radar described in this embodiment is a system for implementing the emergency monitoring method for the coordinated use of slope radar and rock-rolling radar described in the embodiment. As the system disclosed in the embodiment corresponds to the method disclosed in the embodiment, the description is relatively simple. For relevant parts, please refer to the description of the method. It will not be repeated here.

Claims

1. An emergency monitoring method that combines slope radar and rockfall radar, characterized in that, The method includes: A slope radar is used to continuously scan the slope to be monitored, acquiring complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, the six-degree-of-freedom pose correction parameters of the radar system in geographic space are calculated through feature matching between simulation and real radar amplitude images. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are transformed to the geographic coordinate system, generating a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients. The rolling stone radar is used to detect rolling stone targets on the slope and obtain the trajectory parameters of the rolling stones. Based on the observation geometric parameters of the rolling stone radar, the detected rolling stone targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, movement speed and normalized kinetic energy index. The normalized kinetic energy index The calculation formula is: ； in, The radar cross section is used to characterize the size of the rolling stone. Let the velocity vector of the rolling stone be... Lithology-band calibration coefficient; The deformation point cloud and dynamic trajectory point cloud are unified to the same geospatial reference; the monitoring area is divided into a three-dimensional voxel grid under the geographic coordinate system; the deformation rate attribute in the deformation point cloud is mapped to the corresponding voxel, and the average deformation rate of each voxel is calculated; based on the rolling stone trajectory in the dynamic trajectory point cloud, the voxel where the collapse source of the rolling stone trajectory is located is determined by kinematic back-inference, and the corresponding normalized kinetic energy index is accumulated to the voxel to obtain the cumulative collapse energy of each voxel; based on the average deformation rate and cumulative collapse energy of each voxel, the deformation-collapse coupling early warning index of the voxel is calculated. Set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; based on the comparison results, issue corresponding levels of early warning information to voxel areas that meet different combination conditions. Based on the collapse source inverted from the dynamic trajectory point cloud, the data blind zone of the deformation point cloud is identified and marked; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, the coherence filtering threshold of the area in the slope radar data processing is adaptively reduced to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

2. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, Feature matching between simulation and real radar amplitude images, specifically including: Simulated amplitude images are generated using digital elevation model (DEM) data. For each polygon in the DEM, the simulated amplitude value is... The calculation formula is: ； in, For the element index of the digital elevation model, The local incident angle between the radar beam and the surface element normal. The slant range from the surface element to the radar is given. It is a composite constant that includes transmit power, antenna gain, and average surface reflectivity. A binary function to characterize whether the surface element is occluded by the terrain; Feature matching between simulated amplitude images and real radar amplitude images is achieved by maximizing mutual information, i.e., solving for the optimal radar pose correction parameters. This makes the true radar amplitude image The simulated amplitude image after transformation by the pose correction parameters The optimization objective is to maximize the mutual information between them, and the optimization objective is expressed as: ； in, For pose correction parameters that include 3D translation and 2D rotation, This represents a geometric transformation operation performed on a simulated image based on pose correction parameters. This is a mutual information function.

3. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, The detected rolling stone targets are converted to a geographic coordinate system, specifically based on the rolling stone radar's own location. Based on the measured distance, azimuth, and elevation angles, the three-dimensional spatial coordinates of the rolling stone point are calculated using the following coordinate transformation formula. : ； in, The slant range from the rolling stone target to the radar. It is the azimuth angle. The pitch angle.

4. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, The voxels at the collapse origin of the rolling stone trajectory are determined by kinematic back-calculation, specifically including: Extract the starting observation point of each rockfall trajectory, and search upwards for the nearest slope point using digital elevation model data to identify the collapse source.

5. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, The deformation-collapse coupling early warning index The calculation formula is: ； in, This corresponds to the average deformation rate within the voxel. This corresponds to the cumulative collapse energy within the voxel. and These are the historical maximum deformation rate and the historical maximum rockfall energy, respectively. and These are the weighting coefficients.

6. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, The deformation rate threshold is dynamically adjusted based on real-time rainfall data. Specifically, the adjusted deformation rate threshold is calculated using an exponential decay model with a lower bound constraint. The calculation formula is as follows: ； in, The initial deformation rate threshold, As the minimum safety threshold, For effective rainfall, This is the lithology sensitivity coefficient. It is a natural constant.

7. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, Adaptively lowering the coherence filtering threshold, specifically including: When the cumulative collapse energy corresponding to a certain voxel is greater than zero and continues to increase, and the coherence coefficient of the region corresponding to the voxel is less than 0.3, the coherence filtering threshold of the voxel in the slope radar data processing is automatically reduced from the normal value to 0.1 to force the extraction of deformation parameters under low signal-to-noise ratio conditions, and the extraction results are updated to the deformation point cloud.

8. The emergency monitoring method for the coordinated use of slope radar and rockfall radar according to claim 1, characterized in that, Issue corresponding levels of early warning information for voxel regions that meet different combinations of conditions, specifically including: A blue alert is issued when the average deformation rate of a voxel is greater than the adjusted deformation rate threshold and the cumulative collapse energy of the voxel is less than or equal to the rockfall energy threshold. A yellow alert is issued when the average deformation rate of a voxel is less than or equal to the adjusted deformation rate threshold and the cumulative collapse energy of the voxel is greater than the rockfall energy threshold. A red alert is issued when the average deformation rate of a voxel exceeds the adjusted deformation rate threshold and the cumulative collapse energy of the voxel exceeds the rockfall energy threshold.

9. An emergency monitoring system that combines slope radar and rockfall radar, characterized in that, The system is used to implement the emergency monitoring method for coordinated use of slope radar and rockfall radar as described in any one of claims 1 to 8, the system comprising: The deformation point cloud construction module is used to continuously scan the slope to be monitored using slope radar to acquire complex image data. Interferometric processing and phase calculation are performed on the complex image data to obtain the deformation parameters and coherence coefficients of each pixel in the radar coordinate system. The deformation parameters include cumulative deformation value, deformation rate, and deformation acceleration. Using digital elevation model data, feature matching between simulation and real radar amplitude images is used to calculate the six-degree-of-freedom pose correction parameters of the radar system in geographic space. Based on the pose correction parameters, the deformation parameters and coherence coefficients of each pixel are converted to the geographic coordinate system to generate a deformation point cloud containing three-dimensional spatial coordinates, deformation rate, deformation acceleration, and coherence coefficients. The trajectory point cloud construction module is used to detect rolling rocks on slopes using rolling rock radar and obtain the motion trajectory parameters of the rolling rocks. Based on the observation geometric parameters of the rolling rock radar, the detected rolling rock targets are transformed into a geographic coordinate system to generate a dynamic trajectory point cloud containing timestamps, three-dimensional spatial coordinates, motion speed and normalized kinetic energy index. The heterogeneous data fusion module is used to unify the deformation point cloud and dynamic trajectory point cloud to the same geospatial reference; divide the monitoring area into a three-dimensional voxel grid under the geographic coordinate system; map the deformation rate attribute in the deformation point cloud to the corresponding voxel, and calculate the average deformation rate of each voxel; based on the rolling stone trajectory in the dynamic trajectory point cloud, determine the voxel where the collapse source of the rolling stone trajectory is located through kinematic back-inference, and accumulate the corresponding normalized kinetic energy index to the voxel to obtain the cumulative collapse energy of each voxel; and calculate the deformation-collapse coupling early warning index of each voxel based on the average deformation rate and cumulative collapse energy of each voxel. The collaborative early warning module is used to set an initial deformation rate threshold and a rockfall energy threshold; dynamically adjust the initial deformation rate threshold based on real-time rainfall data; compare the average deformation rate of each voxel with the adjusted deformation rate threshold, and compare the cumulative collapse energy with the rockfall energy threshold; and issue corresponding levels of early warning information to voxel areas that meet different combination conditions based on the comparison results. The monitoring feedback optimization module is used to identify and mark the data blind spots of the deformation point cloud based on the collapse source inverted from the dynamic trajectory point cloud; based on the high-frequency rockfall area indicated by the dynamic trajectory point cloud, it adaptively reduces the coherence filtering threshold of the area in the slope radar data processing to extract deformation parameters under low signal-to-noise ratio conditions and update the deformation point cloud.

Images