Point cloud simplification algorithm and device based on flatness division

Abstract

The present disclosure is about a point cloud simplification algorithm and device based on flatness division. The method includes: constructing a tree structure to manage point cloud data, and filtering the managed point cloud data; using a surface fitting algorithm to calculate the flatness of each point cloud in the point cloud data, and by comparing the flatness of each point cloud with a first preset threshold, dividing the point cloud data into flat area point cloud data and non-flat area point cloud data; first using a first simplification algorithm to perform a first point cloud simplification on the flat area point cloud data, and then using a second simplification algorithm to perform a second simplification, thereby obtaining flat area feature point data; dividing the non-flat area point cloud data into multiple grids, obtaining non-flat area feature point data; synthesizing the flat area feature point data and the non-flat area feature point data, and then filtering to obtain total feature point data.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing and application technology, and in particular to a point cloud simplification algorithm and device based on flatness division. Background Art

[0002] In recent years, surface model reconstruction from point cloud data has been widely used in reverse engineering, 3D city modeling, computer animation and 3D model reconstruction. The 3D point cloud model obtained by scanning contains millions of point cloud data, but in most cases, high-density point clouds are not necessary for subsequent modeling. There are a large number of redundant points in high-density point cloud data, which makes the calculation complexity very high and seriously limits the subsequent data processing tasks such as model storage, transmission, and drawing.

[0003] In the related technology, there are two main commonly used point cloud simplification methods, one is grid-based simplification, and the other is 3D point set-based simplification. Currently, the grid-based simplification method is more commonly used. This method first needs to construct the grid topology relationship and complete the simplification by evaluating the importance of the sampling points in the grid. However, the disadvantage of this method is that it needs to generate complex and cumbersome grids, and it needs to consume a huge amount of memory to maintain the grid topology information, and the simplification efficiency is low. The simplification methods based on 3D point sets generally include random sampling methods, bounding box methods, etc. The disadvantage of this method is that it often loses a lot of detailed features for models with complex geometric structures, and the accuracy is low.

[0004] Therefore, it is necessary to provide a new technical solution to improve one or more problems existing in the above solutions.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of the present disclosure is to provide a point cloud simplification algorithm and device based on flatness division, thereby overcoming one or more problems caused by limitations and defects of related technologies at least to a certain extent.

[0007] According to a first aspect of the present disclosure, a point cloud simplification algorithm based on flatness division is provided, comprising:

[0008] Constructing a tree structure to manage the point cloud data, and performing filtering processing on the managed point cloud data;

[0009] Calculating the flatness of each point cloud in the point cloud data by using a surface fitting algorithm, and dividing the point cloud data into flat area point cloud data and non-flat area point cloud data by comparing the flatness of each point cloud with a first preset threshold;

[0010] The flat area point cloud data is first simplified by a first simplification algorithm, and then simplified by a second simplification algorithm, so as to obtain flat area feature point data;

[0011] The non-flat area point cloud data is divided into a plurality of grids, and a normal vector angle between each point cloud in each grid and a plane where an adjacent point cloud in the same grid is located is obtained, and a preset feature value of each point cloud is obtained through the normal vector angle of each point cloud and a distance weight value between each point cloud and other point clouds in the same grid, and the preset feature value of each point cloud is compared with a second preset threshold value to simplify the non-flat area point cloud data, thereby obtaining feature point data of the non-flat area;

[0012] The flat area feature point data and the non-flat area feature point data are synthesized and then filtered to obtain total feature point data.

[0013] In the present disclosure, obtaining the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located includes:

[0014] Through the constructed tree structure, select point cloud n i (i∈1~N, N is the amount of data in the point cloud in the grid) perform K nearest neighbor point cloud searches to obtain the nearest neighbor point group m j (j∈1~K);

[0015] Construct a covariance matrix for adjacent point clouds, and solve the eigenvalue and eigenvector of the covariance matrix. The eigenvector corresponding to the minimum eigenvalue is the normal vector of the plane where the adjacent point cloud is located.

