Millimeter-level precision non-contact monitoring method for three-dimensional overall deformation of landslide

Abstract

The invention discloses a landslide three-dimensional overall deformation millimeter-level precision non-contact monitoring method, which comprises the following steps of: firstly, selecting a calculation center point in an original three-dimensional point cloud according to a certain distance, and forming initial point cloud blocks; then, registering the point cloud blocks with the point cloud data in another period by using an average domain vector algorithm, and interpreting current deformation data; and then, the distance between the calculation center points is reduced to obtain a calculation point with finer granularity, the initial deformation amount of the adjacent calculation point is used as the reference of the current calculation point by using the distance weight, the registration step is repeated, and calculation results are superposed. And when the distance is reduced to a specified threshold value or the deformation amount is smaller than a set threshold value, stopping calculation of the current area. In the calculation process, key parameters such as the block radius, the registration search distance and outlier removal are automatically determined through a trained neural network model in combination with Bayesian optimization. According to the method, the calculation process is automatic, the overall displacement deformation field is obtained, and millimeter-level precision is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-dimensional point cloud, and particularly relates to a non-contact monitoring method for millimeter-level precision of three-dimensional overall deformation of landslides. Background Art

[0002] Large-scale surface three-dimensional point cloud datasets can be obtained through three-dimensional laser scanners, airborne LiDars, and UAV aerial photography. Based on two-phase surface three-dimensional point cloud datasets, deformation analysis can be carried out on the target area. Traditional three-dimensional point cloud analysis methods include the shortest distance method, the vertical distance method, the M3C2 method, etc.

[0003] The vertical distance method obtains the displacement change in the vertical direction of the point as the deformation amount of the point. The shortest distance method searches for the point closest to the target point in another phase of the point cloud and uses the distance between them as the deformation amount at the target point. The advantages of the vertical distance method and the shortest distance method are that the operation is relatively simple and the calculation time is short, and the general deformation situation of the landslide can be obtained relatively quickly. However, the vertical distance method cannot reflect the deformation situation of the landslide in the horizontal direction, and the shortest distance method has a large randomness, and the closest point searched by the algorithm may be far from the actual corresponding point before and after the landslide changes. The M3C2 method can obtain the deformation amount of the calculation point along the normal direction by fitting a plane, and its applicability and accuracy are better than the shortest distance method. However, this method cannot reflect the displacement of the calculation point along the slope direction, and in some landslides with large sliding amounts along the slope, the calculation results of this method will have large errors.

[0004] The average domain vector algorithm is a point cloud deformation calculation method based on the ICP algorithm. The ICP algorithm is a point cloud registration algorithm. Before and after surface deformation, they often have similar surface characteristics. Using ICP registration can register the surface before deformation to the surface after deformation. Through the calculation data during point cloud registration, the overall transformation matrix during point cloud registration can be obtained, and thus the deformation information of the point cloud can be obtained. The average domain vector algorithm has high precision in processing locally uniformly deformed point clouds, but in large-area deformation analysis, the average domain vector algorithm has many limitations, such as the deformation analysis results are difficult to refine. When the local analysis area is selected too large, the boundary of the deformation area will not be obvious, and the calculated deformation value will have a large gap with the true deformation value. When the local analysis area is selected too small, the algorithm is prone to fall into a local optimum and obtain incorrect deformation values. At the same time, this algorithm relies on manual operation, which is time-consuming and laborious. The analysis area needs to be artificially delimited, and it is difficult to complete the deformation calculation of large areas. At the same time, the calculation parameters need to be manually debugged, and unreasonable calculation parameters will lead to large errors in the calculation results, and often multiple trials and errors are required to obtain more reasonable calculation parameters.

