Point cloud model optimization method and base station point selection method based on point cloud model feature extraction

Abstract

The invention relates to the technical field of spatial modeling and three-dimensional point cloud data processing, in particular to a point cloud model optimization method and a base station point selection method based on point cloud model feature extraction. The method comprises the steps of firstly performing down-sampling on a point cloud model, and then for each point, acquiring a unit normal vector of a fitting plane of the point and a neighborhood point of the point as a unit normal vector of the point; performing normal vector optimization on each point in the simplified point cloud; and calculating the feature measurement of each point in combination with the optimized unit normal vector of each point in the simplified point cloud, and selecting the points with the feature measurement in a set feature interval to form a feature point cloud. According to the method, data simplification of the point cloud model is realized under the condition that the original features of the point cloud model are reserved.

Description

Technical Field

[0001] The present invention relates to the technical field of space modeling and three-dimensional point cloud data processing, in particular to a point cloud model optimization method and a base station selection method based on point cloud model feature extraction. Background Art

[0002] A point cloud model is a set of spatial points generated by laser scanning, photogrammetry or other 3D measurement technologies. These points have specific coordinates (X, Y, Z) in 3D space and may contain other attribute information such as color, intensity, etc. Point cloud models can accurately represent the geometry of objects or scenes and are widely used in architecture, engineering, geographic information systems (GIS), reverse engineering, and cultural heritage protection. Since point cloud data contains a large amount of 3D information, specific software tools and algorithms are usually required to process and analyze this data. Through point cloud models, users can perform operations such as visualization, measurement, modeling, feature extraction, etc., thereby providing support for subsequent design, analysis, and decision-making. The advantages of point cloud models are their high accuracy and rich information. However, due to the large amount of data, processing and storage also face certain challenges.

[0003] Point cloud scanning base stations refer to fixed positions set during laser scanning or 3D point cloud acquisition. These base stations serve as reference points for scanning equipment to ensure the accuracy and consistency of scanned data. Each base station collects point cloud data of the surrounding environment during the scanning process to form a local 3D model. By performing multiple scans at different base stations and aligning and merging these scan results, a more complete and detailed 3D point cloud model can be generated. The selection of base stations is usually based on the visibility, scanning range, and occlusion of the scene to ensure that all important details and features are covered. In practical applications, the precise positioning and data processing of base stations are crucial to improving the quality of point cloud models and the reliability of subsequent analysis.

[0004] The selection of traditional point cloud data scanning base stations mainly relies on human judgment, randomly selecting base stations and arranging base stations. This method of selecting base stations has not been tested by computer simulation and is too arbitrary. Some areas may not be fully scanned, resulting in missing and incomplete point cloud data. At the same time, if the selected site angle is not appropriate, the quality of the point cloud data will be reduced and the data overlap will be insufficient. Point clouds with low repetitive data scanned at different sites will also cause seams or discontinuities when merging point clouds. Sometimes people set too many base stations to avoid these problems, but in essence, they still rely on human judgment to select base stations. Setting too many base stations will also cause unnecessary waste of resources.

[0005] BIM (Building Information Modeling) is a digital building design and management method that integrates the building's geometric information, physical properties, functional data, and related time and cost information by creating a three-dimensional model of the building. The BIM model is not just a three-dimensional visualization graphic, it also contains rich attribute information, such as materials, structures, equipment, and construction stages, and can support all stages of the building life cycle, including design, construction, operation, and maintenance. The core advantage of BIM lies in its collaborative work ability. Multiple professional teams can collaborate in real time on the same model, discover and resolve design conflicts in a timely manner, and improve work efficiency and accuracy. In addition, BIM can also perform simulation analysis, such as energy consumption analysis, structural analysis, and construction progress simulation, to help decision makers make more scientific choices. With the continuous advancement of the construction industry's digital transformation, BIM is gradually becoming a standard tool for modern building design and management. However, the conversion of BIM models into point cloud models often results in huge data and extremely low processing efficiency. Summary of the invention

[0006] In order to overcome the defects of low computational efficiency and disordered normal vectors of huge point cloud models in the above-mentioned prior art, the present invention proposes a point cloud model optimization method, which can correct the normal vectors of the point cloud model, reduce the data volume and data errors, and thus improve the computing efficiency of point cloud.

