A high-slope displacement unmanned aerial vehicle monitoring method and system

Abstract

The present application relates to the technical field of image processing, and particularly relates to a kind of high slope displacement UAV monitoring method and system, solve the technical problem that point cloud density of high slope displacement area occurs dramatic increase and geometric feature distortion in prior art, prone to error point to match, influence the accuracy of point cloud matching in high slope displacement monitoring process.This method comprises: obtaining the point cloud data of high slope area to be monitored in multiple periods;According to the point cloud data of multiple periods, analyze the laser reflection intensity and geometric shape feature of multiple points in high slope area, determine the displacement complexity of high slope area;Displacement complexity is used to represent the degree of point cloud scattering caused by displacement in the region;According to the displacement complexity, adjust the learning rate of point cloud matching algorithm in the iteration process;Using the point cloud matching algorithm after adjusting learning rate, the point cloud of high slope area in multiple periods is matched, and the displacement monitoring result of high slope area is output according to the matching result.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a method and system for monitoring high slope displacement using unmanned aerial vehicles (UAVs). Background Technology

[0002] With rapid economic development, the construction of public infrastructure such as highways and bridges is accelerating. In densely mountainous construction areas, frequent geological activities such as landslides and mudslides have significantly impacted operations on high rock slopes, posing serious challenges to the safety of construction workers and the protection of building materials. Therefore, precise displacement monitoring and stability analysis of high rock slopes are necessary before and during construction to promptly obtain information on deformation, loosening, and other hazards, and to issue disaster warnings, thus avoiding losses of manpower and resources.

[0003] In the specific technical practice of displacement monitoring of high rock slopes, one of the core links is to compare and analyze the three-dimensional point cloud data of the slope at different times. Usually, ICP series algorithms such as iterative closest point (ICP) and normal iterative closest point (NICP) are used to achieve point cloud matching by minimizing the total distance between all point pairs.

[0004] However, due to the unique deformation characteristics often exhibited by high rock slopes during displacement, such as the toppling of structural planes, the fractured and loosened rocks generate a large amount of new point cloud data. This leads to a dramatic increase in point cloud density and geometric distortions, such as the transformation of planar surfaces into curved surfaces and the bending of straight lines. Consequently, the point cloud matching process is prone to getting trapped in local optima. That is, a point in the displacement region of the source point cloud may be incorrectly matched to another new point in the target point cloud, which is also generated by a collapse. This new point is the closest to the original point but is not its true counterpart. Ultimately, this affects the accuracy of point cloud matching during the comparison of 3D point cloud data, and consequently, the accuracy and reliability of the monitoring results. Summary of the Invention

[0005] To address the technical problem in existing technologies where the point cloud density increases dramatically and geometric features are distorted in high slope displacement areas, leading to erroneous point pair matching and affecting the accuracy of point cloud matching during high slope displacement monitoring, the present invention aims to provide a UAV monitoring method and system for high slope displacement. The specific technical solution adopted is as follows:

[0006] Firstly, a method for monitoring high slope displacement using unmanned aerial vehicles (UAVs) is provided, comprising: acquiring point cloud data of the high slope area to be monitored over multiple periods; analyzing the laser reflection intensity and geometric morphological characteristics of multiple points within the high slope area based on the point cloud data from multiple periods to determine the displacement complexity of the high slope area; the displacement complexity is used to characterize the degree of point cloud dispersion caused by displacement within the area; adjusting the learning rate of the point cloud matching algorithm during the iteration process based on the displacement complexity; using the point cloud matching algorithm with the adjusted learning rate to match the point clouds of the high slope area over multiple periods, and outputting the displacement monitoring results of the high slope area based on the matching results.

[0007] Based on the above technical solution, in the UAV monitoring method for high slope displacement provided by this invention, by analyzing the laser reflection intensity and geometric morphology characteristics in point cloud data from multiple periods, a displacement complexity that can quantify the degree of point cloud dispersion caused by displacement is constructed. Based on this complexity, the iterative learning rate of the point cloud matching algorithm is dynamically adjusted, thereby effectively overcoming the technical obstacle that the algorithm is prone to getting trapped in local optima when matching high-density, highly scattered point clouds caused by slope displacement. This significantly improves the accuracy of point cloud matching, convergence stability, and the accuracy and reliability of the final displacement monitoring results.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the method for analyzing the laser reflection intensity and geometric morphological characteristics of multiple points within a high slope region based on point cloud data from multiple periods to determine the displacement complexity of the high slope region specifically includes: for each point within the high slope region, determining the local intensity, local density, and local divergence of each point based on the corresponding neighborhood point cloud; the local intensity is used to characterize the dispersion of laser reflection intensity within the neighborhood; the local density is used to characterize the density of point distribution within the neighborhood in the geometric morphological features; the local divergence is used to characterize the drastic change of the displacement field within the neighborhood in the geometric morphological features; clustering multiple points within the high slope region based on the three-dimensional coordinates, local intensity, local density, and local divergence of each point, dividing them into at least one displacement region and at least one non-displaced region; extracting features from the displacement regions and non-displaced regions, and determining the displacement complexity.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the method for extracting features from the displacement region and the non-displaced region and determining the displacement complexity specifically includes: extracting spectral features by sorting local intensity sequences according to spatial location based on the local intensity in the displacement region; determining a first feature by combining the local intensity difference between the displacement region and the non-displaced region; extracting a second feature by performing nonlinear feature extraction based on the local intensity, local density, and local divergence in the displacement region; and determining the displacement complexity based on the first feature and the second feature.

