A geological disaster deformation monitoring method and system based on a laser radar

Abstract

The application relates to the technical field of intelligent sensing systems, in particular to a geological disaster deformation monitoring method and system based on a laser radar, which comprises the following steps: using the laser radar to acquire three-dimensional point clouds for a high and steep slope scene that may produce a collapse disaster; performing feature analysis on the acquired three-dimensional point clouds, calculating distribution feature indexes and local feature indexes, obtaining an optimal threshold value of the three-dimensional point clouds for filtering; matching the filtered high and steep slope three-dimensional point clouds with historical high and steep slope three-dimensional point cloud data and modeling, and feeding back after calculating a deformation variable. The application aims to guarantee the accuracy of geological disaster deformation monitoring for high and steep slope areas.

Description

Technical Field

[0001] This application relates to the field of intelligent sensing system technology, specifically to a geological disaster deformation monitoring method and system based on lidar. Background Technology

[0002] In the natural environment of steep slopes, rock masses are often subjected to long-term effects from factors such as gravity, precipitation, earthquakes, and weathering, which may lead to cracks, damage, and deformation, thereby triggering landslides. To prevent geological disasters such as landslides on steep slopes and avoid serious losses, deformation monitoring of these potential landslide areas is particularly important.

[0003] In geological disaster deformation monitoring, lidar is widely used to acquire 3D point cloud data of disaster areas. By modeling and comparing 3D point cloud data collected at different times, the deformation of geological disaster areas at different time points can be analyzed. However, when acquiring 3D point cloud data in steep slope scenarios, interference from suspended dust is often encountered, requiring filtering of the 3D point cloud data.

[0004] Existing technologies typically set thresholds for statistical filtering algorithms based on fixed values, but they fail to consider the significant airborne noise interference present in the 3D point cloud data of steep slopes. This noise can easily lead to residual noise or loss of detail when monitoring the deformation of steep slopes, resulting in serious deviations in the monitoring. This interference severely affects the accuracy of the original 3D point cloud data of steep slopes, potentially causing significant biases when monitoring the deformation of slopes at risk of landslides. Summary of the Invention

[0005] In view of the above, it is necessary to provide a geological disaster deformation monitoring method and system based on lidar to solve the above problems.

[0006] The first aspect of this application provides a method for monitoring geological disaster deformation based on lidar, the method comprising: A triangulation network is constructed based on the 3D point cloud of the area to be monitored. The distribution distance characteristics of all 3D point clouds in the neighborhood of each 3D point cloud in the triangulation network are analyzed to determine the nearest neighbor distribution characteristic value of each 3D point cloud. The distribution similarity characteristics of the nearest neighbor distribution characteristic values ​​of each 3D point cloud and the neighborhood of other 3D point clouds are analyzed to determine the distribution characteristic index of each 3D point cloud. For each 3D point cloud, surface fitting is performed on all 3D point clouds in its neighborhood. The curvature difference distribution of all 3D point clouds in the neighborhood of each 3D point cloud is analyzed. Based on the degree of fitting of the surface fitting, the comprehensive feature value of each 3D point cloud is determined. The difference between the comprehensive feature values ​​of each 3D point cloud and its neighborhood 3D point clouds is compared to determine the local feature index of each 3D point cloud. Based on the distribution and local feature indices of each 3D point cloud, and combined with the distance distribution between the 3D point cloud and all 3D point clouds in its neighborhood, the optimal threshold for each 3D point cloud is determined, and the 3D point cloud is filtered. The filtered results obtained from two adjacent monitoring sessions in the area to be monitored are matched to identify the geological disaster deformation area, and the deformation monitoring results are obtained by using a point cloud data comparison algorithm.

[0007] Specifically, determining the nearest neighbor distribution feature value for each 3D point cloud involves: The two endpoints of each line in the triangular network are treated as a pair of adjacent 3D point clouds; The lengths of the lines connecting all adjacent pairs of 3D point clouds within the neighborhood of each 3D point cloud are obtained. The nearest neighbor distribution feature value of each 3D point cloud is obtained by combining the overall distribution and dispersion of all the line lengths.

