A method for improving the precision of a three-dimensional model laser scan
By analyzing point cloud datasets from benchmark and observation periods, strain matching point pairs of flexible structures are identified and corrected, solving the problem of 3D modeling errors caused by dynamic deformation of structures in existing technologies and achieving higher-precision 3D model generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KAIXIN (NANJING) TECH CO LTD
- Filing Date
- 2026-03-24
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies cannot effectively identify and correct global and systematic geometric errors introduced by dynamic deformation of flexible structures when performing 3D modeling. This results in the destruction of the geometric integrity of the 3D model and an inability to reflect the true state of the structure at any single moment.
By acquiring point cloud datasets from the baseline and observation periods, rigid matching point pairs are analyzed and identified to determine the degree of local strain. The point cloud datasets are then corrected using strain matching point pairs to form a three-dimensional model of the target object.
It improves the accuracy of 3D models, enabling them to more realistically reflect the geometric shape and structural features of the target object, eliminate deformation interference caused by thermal effects, and generate more accurate 3D models.
Smart Images
Figure CN121904286B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of 3D modeling, and specifically to a method for improving the accuracy of laser scanning of 3D models. Background Technology
[0002] 3D laser scanning technology, as a non-contact method for acquiring 3D information, is applied in the digital modeling of special equipment such as bridges and storage tanks. Existing point cloud data processing workflows are all based on a core theoretical foundation: the object being measured is a rigid body with a constant geometric shape throughout the entire scanning cycle. Based on this rigid body assumption, a stitching algorithm is used to achieve data unification by finding rigidity transformations between data from different stations.
[0003] However, the rigid body assumption mentioned above fails when performing long-term scans of flexible structures such as long-span bridges, which can last for several hours. In reality, structures undergo continuous and slow dynamic deformation due to environmental factors such as uneven solar thermal radiation and changes in traffic loads. Engineering practice shows that bridge structures can experience centimeter-level bending or deflection changes due to thermal effects within a single day. When existing stitching software based on the rigid body assumption processes this type of data that includes the structure's own time-varying deformation, the use of adjustment algorithms and Bundle Adjustment (BA) algorithms forces a deforming object to be fitted as a rigid body. This results in a systematic and unrealistic geometric distortion in the final global 3D model. The global geometric integrity of the model is thus compromised, failing to reflect the true state of the structure at any single moment and failing to meet the requirements for subsequent high-precision structural evaluation or deformation analysis. Currently, the industry lacks methods to effectively identify, quantify, and correct these global and systematic geometric errors introduced by the structure's own dynamic deformation within a single scan cycle.
[0004] In other words, existing technologies are not very effective at 3D modeling flexible structures (which are susceptible to deformation due to temperature). Summary of the Invention
[0005] The purpose of this invention is to provide a method for improving the accuracy of laser scanning of three-dimensional models, thereby solving the technical problem of poor modeling effect of existing technologies for three-dimensional modeling of flexible structures.
[0006] In a first aspect, the present invention provides a method for improving the accuracy of laser scanning of three-dimensional models, the method comprising:
[0007] The reference point cloud dataset and the observation point cloud dataset of the target object are obtained. The reference point cloud dataset is the point cloud dataset collected during the reference time period, and the observation point cloud dataset is the point cloud dataset collected during the observation time period. The thermal effect intensity corresponding to the reference time period is less than the thermal effect intensity corresponding to the observation time period.
[0008] The reference point cloud dataset and the observation point cloud dataset are analyzed to obtain multiple sets of rigid matching point pairs. Each set of rigid matching point pairs includes points from the corresponding reference point cloud dataset and points from the corresponding observation point cloud dataset. The rigid matching point pairs are used to indicate the position of the corresponding rigid body in the target object.
[0009] Based on multiple sets of rigid matching point pairs, the local strain degree of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs is analyzed, and multiple strain intensity values are obtained.
[0010] Based on the multiple strain intensity values, multiple strain matching point pairs are identified in the benchmark point cloud dataset and the observation point cloud dataset. Each strain matching point pair includes a point from the corresponding benchmark point cloud dataset and a point from the corresponding observation point cloud dataset. The feature difference value corresponding to the strain matching point is less than its corresponding difference threshold. The difference threshold is positively correlated with the corresponding strain intensity value. The feature difference value is used to indicate the degree of feature difference between the two point cloud points.
[0011] The corrected point cloud dataset is obtained by correcting the observation point cloud dataset based on multiple sets of strain matching point pairs. Then, a 3D model is formed based on the corrected point cloud dataset and the reference point cloud dataset to form a 3D model of the target object.
[0012] The analysis compares the baseline point cloud dataset with the observed point cloud dataset to obtain multiple sets of rigidly matched point pairs, including:
[0013] Spatial aggregation is performed on the benchmark point cloud dataset to obtain the benchmark structure point set, and spatial aggregation is performed on the observation point cloud dataset to obtain the observation structure point set;
[0014] Key point detection is performed on the baseline structural point set to obtain the baseline key point set, and key point detection is performed on the observed structural point set to obtain the observed key point set;
[0015] Analyze the benchmark key point set and the observation key point set to obtain multiple sets of rigidly matched point pairs.
[0016] The analysis benchmark key point set and the observation key point set are used to obtain multiple sets of rigidly matched point pairs, including:
[0017] Among the multiple benchmark keypoints included in the benchmark keypoint set, a nearest neighbor search is performed on the multiple observation keypoints included in the observation keypoint set according to the data features of each benchmark keypoint to determine the first observation keypoint and the second observation keypoint corresponding to each benchmark keypoint. The first observation keypoint is the observation keypoint with the smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints, and the second observation keypoint is the observation keypoint with the second smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints.
[0018] Among the multiple benchmark key points included in the benchmark key point set, the benchmark key points that meet the first preset condition and their corresponding first observation key points are determined as a set of initial screening matching point pairs to obtain multiple sets of initial screening matching point pairs. The first preset condition is: the ratio of the minimum feature distance corresponding to the benchmark key point to its corresponding second smallest feature distance is less than the ratio threshold.
