Reservoir bank slope geological disaster risk identification method based on unmanned aerial vehicle laser radar

Abstract

The invention discloses a reservoir bank slope geological disaster risk identification method based on an unmanned aerial vehicle laser radar, and relates to the technical field of reservoir bank slope prevention and control, an unmanned aerial vehicle is used for carrying laser radar equipment, flight scanning is carried out on a reservoir bank slope area, and three-dimensional point cloud data is obtained. According to the reservoir bank slope identification method based on the unmanned aerial vehicle laser radar detection data, high-precision terrain and geological information of the reservoir bank slope is rapidly acquired in a non-contact manner by using the three-dimensional imaging capability of the unmanned aerial vehicle laser radar technology, so that the data acquisition efficiency and safety are greatly improved; the method also remarkably improves the precision and integrity of data, accurately captures key topographic factors such as topographic relief, gradient, slope direction and the like through laser radar scanning, and provides a solid foundation for the accurate recognition of bank slope types in combination with geological factors such as geological structures, lithology and the like, so that the prediction and evaluation of bank slope disasters are more accurate. Preventive measures can be taken in advance, and disaster risks are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir bank slope prevention and control, and in particular to a reservoir bank slope geological disaster risk identification method based on unmanned aerial vehicle laser radar. Background Art

[0002] As an important part of water conservancy and hydropower projects, the stability of reservoir slopes is directly related to the safe operation of the projects. With the rapid development of water conservancy and hydropower project construction, many water conservancy and hydropower projects have been built in many large river basins, forming a number of high dams and large reservoirs. These projects have been operating for a long time, and the impact of extreme precipitation and earthquakes has made the risk of reservoir slopes more prominent. Therefore, the safety of slopes has become the focus of attention in the water conservancy and hydropower industry. The identification methods of geological disaster risks on reservoir slopes using traditional or high-tech methods have many limitations to a certain extent, such as the difficulty for manual investigators to reach steep areas, and the inability of satellite / low-altitude remote sensing images to interpret dense vegetation obstruction areas. As an emerging low-altitude remote sensing detection technology, UAV laser radar technology has the advantages of high precision, high efficiency, and non-contact measurement, and shows great application potential in the identification of geological disaster risks on reservoir slopes. Therefore, the application of UAV laser radar technology to accurately and efficiently identify and monitor the geological disaster risks on reservoir slopes is of great significance to ensure the safe operation of the project.

[0003] In the identification of geological hazard risks on reservoir slopes, complex terrain and vegetation coverage factors such as steep slopes, high hidden dangers, and dense vegetation in the reservoir area will affect the accuracy of geological hazard risk identification, and thus interfere with the identification and classification of geological hazard risks on reservoir slopes. Therefore, how to weaken the interference of terrain and vegetation coverage factors on the identification and classification of reservoir slopes and improve the identification accuracy is the problem we need to solve. To this end, a reservoir slope geological hazard risk identification method based on UAV lidar is proposed. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: A reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar comprises the following steps:

[0005] Step 1: Use a drone equipped with a laser radar device to perform a flight scan of the reservoir slope area to obtain high-precision three-dimensional point cloud data and pre-process the original point cloud data;

[0006] Step 2: Extract terrain features from the preprocessed point cloud data and produce DSM, DEM, DOM and other products, derive information including slope, aspect, terrain roughness, elevation, etc., and use machine learning algorithms to classify point cloud data, distinguish vegetation from non-vegetation areas, and remove vegetation areas from point cloud data to reduce their interference with reservoir slope identification;

[0007] Step 3: Use clustering algorithm and edge detection algorithm to segment the bank slope area, extract the boundary of the bank slope area, and classify similar point cloud data into one category for subsequent identification and classification;

[0008] Step 4: Analyze the relationship between the spatial distribution of bank slope disasters and topographic and geological factors, and use machine learning algorithms to identify and classify bank slopes to determine areas prone to bank slope disasters;

[0009] Step 5: Verify the results of identification and classification, and match corresponding management measures based on the classification results of the slope.

