CN112233039B - Three-dimensional laser point cloud denoising method based on domain point space characteristics - Google Patents

Abstract

The invention provides a three-dimensional laser point cloud denoising method based on domain point space characteristics, which comprises the following steps: s1: reading and storing original data point clouds into an n-row 3-column point cloud data set O; s2: dividing the domain point and the target point P into subsets Q; s3: performing plane fitting on all point clouds in the subset Q; s4: translating the plane of the three-dimensional plane equation to a target point P along a normal vector to obtain a new space plane; s5: calculating the average distance d from the rest points in the subset Q except the target point P to the space plane, and storing as the 4 th column of the point cloud data set O, S6: traversing all points, and repeating the steps S2-S5; s7: and (4) counting the distribution characteristics of the data in the 4 th row, selecting a threshold value f according to the distribution range and the confidence interval, and keeping the point cloud data smaller than the threshold value f as a de-noised result. The three-dimensional laser point cloud denoising method based on the domain point space characteristics improves the denoising accuracy and efficiency.

Description

A three-dimensional laser point cloud denoising method based on spatial features of domain points

technical field

The invention relates to the fields of civil engineering and surveying and mapping, in particular to a three-dimensional laser point cloud denoising method based on the spatial characteristics of field points.

Background technique

In tunnel construction, tunnel deformation monitoring plays an important role in ensuring construction safety and determining support time. However, due to the extremely low measurement efficiency of traditional deformation measurement technology, deformation data of only a small number of measuring points can be obtained, which cannot fully reflect the deformation of the tunnel. It is necessary to use 3D laser to monitor the whole space. Due to the special environment in the tunnel, the collected 3D point cloud data contains a lot of noise caused by dust. In the past, the data was denoised by frequency domain filtering and empirical judgment. The noise effect is not good and the efficiency is low.

SUMMARY OF THE INVENTION

In view of the above deficiencies in the prior art, the present invention provides a three-dimensional laser point cloud denoising method based on the spatial features of domain points, which improves the accuracy and efficiency of denoising through the method based on spatial features of domain points.

In order to achieve the above purpose, the present invention provides a three-dimensional laser point cloud denoising method based on the spatial characteristics of domain points, comprising the steps of:

S1: Read and store the original data point cloud as a point cloud dataset O with n rows and 3 columns, where n is a natural number greater than or equal to 1;

S2: Select a target point P (x _p , y _p , z _p ) in the point cloud data set O, where x _p , y _p and z _p are the x-axis, y-axis and z of the target point P, respectively The coordinate value of the axis; use the knn algorithm to retrieve the k domain points (k≥3) closest to the point in the point cloud data set O, and divide the domain points and the target point P into a subset Q ;

S3: Perform plane fitting on all point clouds in the subset Q to obtain a three-dimensional plane equation:

Ax+By+Cz+D=0 (1);

where A, B, and C are the projection components of the plane normal vector in the x, y, and z directions, respectively, and D is a constant term;

S4: Translate the plane of the three-dimensional plane equation to the target point P along the normal vector to obtain a new space plane, and the formula of the space plane is:

Ax+By+Cz+D′=0 (2);

Wherein, D′=-(Ax _p +By _p +Cz _p ) (3);

S5: Calculate the average distance d from the other points in the subset Q except the target point P to the spatial plane, and save it as the fourth column of the point cloud data set O:

S6: Traverse all the points in the point cloud data set O, repeat steps S2 to S5, at this time the point cloud data set O is n rows and 4 columns;

S7: Count the distribution characteristics of the data in the fourth column of the point cloud data set O, select a threshold f according to the distribution range and confidence interval of the data in the fourth column of the point cloud data set O, and keep the data in the fourth column smaller than the The point cloud data with threshold f is the result after denoising.

The present invention has the following beneficial effects due to the adoption of the above technical solutions:

The three-dimensional laser point cloud denoising method based on the spatial feature of the domain point of the present invention can quickly and quickly remove the noise points in the three-dimensional point cloud data, provides a basis for the three-dimensional monitoring of the tunnel, and has a good application prospect in the tunnel monitoring.

Description of drawings

1 is a flowchart of a method for denoising a three-dimensional laser point cloud based on spatial features of domain points according to an embodiment of the present invention;

2 is a feature map of the normal vector distribution of all points in a point cloud before denoising according to an embodiment of the present invention;

FIG. 3 is a characteristic diagram of the normal vector distribution of a point cloud after denoising according to an embodiment of the present invention.

Detailed ways

The preferred embodiments of the present invention are given and described in detail below according to the accompanying drawings 1 to 3 , so that the functions and characteristics of the present invention can be better understood.