[0019] Iterative search is performed based on multiple sets of initial screening matching point pairs to determine target transformation information and multiple sets of rigid matching point pairs matching the target transformation information. The iterative search operation includes:
[0020] Randomly select from multiple sets of initial screening matching point pairs to obtain a random point pair set, wherein the number of initial screening matching point pairs included in the random point pair set is a preset number;
[0021] The rigid transformation information of the corresponding iteration is calculated based on the initial screening matching point pairs included in the random point pair set, and the number of point pairs matching the rigid transformation information of the corresponding iteration in multiple initial screening matching point pairs is obtained to obtain the number of transformation matching point pairs of the corresponding iteration. The target transformation information is the rigid transformation information with the largest number of transformation matching point pairs in multiple iterations.
[0022] The iteration termination condition for the iterative search is: the number of iterations performed reaches a set number.
[0023] The method involves analyzing the local strain degree of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs based on multiple sets of rigid matching point pairs, and obtaining multiple strain intensity values, including:
[0024] Based on multiple sets of rigid matching point pairs, a subset to be analyzed is determined in the set of reference key points. The reference key points included in the subset to be analyzed are located in the spatial neighborhood of the reference key points included in the rigid matching point pairs, and the reference key points included in the subset to be analyzed are different from the reference key points included in any set of rigid matching point pairs.
[0025] Based on multiple sets of rigid matching point pairs, the deformation degree of each benchmark key point included in the subset to be analyzed is analyzed, and the deformation gradient estimation matrix and translation estimation vector of each benchmark key point in the subset to be analyzed are determined.
[0026] Based on the deformation gradient estimation matrix corresponding to each benchmark key point in the subset to be analyzed, the strain intensity value corresponding to each benchmark key point in the subset to be analyzed is determined, wherein the strain intensity value is used to indicate the degree of strain intensity in the local area where the corresponding benchmark key point is located.
[0027] The step of analyzing the deformation degree of each reference key point in the subset to be analyzed based on multiple sets of rigid matching point pairs, and determining the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed, includes:
[0028] In multiple sets of rigid matching point pairs, based on the spatial distance between the reference key points included in the rigid matching point pairs and the reference key points in the subset to be analyzed, the set of nearest neighbor point pairs associated with each reference key point in the subset to be analyzed is determined.
[0029] The minimum local affine transformation between each reference key point in the subset to be analyzed and the set of multiple observation key points included in its associated nearest neighbor pair is analyzed to determine the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed.
[0030] The strain intensity value is the Frobenius norm of the Green-Lagrange strain tensor of the corresponding deformation gradient estimation matrix.
[0031] The step of identifying multiple strain matching point pairs in the benchmark point cloud dataset and the observation point cloud dataset based on the multiple strain intensity values includes:
[0032] Affine transformation is performed on each reference key point in the subset to be analyzed based on the corresponding deformation gradient estimation matrix and translation estimation vector to obtain multiple transformed key points;
[0033] In multiple transformation key points, the degree of feature difference between each transformation key point and each of its associated observation key points to be matched is analyzed to obtain multiple feature difference values corresponding to each transformation key point. Among them, the observation key points to be matched are the observation key points in the observation key point set whose distance to the corresponding transformation key point is less than the distance threshold, and the multiple rigid matching point pairs do not include the observation key points to be matched.
[0034] Based on the strain intensity value corresponding to each benchmark key point and the multiple feature difference values corresponding to each benchmark key point, target key points are identified among the multiple benchmark key points included in the subset to be analyzed. The target key points and the observation key points to be matched with the corresponding minimum feature difference values are determined as a set of strain matching point pairs. The subset to be analyzed is supplemented based on the target key points. The minimum feature difference value corresponding to the target key point is less than its corresponding difference threshold, and the difference threshold is positively correlated with the corresponding strain intensity value.
[0035] The method further includes: ending the identification of rigid matching point pairs when all target key points included in the subset to be analyzed have been identified.
[0036] The process of correcting the observation point cloud dataset based on multiple sets of strain matching point pairs to obtain the corrected point cloud dataset includes:
[0037] In the set of observation key points, the deformation vectors of multiple strain matching point pairs adjacent to each observation key point are weighted and calculated to obtain the deformation amount corresponding to each observation key point. The calculation weight of the deformation vector is positively correlated with the strain intensity value of the corresponding strain matching point pair.
[0038] Among the multiple observation key points included in the observation key point set, the positions of the corresponding observation key points are corrected according to the corresponding deformation amount to obtain the corrected point cloud dataset.
[0039] The steps for obtaining multiple strain-matching point pairs near the observation key point include:
[0040] Calculate the distance between the observation key point and the observation key point to be matched in multiple strain matching point pairs to obtain the multiple point distances corresponding to the observation key point;
[0041] Strain matching point pairs whose corresponding point distance is less than the preset point distance threshold are identified as strain matching point pairs adjacent to the corresponding observation key point.
[0042] Secondly, the present invention also provides a system for improving the accuracy of laser scanning of three-dimensional models, the system comprising:
[0043] The dataset acquisition module is used to acquire the reference point cloud dataset and the observation point cloud dataset of the target object. The reference point cloud dataset is the point cloud dataset collected during the reference time period, and the observation point cloud dataset is the point cloud dataset collected during the observation time period. The thermal effect intensity corresponding to the reference time period is less than the thermal effect intensity corresponding to the observation time period.
[0044] The rigidity analysis module is used to analyze the benchmark point cloud dataset and the observation point cloud dataset to obtain multiple sets of rigid matching point pairs. Each set of rigid matching point pairs includes points from the corresponding benchmark point cloud dataset and points from the corresponding observation point cloud dataset. The rigid matching point pairs are used to indicate the position of the corresponding rigid body in the target object.