[0005] The ICP registration algorithm relies heavily on the features contained in the point cloud during the registration process. When the point cloud contains sufficient features, the registration result is often better. Conversely, it is easy to fall into a local optimum during registration. However, in actual 3D point cloud data, although a large area of the point cloud contains more features, the internal deformation is uneven, and fine deformation calculation cannot be achieved. Summary of the Invention

[0006] In order to solve the technical problems existing in the background art, the present invention aims to provide a non-contact monitoring method for the three-dimensional overall deformation of landslides with millimeter-level accuracy. In order to utilize the sufficient features of the large-area point cloud, during the preliminary registration, a sufficiently large point cloud block is delimited, and preliminary deformation information is obtained through registration. Then, the point cloud block is gradually reduced, and at the same time, the deformation value obtained during the previous block registration is assigned according to the weight, so that the point cloud block gradually approaches its correct matching position, and it is not directly to take the smallest point cloud block, which may cause the algorithm to fall into a local optimum due to too few features. After the calculation result reaches the target accuracy, the calculation is completed, and the final transformation matrix of the point cloud block is obtained through the transformation matrix obtained during each step of registration, so as to further interpret and obtain the fine overall deformation field of the analysis area. The entire process of step-by-step block registration is realized through a program, with low workload. At the same time, many calculation parameters are involved in the step-by-step registration calculation process. By using a large amount of step-by-step registration experimental data to train a neural network model and using the Bayesian optimization algorithm to infer the optimal parameters, the cost of human trial and error can be greatly reduced, and the accuracy of the deformation interpretation result can be improved.

[0007] In order to solve the technical problems, the technical solution of the present invention is as follows:

[0008] A non-contact monitoring method for the three-dimensional overall deformation of landslides with millimeter-level accuracy, the method comprising:

[0009] S1: Obtain two-phase surface three-dimensional deformation point cloud data for deformation interpretation, and extract overall features;

[0010] S2: Based on the extracted overall features of the point cloud and the trained neural network model, use the Bayesian optimization algorithm to find the optimum, and obtain the optimum initial calculation point spacing, minimum point cloud search radius, maximum corresponding point search distance, point cloud inclusion point threshold, and root mean square error threshold;

[0011] S3: According to the initial calculation point spacing, obtain calculation points, divide the initial point cloud block, and use the average domain vector algorithm for preliminary registration to obtain the initial transformation matrix, and analyze the translation vector and rotation matrix;

[0012] S4: Reduce the calculation point spacing, obtain new calculation points, and use the inverse distance weighting method to perform weighted averaging on the rotation vector and translation vector to obtain the initial transformation parameters of the new calculation points;

[0013] S5: Rotate and translate the segmented point cloud to a new position according to the transformation parameters of the newly calculated points, register again, obtain a new transformation matrix, and interpret the new translation vector and rotation matrix;

[0014] S6: If the displacement of the calculated points is less than the specified threshold, it is considered that the calculation of its area is completed, and the spacing of the calculated points is gradually reduced until the spacing of the calculated points reaches the minimum value, then it is considered that the overall calculation is completed.

[0015] Further, the step S2 includes:

[0016] The data for training includes point cloud features, calculation parameters, RMSE of corresponding points in the registration result, and the proportion of successfully matched blocks;

[0017] The loss function considers the root mean square error of the registration result and the proportion of successfully matched blocks, and its calculation formula is given by the following formula:

[0018] loss = αRMSE + βP + γ

[0019] Where: MSE represents the average value of the root mean square errors of all block matching results; P represents the proportion of successfully matched blocks; α, β, γ are hyperparameters;

[0020] For the average value of the root mean square errors of all block matching results RMSE, it is calculated by the following formula:

[0021]

[0022] Where, n S represents the total number of successfully registered blocks, and RMSE i represents the root mean square error of the corresponding point pairs after the i-th registration, and its calculation method is:

[0023]

[0024] Where, n i is the number of points included in the i-th block, is the coordinate of the j-th point in the i-th block, is the coordinate of the matching point of the j-th point in the i-th block in another period of point cloud;

[0025] For the proportion of successfully matched blocks P, its calculation method is:

[0026]

[0027] Where n S represents the total number of successfully registered blocks, and n represents the total number of blocks.