[0008] The distribution similarity feature is calculated by the maximum information coefficient of the nearest neighbor distribution feature values ​​of the three-dimensional point clouds between each three-dimensional point cloud and the neighborhood of the other three-dimensional point clouds; the distribution feature index is obtained by the degree of dispersion of the maximum information coefficient of each three-dimensional point cloud and all three-dimensional point clouds in its neighborhood.

[0009] Specifically, determining the comprehensive feature value of each 3D point cloud involves: Based on the curvature difference distribution, the curvature distribution feature value of each three-dimensional point cloud is obtained; Based on the degree of fitting, the fitting feature value of each three-dimensional point cloud is obtained; The normalized curvature distribution feature values ​​are added to the fitted feature values, and the sum is used as the comprehensive feature value of each 3D point cloud.

[0010] Specifically, the curvature distribution characteristic value of each three-dimensional point cloud is as follows: Calculate the curvature difference between each 3D point cloud and all 3D point clouds in its neighborhood, and use the standard deviation of all the differences as the curvature distribution characteristic value of each 3D point cloud.

[0011] The formula for obtaining the fitted feature value of each 3D point cloud is as follows: Obtain the fitting residual values ​​of all 3D point clouds and corresponding points on the fitted surface within the neighborhood of each 3D point cloud, and use the coefficient of variation of all obtained fitting residual values ​​as the fitting feature value of each 3D point cloud.

[0012] Specifically, determining the local feature indices for each 3D point cloud involves: Analyze the degree of disorder in the difference of the comprehensive feature values ​​between each 3D point cloud and all 3D point clouds in its neighborhood; The disorder level and comprehensive feature value of each 3D point cloud are positively fused to obtain the local feature index of each 3D point cloud.

[0013] The specific formula for determining the optimal threshold for each 3D point cloud is as follows: In the formula, This represents the optimal threshold for a 3D point cloud n. and These represent the distribution characteristic index and local characteristic index of the normalized 3D point cloud n, respectively. Let represent the mean Euclidean distance between a 3D point cloud n and all 3D point clouds in its neighborhood. represents the standard deviation of the Euclidean distance between the 3D point cloud n and all 3D point clouds in its neighborhood; 'a' represents the preset weighting coefficient. This indicates the preset base threshold.

[0014] Specifically, identifying geological disaster deformation areas includes: Calculate the spatial coordinate displacement difference between each 3D point cloud after the current monitoring and filtering and the corresponding 3D point cloud after the previous monitoring and filtering. All spatial coordinate displacement differences greater than the preset deformation tolerance are used as input to the clustering algorithm, and the resulting clusters are used as each deformation region.

[0015] Secondly, embodiments of this application also provide a geological disaster deformation monitoring system based on lidar, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0016] This application has at least the following beneficial effects: This application first constructs a triangulation network based on the 3D point cloud of the area to be monitored, providing a basic geometric structure for subsequent analysis. This facilitates the calculation of the spatial positional relationship of each 3D point cloud, laying the foundation for distance feature analysis. Next, it analyzes the neighborhood distribution distance characteristics of each 3D point cloud. By analyzing the neighborhood distribution around each point cloud, potential anomalies can be identified, leading to the inference of possible geological hazard risks. Determining the nearest neighbor distribution characteristic values ​​helps establish a baseline between normal and abnormal states, thereby improving monitoring sensitivity. Finally, it analyzes the similarity of the nearest neighbor distribution characteristic values ​​of adjacent 3D point clouds. By comparing the distribution characteristics between point clouds, it can discover the changing trends within local areas, thus identifying possible deformation areas. Finally, it performs surface fitting and curvature difference analysis on all 3D point clouds within the neighborhood. Surface fitting can reveal the local geometric shape characteristics of the point clouds, thereby helping to identify geological hazards. Analyzing minute changes in shape and variations in curvature can effectively capture changes in surface features during deformation, playing a crucial role in early warning of phenomena such as landslides and collapses. Comparing differences in comprehensive characteristic values ​​to determine local characteristic indicators can quantify the geological stability of different regions, thereby identifying high-risk areas. Considering the differences in discrete and aggregated features between airborne dust and the three-dimensional point clouds at the edges of rocks and cracks in steep slopes, further combining the distribution characteristics of dust noise with the curvature and slope of the edges of rocks and cracks in steep slopes allows for more effective filtering of dust noise interference in steep slope areas, improving data quality and enhancing the accuracy of subsequent analysis. Matching the three-dimensional point clouds after two filters to identify geological disaster deformation areas, and using point cloud data comparison algorithms to obtain deformation monitoring results, can provide a scientific basis for disaster early warning and response. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the steps of a geological disaster deformation monitoring method based on lidar, provided as an embodiment of this application; Figure 2 A flowchart illustrating the process of obtaining the optimal threshold in one embodiment of this application. Detailed Implementation

