A slope deformation identification and early warning method and system based on vegetation deformation parameters

Abstract

The present application relates to geological disaster early warning and monitoring technical field, specifically to a kind of slope deformation identification early warning method and system based on vegetation deformation parameter, comprising: the vegetation image and topographic data of collection slope area, extract vegetation coverage, tree bending parameter and environmental parameter;Extract bending feature, calculate bending mutation rate, if it is not less than adaptive threshold, then determine as disaster bending;Extract disaster bending feature, group disaster bending feature by clustering algorithm, calculate slope disaster period;For the tree determined as disaster bending, calculate the maximum thrust of single period;According to vegetation coverage, bending parameter, environmental parameter and slope disaster period, through the slope disaster evaluation model constructed, analyze slope deformation risk grade, carry out early warning by grade.This scheme assesses slope deformation risk by quantifying vegetation change, solves the problem of missing detection and missing exploration of disaster slope and disaster period analysis, and improves the reliability of disaster identification and early warning.

Description

Technical Field

[0001] This invention relates to the field of geological disaster early warning and monitoring technology, specifically to a slope deformation identification and early warning method and system based on vegetation deformation parameters. Background Technology

[0002] Currently, slope disaster identification, monitoring, and early warning rely heavily on manual inspections, sensor networks, or non-invasive methods. Manual inspections suffer from low efficiency, limited field of view, and insufficient accuracy in identifying potential deformations. Fixed sensors are deployed only for individual slopes with poor stability, rarely providing full coverage monitoring of strip-shaped slopes, resulting in blind spots in human monitoring. Non-invasive methods (drones and remote sensing technology) offer low cost per monitoring session and can monitor scattered slopes, but suffer from poor timeliness and limited accuracy.

[0003] Studies have shown a significant correlation between slope vegetation cover, tree bending parameters, and slope disaster characteristics (soil pushing → vegetation bending). Vegetation, as a natural attachment to the slope surface, serves as a "natural sensor" for slope disasters. Using it as an object for identification and monitoring offers numerous advantages, including being economical, convenient, requiring no drilling or specialized monitoring equipment, and providing real-time monitoring.

[0004] However, existing slope disaster identification and monitoring technologies do not yet utilize characteristics such as "disordered", "bent", and "fallen" slope vegetation as indicators for disaster identification, early warning, or monitoring, and lack quantitative analysis of the relationship between vegetation coverage, tree morphology parameters, and slope disasters.

[0005] Therefore, there is an urgent need for a slope deformation identification and early warning method and system based on vegetation deformation parameters. By quantifying vegetation changes, the risk of slope deformation can be assessed, the problems of missed detection and investigation of disaster-prone slopes and analysis of disaster phases can be solved, and the reliability of disaster identification and early warning can be improved. Summary of the Invention

[0006] One of the objectives of this invention is to provide a slope deformation identification and early warning method based on vegetation deformation parameters. By quantifying vegetation changes, the method assesses slope deformation risks, solves the problems of missed detection and investigation of disaster-prone slopes and analysis of disaster phases, and improves the reliability of disaster identification and early warning.

[0007] The basic solution provided by this invention is a slope deformation identification and early warning method based on vegetation deformation parameters, which includes the following: Data acquisition steps: Collect vegetation images and topographic data of the slope area and perform preprocessing; Parameter extraction steps: Extract vegetation coverage, tree bending parameters, and environmental parameters from the preprocessed vegetation images and terrain data; Catastrophic bending detection steps: Extract bending features and calculate bending abrupt change rate. Determine whether the bending abrupt change rate is less than an adaptive threshold. If yes, it is determined to be a natural bend; otherwise, it is determined to be a catastrophic bend. Steps for determining the catastrophic period: Extract catastrophic curve features from the catastrophic curve, group the catastrophic curve features using a clustering algorithm, with each group representing an independent slope catastrophic event, and calculate the slope catastrophic period. ; Disaster identification steps: Based on vegetation coverage, bending parameters, environmental parameters, and slope disaster phases, the slope deformation risk level is analyzed using the constructed slope disaster evaluation model. Risk warning steps: Based on the slope deformation risk level, conduct graded warnings.

[0008] Beneficial effects: This solution involves multiple data collection, including vegetation images and topographic data, enabling comprehensive data collection of slope areas and avoiding omissions in investigation and survey. Combining vegetation images and terrain data allows for the analysis of various parameters to quantify vegetation changes, including vegetation cover, tree curvature parameters, and environmental parameters. Curvature features are then extracted from these parameters, and the curvature mutation rate is calculated to determine whether the current tree curvature is natural or catastrophic, thus reducing the impact of natural changes on identification accuracy and making the results more accurate. For catastrophic slope bending, multi-parameter fusion analysis is used, combining vegetation coverage, tree bending parameters, and environmental parameters to analyze the phases of slope catastrophic events. Furthermore, all parameters are input into the constructed slope catastrophic evaluation model to analyze the slope deformation risk level for graded early warning. This reflects the possibility of slope catastrophic events from multiple perspectives, effectively improving the reliability of disaster identification and early warning.

