Automatic positioning detection method and system for floating objects in tilted real-life model

Abstract

The present disclosure proposes a method and system for automatically positioning and detecting floating objects in an inclined real-life model, including the following steps: analyzing the acquired inclined real-life model to generate point cloud data; determining a finite variable cylindrical neighborhood to process the point cloud data, and performing rapid regional growth through seed points in the outer layer of the cylindrical neighborhood to divide the point cloud data into multiple clusters; screening the clusters according to the set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics; based on the clusters obtained after screening, extracting the geometric center of gravity of the clusters to mark the position of the floating objects, and realizing automatic positioning and detection of model floating objects. Point cloud clustering is performed through a finite variable cylindrical neighborhood to avoid the situation where floating objects are missed when the distance between the floating objects and the terrain and the surface of the object is too close, and using the seed points outside the neighborhood for growth can reduce the repeated search of neighboring points, improve the clustering efficiency, and improve the detection efficiency and accuracy of model floating objects.

Description

Technical Field

[0001] The present disclosure relates to the technical field related to three-dimensional solid model detection, and more specifically, to a method and system for automatically positioning and detecting floating objects of an inclined real-scene model. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The two-dimensional world is a flat world with only length and width, while the three-dimensional world is a three-dimensional world with length, width and height, describing the world in three dimensions. In the past, due to the limitations of technology, means, storage and computing power, most of the objective world could only be expressed based on a two-dimensional plane. With the rapid development of "air-ground-space" three-dimensional data acquisition technologies such as high-temporal and spatial resolution mapping satellites, oblique photography, lidar measurement and mobile measurement, as well as new technologies such as 5G networks, blockchain, cloud computing, big data and artificial intelligence, sufficient conditions have been met to describe and manage the real world through three-dimensional means.

[0004] As one of the important components of real-life 3D construction, 3D entity data plays an important role in urban and rural planning, land development and utilization, resource survey, environmental monitoring and other industries. However, many problems arise in the use of data, such as the quality inspection efficiency of model data. Currently, a large amount of 3D entity data has put forward new requirements for visual quality inspection. Among them, the problem of automatic positioning and detection of model floating objects has attracted the attention and research of many relevant practitioners.

[0005] The inventors found that the three-dimensional model detection has the following deficiencies: In the past, in order to solve the problem of floating objects in the three-dimensional solid model, the model was mostly displayed by software, and then the floating objects were found and located through human-computer interaction. In fact, it was manual screening, which was inefficient. When the three-dimensional solid data was massive, it consumed a lot of manpower and time. Some automatic positioning detection methods for floating objects have the problem of low accuracy or even errors, and remain in the research stage, which is difficult to meet the actual industrial needs. Summary of the invention

[0006] In order to solve the above problems, the present disclosure proposes a method and system for automatically positioning and detecting floating objects in an inclined real-life model, which can realize automatic positioning and detection of model floating objects.

[0007] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0008] One or more embodiments provide a method for automatically positioning and detecting floating objects in a tilted real-life model, comprising the following steps:

[0009] Analyze the acquired tilted real scene model and generate point cloud data;

[0010] Determine a finite variable cylindrical neighborhood to process point cloud data, perform fast region growing through seed points on the outer layer of the cylindrical neighborhood, and divide the point cloud data into multiple clusters;

[0011] The clusters are screened according to the set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics;

[0012] According to the clusters obtained after screening, the geometric center of gravity of the clusters is extracted to mark the position of the floating objects, thereby realizing automatic positioning and detection of model floating objects.

[0013] One or more embodiments provide a system for automatically positioning and detecting floating objects of a tilted real-life model, including:

[0014] Point cloud generation module: configured to parse the acquired tilted real scene model and generate point cloud data;

[0015] Cluster segmentation module: configured to determine a finite variable cylindrical neighborhood for processing point cloud data, and to perform fast region growing through seed points in the outer layer of the cylindrical neighborhood to segment the point cloud data into multiple clusters;

[0016] Floating object identification module: configured to screen clusters according to set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics;

[0017] Floating object positioning module: It is configured to extract the geometric center of gravity of the clusters obtained after screening to mark the position of the floating objects, thereby realizing automatic positioning detection of the model floating objects.