[0045] The strain analysis module is used to analyze the local strain of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs, based on multiple sets of rigid matching point pairs, and obtain multiple strain intensity values.
[0046] The strain point pair identification module is used to identify multiple strain matching point pairs in the benchmark point cloud dataset and the observation point cloud dataset based on the multiple strain intensity values. Each strain matching point pair includes a point in the corresponding benchmark point cloud dataset and a point in the corresponding observation point cloud dataset. The feature difference value corresponding to the strain matching point is less than its corresponding difference threshold. The difference threshold is positively correlated with the corresponding strain intensity value. The feature difference value is used to indicate the degree of feature difference between the two point cloud points.
[0047] The modeling module is used to correct the observation point cloud dataset based on multiple sets of strain matching point pairs, obtain the corrected point cloud dataset, and perform 3D modeling based on the corrected point cloud dataset and the reference point cloud dataset to form a 3D model of the target object.
[0048] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect.
[0049] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.
[0050] The present invention has the following beneficial effects:
[0051] This invention analyzes point cloud datasets collected at different stages of thermal effect intensity within a single scanning cycle to obtain multiple sets of rigid matching point pairs, i.e., to determine the point clouds in the point cloud dataset that indicate the rigid body part of the target object. Then, based on the multiple sets of rigid matching point pairs, it analyzes the local strain degree of points in the reference point cloud dataset that are not included in the rigid matching point pairs to determine the strain intensity of the point clouds in the point cloud dataset that indicate the non-rigid body part of the target object. Based on this, it selects reliable strain matching point pairs from the point cloud dataset to guide the point cloud position transformation. Finally, it uses the strain matching points to adaptively correct the position of the point clouds in the point cloud dataset that have deformed due to thermal effects, so as to dynamically compensate for the position deviation of each position of the target object caused by the influence of thermal effects. The reference point cloud dataset and the observed point cloud dataset complement each other, while eliminating the data deviation caused by the deformation of the target object during the point cloud dataset acquisition, and suppressing the interference of temporary thermal deformation as much as possible, so that the final generated three-dimensional model of the target object can more realistically reflect its inherent geometric shape and structural features. Attached Figure Description
[0052] 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.
[0053] Figure 1 This is a flowchart illustrating a method for improving the accuracy of laser scanning of three-dimensional models provided by the present invention;
[0054] Figure 2 This is a schematic diagram of the structure of a system for improving the accuracy of laser scanning of three-dimensional models provided by the present invention;
[0055] Figure 3 This is a schematic diagram of an electronic device provided by the present invention. Detailed Implementation
[0056] 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 method for improving the laser scanning accuracy of a three-dimensional model according to the present invention. In the following description, different embodiments or different embodiments 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.
[0057] 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.
[0058] The following description, in conjunction with the accompanying drawings, details a specific scheme for improving the laser scanning accuracy of a three-dimensional model provided by the present invention.
[0059] In one embodiment, the present invention provides a method for improving the accuracy of laser scanning of three-dimensional models, such as... Figure 1 As shown, the method includes:
[0060] Step S1: Obtain the reference point cloud dataset and the observation point cloud dataset of the target object.
[0061] The reference point cloud dataset is a point cloud dataset collected during the reference time period, and the observation point cloud dataset is a point cloud dataset collected during the observation time period. The thermal effect intensity corresponding to the reference time period is less than the thermal effect intensity corresponding to the observation time period.
[0062] In this invention, the target object can be understood as any object that is easily affected by environmental thermal effects and is prone to bending or deflection changes, such as a bridge.
[0063] In this invention, the average ambient temperature is used to characterize the thermal effect intensity of the target object during a corresponding time period. The fact that the thermal effect intensity corresponding to a reference time period is less than that corresponding to the observation time period can be understood as: the average ambient temperature monitored during the reference time period is less than the average ambient temperature monitored during the observation time period.
[0064] In this invention, the reference time period indicates the time period covered by one scanning job cycle, and the observation time period indicates the time period covered by another scanning job cycle. The specific cycle length of the scanning job cycle can be adaptively set according to the actual scanning time, and this invention does not limit it.
[0065] In one example, a baseline point cloud dataset can be collected during the time period corresponding to the design baseline temperature of the target object (minimizing the environmental thermal effects on the target object, such as 20 degrees Celsius). This is to minimize the interference of environmental thermal effects on the scanning operation during the baseline time period, thereby obtaining more rigid matching point pairs for subsequent use. This enhances the rigid constraints in the subsequent processing, thereby reducing the data errors introduced by the deformation correction process, and enabling the final modeling results to more realistically reflect the inherent geometric shape and structural characteristics of the target object.
[0066] Step S2: Analyze the benchmark point cloud dataset and the observation point cloud dataset to obtain multiple sets of rigidly matched point pairs.
[0067] Each pair of rigid matching points includes points from the corresponding reference point cloud dataset and points from the corresponding observation point cloud dataset. The rigid matching point pairs are used to indicate the position of the corresponding rigid body in the target object.
[0068] Specifically, the analysis of the baseline point cloud dataset and the observed point cloud dataset yields multiple sets of rigidly matched point pairs, including:
[0069] Spatial aggregation is performed on the benchmark point cloud dataset to obtain the benchmark structure point set, and spatial aggregation is performed on the observation point cloud dataset to obtain the observation structure point set;
[0070] Key point detection is performed on the baseline structural point set to obtain the baseline key point set, and key point detection is performed on the observed structural point set to obtain the observed key point set;
[0071] Analyze the benchmark key point set and the observation key point set to obtain multiple sets of rigidly matched point pairs.
[0072] In the above setup, by sequentially executing spatial aggregation and key point detection, the amount of point cloud data to be processed is adaptively reduced while preserving the inherent geometric shape and structural features of the target object, thereby improving modeling efficiency.
[0073] In applications, before performing spatial aggregation operations, the Statistical Outlier Removal (SOR) algorithm can be used to process the baseline point cloud dataset and the observation point cloud dataset separately to identify and remove isolated noise points on the structural surface of the target object that are caused by factors such as measurement errors or airborne dust.