[0018] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application's specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0020] It should also be noted that the terms "first" and "second" in this application and its accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The methods disclosed in the embodiments of this application or the methods shown in the flowcharts include one or more steps for implementing the method. Without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged, and some steps can also be deleted.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0022] The following description, in conjunction with the accompanying drawings, details the specific scheme of the geological disaster deformation monitoring method and system based on lidar provided in this application.

[0023] Please see Figure 1 The diagram illustrates a flowchart of a geological disaster deformation monitoring method based on lidar according to an embodiment of this application. The method includes the following steps: The first step: Obtain the 3D point cloud of the area to be monitored.

[0024] A lidar is deployed in the area to be monitored, and the lidar is used to collect three-dimensional point cloud data of the area to be monitored. This application adopts a non-repeating scanning mode for scanning sampling, sets the sampling frequency to 160kHz, sets the horizontal angular resolution to 0.1°, sets the vertical field of view to 40°, and sets the scanning line beam to 64. Then the vertical angular resolution is the ratio of the vertical field of view to the scanning line beam.

[0025] The second step is to construct a triangulation based on the 3D point cloud, analyze the distribution distance characteristics of all 3D point clouds in the neighborhood of each 3D point cloud in the triangulation, and determine the nearest neighbor distribution characteristic value of each 3D point cloud; analyze the distribution similarity characteristics of the nearest neighbor distribution characteristic values ​​of each 3D point cloud with the neighborhood of the other 3D point clouds, and determine the distribution characteristic index of each 3D point cloud.

[0026] In steep slope areas, due to the abundance of soil and rock masses and the frequent influence of natural winds, a large amount of suspended dust is typically present in the air. When these tiny particles are used to acquire 3D point clouds using lidar, they cause laser pulse reflections, generating numerous random and discrete false 3D point clouds, thus creating background noise resembling "snowflakes" or "fog." Therefore, to improve the quality of the acquired 3D point cloud data, filtering and denoising processing are essential.

[0027] Traditional filtering and denoising of 3D point cloud data is often affected by variations in point cloud density. In steep slope scenes, dust interference may cause dust to be misidentified as valid rock edge features during denoising, thus affecting the filtering effect. Therefore, this application employs a statistical filtering method to denoise the acquired 3D point cloud. The specific operation steps are as follows: When using statistical filtering for denoising, this application requires setting the number of neighboring points in the algorithm. In this embodiment, the number of neighboring points is set to 50, that is, the 50 nearest 3D point clouds to each 3D point cloud are used as the set of nearest neighbor 3D point clouds for statistical filtering. Implementers can adjust this according to the actual situation.

[0028] Considering that in steep slope scenarios, the distribution of dust in the air is random, there are usually a large number of suspended points randomly and discretely distributed in space, and the overall distribution is discontinuous. At the same time, since dust usually presents a cloud-like distribution, the spatial distribution characteristics of different dust areas are different. Although the rock and soil areas have different shape characteristics, rough surfaces, and multi-scale protrusions, their three-dimensional point cloud distribution shows aggregation characteristics. In addition, neighboring areas usually have a high degree of spatial correlation due to the similarity of rock and soil characteristics.