[0010] In conjunction with the first aspect mentioned above, in one possible implementation, the method for determining the local intensity, local density, and local divergence of each point within a high slope area based on the corresponding neighborhood point cloud specifically includes: determining the local intensity of each point based on the laser reflection intensity of multiple points within the neighborhood; determining the local density of each point based on the distance between multiple points within the neighborhood; performing preliminary matching of point clouds from multiple periods to obtain the displacement vector of each point within the high slope area; and determining the local divergence of each point based on the fitting relationship between the three-dimensional coordinates and displacement vectors of multiple points within the neighborhood.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the method for extracting the second feature based on the local intensity, local density, and local divergence of points in the displacement region using nonlinear features specifically includes: constructing a feature matrix for each displacement region based on the normalized results of the local density, local divergence, and local intensity of points in each displacement region; performing nonlinear feature extraction on the feature matrix of each displacement region to obtain an embedding matrix, and determining the global dispersion degree of the embedding matrix; the global dispersion degree includes any one of the coefficient of variation of the distance between points in the matrix, the distribution entropy, or the difference degree from a preset uniform distribution matrix; and determining the mean of the global dispersion degree corresponding to at least one displacement region as the second feature.

[0012] In conjunction with the first aspect above, in one possible implementation, the method of clustering multiple points within a high slope area based on the three-dimensional coordinates, local intensity, local density, and local divergence of each point to divide them into at least one displacement region and at least one non-displacement region specifically includes: clustering based on the normalized results of the three-dimensional coordinates, local intensity, local density, and local divergence of each point to obtain multiple clusters; and determining a comprehensive threshold based on the normalized results of the local density and local divergence of points in the multiple clusters to divide each cluster into a displacement region or a non-displacement region.

[0013] In conjunction with the first aspect above, in one possible implementation, the method of adjusting the learning rate of the point cloud matching algorithm during the iteration process based on the displacement complexity specifically includes: dynamically adjusting the learning rate of the point cloud matching algorithm based on the ratio of the current iteration number to the total iteration number; the adjusted learning rate decreases as the degree of point cloud scattering represented by the displacement complexity increases.

[0014] In conjunction with the first aspect above, in one possible implementation, the method for obtaining point cloud data of the high slope area to be monitored over multiple periods specifically includes: using a drone equipped with a lidar to fly along a preset path in each period to collect raw data of the high slope area; and performing filtering and noise reduction processing on the collected raw data to obtain point cloud data.

[0015] In conjunction with the first aspect above, in one possible implementation, the method for outputting displacement monitoring results of high slope areas based on matching results specifically includes: obtaining the translation vector between each set of matching points at different times and determining the magnitude of the translation vector; if the magnitude of the translation vector is greater than a preset displacement threshold, determining that the location corresponding to the matching point has experienced toppling failure; if the magnitude of the translation vector is less than or equal to the preset displacement threshold, determining that the location corresponding to the matching point has not experienced toppling failure.

[0016] Secondly, a UAV monitoring system for high slope displacement is provided, comprising: a data acquisition module, an algorithm optimization module, and a displacement monitoring module. The data acquisition module is used to acquire point cloud data of the high slope area to be monitored over multiple periods. The algorithm optimization module is used to analyze the laser reflection intensity and geometric morphological characteristics of multiple points in the high slope area based on the point cloud data over multiple periods, and to determine the displacement complexity of the high slope area. The displacement complexity is used to characterize the degree of point cloud dispersion caused by displacement in the area. The algorithm optimization module is also used to adjust the learning rate of the point cloud matching algorithm during the iteration process based on the displacement complexity. The displacement monitoring module is used to match the point cloud data of the high slope area over multiple periods using the point cloud matching algorithm with the adjusted learning rate, and output the displacement monitoring results of the high slope area based on the matching results.

[0017] Thirdly, a drone-based monitoring device for high slope displacement is provided, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is used to execute the instructions to perform the actions described in the first aspect and any possible implementation thereof. This drone-based monitoring device for high slope displacement can be an electronic device or a chip within an electronic device.

[0018] Fourthly, a computer-readable storage medium is provided, which stores instructions that, when executed on a high slope displacement unmanned aerial vehicle (UAV) monitoring device, cause the UAV monitoring device to perform actions as described in the first aspect and any possible implementation thereof.

[0019] Fifthly, a computer program product containing instructions is provided that, when the computer program product is run on a high slope displacement unmanned aerial vehicle (UAV) monitoring device, causes the UAV monitoring device to perform the actions described in the first aspect and any possible implementation thereof.

[0020] The present invention has the following beneficial effects:

[0021] By analyzing the laser reflection intensity and geometric morphology features in point cloud data from multiple periods, a displacement complexity is constructed that can quantify the degree of point cloud dispersion caused by displacement. Based on this complexity, the iterative learning rate of the point cloud matching algorithm is dynamically adjusted, thereby effectively overcoming the technical obstacle that the algorithm is prone to getting trapped in local optima when matching high-density, highly scattered point clouds caused by slope displacement. This significantly improves the accuracy of point cloud matching, convergence stability, and the accuracy and reliability of the final displacement monitoring results. Attached Figure Description

[0022] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A system structure diagram of an unmanned aerial vehicle (UAV) monitoring system for high slope displacement is provided in one embodiment of the present invention;

[0024] Figure 2 A flowchart illustrating a method for monitoring high slope displacement using an unmanned aerial vehicle (UAV) according to an embodiment of the present invention;

[0025] Figure 3 A flowchart illustrating another method for monitoring high slope displacement using an unmanned aerial vehicle (UAV) according to an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of the hardware structure of a drone monitoring device for high slope displacement, provided as an embodiment of the present invention. Detailed Implementation

[0027] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a UAV monitoring method and system for high slope displacement proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0028] 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 invention pertains.

[0029] The following description, in conjunction with the accompanying drawings, details the specific scheme of the UAV monitoring method and system for high slope displacement provided by this invention.