[0074] In this invention, a voxel grid downsampling method is used (using a 5mm cubic grid for voxel separation, and replacing the point cloud with the centroids of all points within each voxel occupied by the point cloud) to spatially aggregate the reference point cloud dataset and the observed point cloud dataset after noise point filtering. This method divides the three-dimensional space containing the point cloud into a series of 5mm cubic grids (voxels) to significantly reduce the computational load of subsequent steps while preserving the key geometric details of the target object's structural surface.
[0075] For example, the Intrinsic Shape Signatures (ISS) algorithm can be used to perform key point detection operations on the reference structure point set and the observation structure point set. This allows for the extraction of a reference key point set from the reference structure point set, identifying geometrically prominent locations such as bolt heads, stiffening rib edges, and weld intersections in the surface structure of the target object. Similarly, it allows for the extraction of an observation key point set from the observation structure point set, identifying geometrically prominent locations such as bolt heads, stiffening rib edges, and weld intersections in the surface structure of the target object.
[0076] Furthermore, the analysis benchmark key point set and the observation key point set are used to obtain multiple sets of rigidly matched point pairs, including:
[0077] Among the multiple benchmark keypoints included in the benchmark keypoint set, a nearest neighbor search is performed on the multiple observation keypoints included in the observation keypoint set according to the data features of each benchmark keypoint to determine the first observation keypoint and the second observation keypoint corresponding to each benchmark keypoint. The first observation keypoint is the observation keypoint with the smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints, and the second observation keypoint is the observation keypoint with the second smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints.
[0078] Among the multiple benchmark key points included in the benchmark key point set, the benchmark key points that meet the first preset condition and their corresponding first observation key points are determined as a set of initial screening matching point pairs to obtain multiple sets of initial screening matching point pairs. The first preset condition is: the ratio of the minimum feature distance corresponding to the benchmark key point to its corresponding second smallest feature distance is less than the ratio threshold.
[0079] Iterative search is performed based on multiple sets of initial screening matching point pairs to determine target transformation information and multiple sets of rigid matching point pairs matching the target transformation information. The iterative search operation includes:
[0080] Randomly select from multiple sets of initial screening matching point pairs to obtain a random point pair set, wherein the number of initial screening matching point pairs included in the random point pair set is a preset number;
[0081] The rigid transformation information of the corresponding iteration is calculated based on the initial screening matching point pairs included in the random point pair set, and the number of point pairs matching the rigid transformation information of the corresponding iteration in multiple initial screening matching point pairs is obtained to obtain the number of transformation matching point pairs of the corresponding iteration. The target transformation information is the rigid transformation information with the largest number of transformation matching point pairs in multiple iterations.
[0082] The iteration termination condition for the iterative search is: the number of iterations performed reaches a set number.
[0083] In this invention, after obtaining the set of benchmark key points, the Fast Point Feature Histogram (FPFH) algorithm is used to obtain the feature vector of each benchmark key point in the set of benchmark key points. Similarly, after obtaining the set of observation key points, the FPFH algorithm is also used to obtain the feature vector of each observation key point in the set of observation key points.
[0084] Nearest neighbor search is performed based on the feature vectors of the baseline key point and the feature vectors of the observed key point. The feature distance can be understood as 1 minus the cosine similarity between the two corresponding feature vectors.
[0085] It should be noted that in practical applications, other local shape description algorithms (such as the SHOT (Signature of Histograms of Orientations) algorithm) can also be used to obtain the feature vectors of the reference key points and observation key points.
[0086] It should be noted that when the ratio of the minimum feature distance to the second smallest feature distance corresponding to the benchmark keypoint is less than the proportional threshold, it can be considered that the benchmark keypoint has an observation keypoint in the set of observation keypoints that is significantly similar to it in terms of geometric features (i.e., its corresponding first observation keypoint). The benchmark keypoint and its corresponding first observation keypoint not only indicate the same position on the surface structure of the target object, but this position also has a high probability of being the rigid body position (a position that is not easily deformed) of the target object. By integrating such benchmark keypoints and their corresponding first observation keypoints into a set of preliminary matching point pairs, a batch of point pairs (i.e., preliminary matching point pairs) with a high probability of indicating the rigid body position of the target object can be initially screened by utilizing the correlation between the benchmark keypoints and observation keypoints in terms of geometric features.
[0087] The above-mentioned ratio threshold is set based on empirical adaptability. In this invention, the ratio threshold is set to 0.8.
[0088] Since the initial screening of multiple sets of matching point pairs may still include incorrect matching point pairs (the point pair indicates a position that is not the rigid position of the target object), the present invention will also perform an iterative search on multiple sets of initial screening matching point pairs to filter out incorrect matching point pairs and retain several sets of initial screening matching point pairs that can accurately indicate the rigid position of the target object to form rigid matching point pairs.
[0089] Since mismatched point pairs contain incorrect point cloud transformation information, while different point pairs indicating the rigid body position of the target object contain similar point cloud transformation information, this invention randomly (e.g., using the Random Sample Consensus (RANSAC) algorithm) extracts a preset number (which can be set to 3 based on experience) of several initial screening matching point pairs in each iteration to calculate the rigid transformation information of the corresponding iteration. The general applicability of the rigid change information of the corresponding iteration is evaluated by the number of initial screening matching point pairs matched by the rigid change information of the corresponding iteration. In order to quickly identify multiple sets of rigid matching point pairs that can accurately indicate the rigid body position of the target object in the effective number of iterations.
[0090] Among them, the rigid transformation information can be understood as: the rigid transformation matrix calculated based on a preset number of initial screening matching point pairs randomly selected in the corresponding iteration ( The homogeneous matrix contains only rotation and translation components.