[0029] Based on the above analysis, the acquired 3D point cloud data is first used as input. The triangular mesh growth algorithm is used to construct a triangular mesh for each 3D point cloud and its nearest set of 3D point clouds. The output is the triangular mesh formed by each 3D point cloud and the 3D point clouds in its nearest set of 3D point clouds. The triangular mesh growth algorithm is a well-known technology, and the specific steps will not be described in detail. Furthermore, the two endpoints of each line in the triangular network are considered as adjacent 3D point cloud pairs. Considering that dust and noise 3D point clouds are usually randomly distributed, the connection lengths between adjacent 3D point cloud pairs are usually random. However, the 3D point clouds on the surface of steep slopes and soil are relatively dense, and the connection lengths between adjacent 3D point cloud pairs are usually closer. Taking 3D point cloud n as an example, the connection lengths between all adjacent 3D point cloud pairs in 3D point cloud n and its nearest neighbor 3D point cloud set are obtained. The connection lengths are calculated using Euclidean distance; in other embodiments, Manhattan distance can also be used. The overall distribution and dispersion of all connection lengths are analyzed. In this embodiment, the overall distribution is calculated using the mean of all connection lengths, and the dispersion is calculated using the standard deviation of all connection lengths. In other embodiments, the overall distribution of all connection lengths can also be calculated using summation; the dispersion can also be calculated using range, coefficient of variation, etc. The product of the obtained mean and standard deviation is used as the nearest neighbor distribution feature value of 3D point cloud n, denoted as . .when The larger the value, the more discrete the 3D point cloud distribution, and the more dispersed the distance distribution between adjacent 3D point cloud pairs. The smaller the value, the denser the 3D point cloud distribution, and the higher the continuity of the overall 3D point cloud distance.

[0030] Furthermore, the nearest neighbor distribution feature values ​​of all 3D point clouds in the neighborhood of 3D point cloud n are obtained and arranged in ascending order according to their distance features from 3D point cloud n, serving as the feature sequence of 3D point cloud n. The distance features between 3D point clouds are calculated using Euclidean distance; in other embodiments, the distance between 3D point clouds can also be calculated using Manhattan distance. Further, the feature sequences corresponding to all 3D point clouds in the neighborhood of 3D point cloud n are obtained, and the maximum information coefficient of the feature sequence of 3D point cloud n and the corresponding feature sequence of each 3D point cloud in its neighborhood are calculated. The dispersion of all maximum information coefficients obtained for 3D point cloud n is also calculated. In this embodiment, the dispersion is obtained using the coefficient of variation of the maximum information coefficient, serving as a distribution feature index of 3D point cloud n. A larger calculated distribution feature index indicates that the correlation between the neighborhoods of the 3D point clouds is more random and generally smaller.

[0031] The third step is to perform surface fitting on all 3D point clouds in the neighborhood of each 3D point cloud, analyze the curvature difference distribution of all 3D point clouds in the neighborhood of each 3D point cloud, and determine the comprehensive feature value of each 3D point cloud by combining the fitting degree of the surface fitting; compare the difference of the comprehensive feature value of each 3D point cloud with the 3D point clouds in its neighborhood to determine the local feature index of each 3D point cloud.

[0032] Considering the high concentration of dust generated during the collapse in localized spaces, with a three-dimensional point cloud density even exceeding that of some loose rock masses, resulting in a "high-density, high-structure" characteristic that can be confused with the fractured areas on the surface of the unstable rock mass, further analysis is needed. This analysis also considers the discrete local curvature distribution of dust, lacking continuous neighborhood structure; while high curvature is observed at the edges of boulders and cracks. Although different morphologies of rock and soil surfaces exist, and there are fewer abrupt curvature changes, the overall curvature will exhibit characteristics similar to the surface of the rock and soil mass, with small curvature fluctuations and strong spatial aggregation. The slope changes correspond to the geological structure, with multiple slopes existing simultaneously.

[0033] Based on the above analysis, firstly, for each 3D point cloud and its nearest neighbor set of 3D point clouds, a least-squares fitting method is used to fit a surface, obtaining an approximate surface equation for each 3D point cloud. Least-squares surface fitting is a well-known technique, and the specific process will not be elaborated further. Next, the curvature of each 3D point cloud is obtained. Taking 3D point cloud n as an example, the differences in curvature between 3D point cloud n and the 3D point clouds in its nearest neighbor set are calculated, and the standard deviation of all the differences is calculated as the curvature distribution characteristic value of 3D point cloud n. When obtained The larger the value, the more discrete the local curvature distribution of the 3D point cloud. In this embodiment, the difference is calculated by the absolute difference. In other embodiments, it can also be calculated by the square of the difference.