[0018] An electronic device comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps described in the above method are completed.

[0019] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps described in the above method are completed.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The present invention performs point cloud clustering through a finite variable cylindrical neighborhood, which can avoid the situation where floating objects are missed when the distance between floating objects and the terrain or the surface of the object is too close, and uses the seed points outside the neighborhood for growth, which can reduce the repeated search of neighboring points, improve the clustering efficiency, and improve the detection efficiency and accuracy of the model floating objects.

[0022] The advantages of the present disclosure and the advantages of additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and the description thereof are used to explain the present disclosure but do not constitute a limitation of the present disclosure.

[0024] Figure 1 is a flow chart of the detection method of Example 1 of the present disclosure;

[0025] Figure 2 It is a block diagram of the detection system of Example 2 of the present disclosure. DETAILED DESCRIPTION

[0026] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments in the present disclosure and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0029] Example 1

[0030] In the technical solutions disclosed in one or more embodiments, Figure 1 As shown, a method for automatically positioning and detecting floating objects of an inclined real-life model comprises the following steps:

[0031] Step 1: Analyze the acquired tilted real scene model to generate point cloud data;

[0032] Step 2: Determine a finite variable cylindrical neighborhood to process the point cloud data, perform rapid region growing through seed points in the outer layer of the cylindrical neighborhood, and divide the point cloud data into multiple clusters;

[0033] Step 3: Screen the clusters according to the set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics;

[0034] Step 4: Based on the clusters obtained after screening, the geometric center of gravity of the clusters is extracted to mark the position of the floating objects, thereby realizing automatic positioning and detection of the model floating objects.

[0035] In this embodiment, point cloud clustering is performed through a finite variable cylindrical neighborhood, which can avoid the situation where floating objects are missed when they are too close to the terrain or the surface of the object, and using seed points outside the neighborhood for growth can reduce repeated searches of neighboring points and improve clustering efficiency.

[0036] In step 1, optionally, the tilted real scene model data is parsed to obtain triangulated network information, and the triangulated network vertices are used as point cloud data;

[0037] The tilted real scene model is specifically: a three-dimensional model constructed based on the tilted photogrammetry data. The tilted real scene model in step 1 is the model to be detected.

[0038] A further technical solution, in step 1, also includes the step of constructing a point cloud index. Specifically, the point cloud data is organized by constructing an octree index, including the following:

[0039] a1.1) Analyze the triangulated network information and retain the vertices as point cloud data;

[0040] a1.2) Obtain the minimum outer bounding box of the point cloud, set the segmentation depth, and establish an octree index structure;

[0041] Specifically, an octree index is established according to the determined parameters such as the octree segmentation granularity size, and corresponding node information is stored.

[0042] A further technical solution is that in step 1, after generating the point cloud data and before the cluster segmentation, an encryption step is also included: encrypting the point cloud data by local surface fitting interpolation.

[0043] The point cloud data is encrypted by local surface fitting interpolation. Specifically, in view of the point cloud sparse problem, the point cloud neighborhood points are used for surface fitting interpolation to complete the point cloud encryption, including the following steps:

[0044] a2.1) Obtaining the neighborhood point cloud of each point based on the neighborhood information of the point cloud data;

[0045] In this embodiment, the neighborhood point cloud of each point can be obtained according to the neighborhood information provided by the octree index;

[0046] a2.2) Use the moving least squares method to fit the neighborhood point cloud and obtain the local surface polynomial f;

[0047] a2.3) Calculate the coordinates of the points to be interpolated and perform point cloud encryption;

[0048] Specifically, the original point cloud organized by the octree is traversed. When the number of point clouds in the neighborhood of a certain point meets the requirement, the moving least squares method is used to fit the surface. According to the local polynomial f obtained by fitting, the coordinates of the interpolation point are calculated, as shown in formula (1):

[0049] {P(x,y,z)|z=f(x,y)∩x∈ε(x center )∩y∈ε(y center )}(1)

[0050] Among them, P(x, y, z) is the coordinate of the interpolation point, ε(x center ) and ε(y center ) are the neighborhood coordinate ranges of the center point.