[0009] In summary, this solution quantifies vegetation changes, assesses slope deformation risks, addresses the issues of missed detection and investigation of disaster-prone slopes and the analysis of disaster phases, and improves the reliability of disaster identification and early warning.

[0010] The second objective of this invention is to provide a slope deformation identification and early warning system based on vegetation deformation parameters. By quantifying vegetation changes, the system assesses slope deformation risks, solves the problems of missed detection and investigation of disaster-prone slopes and analysis of disaster phases, and improves the reliability of disaster identification and early warning.

[0011] This invention provides a second basic solution: a slope deformation identification and early warning system based on vegetation deformation parameters, used to execute the above-mentioned slope deformation identification and early warning method based on vegetation deformation parameters, comprising: The data acquisition module is used to collect vegetation images and topographic data of the slope area and perform preprocessing. The data analysis module is used to extract vegetation coverage, tree bending parameters, and environmental parameters from preprocessed vegetation images and terrain data. It is also used to extract bending features and calculate bending abrupt change rate. Determine whether the bending abrupt change rate is less than an adaptive threshold. If yes, it is determined to be a natural bend; otherwise, it is determined to be a catastrophic bend. It is also used to analyze catastrophic slope bends, extract catastrophic slope bend features, and group these features using a clustering algorithm. Each group represents an independent slope catastrophic event, and the number of slope catastrophic events is calculated. ; The identification and early warning module is used to analyze the slope deformation risk level based on vegetation coverage, bending parameters, environmental parameters and slope disaster period through the constructed slope disaster evaluation model; It is also used for graded early warning based on the risk level of slope deformation. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating an embodiment of the slope deformation identification and early warning method based on vegetation deformation parameters of the present invention. Detailed Implementation

[0013] The following detailed description illustrates the specific implementation method: Example 1 This embodiment is basically as shown in the appendix. Figure 1 As shown, a slope deformation identification and early warning method based on vegetation deformation parameters is provided, including the following: Data acquisition steps: Collect vegetation images and topographic data of the slope area and perform preprocessing; Specifically, a drone equipped with a multispectral camera (resolution ≥ 4K) and a lidar (scanning frequency ≥ 20Hz) is used to acquire vegetation images (multispectral images) and terrain data (point clouds, specifically LiDAR point clouds); and a ground control station is used to plan the flight path and synchronize the timestamp and geographic coordinates. The flight parameters of the UAV include: flight altitude, data acquisition frequency, and geographic coordinate synchronization. The specific settings in this embodiment are as follows: Flight altitude: The drone's flight altitude is adjusted according to the slope of the slope area. When the slope is less than 30°, the flight altitude is 200m; when the slope is between 30° and 60°, the flight altitude is reduced to 150m; and when the slope is greater than 60°, the flight altitude is reduced to 100m to ensure clear acquisition of tree trunk outlines (resolution ≥ 0.1m / pixel) and an overlap rate ≥ 80%. Sampling frequency: Multispectral images: collected monthly during the rainy season and quarterly during the non-rainy season to avoid image noise caused by rain obstruction; Point cloud: Collected once a month, with a point cloud density of ≥8 points / m². 2In key areas such as the toe and crest of the slope, the density of data points should be increased to ≥12 points / m. 2 To ensure the accuracy of terrain elevation data (error <0.2m); Geographic coordinate synchronization: The ground control station uses GPS+BeiDou dual-mode positioning with a timestamp synchronization accuracy of ≤1ms to ensure the spatial coordinate consistency of multi-source data.

[0014] Preprocessing of vegetation images and terrain data includes: The vegetation image is preprocessed as follows: the multispectral image is first radiometrically corrected to eliminate illumination differences, then geometrically corrected to eliminate terrain distortion, and then grayscale and Gaussian filtering are performed to remove noise, so as to ensure the accuracy of subsequent tree outline extraction; in this embodiment, the trees include: trees and shrubs. Point cloud preprocessing: Noise points are removed from the point cloud using statistical filtering, and then a three-dimensional terrain model is generated through an irregular triangular mesh for calculating slope height and slope.