[0016] Calculate the normal vector angle λ between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located.

[0017] In the present disclosure, the preset feature value of each point cloud is obtained by the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid. k ,in .

[0018] In the present disclosure, the tree structure is a tree structure based on KD-Tree.

[0019] In the present disclosure, the managed point cloud data is filtered using a Gaussian smoothing filter.

[0020] In the present disclosure, the surface fitting algorithm is a least squares surface fitting algorithm.

[0021] In the present disclosure, the first simplification algorithm is a QEM three-dimensional model simplification algorithm, and the second simplification algorithm is a random downsampling algorithm or a K-means clustering simplification algorithm or a voxel grid simplification algorithm.

[0022] In the present disclosure, the bounding box method is used to divide the non-flat area point cloud data into multiple grids.

[0023] In the present disclosure, the flat region feature point data and the non-flat region feature point data are synthesized and then filtered using a statistical filter to obtain total feature point data.

[0024] According to a second aspect of an embodiment of the present disclosure, a point cloud simplification device based on flatness division is provided, comprising:

[0025] Constructing a filtering module, which is used to construct a tree structure to manage the point cloud data, and to perform filtering processing on the managed point cloud data;

[0026] A region division module, which uses a surface fitting algorithm to calculate the flatness of each point cloud in the point cloud data, and divides the point cloud data into flat region point cloud data and non-flat region point cloud data by comparing the flatness of each point cloud with a first preset threshold;

[0027] A flat area simplification module, used to perform a first point cloud simplification on the flat area point cloud data by using a first simplification algorithm, and then perform a second simplification on the flat area point cloud data by using a second simplification algorithm, so as to obtain flat area feature point data;

[0028] A non-flat area simplification module is used to divide the non-flat area point cloud data into multiple grids, obtain the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located, obtain the preset feature value of each point cloud through the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid, compare the preset feature value of each point cloud with a second preset threshold value to simplify the non-flat area point cloud data, so as to obtain the non-flat area feature point data;

[0029] The synthesis module is used to synthesize the flat area feature point data and the non-flat area feature point data and then perform filtering to obtain total feature point data.

[0030] The technical solution provided by the present disclosure may have the following beneficial effects:

[0031] In the present disclosure, through the above-mentioned point cloud simplification algorithm based on flatness division, on the one hand, this method divides the spatial area according to the flatness of the point cloud, and uses different processing algorithms for spaces with different characteristics, which is more targeted and integrates the advantages of two major types of point cloud simplification algorithms, ensuring accuracy while avoiding consuming a large amount of memory usage; on the other hand, this method will introduce a distance weighted grid feature point extraction algorithm for feature-rich non-flat area spaces, and the introduction of this algorithm can well retain strong feature points.

[0032] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The flowchart of the point cloud simplification algorithm based on flatness division in the exemplary embodiment of the present disclosure is schematically shown;

[0034] Figure 2 The schematic diagram of the structure of the point cloud simplification device based on flatness division in the exemplary embodiment of the present disclosure is shown schematically. DETAILED DESCRIPTION

[0035] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0036] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0037] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present invention, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.

[0038] In this example implementation, a point cloud simplification algorithm based on flatness division is first provided. Figure 1 As shown in , the method may include:

[0039] Step S101: constructing a tree structure to manage point cloud data, and performing filtering processing on the managed point cloud data;

[0040] Step S102: Calculating the flatness of each point cloud in the point cloud data by using a surface fitting algorithm, and dividing the point cloud data into flat area point cloud data and non-flat area point cloud data by comparing the flatness of each point cloud with a first preset threshold;

[0041] Step S103: the flat area point cloud data is first simplified using a first simplification algorithm, and then simplified using a second simplification algorithm, so as to obtain flat area feature point data;

[0042] Step S104: Divide the non-flat area point cloud data into a plurality of grids, obtain the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located, obtain the preset feature value of each point cloud through the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid, compare the preset feature value of each point cloud with a second preset threshold value to simplify the non-flat area point cloud data, thereby obtaining the non-flat area feature point data;

[0043] Step S105: synthesizing the flat area feature point data and the non-flat area feature point data and then filtering them to obtain total feature point data.