[0010] Preferably, in step 1, the process of acquiring and preprocessing the three-dimensional point cloud data is as follows:

[0011] Plan the flight route of the target area according to the topography and area of ​​the reservoir slope area to ensure coverage of the entire target area, and set the flight altitude, speed and scanning density according to the performance of the LiDAR equipment and the characteristics of the target area;

[0012] The laser radar device is mounted on a drone, and the drone is operated to fly along a predetermined flight route for flight scanning. At the same time, the laser radar device starts to work, emitting laser pulses and receiving reflected signals, recording the three-dimensional coordinates of each point, thereby obtaining the original point cloud data of the target object surface;

[0013] After the flight mission is completed, the raw point cloud data from the UAV and LiDAR equipment is transmitted to the ground processing workstation;

[0014] The original point cloud data is preprocessed, including data decoding, POS data processing and point cloud data solution. The original scanning data of each flight is decoded, and the ground GNSS station and the UAV GNSS are precisely and dynamically post-processed to obtain the UAV GNSS three-dimensional coordinate data. The UAV GNSS three-dimensional coordinate data, IMU data and eccentricity component value are combined to generate a track file, and the point cloud data is solved and produced through the track file, decoded data and system calibration data.

[0015] Preferably, in step 2, the process of extracting terrain features is:

[0016] Extract height, intensity, color and shape features from the preprocessed point cloud data, train the classification model using the support vector machine algorithm, and run the classification model to automatically classify the point cloud data to distinguish between ground points, vegetation points and other object points;

[0017] Extract ground points from the classified point cloud data, use the extracted ground points and other object points (including vegetation points) to generate a digital surface model of the reservoir slope area, remove vegetation and other non-ground features from the digital surface model, retain only the ground points, generate a digital elevation model, provide accurate elevation information for subsequent analysis, and smooth and correct the digital elevation model to improve the accuracy and usability of the model;

[0018] Using digital elevation model data, calculate the slope (slope size) and aspect (slope direction) of each point, and read the elevation value of each point directly from the digital elevation model;

[0019] The height value of each point is extracted as a preliminary feature to distinguish vegetation from non-vegetation, and the reflection intensity value of each point is extracted using the reflection intensity information recorded by the lidar equipment.

[0020] Preferably, in step 2, the process of removing vegetation areas in the point cloud data is:

[0021] Extract the color information of each point in the point cloud data and analyze the intensity of the green channel. Vegetation is usually rich in chlorophyll and reflects green light strongly. Calculate the local neighborhood features of the point cloud normal, curvature, and point cloud density. Simultaneously calculate the normalized difference vegetation index to distinguish the effective features of vegetation and non-vegetation.

[0022] Label the feature data extracted from the point cloud data as training samples, use the training samples as input features, train a vegetation classification model based on a decision tree algorithm, run the vegetation classification model, classify the point cloud data, and distinguish between vegetation and non-vegetation areas;

[0023] According to the classification results, vegetation points are separated and removed from the point cloud data, and only non-vegetation points are retained to reduce their interference with reservoir slope identification.

[0024] Preferably, in step 3, the classification process of similar point cloud data is:

[0025] Apply edge detection algorithm, input preprocessed point cloud data, calculate curvature and normal vector geometric features of point cloud, identify the geometric feature change area in point cloud, i.e. edge area, according to the change of curvature and normal vector, and output a set of edge points based on the change of curvature and normal vector;

[0026] By searching for adjacent edge points, the adjacent edge points are connected using a neighborhood search algorithm to form a complete boundary, and several complete boundary line segments are output;

[0027] Apply the KMeans clustering algorithm, input all point cloud data (including edge points and internal points), select random points to initialize the cluster center, assign points to the nearest cluster center according to the distance metric, update the cluster center to the mean of all points in the cluster, repeat the clustering process until the cluster center no longer changes, and output multiple clusters, each cluster contains a set of similar point cloud data;

[0028] The clustering results are post-processed to remove isolated clusters and merge similar clusters, and the optimized clustering results are output to obtain more accurate and complete slope area segmentation results;

[0029] According to the clustering results, the convex hull algorithm is used to calculate the boundary points of each cluster, and the boundary points between clusters are determined according to the spatial relationship between clusters. The boundary points are connected into a smooth curve using the least squares curve fitting algorithm, and then the boundary curve of the slope area is output;

[0030] After the slope area segmentation and boundary extraction are completed, the PointCNN deep learning model is used to input the point cloud data of each cluster, extract features through the convolution layer, use the fully connected layer for classification, and output the semantic label of each point. According to the semantic label of each point, the entire point cloud data is semantically segmented, and then the classification results of different object features (such as vegetation, soil, rock, etc.) are output.

