CN112419164A - Curvature and neighborhood reconstruction-based weighted guide point cloud model denoising method - Google Patents

Abstract

The invention discloses a weighted guidance point cloud model denoising method based on curvature and neighborhood reconstruction. The method is: first, calculate the curvature information of each point in the point cloud model, and extract the feature points in the model according to the set threshold; The neighborhood points are reconstructed on the basis of the domain, and the reconstructed neighborhood is on one surface; then, according to the reconstructed neighborhood, the three-dimensional position information of each point is used as a guiding signal, and the curvature information is used as a guide signal. The weighted signal is added to the position guidance signal, so that each point in the point cloud model is linearly transformed; finally, each point is linearly transformed according to the calculated linear transformation coefficient to achieve denoising of the point cloud model. The invention maintains the feature information of the point cloud model while denoising the point cloud model, and has better robustness to different degrees of noise.

Description

A Weighted Guided Point Cloud Model Denoising Method Based on Curvature and Neighborhood Reconstruction

technical field

The invention relates to the technical field of computer graphics and three-dimensional point cloud denoising, in particular to a weighted guidance point cloud model denoising method based on curvature and neighborhood reconstruction.

Background technique

In recent years, due to the emergence and promotion of applications such as virtual reality and augmented reality, point cloud model processing technology has become one of the hot spots in the field of computer graphics research. The denoised point cloud model can be used as the basis for applications such as animation, rendering, and 3D reconstruction. With the increasing popularity of various scanning devices, especially consumer-grade depth sensors such as Kinect, the design of robust point cloud denoising methods becomes more and more important, and the main technical challenge of point cloud denoising is how to Effectively preserve the features of the model while removing noise.

Currently commonly used point set filtering methods, such as local optimal projection (LOP), robust implicit moving least squares (RIMLS), weighted LOP (WLOP), edge-aware resampling (EAR-aware resampling) and continuous LOP (CLOP) ), all have significant advantages. However, these point set filtering methods either fail to preserve clear features or are less robust to noise removal. Among them, LOP, WLOP and CLOP are all LOP-based methods, which can remove noise and outliers well, but cannot retain their clear features due to their inherent isotropic properties; EAR is an extended LOP-based method. method, while preserving geometric features, may erase fine-scale geometric features because it needs to exploit a rather large neighborhood size; RIMLS also preserves features, but differs from the Compared to LOP methods, it is generally more sensitive to outliers and noise. The above problems severely limit the robustness and effectiveness of these methods in point group denoising.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a weighted guide point cloud model denoising method based on curvature and neighborhood reconstruction, which has high denoising accuracy, can retain model features, and has good robustness to different degrees of noise.

The technical solution for realizing the object of the present invention is: a weighted guidance point cloud model denoising method based on curvature and neighborhood reconstruction, comprising the following steps:

Step 1: Calculate the curvature information of each point in the point cloud model, and extract the feature points in the model according to the set threshold;

Step 2, according to the extracted feature points, reconstruct the neighborhood points on the basis of the neighborhood obtained by the K-nearest neighbor method, and make the reconstructed neighborhood on one surface;

Step 3: According to the reconstructed neighborhood, use the three-dimensional position information of each point as a guide signal, and add the curvature information as a weighted signal to the position guide signal, so as to perform linear transformation on each point in the point cloud model. ;

Step 4: Perform linear transformation on each point according to the linear transformation coefficients calculated in Step 3 to realize denoising of the point cloud model.

Further, in the calculation of the curvature information of each point in the point cloud model described in step 1, the feature points in the model are extracted according to the set threshold, as follows:

Step 1.1. For each point p _i , calculate its curvature value σ(N _i ) corresponding to the neighborhood N _i as

Among them, λ ₀ , λ ₁ , λ ₂ are the singular _values of the covariance matrix of Ni, and λ ₀ <λ ₁ <λ ₂ , reflecting the distribution of three orthogonal _singular vectors of Ni;

Step 1.2. Set the threshold t, the points with the curvature value greater than the threshold t are the feature points, and the points with the curvature value less than the threshold t are the non-feature points.