[0051] In this embodiment, the triangulated network obtained by analyzing the panoramic data has uneven density. Therefore, the point cloud data composed of the triangulated network with uneven density also has uneven density. The point cloud at some locations is sparse and the spacing is abnormally large, which is not conducive to the setting of the clustering distance threshold. Therefore, the point cloud data is encrypted so that the point cloud can be better clustered.

[0052] In step 2, a finite variable cylindrical neighborhood is determined to process the point cloud data. Through the determined cylindrical neighborhood, seed points in the outer layer of the neighborhood are selected for rapid region growing to segment the point cloud data into multiple clusters.

[0053] In this embodiment, a finite variable cylindrical neighborhood is used to analyze the connectivity of the point cloud data, the cylindrical neighborhood is determined, and the detection accuracy is improved based on the cylindrical neighborhood.

[0054] Among them, a finite variable cylindrical neighborhood is used to analyze the connectivity of the point cloud data to determine the cylindrical neighborhood. Optionally, according to the local characteristics of the point cloud, the principal component analysis method is used to adaptively determine the direction and size of the cylindrical neighborhood to determine the cylindrical neighborhood. Specifically:

[0055] b1.1) Manually set the initial range parameters of the cylinder neighborhood, namely, the initial value of radius r, the initial value of height h, and the initial value of axial direction v;

[0056] b1.2) For any point x∈{x1,x2,…x m}, use the point cloud in the initial cylinder neighborhood to perform principal component analysis and calculate the eigenvalues ​​λ0, λ1, λ2 and eigenvectors V0, V1, V2;

[0057] b1.3) Take the eigenvector V0 corresponding to the minimum eigenvalue λ0 as the axial parameter v of the cylinder x , according to the surface curvature δ described by the eigenvalue, the cylinder neighborhood size parameter r of the point is determined x 、h x ;

[0058]

[0059] Among them, λ0<λ1<λ2 is the characteristic value describing the local distribution of the point cloud.

[0060] Based on the determined cylindrical neighborhood, the outer seed points of the determined cylindrical neighborhood are selected for fast region growing to obtain a cluster, comprising the following steps:

[0061] b2.1) Select any point from the point cloud data as the starting point and create a new cluster C;

[0062] b2.2) Set the growth queue S corresponding to the cluster C c1 and the seed queue S c2 , and add the starting point to the growth queue S c1 and the seed queue S c2 ;

[0063] b2.5) When S c2 When no new points are added, stop clustering C;

[0064] b2.3) Add the point cloud of the inner layer of the cylindrical neighborhood of the starting point to S c1 , the outer point cloud is added to S c2 ;

[0065] b2.4) Use S c2 The newly added point cloud is grown and its neighborhood points are processed according to b2.3);

[0066] b2.5) When S c2 When no new points are added, stop clustering C;

[0067] b2.6) Select a new point from the remaining points that have not been clustered, and repeat the above steps 2.1-2.5 until all points are clustered and the point cloud is finally divided into multiple clusters.

[0068] In the above steps, the seed point is the reference point for regional growth. Growth refers to the method of clustering the seed point and the points in its neighborhood with the seed point as the reference. After a point starts clustering, all its cylindrical neighborhood points are added to the cluster. The outer points are used as seed points to find neighboring points to expand the cluster. The inner points do not perform this operation.

[0069] In step 3, the floating object evaluation index features may include features such as relative ground height, bounding box volume, and neighborhood target distribution.

[0070] Optionally, the clusters are traversed to determine whether the clusters are floating objects based on the clusters' relative height to the ground, bounding box volume, neighborhood target distribution and other characteristics, and the locations of the floating objects are further determined by the geometric center of gravity, as follows:

[0071] c 1) Feature identification of clusters;

[0072] The features of the cluster body, such as relative ground height, bounding box volume, and neighborhood target distribution, are identified. The steps of feature identification are as follows:

[0073] c 1.1) Calculate the lowest point elevation of the cluster z The elevation difference between the grid value G at the corresponding position of the DEM is taken as the relative ground height h of the cluster. cluster ;

[0074] Among them, DEM is a digital elevation model, which realizes the digital simulation of the ground terrain through limited terrain elevation data (that is, the digital expression of the terrain surface morphology). It is a physical ground model that represents the ground elevation in the form of a set of ordered numerical arrays.