[0044] The above-mentioned point cloud simplification algorithm based on flatness division, on the one hand, divides the spatial area according to the flatness of the point cloud, and uses different processing algorithms for spaces with different characteristics, which is more targeted and integrates the advantages of two major types of point cloud simplification algorithms, ensuring accuracy while avoiding consuming a large amount of memory usage; on the other hand, this method will introduce a distance weighted grid feature point extraction algorithm for feature-rich non-flat area spaces, and the introduction of this algorithm can well retain strong feature points.

[0045] Next, we will refer to Figure 1 Each step of the above method in this example implementation is described in more detail.

[0046] In one embodiment, obtaining the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located includes:

[0047] Through the constructed tree structure, select point cloud n i (i∈1~N, N is the amount of data in the point cloud in the grid) perform K nearest neighbor point cloud searches to obtain the nearest neighbor point group m j (j∈1~K);

[0048] Construct a covariance matrix for adjacent point clouds, and solve the eigenvalue and eigenvector of the covariance matrix. The eigenvector corresponding to the minimum eigenvalue is the normal vector of the plane where the adjacent point cloud is located.

[0049] Calculate the normal vector angle λ between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located.

[0050] Specifically, for one of the point clouds, K neighboring points are searched to form a neighboring point group, the normal vectors of the planes where the neighboring point clouds are located are obtained by calculation, and the angle λ between the normal vector of the point cloud and the normal vector of the plane where the neighboring point clouds are located is calculated, wherein the normal vector of each point cloud has been known after step S102.

[0051] In one embodiment, the preset feature value of each point cloud is obtained by the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid. k ,in, .

[0052] In one embodiment, the tree structure is a tree structure based on KD-Tree (k-dimensional tree). Specifically, KD-Tree is a data structure for partitioning k-dimensional data space, mainly used for searching key data in multidimensional space (such as range search and nearest neighbor search), and KD tree is a special case of binary space partition tree.

[0053] In one embodiment, the managed point cloud data can be filtered using a Gaussian smoothing filter. Specifically, Gaussian filtering is a linear smoothing filter that is suitable for eliminating Gaussian noise and is widely used in the noise reduction process of image processing. In layman's terms, Gaussian filtering is the process of weighted averaging the entire image. The value of each pixel is obtained by weighted averaging itself and other pixel values ​​in the neighborhood. The specific operation of Gaussian filtering is: use a template (or convolution, mask) to scan each pixel in the image, and replace the value of the center pixel of the template with the weighted average grayscale value of the pixels in the neighborhood determined by the template.

[0054] In one embodiment, the surface fitting algorithm may be a least squares surface fitting algorithm. Specifically, the calculation result of the least squares surface fitting algorithm has the best accuracy.

[0055] In one embodiment, the first simplification algorithm may be a QEM (Quadric Error Mactrics) three-dimensional model simplification algorithm, and the second simplification algorithm may be a random downsampling algorithm or a K-means clustering simplification algorithm or a voxel grid simplification algorithm. Specifically, the second simplification algorithm is preferably a random downsampling method, which takes the shortest time and has the highest operating efficiency. The random downsampling method specifically sets a reasonable entropy threshold according to requirements, and performs sampling by comparing the entropy value of the point cloud with the entropy threshold.

[0056] In one embodiment, the non-flat area point cloud data can be divided into multiple grids using a bounding box method. Specifically, a bounding box is an algorithm for solving the optimal bounding space of a discrete point set. The basic idea is to use a geometric body (called a bounding box) with a slightly larger volume and simpler characteristics to approximately replace complex geometric objects; common bounding box algorithms include AABB bounding box, bounding sphere, directional bounding box OBB, and fixed directional convex hull FDH.