[0028] Further, the step S3 includes:

[0029] Obtain the initial calculation points according to the initial calculation point spacing, obtain the initial point cloud segmentation according to the point cloud search radius, and register the initial point cloud segmentation with another period of point cloud using the average domain vector algorithm with the determined calculation parameters to obtain the transformation matrix M0 of the point cloud segmentation in the first registration;

[0030]

[0031] Meanwhile, interpret the transformation matrix to obtain the translation vector of the current calculation point and the rotation matrix

[0032] Furthermore, the step S4 includes:

[0033] Reduce the spacing between the calculation points to obtain new calculation points. For each new calculation point, find the five nearest previous calculation points around it, and convert the rotation matrix into a rotation vector represented by axis-angle Use the inverse distance weighting method to perform weighted averaging on its translation vector and the rotation vector represented by axis-angle to obtain the initial translation amount of this calculation point and the initial rotation vector represented by axis-angle

[0034]

[0035] where w i is the weight corresponding to the i-th nearest calculation point, and its calculation method is:

[0036]

[0037] where d i is the distance from the i-th nearest calculation point to this calculation point;

[0038] Convert the initial rotation vector into the initial rotation matrix

[0039] Furthermore, the step S5 includes:

[0040] Convert the initial rotation vector into the initial rotation matrix and transform the point cloud segmentation A corresponding to the calculation point to the new position A through rotation and translation * ,

[0041]

[0042] Use the average domain vector algorithm to register the point cloud blocks again to obtain a new transformation matrix M2, interpret the new translation vector and rotation matrix, and superimpose the translation vector and rotation matrix obtained in the previous step to obtain the current total deformation.

[0043] Further, the step S6 includes:

[0044] Repeat steps S4 and S5. When the displacement of a certain calculation point is less than the specified threshold after a certain registration, assuming the current distance between calculation points is d, the range within d / 2 of this calculation point is regarded as completed calculation. When reducing the distance between calculation points next time, no new calculation points will be taken in the current area, but the calculation points from the previous round will be directly used; at the same time, when the distance between calculation points is reduced to the specified threshold, it is regarded as the completion of the calculation at the end of the current round.

[0045] Further, after the step S6, the method further includes:

[0046] According to all the calculation points and their corresponding transformation matrices in the last round of calculation, interpret the displacement vectors corresponding to the calculation points And the initial displacement vector of the current calculation round Superimpose to obtain the total displacement vector of this calculation point

[0047]

[0048] Thus, the displacement deformation field of the overall calculation area can be obtained.

[0049] A non-contact monitoring system for three-dimensional overall landslide deformation with millimeter-level accuracy, the system is applied to the method described in any one of the above, and the system includes:

[0050] Point cloud data acquisition and feature extraction module: Acquire two-phase surface three-dimensional deformation point cloud data for deformation interpretation and extract overall features;

[0051] Parameter optimization module: Based on the extracted overall features of the point cloud and the trained neural network model, use the Bayesian optimization algorithm to find the optimal initial calculation point spacing, minimum point cloud search radius, maximum corresponding point search distance, point cloud inclusion point number threshold, and root mean square error threshold;

[0052] Initial calculation and registration module: According to the initial calculation point spacing, obtain calculation points, divide the initial point cloud blocks, and use the average domain vector algorithm for preliminary registration to obtain the initial transformation matrix, and analyze the translation vector and rotation matrix;

[0053] Refined calculation point and weighted average module: Reduce the distance between calculation points, obtain new calculation points, and use the inverse distance weighting method to perform weighted averaging on the rotation vector and translation vector to obtain the initial transformation parameters of the new calculation points;

[0054] Point cloud segmentation transformation and re-registration module: Rotate and translate the segmented point cloud to a new position according to the transformation parameters of the newly calculated points, re-register, obtain a new transformation matrix, and interpret the new translation vector and rotation matrix;

[0055] Convergence judgment and calculation control module: If the displacement of the calculated points is less than the specified threshold, it is considered that the regional calculation is completed, and the spacing of the calculated points is gradually reduced until the spacing of the calculated points reaches the minimum value, then it is considered that the overall calculation is completed;

[0056] Final deformation calculation module: According to all the calculated points and their corresponding transformation matrices in the last round of calculation, interpret the displacement vectors corresponding to the calculated points, and superimpose them with the initial displacement vectors of the current calculation round to obtain the total displacement vectors of the calculated points; thus, the displacement deformation field of the overall calculation area can be obtained.