[0034] Furthermore, considering that the dust noise 3D point cloud region lacks a continuous neighborhood structure and usually exhibits clumping aggregation characteristics, dust noise typically cannot appear simultaneously on a single surface or plane. This results in a low overall overlap between the dust region's 3D point cloud and the fitted surface, with a large number of discrete noise 3D point clouds. The fitting residual values ​​of these 3D point clouds are generally large and random. On the other hand, steep slopes and soil typically exhibit various slope characteristics, with relatively smooth surfaces, and usually show a high degree of agreement with the fitted surface. Therefore, taking a 3D point cloud n as an example, the fitting residual values ​​of the 3D point cloud and its nearest neighbor 3D point clouds are obtained, and the coefficients of variation of all obtained fitting residual values ​​are used as the fitting feature values ​​corresponding to the 3D point cloud n. Calculated and obtained The larger the value, the better the overlap between the 3D point cloud n and the 3D point cloud set in its nearest neighbor set and the fitted surface, and the smaller the overall difference.

[0035] Furthermore, the curvature distribution feature value and fitted feature value corresponding to each 3D point cloud are calculated and obtained. These feature values ​​and fitted feature values ​​are then normalized using the Z-score method. The normalized curvature distribution feature values ​​and fitted feature values ​​are added together, and the sum is used as the comprehensive feature value of each 3D point cloud. The difference between the comprehensive feature values ​​of each 3D point cloud and each 3D point cloud in its nearest neighbor set is calculated, and Shannon entropy is calculated for all the differences to characterize the degree of disorder. Shannon entropy calculation is a well-known technique and will not be elaborated further. The product of the Shannon entropy and the comprehensive feature value of each 3D point cloud is used as the local feature index of each 3D point cloud. Taking 3D point cloud n as an example, the local feature index corresponding to 3D point cloud n is obtained based on the above method. When the calculation is obtained The larger the value, the more discrete the curvature of the three-dimensional point clouds in the set of neighboring three-dimensional point clouds, the weaker the surface fitting degree, and the more chaotic the feature distribution differences between different three-dimensional point clouds in the local neighborhood region.

[0036] The fourth step: Based on the distribution characteristic index and local characteristic index of each 3D point cloud, and combined with the distance distribution between the 3D point cloud and all 3D point clouds in its neighborhood, determine the optimal threshold for each 3D point cloud and filter the 3D point cloud.

[0037] When using statistical filtering to denoise 3D point clouds in steep slope areas, the standard deviation threshold typically ranges from 0.5 to 1.5 times. This application considers the varying impacts of local dust interference on the results during monitoring and analysis. In areas with severe dust noise, effective filtering measures are required; while in 3D point cloud areas closely resembling the original steep slope characteristics, these features should be preserved to ensure data accuracy and reliability. Furthermore, the constructed distribution feature indicators and local feature indicators are normalized, and the optimal threshold for statistical filtering is obtained by combining the normalized distribution feature indicators and local feature indicators. Taking a 3D point cloud n as an example, the obtained values ​​are normalized, and a corresponding optimal threshold is set for this 3D point cloud. The specific formula is as follows: In the formula, This represents the optimal threshold for a 3D point cloud n. and These represent the normalized distribution characteristic index and the normalized local characteristic index of the 3D point cloud, respectively, where the normalization method adopts the maximum-minimum value normalization method. This represents the mean Euclidean distance between the 3D point cloud n and all its neighboring 3D point clouds, where a represents a preset weighting coefficient, which is 0.5 in this embodiment. The standard deviation of the Euclidean distance between the 3D point cloud n and all 3D point clouds in its neighborhood is represented. This represents the preset base threshold, set to 0.5. Using 0.5 as the base threshold for the standard deviation ensures that the standard deviation threshold adaptively adjusts with the 3D point cloud features. If dust and noise interference are not significant, a smaller threshold is set to preserve details. The smaller the value, the better. If dust and noise interference is severe, a larger threshold is usually used for effective filtering and noise reduction. The value is relatively large.