[0057] In one embodiment, the flat area feature point data and the non-flat area feature point data are synthesized and then filtered using a statistical filter to obtain the total feature point data. Specifically, the statistical filter measures the local statistical features of the image, and then uses its statistical value as the pixel value of the output image. It mainly eliminates noise or extracts texture features by calculating the statistical values ​​in a small neighboring area (within the window range), such as changes in its mean, median, maximum value, minimum value, etc. In this embodiment, after the flat area feature point data and the non-flat area feature point data are synthesized, it is only necessary to remove the obvious outliers in the feature point data, and the use of a statistical filter has a better effect on removing outliers and can better meet the removal requirements.

[0058] According to a second aspect of the embodiment of the present disclosure, referring to Figure 2 As shown in , a point cloud simplification device based on flatness division is provided, and the device may include:

[0059] Constructing a filtering module, which is used to construct a tree structure to manage the point cloud data, and to perform filtering processing on the managed point cloud data;

[0060] A region division module, which uses a surface fitting algorithm to calculate the flatness of each point cloud in the point cloud data, and divides the point cloud data into flat region point cloud data and non-flat region point cloud data by comparing the flatness of each point cloud with a first preset threshold;

[0061] A flat area simplification module, used to perform a first point cloud simplification on the flat area point cloud data by using a first simplification algorithm, and then perform a second simplification on the flat area point cloud data by using a second simplification algorithm, so as to obtain flat area feature point data;

[0062] A non-flat area simplification module is used to divide the non-flat area point cloud data into multiple grids, obtain the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located, obtain the preset feature value of each point cloud through the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid, compare the preset feature value of each point cloud with a second preset threshold value to simplify the non-flat area point cloud data, so as to obtain the non-flat area feature point data;

[0063] The synthesis module is used to synthesize the flat area feature point data and the non-flat area feature point data and then perform filtering to obtain total feature point data.

[0064] The above-mentioned point cloud simplification device based on flatness division, on the one hand, divides the spatial area according to the flatness of the point cloud, and uses different processing algorithms for spaces with different characteristics, which is more targeted and integrates the advantages of two major types of point cloud simplification algorithms to ensure accuracy while avoiding consuming a large amount of memory; on the other hand, a distance weighted-based grid feature point extraction algorithm is introduced for feature-rich non-flat area spaces. The introduction of this algorithm can well retain strong feature points.

[0065] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0066] It should be noted that although several modules or units of the device for executing actions are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of a module or unit described above can be further divided into multiple modules or units for concretization. The components displayed as modules or units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. Those of ordinary skill in the art can understand and implement it without paying creative work.

[0067] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0068] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0069] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A point cloud simplification method based on flatness division, characterized in that: include: Constructing a tree structure to manage the point cloud data, and performing filtering processing on the managed point cloud data; Calculating the flatness of each point cloud in the point cloud data by using a surface fitting algorithm, and dividing the point cloud data into flat area point cloud data and non-flat area point cloud data by comparing the flatness of each point cloud with a first preset threshold; The flat area point cloud data is first simplified by a first simplification algorithm, and then simplified by a second simplification algorithm, so as to obtain flat area feature point data; The non-flat area point cloud data is divided into a plurality of grids, and a normal vector angle between each point cloud in each grid and a plane where an adjacent point cloud in the same grid is located is obtained, and a preset feature value of each point cloud is obtained through the normal vector angle of each point cloud and a distance weight value between each point cloud and other point clouds in the same grid, and the preset feature value of each point cloud is compared with a second preset threshold value to simplify the non-flat area point cloud data, thereby obtaining feature point data of the non-flat area; The flat area feature point data and the non-flat area feature point data are synthesized and then filtered to obtain total feature point data; Among them, obtaining the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located includes: Through the constructed tree structure, select point cloud n i (i∈1~N, N is the amount of data in the point cloud in the grid) perform K neighbor point cloud searches to obtain the neighbor point group m j (j∈1~K); Construct a covariance matrix for adjacent point clouds, and solve the eigenvalue and eigenvector of the covariance matrix. The eigenvector corresponding to the minimum eigenvalue is the normal vector of the plane where the adjacent point cloud is located. Calculate the angle λ between the normal vector of each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located; Among them, for one of the point clouds, search for K neighboring points to form a neighboring point group, obtain the normal vector of the plane where the neighboring point cloud is located by calculation, and calculate the angle λ between the normal vector of the point cloud and the normal vector of the plane where the neighboring point cloud is located; The preset characteristic value κ of each point cloud is obtained by the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid. in, The first simplification algorithm is a QEM three-dimensional model simplification algorithm, and the second simplification algorithm is a random downsampling algorithm or a K-means clustering simplification algorithm or a voxel grid simplification algorithm.