[0030] Please see Figure 1 The diagram illustrates a system structure of a UAV monitoring system for high slope displacement according to an embodiment of the present invention. The UAV monitoring system for high slope displacement includes: a data acquisition module 1, an algorithm optimization module 2, and a displacement monitoring module 3.

[0031] Among them, the data acquisition module 1 is the basic data support unit of the system, which is responsible for collecting and preprocessing point cloud data of high slopes in multiple periods, providing a clean and effective data source for the subsequent algorithm optimization module 2.

[0032] In some implementations, the data acquisition module 1 includes a path planning submodule 11, a data collection submodule 12, and a data preprocessing submodule 13.

[0033] The path planning submodule 11 is responsible for planning the UAV's flight path. By setting multiple path points over the high slope area to be monitored, it ensures that all path points can completely cover the monitored area. The distribution density of the path points is flexibly adjusted according to the terrain complexity of the high slope to avoid blind spots in data collection. The data acquisition submodule 12 is equipped with a lidar device and controls the UAV to fly along the preset path determined by the path planning submodule 11. It performs point cloud acquisition tasks at different times, and strictly maintains consistent UAV equipment parameters and environmental conditions during the acquisition process to avoid point cloud deviations caused by accidental factors. The data preprocessing submodule 13 processes the raw data acquired by the data acquisition submodule 12. First, a preset filtering algorithm is used to eliminate point cloud noise caused by UAV flight vibration. Then, a preset point cloud optimization algorithm is used to delete erroneous points in the high slope tilting area. Finally, the preprocessed point cloud data is output and transmitted to the algorithm optimization module 2 to lay the foundation for feature analysis and displacement complexity calculation.

[0034] Algorithm optimization module 2 is the core processing unit of the system. It receives the point cloud data output by data acquisition module 1 and optimizes the learning rate of the point cloud matching algorithm through feature extraction, clustering, and complexity calculation, providing accurate algorithm support for displacement monitoring module 3.

[0035] In some implementations, the algorithm optimization module 2 includes a feature extraction submodule 21, a clustering submodule 22, a displacement complexity calculation submodule 23, and a learning rate adjustment submodule 24.

[0036] The feature extraction submodule 21 receives the point cloud data output by the data acquisition module 1. For each point in the high slope area, it uses a preset nearest neighbor search algorithm to obtain its neighborhood point cloud. It determines the local intensity of each point (characterizing the dispersion of laser reflection intensity in the neighborhood) based on the laser reflection intensity of multiple points in the neighborhood, and determines the local density of each point (characterizing the density of distribution of points in the neighborhood) based on the distance between multiple points in the neighborhood. After performing preliminary matching on the point clouds of multiple periods to obtain the displacement vector of each point, it determines the local divergence of each point (characterizing the drastic change of displacement field in the neighborhood) by combining the fitting relationship between the three-dimensional coordinates of multiple points in the neighborhood and the displacement vector. The extracted local intensity, local density, local divergence and point cloud three-dimensional coordinates are then transmitted to the clustering and partitioning submodule 22.

[0037] The clustering submodule 22 normalizes the three-dimensional coordinates, local intensity, local density, and local divergence output by the feature extraction submodule 21, and then uses a preset clustering algorithm to cluster the normalized point cloud to obtain multiple clusters. Subsequently, a comprehensive threshold is determined based on the normalization results of the local density and local divergence of points in all clusters. Based on this threshold, each cluster is divided into a displacement region or a non-displacement region, and the division result is passed to the displacement complexity calculation submodule 23.

[0038] The displacement complexity calculation submodule 23, based on the partitioning results of the clustering submodule 22, performs feature extraction on the displacement region and the non-displaced region: for the local intensity of the displacement region, it sorts the local intensity sequence according to the spatial location and performs spectral feature extraction, and calculates the local intensity difference between the displacement region and the non-displaced region. The two are combined to determine the first feature; for the local intensity, local density and local divergence of the displacement region, it first constructs the feature matrix of each displacement region and performs nonlinear feature extraction to obtain the embedding matrix, and then determines the global dispersion of the embedding matrix by means of the coefficient of variation of the distance between points in the matrix, the distribution entropy or the difference from the preset uniform distribution matrix, and takes the mean of the global dispersion of all displacement regions as the second feature; finally, it performs a weighted sum of the first feature and the second feature according to the preset weight coefficient to obtain the displacement complexity that represents the dispersion of the point cloud caused by displacement in the region, and sends it to the learning rate adjustment submodule 24.

[0039] The learning rate adjustment submodule 24 receives the displacement complexity output by the displacement complexity calculation submodule 23, and dynamically adjusts the original learning rate of the algorithm by combining the ratio of the current iteration number to the total iteration number of the point cloud matching algorithm. The adjusted learning rate decreases as the degree of point cloud dispersion represented by the displacement complexity increases, ensuring that the algorithm is more robust in areas with high point cloud dispersion. Finally, the adjusted learning rate is transmitted to the displacement monitoring module 3.

[0040] The displacement monitoring module 3 is the system's result output unit. Relying on the preprocessed point cloud of the data acquisition module 1 and the adjusted learning rate of the algorithm optimization module 2, it completes point cloud matching and displacement judgment, and outputs the final monitoring results.

[0041] In some implementations, the displacement monitoring module 3 includes a point cloud matching submodule 31, a displacement judgment submodule 32, and a result output submodule 33.