[0057] A computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements a non-contact monitoring method for three-dimensional overall deformation of landslides with millimeter-level accuracy as described in any one of the above.

[0058] A computer-readable storage medium stores a computer program, and when the program is executed by a processor, it implements a non-contact monitoring method for three-dimensional overall deformation of landslides with millimeter-level accuracy as described in any one of the above.

[0059] Compared with the prior art, the advantages of the present invention are:

[0060] 1. High-precision three-dimensional deformation interpretation

[0061] By gradually reducing the spacing of the calculation center points, the deformation interpretation from rough to fine is realized, and the accuracy of the final interpretation result is improved.

[0062] The average domain vector algorithm is used for point cloud segmentation registration, reducing the influence of local noise and ensuring the stability and accuracy of point cloud matching.

[0063] Through the method of iterative optimization and deformation superposition, the continuity of the point cloud is fully considered, making the calculation results smoother and more reasonable.

[0064] 2. Adaptive parameter optimization to improve calculation efficiency

[0065] Combining a neural network model and a Bayesian optimization algorithm to automatically optimize key parameters (such as search radius, point number threshold, root mean square error threshold, etc.), avoiding manual empirical settings and improving the intelligent level of calculation.

[0066] Through training with a large amount of experimental data, it can dynamically adjust calculation parameters according to different point cloud features (such as density, size, deformation amount, terrain type, etc.), enabling the algorithm to be applicable to various terrains and data acquisition methods.

[0067] 3. The calculation point initialization method is scientific, enhancing stability

[0068] The inverse distance weighting method (IDW) is used to estimate the initial deformation amount of new calculation points, enabling the calculation points to have good smoothness and continuity during the iteration process and avoiding the influence of local outliers.

[0069] Through the weighted interpolation of the deformation information of neighboring points, the interference of noise during the registration process is effectively reduced, and the stability of deformation interpretation is improved.

[0070] 4. The calculation termination mechanism is reasonable, reducing redundant calculations

[0071] Set double termination conditions:

[0072] When the distance between calculation points shrinks to the minimum threshold, the waste of computing resources is avoided;

[0073] When the change in deformation amount is less than the set threshold, the calculation is terminated in advance to improve the calculation efficiency.

[0074] Adopt a regional adaptive termination strategy, so that the regions that have converged stably are no longer recalculated, further reducing the calculation burden and improving the overall interpretation efficiency.

[0075] 5. The deformation field is constructed completely and applicable to various applications

[0076] Based on the deformation data of the final calculation points, the overall surface displacement deformation field is reconstructed, which can be used in fields such as geological disaster monitoring and engineering deformation monitoring.

[0077] Applicable to various sources of point cloud data (such as LiDAR, photogrammetry, UAV point cloud, etc.), it has strong generalization ability. Brief Description of the Drawings

[0078] Figure 1 It is a schematic diagram of the step-by-step calculation steps of the present invention;

[0079] Figure 2 It is a schematic diagram of the network training steps of the present invention;

[0080] Figure 3 It is a diagram of the interpretation result of the quarry slope deformation field. Detailed Embodiments

[0081] The following describes the specific embodiments of the present invention in conjunction with the embodiments:

[0082] It should be noted that the structures, ratios, sizes, etc. shown in this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the implementation conditions of the present invention. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0083] At the same time, the terms such as "upper", "lower", "left", "right", "middle", and "one" cited in this specification are only for the convenience of clear narration, and are not used to limit the scope of implementation of the present invention. The change or adjustment of their relative relationships, without substantial change in the technical content, should also be regarded as the scope of implementation of the present invention.

[0084] Embodiment 1:

[0085] As Figure 1-2 shown, a non-contact monitoring method for three-dimensional overall deformation of landslides with millimeter-level accuracy, the specific technical solution is as follows;

[0086] (1) Prepare two-phase surface three-dimensional deformation point cloud data for deformation interpretation, and the areas included in the two-phase point cloud data should be corresponding.