[0007] In order to overcome the problem of redundant base station settings or unreasonable distribution, the present invention also proposes a base station site selection method based on point cloud model feature extraction, which uses the BIM model to predict in advance the optimal site for point cloud collection, effectively solving the problem of redundant base station settings and missing and incomplete point cloud data.

[0008] A point cloud model optimization method proposed in the present invention comprises the following steps:

[0009] S1, downsample the point cloud model to obtain a simplified point cloud;

[0010] S2. For each point in the simplified point cloud, obtain the unit normal vector of the fitting plane between the point and its neighboring points as the unit normal vector of the point;

[0011] S3. Optimize the normal vector of each point in the simplified point cloud. The optimization goal is:

[0012]

[0013]

[0014] Among them, w q (n i , n j) represents the simplified point cloud midpoint p i and point p j The weight in the normal domain, w d (p i , p j ) represents the simplified point cloud midpoint p i and point p j Weight in the spatial domain; n i For point p i The unit normal vector, n j For point p j The unit normal vector of N(i) is the point p in the simplified point cloud. i The neighborhood point set of ; min means taking the minimum value; σ s is the Gaussian bandwidth of the set space domain, σ n is the Gaussian bandwidth of the set normal domain; exp is the exponential function;

[0015] S4. Calculate the feature measure of each point in the simplified point cloud based on the optimized unit normal vector of each point, and select points whose feature measure is in the set feature interval to form a feature point cloud.

[0016] Preferably, in step S4, the simplified point p in the point cloud i The characteristic measure of is obtained as follows:

[0017] First, construct the points p i The neighborhood point tensor voting matrix and the neighborhood normal vector tensor voting matrix

[0018]

[0019] Among them, η(p i , p j , n j ) is the attenuation coefficient, N′(i) represents the point p i The set of neighboring points of is the center point of the neighborhood corresponding to N′(i); the superscript T indicates the matrix transpose; p j Indicates p i Neighborhood points; n j For point p j The unit normal vector of ; the unit normal vector is expressed as a three-dimensional column vector;

[0020] Then the neighborhood point tensor voting matrix and the neighborhood normal vector tensor voting matrix Perform feature decomposition and extract and The three-dimensional features of The three-dimensional eigenvalues ​​of and and make The three-dimensional eigenvalues ​​of and and

[0021] Calculate point p i The characteristic measure f i :

[0022]

[0023] in, For point p i The neighborhood feature degree of Represents point p i The neighborhood point characteristic degree of the neighborhood normal vector.

[0024] Preferably, the attenuation coefficient η(p i , p j , n j ) is calculated as:

[0025]

[0026] Among them, point p j For point p i Neighborhood point, R is point p i Point to point p j The unit vector, n j Represents point p j The unit normal vector of s is the Gaussian bandwidth; ||.||2 represents the second norm.

[0027] Preferably, point p i The neighborhood point set N(i) coincides with N′(i).

[0028] Preferably, the method for downsampling the point cloud model in S1 is: using voxel grids to segment the point cloud model, and using the voxel grids where sampling points exist as mass points to construct a simplified point cloud.

[0029] Preferably, point p i The neighborhood point set N(i) corresponds to the neighborhood of point p i is a spherical domain with r1 as the radius at the center; the side length of the voxel grid of the segmented point cloud model is r, and r≤r1≤2r is set.

[0030] The present invention proposes a method for selecting a base station point based on feature extraction of a point cloud model, comprising the following steps:

[0031] St1. Build the BIM model of the target area and convert the BIM model into a point cloud model;

[0032] St2. Set a feature interval, and execute the point cloud model optimization method according to any one of claims 1 to 6 on the point cloud model to obtain a feature point cloud;

[0033] St3, set the number of clusters and cluster the feature point cloud;

[0034] St4. After the clustering is completed, the center of each cluster is used as the base site.

[0035] Preferably, the upper limit value of the characteristic interval is in the range of [0.3, 0.5], and the lower limit value of the characteristic interval is in the range of [0.01, 0.1].