[0042] The point cloud matching submodule 31 calls the adjusted point cloud matching algorithm, taking the multi-period preprocessed point cloud output by the data acquisition module 1 as input, and performs three-dimensional point cloud matching to obtain the translation vector between each adjacent path point (adjacent path points refer to two points with consecutive spatial positions on the UAV flight path). The translation vector is then transmitted to the displacement judgment submodule 32. The displacement judgment submodule 32 calculates the magnitude of the translation vector and compares it with a preset displacement threshold. If the magnitude is greater than the preset displacement threshold, it is determined that the corresponding high slope area has experienced toppling failure; if the magnitude is less than or equal to the preset displacement threshold, it is determined that the area has not experienced toppling failure, forming a preliminary monitoring result. The result output submodule 33 organizes the preliminary monitoring results from the displacement judgment submodule 32 and outputs the final high slope displacement monitoring results in a clear and intuitive form, providing direct evidence for relevant personnel to conduct disaster early warning.

[0043] Please see Figure 2 The diagram illustrates a flowchart of a method for monitoring high slope displacement using a drone, according to an embodiment of the present invention. This method includes:

[0044] S1. Obtain point cloud data of the high slope area to be monitored at multiple periods.

[0045] In some implementations, a drone equipped with a lidar is used to fly along a preset path at each time period to collect raw data of the high slope area. The collected raw data is then filtered and denoised to obtain point cloud data.

[0046] First, flight path planning is conducted for the high rock slope area to be monitored. Taking into account the terrain complexity, slope, and extent of the high slope, multiple path points are evenly distributed over the area (e.g., a spacing of 5-10 meters between path points in complex areas, and 50-100 meters between path points in larger areas). The path points are then connected to form the UAV's flight path. The distribution of all path points must achieve full coverage of the monitored area, and the density of path points is dynamically adjusted according to the terrain complexity. The density of path points is appropriately increased in areas with significant terrain undulations to ensure no data blind spots during subsequent data collection.

[0047] Secondly, drones equipped with lidar were selected as the data acquisition carrier, and point cloud acquisition was performed using drone oblique flight technology. The acquisition process needed to cover multiple key periods, preferably different time nodes before the construction of high slopes. During each acquisition, the drone's flight altitude, lidar scanning frequency, resolution, and other equipment parameters were strictly kept consistent. At the same time, adverse weather conditions such as strong winds, heavy rain, and dense fog were avoided, and acquisition was carried out in environmental conditions with good visibility and low wind speeds to avoid physical deviations in the raw data collected at different times due to fluctuations in equipment parameters or environmental interference, thus ensuring the comparability of data from multiple periods.

[0048] Subsequently, the raw data collected each time is filtered and denoised using a preset filtering algorithm (such as voxel grid filtering, conditional filtering, statistical filtering, etc.) to effectively eliminate point cloud noise caused by external factors such as vibration and airflow interference during UAV flight, thereby improving the purity of the raw data.

[0049] S2. Based on point cloud data from multiple periods, analyze the laser reflection intensity and geometric morphology characteristics of multiple points within the high slope area to determine the displacement complexity of the high slope area.

[0050] Among them, displacement complexity is used to characterize the degree of point cloud dispersion caused by displacement within a region.

[0051] In one possible implementation, combining Figure 2 ,like Figure 3 As shown, the method in S2 above can be specifically implemented through the following steps S21 to S23, which are explained in detail below:

[0052] S21. For each point within the high slope area, determine the local intensity, local density, and local divergence of each point based on the corresponding neighborhood point cloud.

[0053] First, determine the neighborhood point cloud range for each point. For any point within the high slope area (denoted as the i-th point), first set the range of the number of neighbors to 15-25 (balancing the completeness of neighborhood information and computational efficiency), and then use the K-nearest neighbors (KNN) algorithm to select the K (K is the number of neighbors) points that are closest (e.g., Euclidean distance) to the i-th point to form the neighborhood of the i-th point.

[0054] In some implementations, the laser reflection intensity of multiple points in the neighborhood is retrieved, and the standard deviation is statistically analyzed as the local intensity of the center point. This can quantify the dispersion of the laser reflection intensity in the neighborhood. The point cloud of the high slope collapse and damage area often shows more obvious fluctuations in laser reflection intensity due to terrain deformation, and thus exhibits higher local intensity.

[0055] In some implementations, the spatial distance between all points in the neighborhood (such as Euclidean distance) is calculated, and the inverse of the mean of the spatial distances between all points is used as the local density of the center point. This can quantify the density of the distribution of points in the neighborhood. In areas where high slopes have shifted, point clouds may locally aggregate due to terrain compression or collapse. As the spatial distance between points decreases, the corresponding local density will increase.

[0056] In some implementations, the point cloud acquired in the nth acquisition is used as the baseline point cloud, and the point cloud acquired in the (n+1)th acquisition is used as the comparison point cloud. An initial number of iterations is set (an empirical value ranging from [10, 50]). The ICP series algorithm is used to perform preliminary 3D matching between the two sets of point clouds. Then, the displacement vector (containing displacement components in the x, y, and z axes) of each point in the (n+1)th acquisition relative to the nth acquisition is calculated. Here, to ensure that the local divergence reflects the latest displacement field changes and to improve the timeliness and accuracy of the monitoring results, n is limited to the penultimate sampling, and n+1 is limited to the last sampling.

[0057] When a high slope collapses, all point cloud displacements occur from the top to the bottom of the slope. This means that if a high slope collapses, the neighborhood displacement field of each point cloud in the collapsed area is smooth. Therefore, the (n+1)th time the three-dimensional coordinates and displacement vectors of all point clouds in the neighborhood are used as input, and a linear fitting algorithm is used for fitting. The fitting formula is:

[0058]

[0059] In the formula, This represents the displacement vector of the j-th point in the neighborhood. This represents the displacement vector of the i-th point (center point). Represents the three-dimensional coordinates of the j-th point. Represents the three-dimensional coordinates of the i-th point. This indicates that a 3×3 displacement matrix is ​​fitted, where each row of matrix G represents the rate of change of displacement of the j-th point in the x, y, and z axes, respectively.