[0038] The flowchart for obtaining the optimal threshold is as follows: Figure 2 As shown.

[0039] It should be understood that existing technologies typically set thresholds for statistical filtering algorithms based on fixed values, but do not consider the large amount of suspended dust and noise interference in the air within the three-dimensional point cloud of steep slopes. This can easily lead to noise residue or loss of detail when monitoring the deformation of steep slopes, resulting in serious deviations. Therefore, this application fully combines the feature differences of three-dimensional point clouds within a local range and performs feature weighting on the thresholds corresponding to different three-dimensional point clouds to obtain the optimal threshold for three-dimensional point clouds, thereby reducing the error in monitoring the deformation of steep slopes.

[0040] After obtaining the optimal threshold, the acquired 3D point cloud of the steep slope is used as input. The optimal threshold is then applied to the 3D point cloud using a statistical filtering algorithm to remove noise, outputting the filtered and denoised 3D point cloud of the steep slope region. The statistical filtering algorithm is a well-known technique in this field, and its specific operation steps will not be elaborated further.

[0041] The fifth step is to match the filtered results obtained from two adjacent monitoring sessions of the area to be monitored, identify the geological disaster deformation area, and use a point cloud data comparison algorithm to obtain the deformation monitoring results.

[0042] After obtaining the filtered and denoised 3D point cloud of the steep slope, in order to detect deformation of the steep slope that may cause landslides, it is necessary to use the 3D point cloud obtained by the current monitoring and filtering and the 3D point cloud obtained by the previous monitoring and filtering as input, and use the ICP 3D point cloud registration algorithm to register the 3D point cloud. The ICP 3D point cloud registration algorithm is a well-known technology, and the specific steps will not be described in detail. It should be noted that the 3D point cloud of the first monitoring only needs to be filtered. After the 3D point cloud data of the next monitoring is filtered, the geological disaster deformation identification of the monitoring area can continue.

[0043] Furthermore, by comparing the registered data from the two periods, the spatial coordinate displacement difference between each 3D point cloud in the current monitored and filtered 3D point cloud and the corresponding 3D point cloud in the previous monitored and filtered period is calculated. All spatial coordinate displacement differences greater than the preset deformation tolerance are used as inputs to the DBSCAN density clustering algorithm to perform spatial clustering on the 3D point clouds that have undergone deformation and obtain the deformation region. In this embodiment, the preset deformation tolerance is 3cm. The DBSCAN density clustering algorithm is a well-known technology and will not be described in detail here.

[0044] After identifying the deformation area, quantitative analysis of the deformation characteristics is performed. Using the 3D point cloud data of the deformation area as input, the M3C2 (multi-scale model to model 3D point cloud comparison) algorithm is employed to obtain the monitoring results of deformation points within each deformation area. These monitoring results include the 3D displacement of each deformation point, and output the deformation amount, deformation direction (normal direction), and confidence level for each point. The detailed processing procedure of the M3C2 algorithm is well-known to those skilled in the art and will not be elaborated further. This completes the geological disaster deformation monitoring based on lidar.

[0045] Based on the same inventive concept as the above methods, this application also provides a geological disaster deformation monitoring system based on lidar, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the methods described above.

[0046] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description; sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0047] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some technical features, without causing the essence of the corresponding technical solutions to deviate from the scope of the technical solutions in the embodiments of this application, should all be included within the protection scope of this application.