[0015] Parameter extraction steps: Extract vegetation coverage, tree bending parameters, and environmental parameters from the preprocessed vegetation images and terrain data; The bending parameters of trees include: bending radius, bending rate, and bending percentage; The environmental parameters include: rock slope markings and slope height; The specific process is as follows: For the preprocessed vegetation image, identify the curved region, extract the outline of the tree in the curved region, fit and obtain the curvature radius and curvature direction angle, and calculate the curvature rate based on the curvature radius; calculate the curvature ratio based on the curved vegetation added in the curved region within a preset time period. Specifically, it includes: For the preprocessed vegetation image, ROI is detected by YOLOv5-OB and marked as curved region. If the ROI in the vegetation image collected at different time points changes, the catastrophic curvature judgment is triggered, that is, the catastrophic curvature judgment step is executed. For curved regions, the tree outlines are extracted through edge detection; the outline is the outline of each tree, and the outline of each tree needs to be calculated independently in the future, and then the curvature ratio P of the curved trees is calculated. The contours are simplified using the Douglas-Puk algorithm; For the simplified contour, the least squares method is used to fit the arc to obtain the arc radius. The radius of curvature is taken as the bending radius; and if the goodness of fit is less than the preset goodness of fit threshold, it is determined to be a regular bend, and the radius corresponding to the maximum local curvature of the arc is taken as the bending radius. ; Calculate the curvature based on the bending radius. : ; The Hough transform is used to detect local extrema of the profile and to fit the bending direction angle. : ; in This refers to the coordinate difference of local extrema detected by the Hough transform; specifically, local extrema are the points where the curvature of the arc in the contour is at its maximum (bending inflection points), which are obtained by detecting abrupt changes on the contour line using the Hough transform, and the coordinates of adjacent extrema are extracted. and Calculate the difference , Then calculate the bending direction angle; The percentage of bends is calculated based on the increase in bend vegetation within a preset time period in the bend area. : in The percentage of trees that have bent in the past N days (%). This represents the number of newly added bent trees (in trees) in the past N days. The total number of trees (in trees); specifically, by the change in curvature. The system determines whether a tree is newly bent if the change in its curvature within a preset time period exceeds a preset curvature change threshold. In this embodiment, the preset curvature change threshold is 0.1m. -1 And within a preset time period, that is, within a preset time period, the change in curvature is within a preset time period, such as This avoids interference with natural growth.

[0016] For the preprocessed vegetation images, the soil and rock types (soil / rock) are identified through spectral analysis, and rock slopes are identified by combining point clouds, generating rock slope identifiers and calculating slope height and vegetation coverage. Specifically, For the preprocessed vegetation images, the soil and rock types are classified and identified through spectral analysis. In other embodiments, ground-penetrating radar can also be set up to classify and identify soil and rock types through geodetic radar data. Based on the soil and rock type and combined with point cloud data, rock slopes are identified and rock slope markers are generated. Specifically, if the spectral characteristics show high reflectivity and the point cloud reveals dense fissures, it is marked as 1 (rock slope); if the spectral characteristics match the soil absorption band and the terrain is smooth, it is marked as 0 (non-rock slope, i.e., soil slope). Calculate slope height using point cloud data. ; The vegetation coverage rate is calculated using an image segmentation algorithm on the preprocessed vegetation image. In other embodiments, vegetation types can be distinguished and coverage density can be calculated based on morphological algorithms. The vegetation coverage rate refers to the coverage rate of trees (trees and shrubs), excluding herbaceous plants. Vegetation includes herbaceous plants, trees, and shrubs, but the bending parameter is only for trees (trees / shrubs). Herbaceous plants are easily affected by the season and have poor stability, so they are not included in the main monitoring indicators. The vegetation coverage rate is equal to the area of ​​trees in the slope area divided by the area of ​​the slope area.

[0017] Catastrophic bending detection steps: Extract bending features and calculate bending abrupt change rate. Determine whether the bending abrupt change rate is less than an adaptive threshold. If yes, it is determined to be a natural bend; otherwise, it is determined to be a catastrophic bend. The specific process is as follows: Extract bending features and calculate bending abrupt change rate. : ; in The bending mutation rate (cm / day); Let be the current bending characteristic value (cm) of the i-th tree. is the historical bending characteristic value (cm) of the i-th tree; The time interval (in days) between the current moment and a historical moment. in The mean of the bending feature values ​​collected from the last K collections is taken. If the number of collections is less than K, the bending feature value collected from the first collection is taken. In this embodiment, K=3. The bending features are divided into geometric features, statistical features and environmental features. The geometric features are bending parameters, the statistical features are the bending ratio, and the bending mutation rate can also be used as a statistical feature after calculation. The environmental features are environmental parameters. One or more bending features can be selected according to the needs. In this embodiment, the bending ratio value is selected as the bending feature value.

[0018] Bending characteristics are time-series characteristics and require continuous monitoring (e.g., once a month) to compare with historical data and identify abrupt changes. Based on the actual time difference of acquisition, if the interval between two acquisitions of bending feature values ​​exceeds the preset interval, interpolation calculation (linear interpolation) is added for the intermediate period to avoid distortion of the mutation rate caused by excessively long intervals; in this embodiment, the preset interval is 3 months.