[0036] The present invention proposes a target point selection system based on point cloud model feature extraction, comprising:

[0037] A point cloud acquisition module is used to connect an external device to acquire a point cloud model; the external device is a modeling device or a three-dimensional scanning device;

[0038] A point cloud optimization module stores a computer program, which is used to implement the point cloud model optimization method when executed to obtain a feature point cloud;

[0039] The clustering module is used to cluster the feature point cloud. The initial center point of the clustering is the feature point extracted from the feature point cloud. When the clustering module outputs the center point of each cluster as the target point.

[0040] The present invention provides a storage medium storing a computer program, which is used to implement the point cloud model optimization method when executed.

[0041] The advantages of the present invention are:

[0042] (1) The present invention proposes a point cloud model optimization method, which first simplifies the point cloud by downsampling, and then extracts feature points by combining the fitting plane and normal vector to construct a feature point cloud, thereby greatly simplifying the point cloud density and retaining the point cloud features. The data simplification of the point cloud model is achieved while retaining the original features of the point cloud model.

[0043] (2) In the present invention, the feature point extraction process relies on the normal vector optimization to calculate the feature measure of the data point. The normal vector optimization process simultaneously considers the normal domain and the space domain, so that the feature measure of the data point takes into account multiple features, thereby ensuring the comprehensive screening of high-characteristic data points.

[0044] (3) In the present invention, the feature measure is calculated by combining the eigenvalues ​​of the double voting matrix, and the domain tensor and the neighborhood normal vector are considered, so that the selected feature points can cover the neighborhood, thereby improving the representativeness of the feature points. In the present invention, the voxel grid method is used to downsample the point cloud model, so that the points in the simplified point cloud are more spatially representative, and the data points are more evenly distributed, which is convenient for subsequent calculation and processing.

[0045] (4) The method for selecting base station points based on feature extraction of point cloud models proposed in the present invention first obtains the BIM model of the target and converts it into a point cloud model, then optimizes the point cloud model, extracts simplified point clouds and feature point clouds, and then clusters the feature point clouds. The number of clusters and the location of each cluster can be used as the basis for selecting base station points. The present invention proposes to use the BIM model to predict the base station points in advance, which can reasonably reduce the workload of setting up on-site base stations and collecting point cloud data.

[0046] (5) The present invention uses the BIM model to predict the scanning base station of the point cloud model in advance, and uses machines to replace manual setting of scanning points, which can effectively reduce the excessive setting of base stations and cause unnecessary waste of resources. Moreover, the machine replaces the manually determined scanning base station, which can comprehensively scan the entire building, and the scanned point cloud model will have a certain degree of repeatability, which greatly reduces the work of point cloud model registration. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flow chart of a point cloud model optimization method;

[0048] Figure 2 It is a flow chart of the characteristic measurement calculation method;

[0049] Figure 3 It is a flow chart of a base station selection method based on point cloud model feature extraction;

[0050] Figure 4 It is the BIM model drawing;

[0051] Figure 5 It is a point cloud model diagram;

[0052] Figure 6 is the feature point cloud;

[0053] Figure 7 Schematic diagram of feature point cloud clustering;

[0054] Figure 8 Some base points and feature point clouds. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Reference Figure 1 , a point cloud model optimization method proposed in this embodiment includes the following steps S1-S4.

[0057] S1, downsample the point cloud model to obtain a simplified point cloud;

[0058] In this step, the minimum bounding box that can enclose the point cloud model can be calculated first; the maximum coordinate value X of the minimum bounding box on the X axis is defined max , minimum coordinate value X min and side length L x ; The maximum coordinate value Y of the minimum bounding box on the Y axis max , minimum coordinate value Y min and side length L y ; The maximum coordinate value Z of the minimum bounding box on the Z axis max , minimum coordinate value Z min and side length L z ; L x =X max -X min ; L y =Y max -Y min ; L z =Z max -Z min ; Then get the mapping relationship between the point cloud model and the simplified point cloud:

[0059] (x,y,z)→(h x ,h y ,h z )

[0060]

[0061] Among them, (x, y, z) is a point in the point cloud model, (h x ,h y ,h z ) is the point (x, y, z) in the point cloud model corresponding to the point in the simplified model;

[0062] The length of the minimum bounding box of the simplified model on the X axis is denoted by D x , the length of the side on the Y axis is denoted as D y , the length of the side on the Z axis is denoted as D z;have:

[0063]

[0064]

[0065] Indicates rounding down. Indicates rounding up.