[0060] In this embodiment, the least squares method is used to fit the matrix G. The trace of the displacement matrix G (i.e., the sum of the diagonal elements of the matrix) is taken as the local divergence of the center point (the i-th point). The absolute value of the local divergence can quantify the degree of change of the displacement field in the neighborhood. The larger the absolute value of the local divergence, the more drastic the change of the displacement field in the neighborhood, and the higher the degree of collapse failure of the corresponding high slope.

[0061] For example, taking one point within the neighborhood as an example, let the coordinates of the centerline point be... (2, 3, 4), displacement vector (0.02, 0.01, 0.03); the coordinates of a neighboring point are (2.2, 3.3, 4.4), displacement vector (0.025, 0.018, 0.038). Substitute into the formula. = (0.005, 0.008, 0.008), = (0.2, 0.3, 0.4). Matrix G is a 3×3 matrix, in the form:

[0062]

[0063] Substitute the above differences into the fitting formula and solve for the matrix elements one by one:

[0064]

[0065]

[0066]

[0067] In actual calculations, the optimal matrix G needs to be solved using the least squares method by combining the system of difference equations for all neighboring points. Assume that after fitting the matrix to all points, the complete matrix is ​​obtained:

[0068]

[0069] Finally, the trace of matrix G (the sum of its diagonal elements) is taken as the local divergence. =0.09.

[0070] S22. Based on the three-dimensional coordinates, local intensity, local density and local divergence of each point, cluster multiple points in the high slope area to divide them into at least one displacement area and at least one non-displacement area.

[0071] In some implementations, clustering is performed based on the normalized results of the three-dimensional coordinates, local intensity, local density, and local divergence of each point to obtain multiple clusters. Then, a comprehensive threshold is determined based on the normalized results of the local density and local divergence of points in multiple clusters to classify each cluster as a displaced region or a non-displaced region.

[0072] Specifically, the three-dimensional coordinates, local intensity, local density, and absolute value of local divergence of each point (excluding the interference of displacement direction) are used as processing objects. A preset normalization algorithm (such as minimum-maximum (min-max) normalization) is used to map the feature indicators of different dimensions to a unified numerical range of [0,1], eliminate the dimensional differences between different features, ensure that each feature has equal weight in subsequent cluster analysis, and avoid the clustering results deviating from the actual feature law due to dimensional imbalance.

[0073] Next, the K-medoids algorithm is used for point cloud clustering: First, the number of cluster centers is set (the number of centers can be set to 20, balancing clustering precision and computational efficiency). Using the normalized 3D coordinates, local intensity, local density, and local divergence of each point as input, a weighted sum of Euclidean distances between each point and the candidate cluster centers is calculated (i.e., the sum of the products of each normalized feature value and its weight; the sum of the weights for all features is 1, for example, 3D coordinates have a weight of 0.4, local intensity has a weight of 0.2, local density has a weight of 0.2, and local divergence has a weight of 0.2). This weight is used as a metric to measure the similarity between a point and its cluster centers. Finally, the K-medoids algorithm is run to obtain multiple clusters.

[0074] Subsequently, focusing on the core characteristics of displacement, a comprehensive threshold is determined to distinguish between displaced and non-displaced regions. Among these characteristics, local density and local divergence directly reflect the spatial distribution and displacement field anomalies caused by displacement, and are unrelated to other non-displacement factors (such as illumination and surface material). Therefore, they are used to determine the threshold to focus on the specific core characteristics of displacement and eliminate non-displacement interference. For a single cluster, the mean of the normalized local density and the mean of the normalized local divergence (absolute value) of all points within the cluster are statistically analyzed. The weights are then re-allocated (the sum of the weights is 1, for example, each is 0.5), and the weighted sum of the two is calculated, which is recorded as the weighted sum of the mean of the cluster. This integrates the local density and local divergence characteristics within the cluster into a single quantitative index. The arithmetic mean of the weighted sum of the mean of all clusters is used as the comprehensive threshold for dividing the region.

[0075] Finally, the weighted sum of the mean of each cluster is compared with the comprehensive threshold one by one. If the weighted sum of the mean of a cluster is less than the comprehensive threshold, it means that the local density and local divergence of the points in the cluster are generally low (which matches the characteristic pattern of the high slope displacement area), and the high slope area corresponding to the cluster is determined to be a displacement area; otherwise, it means that the local density and local divergence of the points in the cluster are generally high, and it is determined to be a non-displaced area.

[0076] S23. Extract features from the displaced and undisplaced regions and determine the displacement complexity.

[0077] In some implementations, the first step is to extract spectral features by sorting the local intensity in the displacement region according to its spatial location and then combining the local intensity difference between the displacement region and the non-displaced region to determine the first feature.

[0078] Specifically, for each cluster of displacement regions, the local intensity of each point is sorted in ascending order of spatial location (e.g., using the X coordinate as the primary dimension, then the Y coordinate if X is the same, and the Z coordinate if both X and Y are the same), forming a local intensity sequence. This maintains the correlation between the sequence and the spatial structure of the high slope, avoiding feature disorder. Then, a Discrete Fourier Transform is performed on the local intensity sequence to obtain its frequency domain representation. The frequency domain representation is equivalent to breaking down the corresponding fluctuations in the sequence into smaller waves of varying speeds, with each smaller wave corresponding to a frequency domain component. Next, to measure the strength of each fluctuation, the square of the modulus of each frequency domain component is calculated as its energy. The quotient of this energy to the total energy (the sum of the energies of all frequency domain components) is then calculated to obtain the proportion of each fluctuation's intensity in the total intensity. Finally, based on the intensity proportion of each fluctuation, the Shannon entropy formula is used to calculate the entropy, which measures the degree of disorder in the local intensity and is denoted as the spectral entropy of each displacement region cluster.