[0019] Determine if the bending abrupt change rate is less than an adaptive threshold. If yes, it is determined to be natural bending; otherwise, it is determined to be catastrophic bending. Vegetation bending is affected by the driving mechanism and evolution rate. Natural bending is characterized by a large bending radius and a small bending rate. Its bending is a gradual and reversible environmental adaptation process. Catastrophic bending, on the other hand, is characterized by a small bending radius and a large bending rate. It is a sudden and catastrophic mechanical response to soil deformation. Therefore, it is necessary to distinguish between natural bending and catastrophic bending. If the bending mutation rate is less than the adaptive threshold, it is determined to be natural bending. If the bending mutation rate is greater than the adaptive threshold, it is determined to be catastrophic bending caused by soil deformation.

[0020] Adaptive threshold Dynamically adjusted using the sliding window standard deviation method: in The threshold benchmark value is the arithmetic mean of the bending mutation rate of all historical periods, reflecting the average mutation rate of tree bending in the natural evolution stage of the bending area, and serving as the benchmark rate level for judging natural bending. Standard deviation; This is an adjustment coefficient that balances threshold sensitivity and stability, used for coordination. (Reflecting the laws of natural evolution) and To reflect the relationship between the amplitude of natural fluctuations and the disaster, and to prevent misjudgment or omission of disasters, 1.5 is used in this embodiment.

[0021] In this embodiment, the adaptive threshold Calculations, including: Sliding window settings: The sliding window size is 6 months (including 6 LiDAR point cloud data acquisitions). The window slides forward one unit for each new data acquisition, dynamically updating. and ; Adjustment coefficient Adjustments are made based on slope type: k=1.2 for rock slopes (reducing sensitivity and minimizing false alarms), k=1.8 for soil slopes (increasing sensitivity and preventing false negatives), and k=1.5 for mixed slopes (default value). Mixed slopes are treated as non-rock slopes (…). If the slope is a mixture of soil and rock and there is no obvious dominant type, it will be treated as a soil slope by default. This is because soil slopes deform more gradually and the warning threshold needs to be more sensitive (k=1.5). The bending mutation rate in this embodiment Calculations, including: Initialize adaptive threshold: Calculate the average monthly rate of change using historical data from the past 3 years. and standard deviation Initialize adaptive threshold If the historical data is less than 3 years, an adaptive threshold based on the adjacent similar slopes is used for initialization. Dynamic adjustment: Smoothing coefficient of EWMA model (Exponentially Weighted Moving Average model) =0.3 (Assigning 30% weight to recent data and 70% weight to historical data), the threshold is updated monthly. If there are no warnings for three consecutive months, Adjust to 0.2 (reduce the weight of recent data); if one warning occurs, Adjusted to 0.4 (increasing the weight of recent data); Critical threshold: Set 15% as the critical threshold for vegetation coverage. When the vegetation coverage is below this value, the system will automatically switch to the mode dominated by InSAR point cloud and stop extracting invalid parameters.

[0022] Adaptive thresholds can better adapt to different environments; Steps for determining the catastrophic period: Extract catastrophic curve features from the catastrophic curve, group the catastrophic curve features using a clustering algorithm, with each group representing an independent slope catastrophic event, and calculate the slope catastrophic period. ; The catastrophic bending characteristics include: total bending strength. Consistency of bending direction (through) Calculations, bending radius distribution, bending abrupt change rate Time interval ( <1 year is considered the same event); bending characteristics include the bending characteristics of all trees (bending data includes natural bending), and catastrophic bending characteristics, which only filter the bending characteristics of trees that have undergone catastrophic bending to highlight sudden deformation; and new parameters have been added to the catastrophic bending characteristics, such as total bending intensity and consistency of bending direction; catastrophic bending characteristics are time-series characteristics, and need to be combined with data from multiple periods to determine whether they belong to the same catastrophic event (e.g., Δt < 1 year is considered the same event). The catastrophic slope characteristics are grouped according to spatial and temporal distribution using a clustering algorithm, with each group representing an independent slope catastrophic event; The clustering algorithm clusters trees based on the higher bending intensity (the difference between the curvature or bending ratio of two bends is within a preset range), and determines them to be products of the same catastrophic event. In addition to the number of catastrophic bends, the clustering results also need to be optimized by combining parameters such as vegetation type (herbaceous, shrub, tree) and landslide type (shallow, deep), consistency of bending direction, and time interval. Quantifying the scale of a disaster event: The larger the value, the more severe the bending of the trees involved in the catastrophic event, and the larger the scale.

[0023] Calculate the number of slope disaster periods : ; in This refers to the phases of slope disaster; Total bending strength (m) -1 ·Strain -1 ), representing the total number of bent trees in a single disaster event, used to quantify the scale of the disaster event; The spatial neighborhood radius (m) of the clustering algorithm (DBSCAN) needs to be calibrated using point cloud density (default). =1.5m); For clustering algorithms, this is a parameter representing the minimum number of samples required for the core objects; The weighting coefficient for time continuity; The weighting coefficients represent the consistency of direction.

