CN-121999146-A - Point cloud surface reconstruction method and device, electronic equipment and readable storage medium

CN121999146ACN 121999146 ACN121999146 ACN 121999146ACN-121999146-A

Abstract

The application provides a point cloud surface reconstruction method, a device, electronic equipment and a readable storage medium, wherein the method comprises the steps of calculating bilateral weights of all neighborhood points for each query point according to an included angle between a normal vector of the query point and a normal vector of each neighborhood point of each query point in three-dimensional point cloud data of a target object, generating composite weights of all neighborhood points for the query point according to the bilateral weights, substituting the composite weights into a predefined target function, determining hidden function values of the query point by minimizing the inconsistency degree between signed distances from the query point to tangential planes of all neighborhood points and hidden function values of the query point through weighted least square, solving the target function to obtain hidden function values of the query point, extracting zero level sets according to the hidden function values of all query points, and generating the reconstructed target object surface. By this method, a high quality surface is reconstructed while retaining the sharp corners of the object.

Inventors

FENG KAIYONG

Assignees

深圳市其域创新科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. A method for reconstructing a point cloud surface, comprising: Acquiring three-dimensional point cloud data of a target object and normal vectors of each point in the three-dimensional point cloud data; aiming at each query point in the three-dimensional point cloud data, calculating bilateral weights of all the neighborhood points for the query point according to the included angle between the normal vector of the query point and the normal vector of each neighborhood point of the query point, wherein the bilateral weights and the possibility that the query point and the neighborhood point are positioned on the same surface of the target object are in positive correlation; generating composite weights of all neighborhood points for the query point according to the bilateral weights; The target function is used for minimizing the inconsistency degree between signed distances from the query point to the tangent plane where each neighborhood point is located and the hidden function value of the query point through weighted least square so as to determine the hidden function value of the query point; Solving the objective function to obtain an implicit function value of the query point, wherein the implicit function value is zero when the query point is positioned on the surface of the objective object; and extracting a zero level set according to the hidden function value of each query point to generate the reconstructed target object surface.
2. The method of claim 1, wherein the objective function is: Wherein, the Representing query points Is a hidden function value of (2); Is a set of neighborhood points; Is the location of the query point; is the position of the neighborhood point; The normal vector of the neighborhood point; the method comprises the steps of (1) obtaining a variable to be solved; and the composite weight of the neighborhood point to the query point is obtained.
3. The method of claim 1, wherein before generating the composite weight of each neighboring point to the query point according to the bilateral weight, the method further comprises: Calculating the space distance between the query point and the neighborhood point, and calculating the space weight of the neighborhood point to the query point according to the space distance, wherein the space distance and the space weight are in a negative correlation; Calculating residual weights used for representing whether the neighborhood points are outliers or not, wherein the magnitude of the residual weights and the possibility that the neighborhood points are outliers are in negative correlation; The method comprises the steps of calculating normal consistency weights used for representing the neighborhood points relative to the query points, wherein the normal consistency weights are used for representing the consistency of the neighborhood points and normal vectors of the query points corresponding to hidden function values currently calculated by the objective function, and the consistency degree and the normal consistency weights are in positive correlation.
4. The method of claim 3, wherein generating the composite weight of each neighboring point to the query point according to the bilateral weight comprises: and calculating the product of the spatial weight, the residual weight, the normal consistency weight and the bilateral weight of each neighborhood point of the query point, and taking the product result as the composite weight of the neighborhood point to the query point.
5. The method of claim 1, wherein for each query point in the three-dimensional point cloud data, calculating the bilateral weight of each neighbor point for the query point based on the included angle between the normal vector of the query point and the normal vector of each neighbor point of the query point, comprises: smoothing normal vectors of all neighborhood points of each query point in the three-dimensional point cloud data to obtain a reference normal vector of the query point; and calculating the bilateral weight of each neighborhood point to the query point through a soft transition function according to the dot product of the reference normal vector of the query point and the normal vector of each neighborhood point.
6. The method of claim 5, wherein the smoothing the normal vector of each neighboring point of the query point for each query point in the three-dimensional point cloud data to obtain the reference normal vector of the query point comprises: For each query point in the three-dimensional point cloud data, smoothing the normal vector of each neighborhood point of the query point by the following formula: Wherein, the Representing query points Is defined by the reference normal vector of (a); For the query point Is a set of neighborhood points; Representing query points Is a normal vector to the neighborhood of points.
7. The method according to claim 5, wherein calculating the bilateral weight of each neighborhood point to the query point by the soft transition function according to the dot product of the reference normal vector of the query point and the normal vector of each neighborhood point comprises: calculating bilateral weights of each neighborhood point for the query point through the following soft transition functions: Wherein, the The reference normal vector representing a query point; representing normal vectors of the neighborhood points; Is a transition center point; for controlling the degree of steepness of the transition; Bilateral weights representing query points.
8. A point cloud surface reconstruction device, comprising: the acquisition module is used for acquiring three-dimensional point cloud data of the target object and normal vectors of each point in the three-dimensional point cloud data; The first calculation module is used for calculating bilateral weights of all the neighborhood points for each query point in the three-dimensional point cloud data according to the included angle between the normal vector of the query point and the normal vector of each neighborhood point of the query point, wherein the bilateral weights and the possibility that the query point and the neighborhood point are positioned on the same surface of the target object are in positive correlation; the generating module is used for generating the composite weight of each neighborhood point to the query point according to the bilateral weight; The substituting module is used for substituting the composite weight into a predefined objective function, wherein the objective function is used for determining the hidden function value of the query point by minimizing the degree of inconsistency between the signed distance from the query point to the tangent plane where each neighborhood point is located and the hidden function value of the query point through weighted least square; The solving module is used for solving the objective function to obtain the hidden function value of the query point, wherein the hidden function value is zero when the query point is positioned on the surface of the target object; And the reconstruction module is used for extracting a zero level set according to the hidden function value of each query point and generating the reconstructed target object surface.
9. An electronic device comprising a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication via the bus when the electronic device is in operation, the machine-readable instructions when executed by the processor performing the steps of the method of any of claims 1 to 7.
10. A computer-readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, performs the steps of the method according to any of claims 1 to 7.