[0031] Preferably, in step 4, the process of determining the area prone to bank slope disasters is as follows:

[0032] Based on the extracted slope area boundaries, collect slope disaster data under similar terrain and geological conditions, analyze the spatial distribution of slope disasters, and identify areas affected by terrain and geological factors;

[0033] Collect relevant data on topographic factors and geological factors in the slope area. Topographic factors include topographic relief, slope gradient and aspect, and hydrological data. Geological factors include geological structure, lithology, and strata.

[0034] According to the factors affecting bank slope hazards, image processing technology is used to extract terrain texture features, and geographic information system is used for spatial analysis to extract spatial features, including terrain undulation, slope, slope direction, geological structure, and lithology;

[0035] The extracted features are combined with the neural network model to train the bank slope classification model, and the model is optimized using the cross-validation method to improve the accuracy and generalization ability of the model. The trained bank slope classification model is then used to identify and classify the bank slope and output the classification results.

[0036] Based on the classification results, the bank slope is further analyzed and evaluated in combination with hydrological data, and the bank slope risk identification coefficient is calculated based on the comprehensive characteristic information to determine the areas prone to bank slope disasters, and then the types of bank slope disasters are analyzed.

[0037] Preferably, the calculation process of the bank slope risk identification coefficient is:

[0038] Traverse the terrain data, geological data and hydrological data, and calculate the terrain relief index T, geological structure risk index G and hydrological impact index H respectively;

[0039] For terrain data, collect elevation data of the bank slope area, calculate the slope of each point from the elevation data, and determine the maximum value Max in the slope data. ( S ) and minimum value Min ( S ) , calculate the average slope S avg , the terrain relief index T is obtained by comprehensive analysis, where the expression of the terrain relief index is:

[0040] For geological data, geological structure data are collected, fault zone density D and seismic activity index E are determined, and the geological structure risk index G is calculated, where the expression of the terrain relief index is:

[0041] For hydrological data, collect water level change and flow data, and determine the maximum value of water level change Max from the hydrological data. ( W ) and minimum value Min ( W ) , calculate the average water level W avg , the hydrological impact index H is obtained through comprehensive analysis, where the expression of the hydrological impact index is:

[0042] The weight coefficients of terrain undulation index, geological structure risk index and hydrological impact index were determined by historical data analysis, and the slope risk identification coefficient R was obtained through comprehensive analysis.

[0043] Preferably, the calculation formula of the bank slope risk identification coefficient is:

[0044]

[0045] Among them, R is the slope identification coefficient, T is the terrain undulation index, G is the geological structure risk index, and H is the hydrological impact index, which is calculated based on water level changes and flow rates. a is the weight coefficient of the terrain undulation index, b is the weight coefficient of the geological structure risk index, and c is the weight coefficient of the hydrological impact index. The value range of R is between 0 and 1.

[0046] Preferably, in step 5, the process of determining the risk level and matching the management measures is as follows:

[0047] Collect samples of known slope types and historical slope disaster data. The samples have clear disaster history and stability assessment results. Use the samples of known slope types and historical slope disaster data to verify the trained slope classification model and evaluate the accuracy and reliability of the model.

[0048] Compare the output of the slope classification model with the known slope types and analyze the accuracy of the model output results;

[0049] The classification results are analyzed in depth to clarify the stability of various types of slopes and the risk of disasters, and slope disasters are divided into different types, namely collapse geological disasters, landslide geological disasters, and crack geological disasters;

[0050] Combined with the slope risk identification coefficient and the classification type of slope disasters, corresponding management measures and emergency plans should be formulated for each identified type of slope disasters. For slopes with collapse geological disasters, monitoring and early warning should be strengthened, and engineering control measures should be taken to reduce disaster risks, such as reinforcing slopes and improving drainage systems. For slopes with landslide geological disasters, the monitoring frequency should be increased, a more detailed analysis of geological and hydrological characteristics should be conducted, and emergency plans should be formulated. For slopes with crack geological disasters, attention should be paid to them and regular monitoring and maintenance measures should be taken, such as regular geological exploration and hydrological monitoring.