[0024] Total bending strength , which quantifies the overall bending degree of all bent trees in a single catastrophic event, is one of the core parameters of the DBSCAN clustering algorithm, and its calculation formula is as follows: in For the first The bending radius (m) of a tree.

[0025] in and The optimal weight allocation is achieved using a fusion of information entropy and principal component analysis (PCA), addressing the issues of subjective weighting or the limitations of single-method approaches in traditional methods. The specific process is as follows: Perform parameter standardization: Z-score standardization affects time continuity ( ), directional consistency ( The parameters are standardized to eliminate dimensional differences, and the calculation formula is as follows: in For the first The sample in the first For each variable, the value after Z-score standardization (also known as "standard score" or "Z score"); where the sample refers to a single monitoring unit (such as a slope area covered by a single drone aerial photograph) or time-series monitoring data of a single tree; the variable refers to the monitoring parameters that need to be standardized, including: vegetation coverage, tree curvature parameters, and environmental parameters, i.e., curvature. Bending direction angle Bending ratio Vegetation coverage Slope height , ; In the original dataset, the first The sample at the th Observations on each variable; For the first The mean of all samples (or population) of a variable reflects the central tendency of that variable; For the first A measure of the dispersion of a variable across all samples (or the population), reflecting the magnitude of the fluctuation in the value of that variable.

[0026] PCA dimensionality reduction and principal component extraction were performed. PCA dimensionality reduction and principal component extraction are employed to eliminate redundant correlations in temporal continuity and directional consistency, thereby reducing the dimensionality complexity of subsequent information entropy calculations. Specifically, this includes: Calculate the covariance matrix: analyze the correlation between parameters; Solving for eigenvalues ​​and eigenvectors: Extracting principal components (PC1, PC2, ...); Principal components are selected: Principal components with a cumulative variance contribution rate ≥ a preset percentage (such as PC1 and PC2) are retained. In this embodiment, the preset percentage is 85%. Principal component weights are assigned using the information entropy method: Calculate information entropy : ; in For the first The information entropy of the principal component reflects the information entropy of the first principal component. The degree of confusion or discriminative power of the information carried by each principal component; For the first The first sample The probability distribution of each principal component is used to normalize the principal component scores so that the sum of the probabilities of all samples under the same principal component is 1. ; Calculate normalized weights : ; in For the first The normalized weights of each principal component are used in multi-principal component analysis to represent the proportion of contribution of that principal component to the overall result. The larger the weight, the stronger the influence of the principal component on the analysis result. This refers to the total number of principal components involved in the analysis, i.e., the number of principal components extracted from the original data. For the first The information entropy of the principal component reflects the information entropy of the first principal component. The degree of confusion or distinguishability of information carried by each principal component.

[0027] Map the principal component weights back to the original parameters to obtain the optimal weights: The global weight allocation of PCA is decomposed into the original parameters to ensure that the weights with temporal continuity and directional consistency are consistent with the global optimal solution. Specifically, this includes: Calculate the principal component loadings: ; ; in The first principal component obtained after PCA dimensionality reduction represents the main direction of variation (the direction with the largest variance) of the original data. The second principal component obtained after dimensionality reduction by PCA represents the main direction of variation (the direction with the largest variance) of the original data. for The loading factor in the time dimension reflects the relationship between time and... The degree of relevance; for The load factor in the mid-direction dimension reflects the direction relative to the load factor in the mid-direction dimension. The degree of relevance; for The loading coefficient in the time dimension reflects the degree of correlation between time and PC2; for The load factor in the mid-direction dimension reflects the degree of correlation between the direction and PC2.

[0028] Perform reverse weighting to obtain and Obtain the optimal weight value: ; ; in The first principal component The weights; The second principal component The weight.

[0029] The weights are adaptively adjusted based on geological conditions to adapt to complex slope scenarios.

[0030] Maximum thrust calculation steps: For trees identified as catastrophically bent, calculate the maximum thrust in a single period to quantify the mechanical effects on vegetation during soil deformation; the maximum thrust in a single period is not directly used in the slope catastrophic assessment model, its purpose is independent, and its core purpose is to back-calculate soil shear strength parameters. , It serves engineering design (such as anti-slide pile design); it quantifies the intensity of a single disaster and assists in assessing the disaster level (such as a severe disaster if the thrust exceeds the preset thrust threshold); its relationship with the slope disaster assessment model is that the slope disaster assessment model focuses on macro risks (based on coverage, bending ratio, etc.), while thrust estimation focuses on mechanical mechanisms. The two are complementary but do not overlap.