[0066] S2, calculate the normal vector of each sampling point in the simplified point cloud;

[0067] First, set the radius r1, which is greater than the side length r of the voxel grid. Specifically, r≤r1≤2r can be set; define the simplified point cloud with point p i The area with the center and radius r1 is point p i The first neighborhood of , simplifies the point p in the point cloud that is in the first neighborhood j The average point p i The first neighbor point of

[0068] Then the fitting point p i and all its first neighboring points p j plane, and obtain the unit normal vector of the plane as point p i Specifically, in this step, we first define the fitting point p i and all its first neighboring points p j The set of G(i) is denoted as G(i), let g i ∈G(i), calculate the center of mass and the covariance matrix M i , then the covariance matrix M i Perform feature decomposition to obtain M i The three-dimensional eigenvalues ​​of and and M i The three-dimensional feature vector of and

[0069]

[0070] make but The corresponding eigenvector That is point p i The normal vector of , taking the unit value of the normal vector, is the unit normal vector.

[0071] It is worth noting that the currently obtained unit normal vector is ambiguous, that is, only the straight line where the normal vector is located is obtained, but the direction of the end of the straight line is not determined as the direction of the normal vector; in this case, traverse the points in the simplified point cloud, specify the normal vector direction of the first point p1, and then calculate the normal vector direction of the first point p1 according to n. i·n i+1 The condition of ≤0 determines the direction of the unit normal vector of each point in the simplified point cloud in turn; n i For point p i The unit normal vector, n i+1 For point p i+1 Unit normal vector; this step can be specifically implemented using Kd-tre e The search algorithm traverses the simplified point cloud.

[0072] S3, optimize and simplify the unit normal vector of each point in the point cloud;

[0073] Point p i The optimization goal of the unit normal vector is:

[0074]

[0075] Among them, w q (n i , n j ) represents the simplified point cloud midpoint p i and point p j The weight in the normal domain, w d (p i , p j ) represents the simplified point cloud midpoint p i and point p j Weight in the spatial domain; n i For point p i The unit normal vector, n j For point p j The unit normal vector of N(i) is the point p in the simplified point cloud. i The first neighborhood point set of ; min means taking the minimum value; σ s is the Gaussian bandwidth of the set space domain, σ n is the Gaussian bandwidth of the normal domain, which can be set by s and σ s The value is r1; exp is the exponential function.

[0076] S4. Calculate the feature measure of each point in the simplified point cloud based on the optimized unit normal vector of each point, and select points whose feature measure is in the set feature interval to form a feature point cloud.

[0077] Specifically, the unit normal vector is a column vector representing the direction, with point p j As an example, the unit normal vector of j ={characteristic value of the angle with the x-axis; characteristic value of the angle with the y-axis; characteristic value of the angle with the z-axis} T .

[0078] Reference Figure 2 , S4 specifically includes the following sub-steps:

[0079] S41, constructing a neighborhood point tensor voting matrix;

[0080]

[0081] Among them, p i represents the i-th sampling point in the simplified point cloud, For p i The neighborhood point tensor voting matrix of i The point set in the second neighborhood divided by the center and radius r2; j Indicates p i Neighborhood points; n j For point p j The unit normal vector of For p i The center point of the second neighborhood of i , p j , n j ) is the attenuation coefficient;

[0082]

[0083] Where R is point p i Point to point p j The unit vector, n j Represents point p j The unit normal vector of s is the Gaussian bandwidth; ||.||2 represents the second norm; Represents point p i and p j The square of the distance.

[0084] S42, constructing a neighborhood normal vector tensor voting matrix;

[0085]

[0086] in, For point p i The neighborhood normal vector tensor voting matrix of n j For point p j The unit normal vector of .

[0087] S43. Extraction The characteristic value of

[0088] because are all symmetric semi-positive matrices, they can be eigendecomposed and expressed in terms of spectral components as

[0089]

[0090] in The matrices are The eigenvalues ​​and eigenvectors of are arranged in ascending order, that is, The matrices are The eigenvalues ​​and eigenvectors of , the same eigenvalues ​​are arranged in ascending order, that is

[0091] S44. Calculate point p based on the double tensor voting matrix i Feature measurement;

[0092]

[0093] in, For point p i The neighborhood feature degree of Represents point p i The neighborhood point characteristic degree of the neighborhood normal vector;

[0094] Calculate the point p i Feature measure f i Then compare it with the set threshold μ to extract f i Points with μ ≥ μ are regarded as feature points.