[0079] For clusters of undisplaced regions, local intensity sequences are constructed according to the same rules. Then, a similarity algorithm applicable to sequences of different lengths is used to calculate the difference in local intensity sequences between displaced and undisplaced regions (such as the standard value of dynamic time warping (DTW) distance, which is dimensionless, where the standard value = |variable - mean| / standard deviation), quantifying the difference in local intensity fluctuations between displaced and undisplaced regions and highlighting the unique characteristics of displaced regions.

[0080] Then, the spectral entropy of all displacement region clusters and the difference between all sequence pairs of displacement and non-displacement regions are fused to obtain the first feature. , is represented as:

[0081]

[0082] In the formula, The mean spectral entropy of all clusters within the displacement region is represented by the value of the spectral entropy. The larger the value of the spectral entropy, the more chaotic the fluctuation of the local intensity in the displacement region. This represents the mean difference between all sequence pairs of shifted and unshifted regions. A larger difference indicates a more unique and disordered fluctuation in the shifted region. (First feature) By integrating the intensity fluctuation characteristics of the displacement region itself with the intensity difference characteristics between the displacement and non-displaced regions, the complexity of the displacement region is characterized from the frequency domain dimension.

[0083] The second step involves nonlinear feature extraction based on local intensity, local density, and local divergence in the displacement region to obtain the second feature. Specifically, this includes:

[0084] Based on the normalized results of the local density, local divergence, and local intensity of points in each displacement region, a feature matrix is ​​constructed for each displacement region. The row dimension of the feature matrix is ​​the normalized result of the local density, local divergence (absolute value), and local intensity of the points, and the column dimension is sorted in ascending order according to the spatial position of each point (e.g., using the X coordinate as the main dimension, if X is the same, then using the Y coordinate, if both X and Y are the same, then using the Z coordinate). Finally, a feature matrix of size M×3 is obtained (M is the number of points in each displacement region cluster).

[0085] Then, nonlinear feature extraction is performed on the feature matrix of each displacement region to obtain the embedding matrix, and the global scrambling degree of the embedding matrix is ​​determined. Because the point cloud in the displacement region exhibits a strong nonlinear response—for example, a small increase in density can lead to a sudden and significant increase in divergence (e.g., during a local rock mass collapse, a small change in displacement can cause the point cloud distribution to suddenly become disordered)—this feature relationship lacks a fixed proportion and has more complex variation patterns. The t-distributed stochastic neighbor embedding (t-SNE) algorithm and kernel principal component analysis algorithm are used to perform nonlinear feature extraction on the M×3 feature matrix, outputting the embedding matrix.

[0086] The global dispersion degree includes any one of the following: the coefficient of variation of the distance between points within the matrix (e.g., Euclidean distance) (the larger the coefficient of variation, the higher the dispersion degree); the distribution entropy (the larger the distribution entropy, the more disordered the point cloud distribution, and the higher the dispersion degree); or the difference degree from the preset uniform distribution matrix (e.g., Euclidean distance, the larger the Euclidean distance, the higher the dispersion degree). The mean of the global dispersion degree of all displacement region clusters is determined as the second feature. , is represented as:

[0087]

[0088] In the formula, R is the total number of displacement region clusters; This represents the global dispersion of the cluster in the r-th displacement region. (Second feature) It integrates the nonlinear characteristics of local intensity, local density, and local divergence of the displacement region, and characterizes the degree of dispersion of the displacement region from the spatial distribution dimension.

[0089] Finally, based on the first feature Second feature Determine the displacement complexity , is represented as:

[0090]

[0091] In the formula, It can be flexibly adjusted according to the rock characteristics of high slopes and the monitoring accuracy requirements, for example... =0.6, =0.4.

[0092] S3. Adjust the learning rate of the point cloud matching algorithm during the iteration process according to the displacement complexity.

[0093] In some implementations, the learning rate of the point cloud matching algorithm is dynamically adjusted based on the ratio of the current iteration number to the total iteration number, taking into account the displacement complexity. The adjusted learning rate decreases as the point cloud scattering represented by the displacement complexity increases, as expressed by:

[0094]

[0095] In the formula, The point cloud matching algorithm is represented in the first... The learning rate adjusted in the next iteration. The point cloud matching algorithm is represented in the first... The original learning rate at the next iteration. Let e ​​represent the total number of iterations, and let e represent the natural exponential function. This represents the displacement complexity.

[0096] The calculation logic of this formula is based on the learning rate decay of displacement complexity and iteration progress.

[0097] This represents the proportion of the q-th iteration to the total number of iterations, i.e., the current iteration progress. All parameters are greater than zero, and the exponent term is negative. It is a decay factor in the range (0, 1].

[0098] When the point cloud is highly scattered (C large) or the iteration enters the later stage ( When the learning rate is large, the decay factor will reduce the learning rate to prevent the algorithm from converging to a local optimum. When the point cloud is relatively regular or the iteration is in the early stage, the decay factor is close to 1 to maintain a large learning rate to improve the iteration efficiency and ultimately improve the matching accuracy of the point cloud matching algorithm in the scattered point cloud of the high slope.

[0099] S4. Using a point cloud matching algorithm with an adjusted learning rate, the point clouds of the high slope area are matched in multiple periods, and the displacement monitoring results of the high slope area are output based on the matching results.