[0075] Specifically, the ground points in the original point cloud data are obtained through ground filtering to generate DEM, the clusters are traversed, the horizontal position x, y of the lowest point of the cluster is calculated, and the grid row and column numbers row and col in the corresponding DEM are obtained. The elevation difference between the lowest point elevation z and the grid value G is taken as the relative ground height h of the cluster. cluster .

[0076] h cluster =zG

[0077] c1.2) Calculate the horizontal center point P of the point cloud in the cluster c ;

[0078] c1.3) Using point cloud P i Construct the covariance matrix C from the horizontal x and y two-dimensional coordinates and solve it;

[0079] c1.4) Take the eigenvector corresponding to the maximum eigenvalue as the long side direction v of the cluster bounding box l , then find the value that is related to v l The vertical direction is the short side direction v of the cluster bounding box. s ;

[0080] c1.5) By projecting the point cloud to the long side and short side directions, the endpoints of the two projected line segments are calculated. The specific calculation method is to select any point from the line segment point cloud, and then calculate the distance from the remaining line segment points to this point, select the point with the longest distance as the first endpoint, and then calculate the distance from the remaining points to this endpoint, and select the point with the longest distance as the second endpoint;

[0081] c1.6) The projection endpoints of the long side are (a1 x , a1 y , a1 z ) and (a2 x , a2 y , a2 z ), the projection endpoints in the short side direction are (b1 x , b1 y , b1 z ) and (b2 x , b2 y , b2 z ), calculate the lengths a and b of the long and short sides, and the height h of the bounding box is determined by the maximum elevation z of the point cloud in the cluster max 、z mim The volume of the bounding box V is obtained by subtracting box ;

[0082]

[0083] Among them, C is the covariance matrix, there are n point clouds in the cluster point cloud set, P i is the coordinate of the i-th point, P c is the center point of the point cloud of the cluster. The projection endpoints of the point cloud in the horizontal long side direction are (a1 x , a1 y , a1 z ) and (a2 x , a2 y , a2 z ), the projection endpoints of the point cloud in the horizontal short side direction are (b1 x , b1 y , b1 z ) and (b2 x , b2 y , b2 z ).

[0084] In this embodiment, an accurate bounding box can be constructed through the above steps to reflect the actual volume of the cluster.

[0085] c1.7) Search for the number and area of ​​other targets within the set neighborhood centered on the cluster. If there are other targets with larger plane areas, they are considered as candidate targets for floating objects and Q(x) is marked as 1.

[0086] Floating objects usually present an isolated spatial distribution pattern. After the previous feature screening, the neighborhood range centered on the cluster is manually set to search for the number and area of ​​other targets within the range. If there are other targets with a larger plane area, the cluster is regarded as a candidate target for floating objects and marked as 1.

[0087]

[0088] Among them, Q(x) is the cluster label, 1 represents a floating object candidate, and 0 represents a non-floating object candidate. N is the number of neighborhood targets, S is the target area, and T is the area constraint threshold.

[0089] c1.8) Construct a floating object judgment model based on the relative ground height of the cluster, the bounding box volume, and the neighborhood target distribution, as follows;

[0090]

[0091] Among them, FO(x) is the mark for judging the cluster category, 1 represents floating objects, 0 represents other targets, and H T and V T They are the manually set relative ground height threshold and cluster bounding box volume threshold respectively.

[0092] The method for judging whether it is a model floating object based on the characteristics of the cluster's relative ground height, bounding box volume and neighborhood target distribution is as follows: the cluster's relative ground height is greater than the height setting value, and the bounding box volume is greater than the volume setting value, and the number and area of ​​other targets within the set neighborhood range centered on the cluster. If there are other targets with a plane area greater than the set area, it is judged that there are floating objects; otherwise, it is judged that there are no model floating objects.