[0051] The present invention provides a reservoir slope geological disaster risk identification method based on UAV laser radar. It has the following beneficial effects:

[0052] 1. This method of identifying geological disaster risks on reservoir slopes based on drone lidar uses the three-dimensional imaging capability of drone lidar technology to quickly and contactlessly obtain high-precision topographic and geological information on reservoir slopes, which not only greatly improves the efficiency and safety of data acquisition, but also significantly improves the accuracy and integrity of the data. Through lidar scanning, it accurately captures key terrain factors such as terrain undulation, slope, and slope direction. Combined with geological factors such as geological structure and lithology, it provides a solid foundation for the accurate identification of slope types, making the prediction and assessment of slope disasters more accurate, which helps to take preventive measures in advance and reduce disaster risks.

[0053] 2. The reservoir slope geological disaster risk identification method based on UAV lidar can quickly identify the changing trend of slope stability by analyzing the monitoring data collected by the drone, providing timely and accurate information support for disaster warning. In addition, lidar technology can also penetrate vegetation coverage and obtain real elevation information of the surface. It can carry out efficient and accurate disaster monitoring in complex terrain and densely vegetated areas. And by utilizing the rapid response and real-time monitoring characteristics of UAV lidar technology, it can carry out large-scale and high-frequency scanning and monitoring of reservoir slopes in a short period of time, which is crucial for timely detection of potential disaster signs such as slope deformation and landslides. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a method flow chart of the reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar of the present invention;

[0055] Figure 2 It is a classification flow chart of similar point cloud data of the present invention;

[0056] Figure 3 This is a flow chart for determining areas prone to bank slope disasters according to the present invention. DETAILED DESCRIPTION

[0057] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0058] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: a reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar, comprising the following steps:

[0059] Step 1: Use a drone equipped with a laser radar device to perform a flight scan of the reservoir bank slope area to obtain high-precision three-dimensional point cloud data, and pre-process the original point cloud data. According to the terrain and area of ​​the reservoir bank slope area, plan the flight route of the target area to ensure that the entire target area is covered. According to the performance of the laser radar equipment and the characteristics of the target area, set the flight altitude, speed and scanning density, carry the laser radar equipment on the drone, operate the drone to fly along the predetermined flight route, and perform flight scanning. At the same time, the laser radar equipment starts working, emitting laser pulses and receiving reflected signals, recording the three-dimensional coordinates of each point, and thus obtaining the surface of the target object. The original point cloud data of the surface. After the flight mission is completed, the original point cloud data in the UAV and LiDAR equipment are transmitted to the ground processing workstation, and the original point cloud data is pre-processed, including data screening, filtering, data synchronization and data solution. Among them, the acquired original point cloud data is checked to remove abnormal data caused by equipment failure, environmental factors, etc., and the point cloud data is smoothed by applying the filtering algorithm to remove noise and redundant information. The point cloud data is synchronized with the GNSS data of the UAV through data synchronization to ensure the accuracy of the spatial position of each point, and the original data is solved, including position solution and attitude solution, to correct the system error;