[0100] First, the earliest point cloud from multiple periods is selected as the baseline point cloud, and point clouds from other periods are used as comparison point clouds. An adjusted point cloud matching algorithm is then invoked, inputting both the baseline and comparison point clouds. The algorithm dynamically adapts the iterative learning rate based on the displacement complexity of different regions. For displacement regions with high point cloud dispersion, a smaller learning rate is used, resulting in a more robust iteration process and avoiding getting trapped in local optima. For non-displaced regions with low point cloud dispersion, a relatively larger learning rate is used to improve matching efficiency. Leveraging the advantages of the adaptive learning rate algorithm, high-precision alignment of point clouds across multiple periods can be achieved. Ultimately, the matching of points between the baseline and comparison point clouds is realized.

[0101] In some implementations, the translation vectors between each pair of matching points at different times are obtained, and the magnitude of the translation vectors is determined. Specifically, the point cloud from the first acquisition period in multiple periods is selected as the baseline point cloud, and the point clouds from subsequent periods are used as comparison point clouds, thus clarifying the time dimension reference system for displacement monitoring. A three-dimensional translation vector is extracted from each pair of matching points (corresponding point pairs matched by the algorithm) in the baseline and comparison point clouds to represent the displacement of the high slope. Then, the magnitude of the translation vector is calculated using the vector magnitude calculation formula, simplifying the three-dimensional displacement into a single scalar.

[0102] Furthermore, considering factors such as the rock type, topographic slope, and engineering safety specifications of high slopes, historical displacement data and disaster cases of similar high slopes are collected. After outliers are eliminated using the 3σ statistical principle, a preset displacement threshold that meets engineering safety requirements is set. For example, the consequences of damage to rock high slopes are severe, and deformation must be strictly controlled; its preset displacement threshold can be set to 0.5 to 1.5 meters.

[0103] The magnitude of each translation vector is compared with a preset displacement threshold one by one: if the magnitude of the translation vector corresponding to a certain matching point is greater than the preset displacement threshold, it means that the displacement of the high slope corresponding to the matching point is large at different time intervals, and it is determined that the collapse has occurred; if the magnitude of the translation vector is less than or equal to the preset displacement threshold, it means that the displacement of the high slope corresponding to the matching point is small at different time intervals, and it is determined that the collapse has not occurred at the location.

[0104] Finally, the judgment results of all matching points are integrated to generate a comprehensive report on high slope displacement monitoring. The report can mark the location, translation vector magnitude, and whether toppling failure has occurred for each set of matching points, and present the overall displacement distribution of the high slope in a visual form (such as different colors marking dangerous areas and safe areas), providing direct and accurate technical support for engineers to formulate protective measures and initiate disaster early warning.

[0105] Based on the above technical solution, by analyzing the laser reflection intensity and geometric morphology characteristics in point cloud data from multiple periods, a displacement complexity that can quantify the degree of point cloud dispersion caused by displacement is constructed. Based on this complexity, the iterative learning rate of the point cloud matching algorithm is dynamically adjusted, thereby effectively overcoming the technical obstacle that the algorithm is prone to getting trapped in local optima when matching high-density, highly scattered point clouds caused by slope displacement. This significantly improves the accuracy of point cloud matching, convergence stability, and the accuracy and reliability of the final displacement monitoring results.

[0106] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0107] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0108] In this embodiment of the invention, the UAV monitoring device for high slope displacement can be divided into functional units according to the above method example. For example, each function can be divided into its own functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or software. It should be noted that the unit division in this embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0109] This invention also provides a schematic diagram of the hardware structure of a drone monitoring device for high slope displacement, see [link / reference]. Figure 4 The high slope displacement drone monitoring device 400 includes a processor 401, and optionally, a memory 402 connected to the processor 401.

[0110] In the first possible implementation, see Figure 4 The UAV monitoring device 400 for high slope displacement also includes a transceiver 403. The processor 401, memory 402, and transceiver 403 are connected via a bus. The transceiver 403 is used to communicate with other devices or communication networks. Optionally, the transceiver 403 may include a transmitter and a receiver. The device in the transceiver 403 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of the present invention. The device in the transceiver 403 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of the present invention.

[0111] Based on the first possible implementation method Figure 4 The structural diagram shown can be used to illustrate the structure of the UAV monitoring device for high slope displacement involved in the above embodiments.

[0112] in, Figure 4 The diagram can also illustrate the system chip in a drone monitoring device for high slope displacement. In this case, the actions performed by the aforementioned drone monitoring device for high slope displacement can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0113] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in this embodiment can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0114] The processor in this invention may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a standalone semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a System-on-a-Chip (SoC), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), programmable logic devices (PLDs), or logic circuits that implement dedicated logic operations.

[0115] The memory in the embodiments of the present invention may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable read-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0116] This invention also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0117] This invention also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0118] This invention also provides a chip, which includes a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0119] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0120] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings and the disclosure, will understand and implement other variations of the disclosed embodiments in carrying out the claimed invention. In this invention, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several of the functions listed in this invention.