[0091] The rigid deformation information of the initial screening matching point pair can be understood as follows: after the benchmark key point in the initial screening matching point pair undergoes rigid transformation based on the corresponding rigid transformation matrix, the Euclidean distance between the obtained rigid transformation key point and the observed key point in the initial screening matching point pair is less than the preset interior point threshold (set to 10mm in this invention; in actual applications, it can be determined according to the nominal accuracy of the point cloud scanning device and the expected lower limit adaptability of the structural deformation of the target object).
[0092] Rigid matching point pairs are specifically the initial screening matching point pairs that match the transformation information of the target.
[0093] Step S3: Based on multiple sets of rigid matching point pairs, analyze the local strain degree of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs, and obtain multiple strain intensity values.
[0094] Furthermore, based on multiple sets of rigid matching point pairs, the local strain degree of points in the benchmark point cloud dataset not included in the rigid matching point pairs is analyzed to obtain multiple strain intensity values, including:
[0095] Based on multiple sets of rigid matching point pairs, a subset to be analyzed is determined in the set of reference key points. The reference key points included in the subset to be analyzed are located in the spatial neighborhood of the reference key points included in the rigid matching point pairs, and the reference key points included in the subset to be analyzed are different from the reference key points included in any set of rigid matching point pairs.
[0096] Based on multiple sets of rigid matching point pairs, the deformation degree of each benchmark key point included in the subset to be analyzed is analyzed, and the deformation gradient estimation matrix and translation estimation vector of each benchmark key point in the subset to be analyzed are determined.
[0097] Based on the deformation gradient estimation matrix corresponding to each benchmark key point in the subset to be analyzed, the strain intensity value corresponding to each benchmark key point in the subset to be analyzed is determined, wherein the strain intensity value is used to indicate the degree of strain intensity in the local area where the corresponding benchmark key point is located.
[0098] Specifically, the spatial neighborhood of the reference key point is a spherical region centered on the reference key point and with a set radius value (which can be set based on experience and adaptability).
[0099] Because the bending or torsional deformation of the main girder of a bridge can be effectively approximated by a linear affine transformation in a small local area (e.g., a bridge deck area of a few meters square), this invention uses rigid matching point pairs corresponding to rigid positions with low deformation as basic data to adaptively evaluate the deformation degree of reference key points in the vicinity of the rigid position. This allows for a relatively accurate quantitative estimation of the deformation degree of the reference key points in the vicinity of the rigid position, ensuring the accuracy of the calculated strain intensity value, deformation gradient estimation matrix, and translation estimation vector.
[0100] Furthermore, the step of analyzing the deformation degree of each reference key point in the subset to be analyzed based on multiple sets of rigid matching point pairs, and determining the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed, includes:
[0101] In multiple sets of rigid matching point pairs, based on the spatial distance between the reference key points included in the rigid matching point pairs and the reference key points in the subset to be analyzed, the set of nearest neighbor point pairs associated with each reference key point in the subset to be analyzed is determined.
[0102] The minimum local affine transformation between each reference key point in the subset to be analyzed and the set of multiple observation key points included in its associated nearest neighbor pair is analyzed to determine the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed.
[0103] The strain intensity value is the Frobenius norm of the Green-Lagrange strain tensor of the corresponding deformation gradient estimation matrix.
[0104] The spatial distance between the reference key points included in the rigid matching point pair and the reference key points in the subset to be analyzed is specifically the Euclidean distance between the reference key points included in the rigid matching point pair and the reference key points in the subset to be analyzed.
[0105] To avoid confusion, the reference key points included in the rigid matching point pair are defined as the first reference key points, and the reference key points included in the subset to be analyzed are defined as the second reference key points. In application, the Euclidean distance between each first reference key point and the same second reference key point can be calculated separately, and sorted in ascending order according to the Euclidean distance. Then, the rigid matching point pairs corresponding to the first few (in this invention, the number is set to 8) first reference key points in the sequence are determined as the nearest neighbor pair set of the corresponding second reference key points.
[0106] The process of analyzing the minimum local affine transformation between each reference key point in the subset to be analyzed and its associated set of nearest neighbor points, to determine the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed, can be understood as follows:
[0107] The optimal transformation solution is calculated in the nearest neighbor pair set associated with each reference key point in the subset to be analyzed. The optimal transformation solution is used to indicate the deformation gradient matrix solution and translation vector solution that minimizes the cumulative sum of the Euclidean distances between the reference key point and the observed key point of each rigid matching point pair in the nearest neighbor pair set.
[0108] For example, the function representing the minimum local affine transformation between the i-th reference keypoint in the subset to be analyzed and the set of multiple observed keypoints included in its associated nearest neighbor pair is expressed as:
[0109]
[0110] in, Let represent the deformation gradient estimation matrix of the i-th reference keypoint in the subset to be analyzed. Let represent the translation estimation vector of the i-th reference key point in the subset to be analyzed. Let represent the reference key point of the nth rigid matching point pair in the nearest neighbor pair set of the i-th reference key point in the subset to be analyzed. This represents the observation key point of the nth rigid matching point pair in the nearest neighbor pair set of the i-th benchmark key point in the subset to be analyzed.
[0111] The above function can be solved by singular value decomposition (SVD), and the specific solution process will not be described here.
[0112] Green-Lagrange strain tensor corresponding to the i-th reference key point in the subset to be analyzed It can be represented as:
[0113]
[0114] in, for identity matrix For matrix The transpose of .
[0115] Among them, the strain intensity value corresponding to the i-th reference key point in the subset to be analyzed It can be represented as:
[0116]
[0117] in, It is a matrix In the Line number The elements of the column.
[0118] Step S4: Based on the multiple strain intensity values, identify multiple pairs of strain matching points in the benchmark point cloud dataset and the observation point cloud dataset.
[0119] Each strain matching point pair includes points from the corresponding benchmark point cloud dataset and points from the corresponding observation point cloud dataset. The feature difference value corresponding to the strain matching point is less than its corresponding difference threshold. The difference threshold is positively correlated with the corresponding strain intensity value. The feature difference value is used to indicate the degree of feature difference between the two point cloud points.