2. The point cloud simplification method based on flatness division according to claim 1, characterized in that: The tree structure is a tree structure based on KD-Tree.

3. The point cloud simplification method based on flatness division according to claim 2, characterized in that: The managed point cloud data is filtered using a Gaussian smoothing filter.

4. The point cloud simplification method based on flatness division according to claim 3 is characterized in that: The surface fitting algorithm is a least squares surface fitting algorithm.

5. The point cloud simplification method based on flatness division according to claim 1, characterized in that: The point cloud data of the non-flat area is divided into a plurality of grids by using a bounding box method.

6. The point cloud simplification method based on flatness division according to claim 1, characterized in that: The flat region feature point data and the non-flat region feature point data are synthesized and then filtered using a statistical filter to obtain total feature point data.

7. A point cloud simplification device based on flatness division, characterized in that: include: Constructing a filtering module, which is used to construct a tree structure to manage the point cloud data, and to perform filtering processing on the managed point cloud data; A region division module, which uses a surface fitting algorithm to calculate the flatness of each point cloud in the point cloud data, and divides the point cloud data into flat region point cloud data and non-flat region point cloud data by comparing the flatness of each point cloud with a first preset threshold; A flat area simplification module, used for first simplifying the flat area point cloud data by a first simplification algorithm, and then simplifying the flat area point cloud data by a second simplification algorithm, so as to obtain flat area feature point data, wherein the first simplification algorithm is a QEM three-dimensional model simplification algorithm, and the second simplification algorithm is a random downsampling algorithm, a K-means clustering simplification algorithm, or a voxel grid simplification algorithm; The non-flat area simplification module is used to divide the non-flat area point cloud data into multiple grids, obtain the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located, obtain the preset feature value of each point cloud through the normal vector angle of each point cloud and the distance weight value between each point cloud and the remaining point clouds in the same grid, and compare the preset feature value of each point cloud with a second preset threshold to simplify the non-flat area point cloud data, thereby obtaining the non-flat area feature point data, wherein obtaining the normal vector angle between each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located includes: Through the constructed tree structure, select point cloud n i (i∈1~N, N is the amount of data in the point cloud in the grid) perform K neighbor point cloud searches to obtain the neighbor point group m j (j∈1~K); Construct a covariance matrix for adjacent point clouds, and solve the eigenvalue and eigenvector of the covariance matrix. The eigenvector corresponding to the minimum eigenvalue is the normal vector of the plane where the adjacent point cloud is located. Calculate the angle λ between the normal vector of each point cloud in each grid and the plane where the adjacent point cloud in the same grid is located; Among them, for one of the point clouds, search for K neighboring points to form a neighboring point group, obtain the normal vector of the plane where the neighboring point cloud is located by calculation, and calculate the angle λ between the normal vector of the point cloud and the normal vector of the plane where the neighboring point cloud is located; The preset characteristic value κ of each point cloud is obtained by the normal vector angle of each point cloud and the distance weight value between each point cloud and other point clouds in the same grid. in, The synthesis module is used to synthesize the flat area feature point data and the non-flat area feature point data and then perform filtering to obtain total feature point data.

Images