Referring to FIG. 1, a method for denoising a 3D laser point cloud based on spatial features of domain points according to an embodiment of the present invention includes steps:

S1: Read and store the original data point cloud as a point cloud dataset O with n rows and 3 columns, where n is a natural number greater than or equal to 1;

S2: Select a target point P (x _p , y _p , z _p ) in the point cloud data set O, where x _p , y _p and z _p are the coordinate values of the x-axis, y-axis and z-axis of the target point P, respectively ;Use the knn algorithm to retrieve the k domain points (k≥3) closest to the point in the point cloud data set O, and divide the domain points and the target point P into a subset Q;

S3: Perform plane fitting on all point clouds in the subset Q to obtain the three-dimensional plane equation:

Ax+By+Cz+D=0 (1);

where A, B, and C are the projection components of the plane normal vector in the x, y, and z directions, respectively, and D is a constant term;

S4: Translate the plane of the three-dimensional plane equation to the target point P along the normal vector to obtain a new space plane. The formula of the space plane is:

Ax+By+Cz+D′=0 (2);

Wherein, D′=-(Ax _p +By _p +Cz _p ) (3);

S5: Calculate the average distance d of the remaining points in the subset Q except the target point P to the spatial plane, and save it as the fourth column of the point cloud data set O:

S6: Traverse all the points in the point cloud data set O, and repeat steps S2 to S5. At this time, the point cloud data set O is n rows and 4 columns;

S7: Count the distribution characteristics of the data in the fourth column of the point cloud data set O, select the threshold f according to the distribution range and confidence interval of the data in the fourth column of the point cloud data set O, and keep the point cloud data whose data in the fourth column is less than the threshold f as the The result after noise.

A method for denoising a 3D laser point cloud based on spatial features of domain points according to an embodiment of the present invention, by calculating the average distance from the domain points of all points in the point cloud space to the fitting surface as a parameter, and setting a threshold according to the statistical results , remove noise points, and achieve the effect of automatic denoising of 3D laser point cloud data.

In the specific implementation, firstly, any point is selected as the target point, and then its domain points are retrieved to form a spatial subset. By performing plane fitting on the points of the spatial subset class, the plane is obtained, and the plane is translated to the target point along the plane's normal vector direction. Calculates the average distance from the domain point to the plane. Traverse all the points in the point cloud and repeat the above steps to obtain the statistical characteristics of the average distance from each point in the field to the plane, and then set the denoising threshold for denoising according to the statistical characteristics.

By comparing Figure 2 and Figure 3, it can be seen that after denoising, the concentration of the distribution is more obvious, and the noise interference is significantly reduced.

The present invention has been described in detail above with reference to the embodiments of the accompanying drawings, and those skilled in the art can make various modifications to the present invention according to the above description. Therefore, some details in the embodiments should not be construed to limit the present invention, and the present invention will take the scope defined by the appended claims as the protection scope of the present invention.

Claims (1)

1. A three-dimensional laser point cloud denoising method based on domain point space features, comprising the steps of: S1: Read and store the original data point cloud as a point cloud dataset O with n rows and 3 columns, where n is a natural number greater than or equal to 1; S2: Select a target point P (x _p , y _p , z _p ) in the point cloud data set O, where x _p , y _p and z _p are the x-axis, y-axis and z of the target point P, respectively The coordinate value of the axis; use the knn algorithm to retrieve the k field points closest to the point in the point cloud data set O, k≥3, and divide the field points and the target point P into a subset Q; S3: Perform plane fitting on all point clouds in the subset Q to obtain a three-dimensional plane equation: Ax+By+Cz+D=0 (1); where A, B, and C are the projection components of the plane normal vector in the x, y, and z directions, respectively, and D is a constant term; S4: Translate the plane of the three-dimensional plane equation to the target point P along the normal vector to obtain a new space plane, and the formula of the space plane is: Ax+By+Cz+D′=0 (2); Wherein, D′=-(Ax _p +By _p +Cz _p ) (3); S5: Calculate the average distance d from the remaining points in the subset Q to the spatial plane except the target point P, and save it as the fourth column of the point cloud data set O:

S6: Traverse all the points in the point cloud data set O, repeat steps S2 to S5, at this time the point cloud data set O is n rows and 4 columns; S7: Count the distribution characteristics of the data in the fourth column of the point cloud data set O, select a threshold f according to the distribution range and confidence interval of the data in the fourth column of the point cloud data set O, and keep the data in the fourth column smaller than the The point cloud data with threshold f is the result after denoising.

Images