[0121] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A method for monitoring high slope displacement using unmanned aerial vehicles (UAVs), characterized in that, include: Acquire point cloud data of the high slope area to be monitored at multiple time periods; For each point within the high slope area, the local intensity, local density, and local divergence of each point are determined based on the corresponding neighborhood point cloud. The local intensity is used to characterize the dispersion of laser reflection intensity within the neighborhood; the local density is used to characterize the density of point distribution within the neighborhood in the geometric features; and the local divergence is used to characterize the drastic change of the displacement field within the neighborhood in the geometric features. Based on the three-dimensional coordinates, local intensity, local density, and local divergence of each point, multiple points in the high slope area are clustered and divided into at least one displacement region and at least one non-displacement region. Based on the local intensity in the displacement region, a local intensity sequence is formed by sorting the local intensity according to its spatial location for spectral feature extraction. The first feature is determined by combining the local intensity difference between the displacement region and the non-displaced region. The second feature is obtained by nonlinear feature extraction based on the local intensity, local density, and local divergence in the displacement region. Based on the first feature and the second feature, the displacement complexity of the high slope area is determined; the displacement complexity is used to characterize the degree of point cloud dispersion caused by displacement within the area. Based on the ratio of the current iteration count to the total iteration count of the point cloud matching algorithm, the learning rate of the point cloud matching algorithm is dynamically adjusted according to the displacement complexity. The adjusted learning rate decreases as the point cloud scattering represented by the displacement complexity increases. The formula for dynamic adjustment is: The point cloud matching algorithm is represented in the first... The learning rate adjusted in the next iteration. The point cloud matching algorithm is represented in the first... The original learning rate at the next iteration. Indicates the current iteration number. Let e ​​represent the total number of iterations, and let e represent the natural exponential function. Indicates the displacement complexity; The point cloud matching algorithm with an adjusted learning rate is used to match the point clouds of the high slope area at multiple time periods, and the displacement monitoring results of the high slope area are output based on the matching results.

2. The UAV monitoring method according to claim 1, characterized in that, For each point within the high slope area, the local intensity, local density, and local divergence of each point are determined based on the corresponding neighborhood point cloud, including: The local intensity of each point is determined based on the laser reflection intensity of multiple points in the neighborhood; The local density of each point is determined based on the distance between multiple points in the neighborhood; Preliminary matching of point clouds from multiple periods is performed to obtain the displacement vector of each point within the high slope region. Based on the fitting relationship between the three-dimensional coordinates and displacement vectors of multiple points in the neighborhood, the local divergence of each point is determined.

3. The UAV monitoring method according to claim 1, characterized in that, Nonlinear feature extraction is performed based on the local intensity, local density, and local divergence in the displacement region to obtain the second feature, including: Based on the normalized results of the local density, local divergence and local intensity of points in each displacement region, a feature matrix for each displacement region is constructed. Nonlinear feature extraction is performed on the feature matrix of each displacement region to obtain an embedding matrix, and the global scrambling degree of the embedding matrix is ​​determined; the global scrambling degree includes any one of the coefficient of variation of the distance between points in the matrix, the distribution entropy, or the degree of difference from the preset uniform distribution matrix; The mean value of the global dispersion corresponding to at least one displacement region is determined as the second feature.

4. The UAV monitoring method according to claim 1, characterized in that, Based on the three-dimensional coordinates, local intensity, local density, and local divergence of each point, multiple points within the high slope area are clustered and divided into at least one displacement region and at least one non-displaced region, including: Clustering is performed based on the normalized results of the three-dimensional coordinates, local intensity, local density, and local divergence of each point to obtain multiple clusters; A comprehensive threshold is determined based on the normalized results of the local density and local divergence of points in multiple clusters to divide each cluster into a displaced region or an undisplaced region.

5. The UAV monitoring method according to claim 1, characterized in that, Obtain point cloud data of the high slope area to be monitored at multiple time periods, including: At each stage, drones equipped with lidar are used to fly along a preset path to collect raw data of the high slope area; The collected raw data is filtered and denoised to obtain point cloud data.

6. The UAV monitoring method according to claim 5, characterized in that, Based on the matching results, the displacement monitoring results for the high slope area are output, including: Obtain the translation vector between each set of matching points at different time periods, and determine the magnitude of the translation vector; If the magnitude of the translation vector is greater than a preset displacement threshold, the position corresponding to the matching point is determined to be a point of collapse and destruction. If the magnitude of the translation vector is less than or equal to a preset displacement threshold, it is determined that no tilting or damage has occurred at the location corresponding to the matching point.

7. A drone-based monitoring system for high slope displacement, characterized in that, include: Data acquisition module, algorithm optimization module, and displacement monitoring module; The data acquisition module is used to acquire point cloud data of the high slope area to be monitored at multiple periods; The algorithm optimization module is used to determine the local intensity, local density, and local divergence of each point within the high slope region based on the corresponding neighborhood point cloud. The local intensity characterizes the dispersion of laser reflection intensity within the neighborhood; the local density characterizes the density of point distribution within the neighborhood in the geometric features; and the local divergence characterizes the drastic change in the displacement field within the neighborhood in the geometric features. Based on the three-dimensional coordinates, local intensity, local density, and local divergence of each point, multiple points within the high slope region are clustered, dividing them into at least one displacement region and at least one non-displaced region. Based on the local intensity in the displacement regions, a local intensity sequence is formed by spatially sorting the local intensity for spectral feature extraction. Combining the local intensity difference between the displacement regions and the non-displaced regions, a first feature is determined. Based on the local intensity, local density, and local divergence in the displacement regions, nonlinear feature extraction is performed to obtain a second feature. Based on the first and second features, the displacement complexity of the high slope region is determined. The displacement complexity characterizes the degree of point cloud dispersion caused by displacement within the region. The algorithm optimization module is further configured to dynamically adjust the learning rate of the point cloud matching algorithm based on the ratio of the current iteration number to the total iteration number, according to the displacement complexity. The adjusted learning rate decreases as the point cloud scattering represented by the displacement complexity increases, and the formula for dynamic adjustment is: The point cloud matching algorithm is represented in the first... The learning rate adjusted in the next iteration. The point cloud matching algorithm is represented in the first... The original learning rate at the next iteration. Indicates the current iteration number. Let e ​​represent the total number of iterations, and let e represent the natural exponential function. Indicates the displacement complexity; The displacement monitoring module is used to match the point cloud data of the high slope area at multiple time periods using a point cloud matching algorithm with an adjusted learning rate, and output the displacement monitoring results of the high slope area based on the matching results.

Images