[0120] Specifically, the step of identifying multiple strain matching point pairs in the benchmark point cloud dataset and the observation point cloud dataset based on the multiple strain intensity values includes:
[0121] Affine transformation is performed on each reference key point in the subset to be analyzed based on the corresponding deformation gradient estimation matrix and translation estimation vector to obtain multiple transformed key points;
[0122] In multiple transformation key points, the degree of feature difference between each transformation key point and each of its associated observation key points to be matched is analyzed to obtain multiple feature difference values corresponding to each transformation key point. Among them, the observation key points to be matched are the observation key points in the observation key point set whose distance to the corresponding transformation key point is less than the distance threshold, and the multiple rigid matching point pairs do not include the observation key points to be matched.
[0123] Based on the strain intensity value corresponding to each benchmark key point and the multiple feature difference values corresponding to each benchmark key point, target key points are identified among the multiple benchmark key points included in the subset to be analyzed. The target key points and the observation key points to be matched with the corresponding minimum feature difference values are determined as a set of strain matching point pairs. The subset to be analyzed is supplemented based on the target key points. The minimum feature difference value corresponding to the target key point is less than its corresponding difference threshold, and the difference threshold is positively correlated with the corresponding strain intensity value.
[0124] The method further includes: ending the identification of rigid matching point pairs when all target key points included in the subset to be analyzed have been identified.
[0125] The feature difference value is specifically the numerical value of 1 minus the cosine similarity of the feature vectors between the corresponding transformed key point and the observed key point to be matched.
[0126] In this invention, the difference threshold corresponding to the i-th benchmark key point in the subset to be analyzed is... It can be represented as:
[0127]
[0128] in, The tolerance for basic geometric deviation when the structure is without strain (can be set based on experience, and is set to 0.1 in this invention). It is a sensitivity coefficient for the effect of strain on geometric deviation, used to adjust the rate of increase of the difference threshold with the strain intensity value. In this invention, it is set to 1.5 (and The upper limit of the value is 0.5).
[0129] With the above settings, the tolerance of basic geometric deviation is adaptively corrected according to the strain intensity value corresponding to each first reference key point to obtain the difference threshold that adapts to its deformation degree, and thereby identify whether there are any observation key points to be matched that are less than the corresponding difference threshold. This can adaptively identify point pairs (i.e. strain matching point pairs) that can be used to accurately guide the point cloud transformation of the non-rigid position of the surface structure of the target object, taking into account dynamic deformation.
[0130] Transformation key points can be understood as the mapping points of the corresponding first reference key points during the observation period.
[0131] In this invention, if the Euclidean distance between the observation key point to be matched and the corresponding transformation key point is less than a distance threshold, it is determined that the observation key point to be matched is associated with the corresponding transformation key point.
[0132] It should be noted that in this invention, for each target key point identified, the subset to be analyzed will be supplemented according to the spatial neighborhood adaptability of the identified target key point. That is, in addition to the reference key points located in the spatial neighborhood of the first reference key point, the subset to be analyzed will also include the reference key points located in the spatial neighborhood of the target key point (which do not belong to rigid matching point pairs and strain matching point pairs).
[0133] The calculation of data such as the deformation gradient estimation matrix (completed based on the nearby rigid matching point pairs and strain matching point pairs), translation estimation vector, and strain intensity value of the reference key points located in the spatial neighborhood of the target key point in the subset to be analyzed, as well as the determination of whether they belong to the target key points, can all be completed by referring to the above process.
[0134] This enables diffusion from rigid locations to non-rigid locations, thereby comprehensively and fully collecting point pairs for guiding the point cloud transformation of non-rigid locations on the surface structure of the target object.
[0135] It should be noted that when a second benchmark key point has been processed through the above process and is not identified as a target key point (its minimum feature difference value is greater than or equal to its corresponding difference threshold), the second benchmark key point will be removed from the subset to be analyzed.
[0136] Step S5: Correct the observation point cloud dataset based on multiple sets of strain matching point pairs to obtain the corrected point cloud dataset, and perform 3D modeling based on the corrected point cloud dataset and the reference point cloud dataset to form a 3D model of the target object.
[0137] The step of obtaining the corrected point cloud dataset by correcting the observation point cloud dataset based on multiple sets of strain matching point pairs includes:
[0138] In the set of observation key points, the deformation vectors of multiple strain matching point pairs adjacent to each observation key point are weighted and calculated to obtain the deformation amount corresponding to each observation key point. The calculation weight of the deformation vector is positively correlated with the strain intensity value of the corresponding strain matching point pair.
[0139] Among the multiple observation key points included in the observation key point set, the positions of the corresponding observation key points are corrected according to the corresponding deformation amount to obtain the corrected point cloud dataset.
[0140] Specifically, the steps for obtaining multiple strain matching point pairs near the observation key point include:
[0141] Calculate the distance between the observation key point and the observation key point to be matched in multiple strain matching point pairs to obtain the multiple point distances corresponding to the observation key point;
[0142] Strain matching point pairs whose corresponding point distance is less than the preset point distance threshold are identified as strain matching point pairs adjacent to the corresponding observation key point.
[0143] The above-mentioned point distance threshold can be adaptively set according to actual needs, and the present invention does not limit it.
[0144] Based on the above settings, the correction accuracy of the corresponding observation key points can be adaptively adjusted according to the deformation degree of the adjacent strain matching point pairs, so as to complete the deformation correction point by point, realize the registration of the benchmark key point set and the observation key point set, ensure that the overall shape of the structure indicated by the benchmark point cloud dataset is close to the overall shape of the structure indicated by the corrected observation key point set, effectively suppress the structural deformation interference caused by environmental influence, and obtain a more accurate three-dimensional model.
[0145] The deformation vector indicates the vector change from the corresponding strain matching point pair to the corresponding target key point, which includes the key points to be matched.