[0095] Generally, points on the corners of the point cloud model (such as corners, endpoints, and edges) will show more significant feature measures f i , so that it is easier to be judged as a feature point. In specific implementation, by setting a threshold, the ratio of feature points can also be controlled, so as to give priority to retaining feature points with larger feature measures, further simplifying the effect of point cloud.

[0096] In subsequent experiments, the feature measure f corresponding to the inflection point of the point cloud model i is 0.25; thus, in order to ensure that all feature points are screened, the feature measure f i The screening interval is set to [0.05, 0.4], which is equivalent to setting μ = 0.05; the setting of 0.4 can clean outliers.

[0097] Reference Figure 3 The present embodiment proposes a base station selection method based on point cloud model feature extraction, comprising the following steps:

[0098] St1, first build the building information model (BIM model) of the target area and convert the building information model into a point cloud model;

[0099] Specifically, in this step, you can import the BIM model into Revit 2020, install the Obj plug-in, and export the imported BIM model to an Obj model through the Obj plug-in; then import the Obj model into the software CloudCompare to convert it into a point cloud model in .pcd or .ply format. Both formats can be interpreted and read by Python, which is convenient for subsequent point cloud processing operations.

[0100] St2. Execute the above steps S1-S4 on the point cloud model to obtain feature point cloud.

[0101] In specific implementation, the feature point cloud can be selected in a targeted manner by setting the feature interval in step S4. For example, the points on the corners of the point cloud model will show more significant feature measurements, and the feature measurement corresponding to the corner is 0.25, so the feature interval can be set to [0.05, 0.4]; this feature interval can include most of the feature points in the simplified point cloud, and the points selected according to the set interval have a lower density in the simplified point cloud.

[0102] St3, set the number of clusters and cluster the feature point cloud. In this step, the K-Means clustering algorithm can be used.

[0103] St4. After the clustering is completed, the center of each cluster is used as the base site.

[0104] The above-mentioned base station selection method based on point cloud model feature extraction is verified in combination with specific embodiments below.

[0105] The BIM model constructed in this embodiment is as follows Figure 4 As shown, the corresponding point cloud model is as follows Figure 5 described.

[0106] Set the voxel grid side length r = 3, r1 = r2 = 5; set the spatial domain Gaussian bandwidth σ s and the normal domain Gaussian bandwidth σ n The first neighborhood radius r1 is used; that is, σ s =σ n =5.

[0107] Table 1. Feature measurement values ​​of each point

[0108]

[0109] It can be seen that the feature measures of the feature points (inflection points and edge points) expected to be selected in this embodiment are 0.251010 and 0.063336 respectively, and the feature measure of the plane point is 0.000005; therefore, in this embodiment, the feature interval of the feature point cloud screened in S4 is set to [0.05, 0.400], and the points with feature measures within the interval [0.05, 0.400] are extracted from the simplified point cloud to form Figure 6 The feature point cloud shown in the figure contains 6913 points, which can fully reflect the Figure 5 Point cloud model features shown.

[0110] After clustering the feature point cloud in this embodiment, the clusters of points in the feature point cloud are as follows: Figure 7 As shown, feature points of the same color belong to the same cluster.

[0111] In this embodiment, some base stations obtained after the feature point cloud clustering is completed are Figure 7 The corresponding relationship of the midpoint set A is as follows Figure 8 As shown in the figure, any point in the point cloud can be covered by two or more base stations, which ensures the uniformity and effectiveness of the base station layout.

[0112] Of course, it will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention. Any reference numeral in a claim should not be regarded as limiting the claim to which it relates.

[0113] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0114] The techniques, shapes, and structural parts not described in detail in the present invention are all well-known techniques.