[0031] Specifically, it includes: For trees identified as catastrophically bent, calculate the soil load that caused the bending. : ; ; Where M is the bending moment (Nm) experienced by the tree; The tree elastic modulus (MPa) was determined by establishing a three-dimensional database of tree species-diameter at breast height-E value through random field sampling and laboratory calibration. I is the moment of inertia of the cross section (m) 4 ), which is related to the shape of the tree's cross-sectional area; The curvature of the tree is obtained through image recognition; in this embodiment... That is, curvature They have the same physical meaning, both reflecting the degree of bending; If the load on the tree is distributed in a triangular pattern, that is... For rock slopes or relatively hard soil, when a triangular load is applied, the maximum thrust is: If the load on a tree is distributed parabolically, that is... For a relatively soft soil slope, using a parabolic load, the maximum thrust is: in The maximum bending moment (Nm) experienced by the tree. The length of the curved section is (m).

[0032] The determination of the elastic modulus of trees includes: Based on the basic E-value classification of tree species, they are categorized according to common slope trees: Hardwoods (such as Pinus tabuliformis and Pinus massoniana): The basic E value is 12-14 GPa. These trees have high wood density and stable mechanical properties. Softwoods (such as Amorpha fruticosa and Vitex negundo): The basic E value is 8-10 GPa. These trees have relatively loose wood and a large range of elastic deformation. Mixed type (no clearly dominant tree species): The basic E value is 10-11 GPa, and the average of the two types is taken and adjusted in combination with on-site sampling.

[0033] The correction factor for determining the baseline E value based on chest diameter: Diameter at breast height < 10cm: The wood of young trees is relatively soft, and the E value is lower than that of mature trees of the same species, with a correction factor of 0.8 to 0.9; For timber with a diameter at breast height (DBH) of 10-28 cm: wood properties are in their optimal range, with an E value close to the baseline value and a correction factor of 1.0–1.1. Diameter at breast height > 28cm: The wood of old trees tends to be denser but less tough, and the E value drops slightly, with a correction factor of 0.9-1.0.

[0034] Perform localized calibration: On-site sampling: 3 to 5 representative trees were randomly selected from each type of slope, and their diameter at breast height (DBH) was measured and core samples were collected. Laboratory testing: The E value of the sample was determined according to the physical and mechanical standards of wood to verify the matching degree between the baseline value and the correction coefficient; Data fitting: Establish a three-dimensional database of local tree species, diameter at breast height (DBH), and E-value, and dynamically update and correct the parameters.

[0035] The calculation of the moment of inertia of the cross section includes: If the tree trunk has a circular cross-section, the diameter... (Measured via image recognition, with an accuracy ≤ 0.1cm), then ; If the tree trunk cross-section is irregular in shape (such as ellipse), the cross-sectional contour is extracted using LiDAR point cloud, and the integral method is used to calculate... The integration step size is ≤0.001m.

[0036] Disaster identification steps: Based on vegetation coverage, bending parameters, environmental parameters, and slope disaster phases, the slope deformation risk level is analyzed using the constructed slope disaster evaluation model. The slope disaster assessment model is as follows: in This represents the slope disaster risk value. , , , , and The weighting coefficients are determined by fusing random forest and AHP subjective weights. The rate of change of vegetation cover; This refers to the slope height; For marking rock slopes; This represents the maximum curvature of the vegetation. The percentage of bending; This refers to the number of catastrophic landslides, i.e., the number of catastrophic slope events.

[0037] Risk warning steps: Based on the slope deformation risk level, conduct graded warnings.

[0038] Specifically, if the first risk threshold ≤ If the risk threshold is less than the second risk threshold, a yellow alert is issued; in this embodiment, the first risk threshold is 0.4 and the second risk threshold is 0.6. A yellow alert is issued via web-based notification. Control personnel must complete on-site verification within 72 hours and record information such as slope surface cracks and vegetation changes. If the third risk threshold ≤ If the risk threshold is less than the fourth risk threshold, an orange alert is issued; in this embodiment, the third risk threshold is 0.7 and the fourth risk threshold is 0.9. An orange alert triggers an audible and visual alarm, which is then sent to control personnel and local emergency departments. On-site verification is completed within 24 hours, and monitoring stakes (spaced 10m apart) are set up to enhance monitoring. If the fifth risk threshold ≤ If the risk threshold is 0.9, a red alert will be issued; in this embodiment, the fifth risk threshold is 0.9. A red alert is issued, and an emergency response is activated. The area surrounding the slope is closed, personnel are evacuated, and emergency departments are required to arrive at the scene within 4 hours. Drones are used to collect data intensively (once every 12 hours).

[0039] The weight coefficients are determined by fusing random forest and AHP subjective weights. In this embodiment, the random forest model parameters are set as follows: Number of decision trees: 150-200; maximum depth: 18 for sample size > 10,000, 15 for sample size 5,000-10,000, and 12 for sample size < 5,000. Feature importance calculation: The reduction in error of out-of-bag data is used, and the calculation is performed three times and the average is taken to ensure the stability of the results.