[0087] (2) Obtain the overall characteristics of the point cloud, including the average density of the point cloud, smoothness, size, surface type, data acquisition method, average deformation prediction value, etc.

[0088] (3) According to the overall characteristics of the point cloud obtained in (2) and the trained neural network model, use the Bayesian optimization algorithm to optimize, and obtain the key parameters required later, such as the optimal initial calculation point spacing, minimum point cloud search radius, maximum corresponding point search distance, point cloud point number threshold, root mean square error threshold, etc. The data used for training includes point cloud features (average density of the point cloud, point cloud size, magnitude of deformation, terrain type, point cloud acquisition method, point cloud smoothness), calculation parameters (minimum point cloud search radius, maximum corresponding point search distance, point cloud point number threshold, root mean square error threshold), and the RMSE of the corresponding points in the registration result, and the proportion of successfully matched blocks. The loss function mainly considers the root mean square error of the registration result and the proportion of successfully matched blocks, and its calculation formula is given by the following formula:

[0089] loss = αRMSE + βP + γ

[0090] Where: MSE represents the average value of the root mean square error of all block matching results; P represents the proportion of successfully matched blocks; α, β, γ are hyperparameters;

[0091] For the average value RMSE of the root mean square error of all block matching results, it is calculated by the following formula:

[0092]

[0093] Among them, n S represents the total number of successfully registered blocks, and RMSE i represents the root mean square error of the corresponding point pairs after the i-th registration, and its calculation method is:

[0094]

[0095] Among them, n i is the number of points included in the i-th block, is the coordinate of the j-th point in the i-th block, is the coordinate of the matching point of the j-th point in the i-th block in another period of point cloud.

[0096] For the proportion P of successfully matched blocks, its calculation method is:

[0097]

[0098] Among them, n S represents the total number of successfully registered blocks, and n represents the total number of blocks.

[0099] (4) Obtain the initial calculation points according to the initial calculated point spacing, obtain the initial point cloud blocks according to the point cloud search radius, and use the average domain vector algorithm to register the initial point cloud blocks with another period of point cloud with the calculation parameters determined in (3) to obtain the transformation matrix M0 of the point cloud blocks in the first registration.

[0100]

[0101] At the same time, interpret the transformation matrix to obtain the translation vector of the current calculation points and the rotation matrix

[0102] (5) Reduce the spacing between the calculation points to obtain new calculation points. For each new calculation point, find the five nearest previous calculation points around it, and convert the rotation matrix to a rotation vector represented by axis-angle Use the inverse distance weighting method to perform weighted averaging on its translation vector and the rotation vector represented by axis-angle to obtain the initial translation amount and the initial rotation vector represented by axis-angle

[0103]

[0104] Among them, w i is the weight corresponding to the i nearest calculation points, and its calculation method is:

[0105]

[0106] where d i is the distance from the i-th nearest calculated point to this calculated point.

[0107] Convert the initial rotation vector into the initial rotation matrix

[0108] (6) Convert the initial rotation vector into the initial rotation matrix and transform the point cloud block A corresponding to the calculated point to the new position A through rotation and translation * ,

[0109]

[0110] Use the average domain vector algorithm to register the point cloud block again to obtain a new transformation matrix M2, interpret the new translation vector and rotation matrix, and superimpose the translation vector and rotation matrix obtained in the previous step to obtain the current total deformation.

[0111] (7) Repeat steps (5) and (6). When, after registration of a certain calculated point, the displacement of the current calculated point is less than the specified threshold, assuming the current calculated point spacing is d, the range within d / 2 of this calculated point is regarded as the calculation completed. When reducing the calculated point spacing next time, no new calculated points will be taken in the current area, but the calculated points of the previous round will be directly used. At the same time, when the calculated point spacing is reduced to the specified threshold, it is regarded as the calculation completed at the end of the current round.