[0093] In step 4, the model floating object is automatically positioned and detected by determining the position of the floating object;

[0094] For the cluster point cloud Pi marked as floating objects, the geometric center of gravity Pc3 is extracted from the n point clouds to mark the position of the floating objects, completing the automatic positioning detection of the floating objects in the inclined real scene model;

[0095]

[0096] In this embodiment, the relative ground height, bounding box volume, neighborhood target distribution and other features of the cluster are identified, and the geometric center of gravity is extracted, so that accurate positioning and detection of floating objects can be achieved.

[0097] Example 2

[0098] Based on Example 1, Figure 2As shown, this embodiment provides an automatic positioning detection system for floating objects of an inclined real-life model, comprising:

[0099] Point cloud generation module: configured to parse the acquired tilted real scene model and generate point cloud data;

[0100] Cluster segmentation module: configured to determine a finite variable cylindrical neighborhood for processing point cloud data, and to perform fast region growing through seed points in the outer layer of the cylindrical neighborhood to segment the point cloud data into multiple clusters;

[0101] Floating object identification module: configured to screen clusters according to set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics;

[0102] Floating object positioning module: It is configured to extract the geometric center of gravity of the clusters obtained after screening to mark the position of the floating objects, thereby realizing automatic positioning detection of the model floating objects.

[0103] Example 3

[0104] This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps described in the method of Embodiment 1 are completed.

[0105] Example 4

[0106] This embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps described in the method of Embodiment 1 are completed.

[0107] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

[0108] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A method for automatically positioning and detecting floating objects of an inclined real-life model, characterized in that: The steps include: Parse the tilted real scene model data to obtain triangulated network information, and use the triangulated network vertices as point cloud data; According to the local features of the point cloud, the principal component analysis method is used to adaptively determine the direction and size of the cylinder neighborhood, and the cylinder neighborhood is determined as follows: b1.1) Manually set the initial range parameters of the cylinder neighborhood, including the initial value of radius r, the initial value of height h, and the initial value of axial direction v; b1.2) For any point x in the point cloud data, use the point cloud in its initial cylindrical neighborhood to perform principal component analysis and calculate the eigenvalues ​​λ0, λ1, λ2 and eigenvectors V0, V1, V2; b1.3) Take the eigenvector V0 corresponding to the minimum eigenvalue λ0 as the axial parameter v of the cylinder x , according to the surface curvature δ described by the eigenvalue, the cylinder neighborhood size parameter r of the point is determined x 、h x ; Among them, λ0<λ1<λ2 is the characteristic value describing the local distribution of the point cloud; Based on a determined cylindrical neighborhood, an outer seed point of the determined cylindrical neighborhood is selected for fast region growing to obtain a clustering body, comprising the following steps: b2.1) Select any point from the point cloud data as the starting point and create a new cluster C; b2.2) Set the growth queue corresponding to cluster C and seed queue ; b2.3) Add the point cloud of the inner layer of the cylindrical neighborhood of the starting point to , the outer point cloud is added ; b2.4) Use Grow the newly added point cloud and process its neighborhood points according to step b2.3; b2.5) When When no new points are added, the clustering of cluster C is stopped; b2.6) Select a new point from the remaining points that have not been clustered, and repeat the above steps b2.1-b2.5 until all points are clustered and the point cloud is finally divided into multiple clusters; The clusters are screened according to the set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics; According to the clusters obtained after screening, the geometric center of gravity of the clusters is extracted to mark the position of the floating objects, thereby realizing automatic positioning and detection of model floating objects.

2. The method for automatically positioning and detecting floating objects of a tilted real-life model according to claim 1, characterized in that: Generating point cloud data also includes the step of constructing a point cloud index. Specifically, the point cloud data is organized by constructing an octree index, including the following: Parse the triangulated network information and retain the vertices as point cloud data; Get the minimum outer bounding box of the point cloud, set the segmentation depth, and establish the octree index structure.

3. The method for automatically positioning and detecting floating objects of a tilted real-life model according to claim 1, characterized in that: After acquiring the point cloud data and before the cluster segmentation, an encryption step is also included: using the point cloud neighborhood points to perform surface fitting interpolation to complete the point cloud encryption.