[0060] Step 2: Extract terrain features from the preprocessed point cloud data and produce DSM, DEM, DOM and other products, derive information including slope, aspect, terrain roughness, elevation, etc., and use machine learning algorithms to classify point cloud data, distinguish vegetation from non-vegetation areas, and remove vegetation areas from point cloud data to reduce their interference with reservoir bank slope identification. Extract height, intensity, color and shape features from the preprocessed point cloud data, use support vector machine algorithm to train the classification model, and run the classification model to automatically classify the point cloud data, distinguish ground points, vegetation points and other object points, extract ground points from the classified point cloud data, use the extracted ground points and other object points (including vegetation points) to generate a digital surface model of the reservoir bank slope area, remove vegetation and other non-ground features from the digital surface model, retain only ground points, generate a digital elevation model, provide accurate elevation information for subsequent analysis, and smooth and correct the digital elevation model to improve the accuracy and availability of the model. Using digital elevation model data, calculate the slope (slope size) and slope direction (slope direction) of each point, and directly read the elevation value of each point from the digital elevation model, extract the height value of each point as the preliminary feature to distinguish vegetation from non-vegetation, use the reflection intensity information recorded by the lidar equipment to extract the reflection intensity value of each point, extract the color information of each point in the point cloud data, analyze the intensity of the green channel, vegetation is usually rich in chlorophyll and reflects green light strongly, calculate the local neighborhood features of the point cloud normal, curvature, and point cloud density, and simultaneously calculate the normalized difference vegetation index to distinguish effective features from vegetation and non-vegetation, annotate the feature data extracted from the point cloud data as training samples, and use the training samples as input features to train a vegetation classification model based on the decision tree algorithm, run the vegetation classification model, classify the point cloud data, distinguish vegetation from non-vegetation areas, separate vegetation points according to the classification results, remove vegetation points from the point cloud data, and only retain non-vegetation points to reduce their interference with reservoir slope identification;

[0061] Step 3: Use clustering algorithm and edge detection algorithm to segment the bank slope area, extract the boundary of the bank slope area, classify similar point cloud data into one category for subsequent identification and classification, apply edge detection algorithm, input preprocessed point cloud data, calculate the curvature and normal vector geometric features of the point cloud, and identify the geometric feature change area in the point cloud, that is, the edge area, according to the change of curvature and normal vector, and output a set of edge points based on the change of curvature and normal vector. By searching for adjacent edge points, the adjacent edge points are connected using the neighborhood search algorithm to form a complete boundary, and several complete boundary line segments are output. The KMeans clustering algorithm is applied, all point cloud data (including edge points and internal points) are input, random points are selected to initialize the cluster center, points are assigned to the nearest cluster center according to the distance metric, and the cluster center is updated to the mean of all points in the cluster. The clustering process is repeated until the cluster center no longer changes. , and output multiple clusters, each cluster contains a group of similar point cloud data, and the clustering results are post-processed to remove isolated clusters and merge similar clusters, and the optimized clustering results are output to obtain more accurate and complete bank slope area segmentation results. According to the clustering results, for each cluster, the convex hull algorithm is used to calculate its boundary points, and the boundary points between clusters are determined according to the spatial relationship between clusters. The boundary points are connected into a smooth curve using the least squares curve fitting algorithm, and then the boundary curve of the bank slope area is output. After the bank slope area segmentation and boundary extraction are completed, the PointCNN deep learning model is used to input the point cloud data of each cluster, extract features through the convolution layer, use the fully connected layer for classification, and output the semantic label of each point. According to the semantic label of each point, the entire point cloud data is semantically segmented, and then the classification results of different object features (such as vegetation, soil, rock, etc.) are output;

[0062] Step 4: Analyze the relationship between the spatial distribution of bank slope disasters and topographic and geological factors, and use machine learning algorithms to identify and classify bank slopes to determine areas prone to bank slope disasters;

[0063] Step 5: Verify the results of identification and classification, and match corresponding management measures based on the classification results of the slope.

[0064] The second embodiment is based on the first embodiment. Figure 3 As shown in Figure 2, in step 4, the process of determining the area prone to bank slope disasters is as follows:

[0065] Based on the extracted slope area boundaries, the data on slope disasters under similar terrain and geological conditions are collected, and the spatial distribution of slope disasters is analyzed. The areas affected by terrain and geological factors are identified, and the relevant data of terrain factors and geological factors in the slope area are collected. The terrain factors include terrain undulation, slope and slope direction, as well as hydrological data. The geological factors include geological structure, lithology and strata. Among them, the greater the terrain undulation, the higher the frequency of disasters. Steep and broken terrain is prone to frequent disasters. For example, mountainous and hilly areas are prone to landslides, mud-rock flows and other disasters. The steeper the slope, the more likely it is to have geological disasters such as landslides and collapses. At the same time, the slope direction also affects the distribution of geological disasters. Sunny slopes are more susceptible to erosion by sunlight and rain. The erosion and accumulation of rivers affect the geological structure, which may cause disasters such as landslides and mud-rock flows. Hydrological factors such as water level changes and water flow scouring may also aggravate the occurrence of geological disasters. Geological structures such as fault zones and earthquake zones The geological characteristics have an important impact on the distribution of geological disasters. Areas with frequent seismic activities are prone to geological disasters. The physical and chemical properties of different lithologies have an impact on the distribution of disasters. The combination and distribution of strata have a decisive role in the type and distribution of disasters. According to the influencing factors of bank slope disasters, image processing technology is used to extract terrain texture features, and a geographic information system is used for spatial analysis to extract spatial features, including terrain undulation, slope, slope direction, geological structure, and lithology. The extracted features are combined with a neural network model to train a bank slope classification model, and the model is optimized using a cross-validation method to improve the accuracy and generalization ability of the model. The trained bank slope classification model is then used to identify and classify the bank slope, and the classification results are output. Based on the classification results, the bank slope is further analyzed and evaluated in combination with hydrological data, and the bank slope risk identification coefficient is calculated based on the comprehensive feature information to determine the area prone to bank slope disasters, and then the type of bank slope disasters is analyzed.

[0066] Furthermore, the calculation process of the slope risk identification coefficient is:

[0067] Traverse the terrain data, geological data and hydrological data, and calculate the terrain relief index T, geological structure risk index G and hydrological impact index H respectively;

[0068] For terrain data, collect the elevation data of the bank slope area, calculate the slope of each point from the elevation data, determine the maximum value Max(S) and the minimum value Min(S) in the slope data, and calculate the average value S of the slope avg , the terrain relief index T is obtained by comprehensive analysis, where the expression of the terrain relief index is:

[0069] For geological data, geological structure data are collected, fault zone density D and seismic activity index E are determined, and the geological structure risk index G is calculated, where the expression of the terrain relief index is:

[0070] For hydrological data, collect water level change and flow data, and determine the maximum value of water level change Max from the hydrological data. ( W ) and minimum value Min ( W ) , calculate the average water level W avg , the hydrological impact index H is obtained through comprehensive analysis, where the expression of the hydrological impact index is:

[0071] The weight coefficients of terrain relief index, geological structure risk index and hydrological impact index were determined by historical data analysis, and the slope risk identification coefficient R was obtained by comprehensive analysis.

[0072] Furthermore, the calculation formula of the slope risk identification coefficient is:

[0073]

[0074] Among them, R is the slope identification coefficient, T is the terrain undulation index, G is the geological structure risk index, and H is the hydrological impact index, which is calculated based on water level changes and flow. a is the weight coefficient of the terrain undulation index, b is the weight coefficient of the geological structure risk index, and c is the weight coefficient of the hydrological impact index. Each weight coefficient is adjusted according to the specific situation and historical data to reflect the influence of different factors on the stability of the slope. When the value of T, G or H increases, -(a·T+b·G+c·H) will decrease, resulting in an increase in the value of R, that is, the risk of slope disasters increases. On the contrary, if the value of T, G or H decreases, the value of R will decrease, indicating that the risk is reduced. The value range of R is between 0 and 1. The closer the value is to 1, the higher the severity of the slope disaster.

[0075] In step 5, the process of determining risk levels and matching management measures is as follows:

[0076] Collect samples of known slope types and historical slope disaster data, which have clear disaster history and stability assessment results. Use samples of known slope types and historical slope disaster data to verify the trained slope classification model, evaluate the accuracy and reliability of the model, compare the results output by the slope classification model with known slope types, analyze the accuracy of the model output results, conduct in-depth analysis of the classification results, clarify the stability of various types of slopes and the risk of disasters, and divide slope disasters into different types, namely collapse geological disasters, landslide geological disasters, and crack geological disasters. Combined with the slope risk identification coefficient and the classification type of slope disasters, formulate corresponding management measures and emergency plans for each identified slope disaster type. For slopes with collapse geological disasters, monitoring and early warning should be strengthened, and engineering management measures should be taken to reduce disaster risks, such as strengthening slopes and improving drainage systems. For slopes with landslide geological disasters, the monitoring frequency should be increased, more detailed geological and hydrological characteristics analysis should be conducted, and emergency plans should be formulated. For slopes with crack geological disasters, attention should be paid to them and regular monitoring and maintenance measures should be taken, such as regular geological exploration and hydrological monitoring.