[0112] (8) According to all the calculated points and their corresponding transformation matrices in the last round of calculation, interpret the displacement vector corresponding to the calculated point and superimpose it with the initial displacement vector of the current calculation round to obtain the total displacement vector of this calculated point

[0113]

[0114] Thus, the displacement deformation field of the overall calculation area can be obtained.

[0115] Embodiment 2:

[0116] A non-contact monitoring method for the three-dimensional overall deformation of a landslide with millimeter-level accuracy in this embodiment takes a landslide analysis case in a certain place in Liangshan Prefecture, Sichuan as an example:

[0117] (1) There are two-phase three-dimensional laser scanning point cloud data of a landslide in a certain place in Liangshan Prefecture, Sichuan. After processing with RiSCAN Pro software, two-phase corresponding three-dimensional point cloud data of the deformation analysis area are obtained.

[0118] (2) Calculate the average density, point cloud size, preliminary estimated average deformation amount, point cloud acquisition method (3D laser scanning), smoothness of the point cloud (obtained through PointNet), and terrain type (mountain with vegetation) of the point cloud data. Convert the above features into corresponding data and input them into the trained neural network model. Obtain the recommended calculation parameters through Bayesian optimization, including the minimum block radius (0.8 m), maximum corresponding point search distance (2.2 m), minimum number of points per block (63), and RMSE threshold (0.19 m).

[0119] (3) Input the above parameters into the step-by-step registration program for progressive calculation. The specific process is as follows:

[0120] Initially, set the calculation point spacing to 1 / 5 of the point cloud's horizontal size, and set the block search deformation to the larger value between 1 / 2 of the calculation point spacing and the minimum block radius (0.8 m). Blocks with fewer than 63 points are regarded as invalid data. During the registration process, those with an RMSE greater than 0.19 m are regarded as incorrect registrations and discarded as interference data. Set the maximum search radius during the ICP registration process to 2.2 m. After the last round of block division, obtain the deformation data of 238,404 calculation points, thereby generating the overall deformation field in this area.

[0121] (4) To verify the quality of the deformation field results obtained in step (3) of this method, calculate the true displacement amount by manually calculating the coordinate differences of the obvious feature points of the original point cloud data. At the same time, use the shortest distance method and the M3C2 method to obtain the deformation calculation results for comparison, as shown in Table 1:

[0122] Table 1 - Comparison table of displacement calculation results of different methods at multiple points (unit: m)

[0123]

[0124] It can be seen that the displacement amounts calculated by this method are generally closer to the true displacement and have a smaller gap, proving that the results have higher accuracy and are better.

[0125] (5) To further verify the accuracy of this method, a quarry slope in a certain place in Aba Prefecture, Sichuan was selected for analysis. Generate the overall deformation field of the analysis area according to the above steps. Select a part of the deformation results in the stable area in the deformation interpretation results for analysis. The results show that the deformation interpretation results in the stable area are all within 1 cm, proving that this method has reached the mm-level accuracy.

[0126] Example 3:

[0127] This embodiment provides a terminal device, which includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function. The processor described in the embodiments of the present invention can be used for the operation of a non-contact monitoring method for three-dimensional overall deformation of landslides with millimeter-level accuracy, including the following steps:

[0128] First, calculate the center points at a certain distance in the original three-dimensional point cloud to form a preliminary point cloud segmentation. Then, use the average domain vector algorithm to register the point cloud segmentation with the point cloud data of another period to interpret the current deformation data. Subsequently, reduce the distance between the calculation center points to obtain finer-grained calculation points, use the distance weight to take the initial deformation amount of adjacent calculation points as a reference for the current calculation point, repeat the registration step, and superimpose the calculation results. When the distance is reduced to a specified threshold or the deformation amount is less than the set threshold, stop the calculation of the current area. During the calculation process, key parameters such as the segmentation radius, registration search distance, and outlier removal are automatically determined through a trained neural network model combined with Bayesian optimization.

[0129] Embodiment 4:

[0130] This embodiment provides a storage medium, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a terminal device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and, of course, the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, and this storage space stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by a processor are stored in this storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory.