[0040] The AHP subjective weight fusion process includes: Expert scoring: Five experts in the field of geological disaster monitoring were invited to score the parameters ( The importance of each ) is scored on a scale of 1 to 9 to form a judgment matrix; Consistency check: Calculate the consistency ratio CR of the judgment matrix. If CR > 0.1, experts need to be invited to score again; if CR ≤ 0.1, the subjective weight is calculated using the eigenvector method. Weighted fusion: Objective weights (random forest) account for 70%, subjective weights (AHP) account for 30%, and the final weight coefficients satisfy the condition that the sum of the weights is 1.

[0041] This solution involves multi-data acquisition, including vegetation and topographic images. Combining these images allows for the analysis of various parameters, including vegetation cover, tree bending parameters, and environmental parameters. Further multi-parameter fusion analysis integrates vegetation cover, tree bending parameters, and environmental parameters, and analyzes the stages of slope disasters to further assess slope deformation risk. This comprehensive analysis reflects the likelihood of slope disasters and effectively improves the reliability of disaster identification and early warning. Furthermore, this solution can use vegetation bending to estimate the maximum load generated during soil deformation, i.e., the maximum thrust in a single period, to assist in analyzing the severity of the disaster. This solution adopts non-contact measurement, based on drones, machine vision, and slope vegetation as natural attachments, avoiding the invasiveness of traditional sensor installation, making monitoring more environmentally friendly, and the monitoring equipment is less affected by external factors, and collecting more comprehensive information. The bending features extracted in this scheme are time-series bending features. Time series analysis is used to distinguish between natural bending (caused by slow growth) and catastrophic bending (caused by sudden displacement).

[0042] In practical applications, lightweight edge computing can be performed, and TensorFlow Lite models can be deployed on drones to calculate curvature in real time (latency <50ms), effectively avoiding the drawbacks of manual inspection and passive monitoring, and enabling disaster identification and early warning.

[0043] Example 2 This embodiment provides a slope deformation identification and early warning system based on vegetation deformation parameters, used to execute the aforementioned slope deformation identification and early warning method based on vegetation deformation parameters, including: The data acquisition module is used to collect vegetation images and topographic data of the slope area and perform preprocessing. The data analysis module is used to extract vegetation coverage, tree bending parameters, and environmental parameters from preprocessed vegetation images and terrain data. It is also used to extract bending features and calculate bending abrupt change rate. Determine whether the bending abrupt change rate is less than an adaptive threshold. If yes, it is determined to be a natural bend; otherwise, it is determined to be a catastrophic bend. It is also used to analyze catastrophic slope bends, extract catastrophic slope bend features, and group these features using a clustering algorithm. Each group represents an independent slope catastrophic event, and the number of slope catastrophic events is calculated. ; It is also used to calculate the maximum thrust in a single period for trees that are determined to be catastrophically bent; The identification and early warning module is used to analyze the slope deformation risk level based on vegetation coverage, bending parameters, environmental parameters and slope disaster period through the constructed slope disaster evaluation model; It is also used for graded early warning based on the risk level of slope deformation.

[0044] In addition, a visualization module, a data storage module, and a power management module can be set up; The visualization module, based on a GIS visualization platform, dynamically renders and enhances heat maps in three dimensions, and provides multi-dimensional risk visualization analysis by combining disaster event playback.

[0045] The heat map is shown below. Mapped to a spatial distribution map; a time series animation of slope deformation is generated based on point cloud data, and multi-role permission management (such as administrator, monitor) is supported through a web-based GIS visualization platform. The data storage module stores multi-source data based on a cloud platform, supports SQL queries and version backtracking; all raw data, intermediate parameters, and model output results are stored in categories of "year-month-date", retaining at least 5 years of data; it supports backtracking monitoring data and early warning results for any time period by time point, and each adjustment of model parameters (such as weights and thresholds) requires recording the version number to facilitate error analysis and model optimization.

[0046] The power management system integrates solar panels (power ≥ 20W) and backup batteries (capacity ≥ 5000mAh) to provide power to other modules.

[0047] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A slope deformation identification and early warning method based on vegetation deformation parameters, characterized in that, Includes the following: Data acquisition steps: Collect vegetation images and topographic data of the slope area and perform preprocessing; Parameter extraction steps: Extract vegetation coverage, tree bending parameters, and environmental parameters from the preprocessed vegetation images and terrain data; Catastrophic bending determination steps: Extract bending features, calculate bending mutation rate, and determine whether the bending mutation rate is less than the adaptive threshold. If yes, it is determined to be natural bending; otherwise, it is determined to be catastrophic bending. Steps for determining the disaster period: Extract the disaster curve features for the disaster curve, group the disaster curve features into groups using a clustering algorithm, with each group representing an independent slope disaster event, and calculate the slope disaster period. Disaster identification steps: Based on vegetation coverage, bending parameters, environmental parameters, and the number of slope disasters, the slope deformation risk level is analyzed using the constructed slope disaster evaluation model.

2. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 1, characterized in that, The parameter extraction step includes: For the preprocessed vegetation image, the curved regions are identified, the outlines of the trees in the curved regions are extracted, the curvature radius and curvature direction angle are obtained by fitting, and the curvature rate is calculated based on the curvature radius; the curvature ratio is calculated based on the curved vegetation added in the curved regions within a preset time period. For the preprocessed vegetation images, the rock and soil types are identified through spectral analysis, and rock slopes are identified by combining topographic data, generating rock slope identifiers and calculating slope height and vegetation coverage.

3. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 2, characterized in that, The fitting process for obtaining the bending radius and bending direction angle includes: For the preprocessed vegetation images, curved regions were detected using YOLOv5-OB; For curved areas, the tree outlines are extracted using edge detection; For the contour, the least squares method is used to fit the arc to obtain the arc radius. The radius of curvature is taken as the bending radius; and if the goodness of fit is less than the preset goodness of fit threshold, it is determined to be a regular bend, and the radius corresponding to the maximum local curvature of the arc is taken as the bending radius. ; The Hough transform is used to detect local extrema of the profile and to fit the bending direction angle. : in The coordinate difference of the local extremum points detected by the Hough transform.

4. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 3, characterized in that, After extracting the tree contours through edge detection for the curved region, the method further includes: The contours are simplified using the Douglas-Puk algorithm; The simplified contour is fitted with a circular arc using the least squares method.

5. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 2, characterized in that, The bending mutation rate : ； in The bending mutation rate; Let be the current bending characteristic value of the i-th tree; Let be the historical bending characteristic value of the i-th tree; This represents the time interval between the current moment and a historical moment. in Take the average of the bending feature values ​​collected from the last K collections. If the number of collections is less than K, then take the bending feature value collected from the first collection. Based on the interval time.

6. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 1, characterized in that, The adaptive threshold It is dynamically adjusted using the sliding window standard deviation method: in This is the threshold baseline value; Standard deviation; This is an adjustment coefficient that balances threshold sensitivity and stability.

7. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 2, characterized in that, The steps for determining the disaster period include: The catastrophic slope characteristics are grouped according to spatial and temporal distribution using a clustering algorithm, with each group representing an independent slope catastrophic event; Calculate the number of slope disaster periods : ； in This refers to the phases of slope disaster; Total bending strength; The spatial neighborhood radius of the clustering algorithm; For clustering algorithms, this is a parameter representing the minimum number of samples required for the core objects; The weighting coefficient for time continuity; The weighting coefficient for directional consistency; Total bending strength : in For the first The bending radius of a tree.

8. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 1, characterized in that, The slope disaster assessment model is as follows: in This represents the slope disaster risk value. , , , , and These are the weighting coefficients; The rate of change of vegetation cover; This refers to the slope height; For marking rock slopes; This represents the maximum curvature of the vegetation. The percentage of bending; This refers to the period of slope disaster.

9. The slope deformation identification and early warning method based on vegetation deformation parameters according to claim 1, characterized in that, Also includes: Maximum thrust calculation steps: For trees identified as catastrophic bending, calculate the maximum thrust in a single period; Specifically, it includes: For trees identified as catastrophically bent, calculate the soil load that caused the bending. : ； ； Where M is the bending moment experienced by the tree; The elastic modulus of trees; I represents the moment of inertia of the cross section, which is related to the shape of the tree's cross-sectional area; The curvature of the tree; If the load on the tree is distributed in a triangular pattern, then the maximum thrust is: If the load on the tree is distributed parabolically, then the maximum thrust is: in The maximum bending moment experienced by the tree; This represents the length of the curved section.

10. A slope deformation identification and early warning system based on vegetation deformation parameters, characterized in that, A method for performing slope deformation identification and early warning based on vegetation deformation parameters as described in any one of claims 1-9, comprising: The data acquisition module is used to collect vegetation images and topographic data of the slope area and perform preprocessing. The data analysis module is used to extract vegetation coverage, tree bending parameters, and environmental parameters from preprocessed vegetation images and terrain data. It is also used to extract bending features, calculate the bending abrupt change rate, and determine whether the bending abrupt change rate is less than an adaptive threshold. If it is, it is determined to be a natural bend; if not, it is determined to be a catastrophic bend. It is also used to extract the features of catastrophic slope bending, and to group the features of catastrophic slope bending through a clustering algorithm. Each group represents an independent slope catastrophic event, and the number of slope catastrophic events is calculated. The identification and early warning module is used to analyze the slope deformation risk level based on vegetation coverage, bending parameters, environmental parameters and slope disaster period through the constructed slope disaster evaluation model; It is also used for graded early warning based on the risk level of slope deformation.

Images