[0077] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A reservoir slope geological disaster risk identification method based on UAV laser radar is characterized by: The following steps are involved: Step 1: Use a drone equipped with a laser radar device to perform a flight scan of the reservoir slope area, obtain three-dimensional point cloud data, and pre-process the original point cloud data; Step 2: Extract terrain features from the preprocessed point cloud data and produce DSM, DEM, and DOM products, derive slope, aspect, terrain roughness, and elevation information, and use machine learning algorithms to classify point cloud data, distinguish between vegetation and non-vegetation areas, and remove vegetation areas from point cloud data; Step 3: Use clustering algorithm and edge detection algorithm to segment the slope area, extract the boundary of the slope area, and classify similar point cloud data into one category; Step 4: Analyze the relationship between the spatial distribution of bank slope disasters and topographic and geological factors, and use machine learning algorithms to identify and classify bank slopes to determine areas prone to bank slope disasters; Step 5: Verify the results of identification and classification, and match corresponding management measures based on the classification results of the slope.

2. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 1 is characterized by: In step 1, the acquisition and preprocessing process of the three-dimensional point cloud data is as follows: Plan the flight route of the target area according to the topography and area of ​​the reservoir bank slope area, and set the flight altitude, speed and scanning density according to the performance of the LiDAR equipment and the characteristics of the target area; The laser radar device is mounted on a drone, and the drone is operated to fly along a predetermined flight route for flight scanning. At the same time, the laser radar device starts to work, emitting laser pulses and receiving reflected signals, recording the three-dimensional coordinates of each point, thereby obtaining the original point cloud data of the target object surface; After the flight mission is completed, the raw point cloud data from the UAV LiDAR device is transmitted to the ground processing workstation; The original point cloud data is preprocessed, including data decoding, POS data processing and point cloud data solution. The track file is generated using the three-dimensional coordinates, IMU data and eccentricity component values ​​solved by the UAV GNSS. The point cloud data is produced through the solution of the track file, decoded data and system calibration data, including position solution and attitude solution, to correct the system error.

3. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 2 is characterized by: In step 2, the process of extracting terrain features is as follows: Extract height, intensity, color and shape features from the preprocessed point cloud data, train the classification model using the support vector machine algorithm, and run the classification model to automatically classify the point cloud data to distinguish between ground points, vegetation points and other object points; Extracting ground points from the classified point cloud data, using the extracted ground points and other object points to generate a digital surface model of the reservoir bank slope area, and removing vegetation and other non-ground features from the digital surface model, retaining only the ground points, and generating a digital elevation model; Using digital elevation model data, calculate the slope and aspect of each point, and directly read the elevation value of each point from the digital elevation model; The height value of each point is extracted as a preliminary feature to distinguish vegetation from non-vegetation, and the reflection intensity value of each point is extracted using the reflection intensity information recorded by the lidar equipment.

4. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 3 is characterized by: In step 2, the process of removing vegetation areas from the point cloud data is as follows: Extract the color information of each point in the point cloud data, analyze the intensity of the green channel, calculate the local neighborhood characteristics of the point cloud normal, curvature, and point cloud density, and simultaneously calculate the normalized difference vegetation index; Label the feature data extracted from the point cloud data as training samples, use the training samples as input features, train a vegetation classification model based on a decision tree algorithm, run the vegetation classification model, classify the point cloud data, and distinguish between vegetation and non-vegetation areas; According to the classification results, vegetation points are separated and removed from the point cloud data, leaving only non-vegetation points.

5. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 4 is characterized by: In step 3, the classification process of similar point cloud data is as follows: Apply edge detection algorithm, input preprocessed point cloud data, calculate curvature and normal vector geometric features of point cloud, identify the geometric feature change area in point cloud, i.e. edge area, according to the change of curvature and normal vector, and output a set of edge points based on the change of curvature and normal vector; By searching for adjacent edge points, the adjacent edge points are connected using a neighborhood search algorithm to form a complete boundary, and several complete boundary line segments are output; Apply the KMeans clustering algorithm, input all point cloud data, select random points to initialize the cluster center, assign points to the nearest cluster center according to the distance metric, update the cluster center to the mean of all points in the cluster, repeat the clustering process until the cluster center no longer changes, and output multiple clusters, each cluster contains a set of similar point cloud data; The clustering results are post-processed to remove isolated clusters and merge similar clusters, and the optimized clustering results are output to obtain a complete slope area segmentation result; According to the clustering results, the convex hull algorithm is used to calculate the boundary points of each cluster, and the boundary points between clusters are determined according to the spatial relationship between clusters. The boundary points are connected into a smooth curve using the least squares curve fitting algorithm, and then the boundary curve of the slope area is output; After the slope area segmentation and boundary extraction are completed, the PointCNN deep learning model is used to input the point cloud data of each cluster, extract features through the convolution layer, use the fully connected layer for classification, and output the semantic label of each point. According to the semantic label of each point, the entire point cloud data is semantically segmented, and then the classification results of different object features are output.

6. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 5 is characterized by: In step 4, the process of determining the area prone to bank slope disasters is as follows: Based on the extracted slope area boundaries, collect slope disaster data under similar terrain and geological conditions, analyze the spatial distribution of slope disasters, and identify areas affected by terrain and geological factors; Collect relevant data on topographic factors and geological factors in the slope area. Topographic factors include topographic relief, slope gradient and aspect, and hydrological data. Geological factors include geological structure, lithology, and strata. According to the factors affecting bank slope hazards, image processing technology is used to extract terrain texture features, and geographic information system is used for spatial analysis to extract spatial features, including terrain undulation, slope, slope direction, geological structure, and lithology; The extracted features are combined with a neural network model to train a bank slope classification model, and the model is optimized using a cross-validation method. The trained bank slope classification model is then used to identify and classify the bank slope, and the classification results are output; Based on the classification results, the bank slope is further analyzed and evaluated in combination with hydrological data, and the bank slope risk identification coefficient is calculated based on the comprehensive characteristic information to determine the areas prone to bank slope disasters, and then the types of bank slope disasters are analyzed.

7. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 6 is characterized by: The calculation process of the slope risk identification coefficient is as follows: Traverse the terrain data, geological data and hydrological data, and calculate the terrain relief index T, geological structure risk index G and hydrological impact index H respectively; For terrain data, collect the elevation data of the bank slope area, calculate the slope of each point from the elevation data, determine the maximum value Max(S) and the minimum value Min(S) in the slope data, and calculate the average value S of the slope avg , the terrain relief index T is obtained by comprehensive analysis, where the expression of the terrain relief index is: For geological data, geological structure data are collected, fault zone density D and seismic activity index E are determined, and the geological structure risk index G is calculated, where the expression of the terrain relief index is: For hydrological data, collect water level change and flow data, and determine the maximum value of water level change Max from the hydrological data. ( W ) and minimum value Min ( W ) , calculate the average water level W avg , the hydrological impact index H is obtained through comprehensive analysis, where the expression of the hydrological impact index is: The weight coefficients of terrain undulation index, geological structure risk index and hydrological impact index were determined by historical data analysis, and the slope risk identification coefficient R was obtained through comprehensive analysis.

8. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 7 is characterized by: The calculation formula of the slope risk identification coefficient is: Among them, R is the slope identification coefficient, T is the terrain undulation index, G is the geological structure risk index, H is the hydrological impact index, a is the weight coefficient of the terrain undulation index, b is the weight coefficient of the geological structure risk index, c is the weight coefficient of the hydrological impact index, and the value range of R is between 0 and 1.

9. The reservoir slope geological disaster risk identification method based on unmanned aerial vehicle laser radar according to claim 8 is characterized by: In step 5, the process of determining the risk level and matching the management measures is as follows: Collect samples of known slope types and historical slope hazard data, and use them to verify the trained slope classification model; Compare the output of the slope classification model with the known slope types and analyze the accuracy of the model output results; The classification results are analyzed in depth to clarify the stability of various types of slopes and the risk of disasters, and slope disasters are divided into different types, namely collapse geological disasters, landslide geological disasters, and crack geological disasters; Based on the slope risk identification coefficient and the classification type of slope disasters, corresponding management measures and emergency plans are formulated for each identified type of slope disasters.

Images