[0146] The specific weight for the deformation vector is the ratio of the strain intensity value of the corresponding strain matching point pair to the total strain intensity value (the sum of the strain intensity values of multiple strain matching point pairs adjacent to the corresponding observation key point).
[0147] For example, the deformation vector of the k-th strain-matching point pair in a plurality of strain-matching point pairs can be expressed as:
[0148]
[0149] in, This represents the key observation point to be matched for the k-th strain matching point pair. The reference key points represent the strain matching point pairs.
[0150] For example, among multiple key observation points, the first Key observation points Key points of correction obtained after modification It can be represented as:
[0151]
[0152] in, Indicates the first Key observation points The corresponding deformation amount.
[0153] In another embodiment, the present invention also provides a system 200 for improving the accuracy of laser scanning of three-dimensional models, such as... Figure 2 As shown, the system 200 for improving the accuracy of laser scanning of three-dimensional models includes:
[0154] The dataset acquisition module 201 is used to acquire the reference point cloud dataset and the observation point cloud dataset of the target object. The reference point cloud dataset is the point cloud dataset collected during the reference time period, and the observation point cloud dataset is the point cloud dataset collected during the observation time period. The reference time period and the observation time period constitute a scanning operation cycle, and the thermal effect intensity corresponding to the reference time period is less than the thermal effect intensity corresponding to the observation time period.
[0155] The rigid analysis module 202 is used to analyze the reference point cloud dataset and the observation point cloud dataset to obtain multiple sets of rigid matching point pairs. Each set of rigid matching point pairs includes points from the corresponding reference point cloud dataset and points from the corresponding observation point cloud dataset. The rigid matching point pairs are used to indicate the position of the corresponding rigid body in the target object.
[0156] The strain analysis module 203 is used to analyze the local strain degree of points in the benchmark point cloud data that are not included in the rigid matching point pairs based on multiple sets of rigid matching point pairs, and obtain multiple strain intensity values.
[0157] The strain point pair identification module 204 is used to identify multiple strain matching point pairs in the reference point cloud dataset and the observation point cloud dataset based on the multiple strain intensity values. Each strain matching point pair includes a point in the corresponding reference point cloud dataset and a point in the corresponding observation point cloud dataset. The feature difference value corresponding to the strain matching point is less than its corresponding difference threshold. The difference threshold is positively correlated with the corresponding strain intensity value. The feature difference value is used to indicate the degree of feature difference between the two point cloud points.
[0158] Modeling module 205 is used to correct the observation point cloud dataset based on multiple sets of strain matching point pairs, obtain a corrected point cloud dataset, and perform 3D modeling based on the corrected point cloud dataset and the reference point cloud dataset to form a 3D model of the target object.
[0159] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the system for improving the laser scanning accuracy of a 3D model provided in the above embodiments and the method for improving the laser scanning accuracy of a 3D model belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.
[0160] The present invention also provides an electronic device. See also: Figure 3 The electronic device may include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and capable of running on the processor 301.
[0161] When program 3021 is executed by processor 301, it can achieve the following: Figure 1 Any steps in the corresponding method embodiments and the achievement of the same beneficial effects will not be repeated here.
[0162] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions, and the program can be stored in a readable medium.
[0163] The present invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described functions. Figure 1 Any step in the corresponding method embodiment can achieve the same technical effect, and will not be repeated here to avoid repetition.
[0164] The computer-readable storage medium of the present invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of a computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
[0165] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.
[0166] The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0167] Computer program code for performing the operations of this invention can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, as well as conventional procedural programming languages such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or terminal. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0168] The present invention also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to achieve the method for improving the laser scanning accuracy of a three-dimensional model provided in the above embodiments.
[0169] 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.
[0170] 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.
Claims
1. A method for improving the accuracy of laser scanning of three-dimensional models, characterized in that, The method includes: The reference point cloud dataset and the observation point cloud dataset of the target object are obtained. The reference point cloud dataset is the point cloud dataset collected during the reference time period, and the observation point cloud dataset is the point cloud dataset collected during the observation time period. The thermal effect intensity corresponding to the reference time period is less than the thermal effect intensity corresponding to the observation time period. The reference point cloud dataset and the observation point cloud dataset are analyzed to obtain multiple sets of rigid matching point pairs. Each set of rigid matching point pairs includes points from the corresponding reference point cloud dataset and points from the corresponding observation point cloud dataset. The rigid matching point pairs are used to indicate the position of the corresponding rigid body in the target object. Based on multiple sets of rigid matching point pairs, the local strain degree of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs is analyzed, and multiple strain intensity values are obtained. Based on the multiple strain intensity values, multiple strain matching point pairs are identified in the benchmark point cloud dataset and the observation point cloud dataset. Each strain matching point pair includes a point from the corresponding benchmark point cloud dataset and a point from the corresponding observation point cloud dataset. The feature difference value corresponding to the strain matching point is less than its corresponding difference threshold. The difference threshold is positively correlated with the corresponding strain intensity value. The feature difference value is used to indicate the degree of feature difference between the two point cloud points. The corrected point cloud dataset is obtained by correcting the observation point cloud dataset based on multiple sets of strain matching point pairs. Then, a 3D model is formed based on the corrected point cloud dataset and the reference point cloud dataset to form a 3D model of the target object.
2. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 1, characterized in that, The analysis compares the baseline point cloud dataset with the observed point cloud dataset to obtain multiple sets of rigidly matched point pairs, including: Spatial aggregation is performed on the benchmark point cloud dataset to obtain the benchmark structure point set, and spatial aggregation is performed on the observation point cloud dataset to obtain the observation structure point set; Key point detection is performed on the baseline structural point set to obtain the baseline key point set, and key point detection is performed on the observed structural point set to obtain the observed key point set; Analyze the benchmark key point set and the observation key point set to obtain multiple sets of rigidly matched point pairs.
3. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 2, characterized in that, The analysis benchmark key point set and the observation key point set are used to obtain multiple sets of rigidly matched point pairs, including: Among the multiple benchmark keypoints included in the benchmark keypoint set, a nearest neighbor search is performed on the multiple observation keypoints included in the observation keypoint set according to the data features of each benchmark keypoint to determine the first observation keypoint and the second observation keypoint corresponding to each benchmark keypoint. The first observation keypoint is the observation keypoint with the smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints, and the second observation keypoint is the observation keypoint with the second smallest feature distance to the corresponding benchmark keypoint among the multiple observation keypoints. Among the multiple benchmark key points included in the benchmark key point set, the benchmark key points that meet the first preset condition and their corresponding first observation key points are determined as a set of initial screening matching point pairs to obtain multiple sets of initial screening matching point pairs. The first preset condition is: the ratio of the minimum feature distance corresponding to the benchmark key point to its corresponding second smallest feature distance is less than the ratio threshold. Iterative search is performed based on multiple sets of initial screening matching point pairs to determine target transformation information and multiple sets of rigid matching point pairs matching the target transformation information. The iterative search operation includes: Randomly select from multiple sets of initial screening matching point pairs to obtain a random point pair set, wherein the number of initial screening matching point pairs included in the random point pair set is a preset number; The rigid transformation information of the corresponding iteration is calculated based on the initial screening matching point pairs included in the random point pair set, and the number of point pairs matching the rigid transformation information of the corresponding iteration in multiple initial screening matching point pairs is obtained to obtain the number of transformation matching point pairs of the corresponding iteration. The target transformation information is the rigid transformation information with the largest number of transformation matching point pairs in multiple iterations. The iteration termination condition for the iterative search is: the number of iterations performed reaches a set number.
4. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 2, characterized in that, The method involves analyzing the local strain degree of points in the benchmark point cloud dataset that are not included in the rigid matching point pairs based on multiple sets of rigid matching point pairs, and obtaining multiple strain intensity values, including: Based on multiple sets of rigid matching point pairs, a subset to be analyzed is determined in the set of reference key points. The reference key points included in the subset to be analyzed are located in the spatial neighborhood of the reference key points included in the rigid matching point pairs, and the reference key points included in the subset to be analyzed are different from the reference key points included in any set of rigid matching point pairs. Based on multiple sets of rigid matching point pairs, the deformation degree of each benchmark key point included in the subset to be analyzed is analyzed, and the deformation gradient estimation matrix and translation estimation vector of each benchmark key point in the subset to be analyzed are determined. Based on the deformation gradient estimation matrix corresponding to each benchmark key point in the subset to be analyzed, the strain intensity value corresponding to each benchmark key point in the subset to be analyzed is determined, wherein the strain intensity value is used to indicate the degree of strain intensity in the local area where the corresponding benchmark key point is located.
5. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 4, characterized in that, The step of analyzing the deformation degree of each reference key point in the subset to be analyzed based on multiple sets of rigid matching point pairs, and determining the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed, includes: In multiple sets of rigid matching point pairs, based on the spatial distance between the reference key points included in the rigid matching point pair and the reference key points in the subset to be analyzed, the set of nearest neighbor point pairs associated with each reference key point in the subset to be analyzed is determined. The minimum local affine transformation between each reference key point in the subset to be analyzed and the set of multiple observation key points included in its associated nearest neighbor pair is analyzed to determine the deformation gradient estimation matrix and translation estimation vector of each reference key point in the subset to be analyzed.
6. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 4, characterized in that, The strain intensity value is the Frobenius norm of the Green-Lagrange strain tensor of the corresponding deformation gradient estimation matrix.
7. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 4, characterized in that, The step of identifying multiple strain matching point pairs in the benchmark point cloud dataset and the observation point cloud dataset based on the multiple strain intensity values includes: Affine transformation is performed on each reference key point in the subset to be analyzed based on the corresponding deformation gradient estimation matrix and translation estimation vector to obtain multiple transformed key points; In multiple transformation key points, the degree of feature difference between each transformation key point and each of its associated observation key points to be matched is analyzed to obtain multiple feature difference values corresponding to each transformation key point. Among them, the observation key points to be matched are the observation key points in the observation key point set whose distance to the corresponding transformation key point is less than the distance threshold, and the multiple rigid matching point pairs do not include the observation key points to be matched. Based on the strain intensity value corresponding to each benchmark key point and the multiple feature difference values corresponding to each benchmark key point, target key points are identified among the multiple benchmark key points included in the subset to be analyzed. The target key points and the observation key points to be matched with the corresponding minimum feature difference values are determined as a set of strain matching point pairs. The subset to be analyzed is supplemented based on the target key points. The minimum feature difference value corresponding to the target key point is less than its corresponding difference threshold, and the difference threshold is positively correlated with the corresponding strain intensity value.
8. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 7, characterized in that, The method further includes: ending the identification of rigid matching point pairs when all target key points included in the subset to be analyzed have been identified.
9. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 2, characterized in that, The process of correcting the observation point cloud dataset based on multiple sets of strain matching point pairs to obtain the corrected point cloud dataset includes: In the set of observation key points, the deformation vectors of multiple strain matching point pairs adjacent to each observation key point are weighted and calculated to obtain the deformation amount corresponding to each observation key point. The calculation weight of the deformation vector is positively correlated with the strain intensity value of the corresponding strain matching point pair. Among the multiple observation key points included in the observation key point set, the positions of the corresponding observation key points are corrected according to the corresponding deformation amount to obtain the corrected point cloud dataset.
10. The method for improving the laser scanning accuracy of a three-dimensional model according to claim 9, characterized in that, The steps for obtaining multiple strain-matching point pairs near the observation key point include: Calculate the distance between the observation key point and the observation key point to be matched in multiple strain matching point pairs to obtain the multiple point distances corresponding to the observation key point; Strain matching point pairs whose corresponding point distance is less than the preset point distance threshold are identified as strain matching point pairs adjacent to the corresponding observation key point.