Claims

1. A method for monitoring geological disaster deformation based on lidar, characterized in that, The method includes the following steps: A triangulation network is constructed based on the 3D point cloud of the area to be monitored. The distribution distance characteristics of all 3D point clouds in the neighborhood of each 3D point cloud in the triangulation network are analyzed to determine the nearest neighbor distribution characteristic value of each 3D point cloud. The distribution similarity characteristics of the nearest neighbor distribution characteristic values ​​of each 3D point cloud and the neighborhood of other 3D point clouds are analyzed to determine the distribution characteristic index of each 3D point cloud. For each 3D point cloud, surface fitting is performed on all 3D point clouds in its neighborhood. The curvature difference distribution of all 3D point clouds in the neighborhood of each 3D point cloud is analyzed. Based on the degree of fitting of the surface fitting, the comprehensive feature value of each 3D point cloud is determined. The difference between the comprehensive feature values ​​of each 3D point cloud and its neighborhood 3D point clouds is compared to determine the local feature index of each 3D point cloud. Based on the distribution and local feature indices of each 3D point cloud, and combined with the distance distribution between the 3D point cloud and all 3D point clouds in its neighborhood, the optimal threshold for each 3D point cloud is determined, and the 3D point cloud is filtered. The filtered results obtained from two adjacent monitoring sessions in the area to be monitored are matched to identify the geological disaster deformation area, and the deformation monitoring results are obtained by using a point cloud matching algorithm.

2. The geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The determination of the nearest neighbor distribution feature value for each 3D point cloud is specifically as follows: The two endpoints of each line in the triangular network are treated as a pair of adjacent 3D point clouds; The lengths of the lines connecting all adjacent pairs of 3D point clouds within the neighborhood of each 3D point cloud are obtained. The nearest neighbor distribution feature value of each 3D point cloud is obtained by combining the overall distribution and dispersion of all the line lengths.

3. The geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The distribution similarity feature is calculated by the maximum information coefficient of the nearest neighbor distribution feature values ​​of the three-dimensional point clouds between each three-dimensional point cloud and the neighborhood of the other three-dimensional point clouds; the distribution feature index is obtained by the degree of dispersion of the maximum information coefficient of each three-dimensional point cloud and all three-dimensional point clouds in its neighborhood.

4. The geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The determination of the comprehensive feature value of each 3D point cloud is specifically as follows: Based on the curvature difference distribution, the curvature distribution feature value of each three-dimensional point cloud is obtained; Based on the degree of fitting, the fitting feature value of each three-dimensional point cloud is obtained; The normalized curvature distribution feature values ​​are added to the fitted feature values, and the sum is used as the comprehensive feature value of each 3D point cloud.

5. A geological disaster deformation monitoring method based on lidar as described in claim 4, characterized in that, The curvature distribution characteristic value of each three-dimensional point cloud is specifically as follows: Calculate the curvature difference between each 3D point cloud and all 3D point clouds in its neighborhood, and use the standard deviation of all the differences as the curvature distribution characteristic value of each 3D point cloud.

6. The geological disaster deformation monitoring method based on lidar as described in claim 4, characterized in that, The fitted feature values ​​for each 3D point cloud are obtained using the following formula: Obtain the fitting residual values ​​of all 3D point clouds and corresponding points on the fitted surface within the neighborhood of each 3D point cloud, and use the coefficient of variation of all obtained fitting residual values ​​as the fitting feature value of each 3D point cloud.

7. The geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The determination of the local feature indices for each 3D point cloud specifically involves: Analyze the degree of disorder in the difference of the comprehensive feature values ​​between each 3D point cloud and all 3D point clouds in its neighborhood; The disorder level and comprehensive feature value of each 3D point cloud are positively fused to obtain the local feature index of each 3D point cloud.

8. The geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The specific formula for determining the optimal threshold for each 3D point cloud is as follows: In the formula, This represents the optimal threshold for a 3D point cloud n. and These represent the distribution characteristic index and local characteristic index of the normalized 3D point cloud n, respectively. Let represent the mean Euclidean distance between a 3D point cloud n and all 3D point clouds in its neighborhood. represents the standard deviation of the Euclidean distance between the 3D point cloud n and all 3D point clouds in its neighborhood; 'a' represents the preset weighting coefficient. This indicates the preset base threshold.

9. A geological disaster deformation monitoring method based on lidar as described in claim 1, characterized in that, The identification of geological disaster deformation areas specifically includes: Calculate the spatial coordinate displacement difference between each 3D point cloud after the current monitoring and filtering and the corresponding 3D point cloud after the previous monitoring and filtering. All spatial coordinate displacement differences greater than the preset deformation tolerance are used as input to the clustering algorithm, and the resulting clusters are used as each deformation region.

10. A geological disaster deformation monitoring system based on lidar, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images