4. The method for automatically positioning and detecting floating objects of a tilted real-life model according to claim 1, characterized in that: The evaluation index characteristics of floating objects include relative height to the ground, bounding box volume and neighborhood target distribution characteristics; Whether it is a model floating object is determined based on the cluster's relative ground height, bounding box volume and neighborhood target distribution characteristics. Specifically, the cluster's relative ground height is greater than the height setting value, and the bounding box volume is greater than the volume setting value, and the number and area of ​​other targets within the set neighborhood range centered on the cluster are identified. If there are other targets with a plane area greater than the set area, it is determined that there are floating objects; otherwise, it is determined that there are no model floating objects.

5. The method for automatically locating and detecting floating objects of an inclined real-life model according to claim 4, characterized in that: Calculate the lowest point elevation of the cluster Grid value at the location corresponding to the DEM The elevation difference between them is taken as the relative ground height of the cluster; Among them, DEM is a digital elevation model, which realizes the digital simulation of ground terrain through limited terrain elevation data. It is a physical ground model that represents the ground elevation in the form of a set of ordered numerical arrays.

6. The method for automatically positioning and detecting floating objects of a tilted real-life model according to claim 4, characterized in that: The method for determining the volume of the bounding box is: Calculate the horizontal center point of the point cloud in the cluster; The covariance matrix is ​​constructed and solved using the horizontal two-dimensional coordinates of the point cloud data; The eigenvector corresponding to the maximum eigenvalue is used as the long side direction of the cluster bounding box, and then the direction perpendicular to the long side direction is obtained as the short side direction of the cluster bounding box; By projecting the point cloud into the long and short side directions, the endpoints of the two projected line segments are calculated; According to the endpoints of the projection line segment and the height of the bounding box, the volume of the bounding box is obtained; The height of the cluster bounding box is the difference between the maximum and minimum elevations of the point cloud in the cluster.

7. An automatic positioning and detection system for floating objects in an inclined real-life model, characterized in that: include: Point cloud generation module: configured to parse the tilted real scene model data to obtain triangulated network information, and use the triangulated network vertices as point cloud data; Cluster segmentation module: It is configured to adaptively determine the direction and size of the cylinder neighborhood based on the local features of the point cloud using the principal component analysis method to determine the cylinder neighborhood, specifically: b1.1) Manually set the initial range parameters of the cylinder neighborhood, including the initial value of radius r, the initial value of height h, and the initial value of axial direction v; b1.2) For any point x in the point cloud data, use the point cloud in its initial cylindrical neighborhood to perform principal component analysis and calculate the eigenvalues ​​λ0, λ1, λ2 and eigenvectors V0, V1, V2; b1.3) Take the eigenvector V0 corresponding to the minimum eigenvalue λ0 as the axial parameter v of the cylinder x , according to the surface curvature δ described by the eigenvalue, the cylinder neighborhood size parameter r of the point is determined x 、h x ; Among them, λ0<λ1<λ2 is the characteristic value describing the local distribution of the point cloud; Based on a determined cylindrical neighborhood, an outer seed point of the determined cylindrical neighborhood is selected for fast region growing to obtain a clustering body, comprising the following steps: b2.1) Select any point from the point cloud data as the starting point and create a new cluster C; b2.2) Set the growth queue corresponding to cluster C and seed queue ; b2.3) Add the point cloud of the inner layer of the cylindrical neighborhood of the starting point to , the outer point cloud is added ; b2.4) Use Grow the newly added point cloud and process its neighborhood points according to step b2.3; b2.5) When When no new points are added, the clustering of cluster C is stopped; b2.6) Select a new point from the remaining points that have not been clustered, and repeat the above steps b2.1-b2.5 until all points are clustered and the point cloud is finally divided into multiple clusters; Floating object identification module: configured to screen clusters according to set floating object evaluation index characteristics to obtain clusters that meet the floating object characteristics; Floating object positioning module: It is configured to extract the geometric center of gravity of the clusters obtained after screening to mark the position of the floating objects, thereby realizing automatic positioning detection of the model floating objects.

8. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps described in any one of the methods of claims 1 to 6 are completed.

9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps described in any one of claims 1 to 6.

Images