Description

Point cloud surface reconstruction method and device, electronic equipment and readable storage medium Technical Field The present application relates to the field of point cloud processing technologies, and in particular, to a method and an apparatus for reconstructing a point cloud surface, an electronic device, and a readable storage medium. Background Point cloud surface reconstruction refers to the process of reconstructing discrete three-dimensional point cloud data into a continuous, smooth surface (curved surface) model after scanning an object with a 3D scanner to obtain three-dimensional point cloud data. In the prior art, the surface reconstruction method based on Implicit Moving Least Squares (IMLS) is widely adopted because of simple mathematics and convenient realization. The standard IMLS method defines an implicit function by performing a weighted least squares fit to the local neighborhood points, the zero level set of which is the reconstructed surface. However, in the conventional IMLS method, since IMLS is essentially a low-pass filter, there are scan points on both sides of the edges of the object, such as the sides of the dice, where the two sides meet during the fitting process. IMLS take into account the points on both sides and then take an average. This smoothes out sharp edges and corners so that the otherwise sharp corners are rounded off, become smooth transitions, and lose the detail of the model. And, near sharp features (such as corners), points from the other side, while spatially close, geometrically belong to the other face. Conventional IMLS also pulls in the "opposite point" to vote, which can lead to distortion and shrinkage of the reconstructed surface at the corners, which is inaccurate. Disclosure of Invention In view of the above, an object of the present application is to provide a method, an apparatus, an electronic device and a readable storage medium for reconstructing a high-quality surface using three-dimensional point cloud data while maintaining a sharp edge of an object. In a first aspect, an embodiment of the present application provides a method for reconstructing a point cloud surface, including: Acquiring three-dimensional point cloud data of a target object and normal vectors of each point in the three-dimensional point cloud data; aiming at each query point in the three-dimensional point cloud data, calculating bilateral weights of all the neighborhood points for the query point according to the included angle between the normal vector of the query point and the normal vector of each neighborhood point of the query point, wherein the bilateral weights and the possibility that the query point and the neighborhood point are positioned on the same surface of the target object are in positive correlation; generating composite weights of all neighborhood points for the query point according to the bilateral weights; The target function is used for minimizing the inconsistency degree between signed distances from the query point to the tangent plane where each neighborhood point is located and the hidden function value of the query point through weighted least square so as to determine the hidden function value of the query point; Solving the objective function to obtain an implicit function value of the query point, wherein the implicit function value is zero when the query point is positioned on the surface of the objective object; and extracting a zero level set according to the hidden function value of each query point to generate the reconstructed target object surface. With reference to the first aspect, the embodiment of the present application provides a first possible implementation manner of the first aspect, wherein the objective function is: Wherein, the Representing query pointsIs a hidden function value of (2); Is a set of neighborhood points; Is the location of the query point; is the position of the neighborhood point; The normal vector of the neighborhood point; the method comprises the steps of (1) obtaining a variable to be solved; and the composite weight of the neighborhood point to the query point is obtained. With reference to the first aspect, an embodiment of the present application provides a second possible implementation manner of the first aspect, where before generating, according to the bilateral weight, a composite weight of each neighboring point to the query point, the method further includes: Calculating the space distance between the query point and the neighborhood point, and calculating the space weight of the neighborhood point to the query point according to the space distance, wherein the space distance and the space weight are in a negative correlation; Calculating residual weights used for representing whether the neighborhood points are outliers or not, wherein the magnitude of the residual weights and the possibility that the neighborhood points are outliers are in negative correlation; The method comprises