[0131] One or more instructions stored in the computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps in the above embodiment regarding a non-contact monitoring method for three-dimensional overall deformation of a landslide with millimeter-level accuracy; one or more instructions in the computer-readable storage medium are loaded and executed by a processor to perform the following steps:

[0132] First, calculate the center points at a certain distance in the original three-dimensional point cloud to form a preliminary point cloud segmentation. Then, use the average domain vector algorithm to register the segmented point cloud with the point cloud data of another period to interpret the current deformation data. Subsequently, reduce the spacing between the calculated center points to obtain finer-grained calculation points, use distance weights to take the initial deformation amounts of adjacent calculation points as references for the current calculation points, repeat the registration step, and superimpose the calculation results. When the spacing is reduced to a specified threshold or the deformation amount is less than the set threshold, stop the calculation of the current area. During the calculation process, key parameters such as the segmentation radius, registration search distance, and outlier removal are automatically determined through a trained neural network model combined with Bayesian optimization.

[0133] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0134] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0135] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means that implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0137] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various changes can be made without departing from the gist of the present invention.

[0138] Many other changes and modifications can be made without departing from the concept and scope of the present invention. It should be understood that the present invention is not limited to a specific embodiment, and the scope of the present invention is defined by the appended claims.

Claims

1. A non-contact monitoring method for three-dimensional overall deformation of landslide with millimeter-level accuracy, characterized in that: The method comprises: S1: Obtain two phases of surface 3D deformation point cloud data for deformation interpretation and extract overall features; S2: Based on the overall features of the extracted point cloud and the trained neural network model, the Bayesian optimization algorithm is used to find the optimal initial calculation point spacing, minimum point cloud search radius, maximum corresponding point search distance, point cloud included point number threshold and root mean square error threshold; S3: according to the initial calculation point spacing, obtain the calculation points, divide the initial point cloud into blocks, use the average domain vector algorithm for preliminary registration, obtain the initial transformation matrix, and parse the translation vector and rotation matrix; S4: Reduce the distance between calculation points, obtain new calculation points, use the inverse distance weight method to perform weighted average on the rotation vector and translation vector, and obtain the initial transformation parameters of the new calculation point; S5: Rotate and translate the point cloud blocks to new positions according to the transformation parameters of the newly calculated points, align them again, obtain the new transformation matrix, and interpret the new translation vector and rotation matrix; S6: If the displacement of the calculation point is less than the specified threshold, the regional calculation is considered to be completed, and the distance between the calculation points is gradually reduced until the distance between the calculation points reaches the minimum value, then the overall calculation is considered to be completed.

2. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 1 is characterized in that: The step S2 comprises: The data used for training includes point cloud features, calculation parameters, RMSE of corresponding points of registration results, and the proportion of successfully matched blocks; The loss function takes into account the root mean square error of the registration result and the proportion of successfully matched blocks. Its calculation formula is given by the following formula: loss = αRMSE + βP + γ Where: MSE represents the average root mean square error of all block matching results; P represents the proportion of successfully matched blocks; α, β, and γ are hyperparameters; The average RMSE of the root mean square error of all block matching results is calculated by the following formula: Among them, n S Indicates the total number of successfully registered blocks, RMSE i It represents the root mean square error of the corresponding point pair after the i-th registration, which is calculated as: Among them, n i is the number of points contained in the i-th block, is the coordinate of the jth point in the i-th block, is the coordinate of the matching point of the jth point in the i-th block in another point cloud; The calculation method for the proportion of successfully matched blocks P is: Where n S It indicates the total number of blocks that are successfully registered, and n indicates the total number of blocks.

3. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 1 is characterized in that: The step S3 comprises: The initial calculation points are obtained according to the initial calculation point spacing, and the initial point cloud blocks are obtained according to the point cloud search radius. The initial point cloud blocks are aligned with another point cloud using the average domain vector algorithm with the determined calculation parameters to obtain the transformation matrix M0 of the point cloud blocks in the first alignment; At the same time, interpret the transformation matrix to obtain the translation vector of the current calculation point and the rotation matrix 4. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 1 is characterized in that: The step S4 comprises: Reduce the distance between the calculation points, obtain new calculation points, and for each new calculation point, find the five nearest calculation points around it, and rotate the matrix Converts a rotation vector to an axis-angle vector Use the inverse distance weighted method to perform a weighted average of the translation vector and the rotation vector represented by the axis angle to obtain the initial translation of the calculation point and the initial rotation vector expressed in axis angle Among them, w i is the weight corresponding to the i nearest calculation points, which is calculated as follows: Among them, d i is the distance from the i-th nearest calculation point to this calculation point; The initial rotation vector Convert to the initial rotation matrix 5. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 1 is characterized in that: The step S5 comprises: The initial rotation vector Convert to the initial rotation matrix The point cloud block A corresponding to the calculated point is transformed to the new position A by rotation and translation. * , The point cloud blocks are again aligned using the average domain vector algorithm to obtain a new transformation matrix M2, which is then interpreted into a new translation vector and rotation matrix. The translation vector and rotation matrix obtained in the previous step are superimposed to obtain the current total deformation.

6. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 1 is characterized in that: The step S6 comprises: Repeat steps S4 and S5. When the displacement of a certain calculation point is less than the specified threshold after a certain registration, assuming that the current calculation point spacing is d, the calculation point d / 2 is considered to be completed. When the calculation point spacing is reduced next time, the current area will no longer take new calculation points, but directly use the calculation points of the previous round. At the same time, when the calculation point spacing is reduced to the specified threshold, the calculation is considered to be completed after the current round ends.

7. The method for non-contact monitoring of three-dimensional overall deformation of landslide with millimeter-level accuracy according to claim 6 is characterized in that: After step S6, the method further includes: According to the transformation matrix corresponding to all the calculation points in the last round of calculation, the displacement vector corresponding to the calculation point is interpreted The initial displacement vector of the current calculation round Superposition, get the total displacement vector of the calculation point Thus, the displacement deformation field of the entire calculation area can be obtained.

8. A non-contact monitoring system for three-dimensional overall deformation of landslides with millimeter-level accuracy, characterized in that: The system is applied to the method according to any one of claims 1 to 7, and the system comprises: Point cloud data acquisition and feature extraction module: obtain two phases of surface three-dimensional deformation point cloud data for deformation interpretation and extract overall features; Parameter optimization module: Based on the overall features of the extracted point cloud and the trained neural network model, the Bayesian optimization algorithm is used to find the optimal initial calculation point spacing, minimum point cloud search radius, maximum corresponding point search distance, point cloud included point number threshold and root mean square error threshold; Initial calculation and registration module: according to the initial calculation point spacing, the calculation points are obtained, and the initial point cloud is divided into blocks, and the average domain vector algorithm is used for preliminary registration to obtain the initial transformation matrix, and the translation vector and rotation matrix are parsed; Refinement of calculation points and weighted average module: reduce the distance between calculation points, obtain new calculation points, use the inverse distance weight method to perform weighted average on the rotation vector and translation vector, and obtain the initial transformation parameters of the new calculation points; Point cloud block transformation and re-registration module: rotate and translate the point cloud blocks to new positions according to the transformation parameters of the newly calculated points, re-register, obtain the new transformation matrix, and interpret the new translation vector and rotation matrix; Convergence judgment and calculation control module: If the displacement of the calculation point is less than the specified threshold, the regional calculation is considered to be completed, and the distance between the calculation points is gradually reduced until the distance between the calculation points reaches the minimum value, then the overall calculation is considered to be completed; Final deformation calculation module: According to all the calculation points in the last round of calculation and their corresponding transformation matrices, the displacement vector corresponding to the calculation point is interpreted and superimposed with the initial displacement vector of the current calculation round to obtain the total displacement vector of the calculation point; thus, the displacement deformation field of the entire calculation area can be obtained.

9. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a non-contact monitoring method for three-dimensional overall deformation of a landslide with millimeter-level accuracy as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a non-contact monitoring method for three-dimensional overall deformation of a landslide with millimeter-level accuracy as claimed in any one of claims 1 to 7.

Images