Claims

1. A point cloud model optimization method, characterized in that: The following steps are involved: S1, downsample the point cloud model to obtain a simplified point cloud; S2. For each point in the simplified point cloud, obtain the unit normal vector of the fitting plane between the point and its neighboring points as the unit normal vector of the point; S3. Optimize the normal vector of each point in the simplified point cloud. The optimization goal is: Among them, w q (n i ,n j ) represents the simplified point cloud midpoint p i and point p j The weight in the normal domain, w d (p i ,p j ) represents the simplified point cloud midpoint p i and point p j Weight in the spatial domain; n i For point p i The unit normal vector, n j For point p j The unit normal vector of N(i) is the point p in the simplified point cloud. i The neighborhood point set of ; min means taking the minimum value; σ s is the Gaussian bandwidth of the set space domain, σ n is the Gaussian bandwidth of the set normal domain; exp is the exponential function; S4. Calculate the feature measure of each point in the simplified point cloud based on the optimized unit normal vector of each point, and select points whose feature measure is in the set feature interval to form a feature point cloud.

2. The point cloud model optimization method according to claim 1, characterized in that: In step S4, the simplified point cloud midpoint p i The characteristic measure of is obtained as follows: First, construct the points p i The neighborhood point tensor voting matrix and the neighborhood normal vector tensor voting matrix Among them, η(p i ,p j ,n j ) is the attenuation coefficient, N'(i) represents the point p i The set of neighboring points of is the center point of the neighborhood corresponding to N'(i); the superscript T indicates the matrix transpose; p j Indicates p i Neighborhood points; n j For point p j The unit normal vector of ; the unit normal vector is expressed as a three-dimensional column vector; Then the neighborhood point tensor voting matrix and the neighborhood normal vector tensor voting matrix Perform feature decomposition and extract and The three-dimensional features of The three-dimensional eigenvalues ​​of and and make The three-dimensional eigenvalues ​​of and and Calculate point p i The characteristic measure f i : in, For point p i The neighborhood feature degree of Represents point p i The neighborhood point characteristic degree of the neighborhood normal vector.

3. The point cloud model optimization method according to claim 2, characterized in that: Attenuation coefficient η(p i ,p i ,n j ) is calculated as: Among them, point p j For point p i Neighborhood point, R is point p i Point to point p j The unit vector, n j Represents point p j The unit normal vector of s is the Gaussian bandwidth; ||.||2 represents the second norm.

4. The point cloud model optimization method according to claim 1, characterized in that: Point p i The neighborhood point set N(i) coincides with N'(i).

5. The point cloud model optimization method according to claim 1, characterized in that: The method for downsampling the point cloud model in S1 is: using voxel grids to segment the point cloud model, and using the voxel grids with sampling points as mass points to construct a simplified point cloud.

6. The point cloud model optimization method according to claim 5, characterized in that: Point p i The neighborhood point set N(i) corresponds to the neighborhood of point p i is a spherical domain with the center and radius r1; The side length of the voxel grid of the segmented point cloud model is r, and r≤r1≤2r is set.

7. A method for selecting base point points based on point cloud model feature extraction using the point cloud model optimization method according to any one of claims 1 to 6, characterized in that: The following steps are involved: St1. Build the BIM model of the target area and convert the BIM model into a point cloud model; St2. Set a feature interval, and execute the point cloud model optimization method according to any one of claims 1 to 6 on the point cloud model to obtain a feature point cloud; St3, set the number of clusters and cluster the feature point cloud; St4. After the clustering is completed, the center of each cluster is used as the base site.

8. The base station selection method based on point cloud model feature extraction according to claim 7, characterized in that: The upper limit value of the characteristic interval ranges from [0.3, 0.5], and the lower limit value of the characteristic interval ranges from [0.01, 0.1].

9. A target point selection system based on point cloud model feature extraction, characterized in that: include: A point cloud acquisition module is used to connect an external device to acquire a point cloud model; the external device is a modeling device or a three-dimensional scanning device; A point cloud optimization module storing a computer program, which is used to implement the point cloud model optimization method according to any one of claims 1 to 6 to obtain a feature point cloud when the computer program is executed; The clustering module is used to cluster the feature point cloud, and the initial center point of the cluster is the feature point extracted from the feature point cloud; When the clustering module outputs clustering, the center point of each cluster is output as the target point.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed, it is used to implement the point cloud model optimization method according to any one of claims 1 to 6.

Images