Further, according to the extracted feature points described in step 2, the neighborhood points are reconstructed on the basis of the neighborhood obtained by the K-nearest neighbor method, and the reconstructed neighborhood is on one plane, as follows:

Step 2.1. Use the K-nearest neighbor method to assign an initial neighborhood N to each feature point p _i ;

初始化为只包含其自身p_ij和特征点p_i两个点；Step 2.2. For each neighborhood point p _ij in the initial neighborhood N, also use the K-nearest neighbor method to obtain the neighborhood N _ij of the neighborhood point p _ij , and at the same time select the candidate neighbor of the feature point p _i at p _ij area

Initialized to include only two points of its own p _ij and feature point p _i ;

中，根据当前邻域的曲率值以及点的位置关系构建下式评判标准：Step 2.3. Scan each point in N _ij to determine whether the point can be added to the candidate neighborhood

, the following evaluation criteria are constructed according to the curvature value of the current neighborhood and the positional relationship of the points:

表示衡量当前邻域的曲率值以及邻域中点的位置关系的标准值，

表示p_ij在邻域

下的曲率值，K代表邻域

中所有点的个数，α和β为用户自定义的控制系数；p_ijk表示邻域

中的第k个点，k＝1,2,…,K，K表示邻域

中点的总数；in,

Represents the standard value for measuring the curvature value of the current neighborhood and the positional relationship of the points in the neighborhood,

Indicates that p _ij is in the neighborhood

The curvature value under , K represents the neighborhood

The number of all points in , α and β are user-defined control coefficients; p _ijk represents the neighborhood

The kth point in , k=1,2,...,K, K represents the neighborhood

the total number of midpoints;

的值减小，则代表该点能够加入到候选邻域

中，并将距离该点最近的5个点也加入到候选邻域

中；Step 2.4. If this point is added, the formula (2) is

The value of is reduced, it means that the point can be added to the candidate neighborhood

, and the 5 points closest to this point are also added to the candidate neighborhood

middle;

的值，其中最小的一个值对应的邻域即为该特征点p_i重构出的邻域N'。Step 2.5. For the remaining points in N, a candidate neighborhood including the feature point p _i is obtained. For each candidate neighborhood, calculate according to formula (2).

, and the neighborhood corresponding to the smallest value is the neighborhood N' reconstructed by the feature point p _i .

Further, according to the reconstructed neighborhood described in step 3, the three-dimensional position information of each point is used as a guide signal, and the curvature information is added to the position guide signal as a weighted signal, so that each point in the point cloud model is used. A point is linearly transformed, as follows:

The cost function E of the weighted guided filtering algorithm is:

γ(i)=(σ-t) ^s(i) +χ (4)

s(i)=-sgn(σ-t)×μ×σ (5)

动态决定。Among them, N(pi ) represents the neighborhood of the current point _pi , p _ij is the point _in the neighborhood, a _i and _bi are the linear transformation coefficients to be obtained, ε is the parameter controlling the filtering effect; σ is the neighborhood The curvature value of the point calculated before the domain reconstruction, t is the threshold for judging the feature point; χ is a positive number, used to prevent the weight γ(i) from being 0; μ is the magnification, given by

Dynamic decision.

Further, according to the linear transformation coefficient calculated in step 3 described in step 4, linear transformation is performed on each point to realize denoising of the point cloud model, as follows:

Step 4.1. The coefficients a _i and b _i of the linear transformation obtained from step 3 are:

in:

为邻域的中心点，ε为控制滤波效果的参数；Among them, |N(pi )| represents the number of points contained in the neighborhood of point _pi , p _ij is a point in the neighborhood N( _{pi )} of _pi ,

is the center point of the neighborhood, and ε is the parameter that controls the filtering effect;

Step 4.2. According to the obtained linear transformation coefficients a _i and b _i , perform linear transformation on each feature point to obtain the position of the denoised point, and obtain the denoised point cloud model after all points are updated.

Compared with the prior art, the present invention has the following significant advantages: (1) the input is simple, only the position information of the point cloud model point needs to be input, and there is no need to rely on the initial normal estimation; (2) the curvature information is introduced as a weighted signal, and the The feature area of the model is processed separately from the flat area, so the sharp features of the model are better preserved; (3) By reconstructing the neighborhood of the feature points, each point obtains a relatively smooth and consistent neighborhood, so the different degrees of noise is robust.

Description of drawings

FIG. 1 is a schematic flowchart of the weighted guided point cloud model denoising method based on curvature and neighborhood reconstruction of the present invention.

FIG. 2 is a flow frame diagram of feature point neighborhood reconstruction in the present invention.

FIG. 3 is a frame diagram of the weighted guided filtering method in the present invention.

4 is a denoising effect diagram in an embodiment of the present invention, wherein (a) is a schematic diagram of an input point cloud model with noise, and (b) is a schematic diagram of an output point cloud model after denoising.

5 is a denoising effect diagram in an embodiment of the present invention, wherein (a) is a schematic diagram of an input point cloud model with noise, and (b) is a schematic diagram of an output point cloud model after denoising.

Detailed ways

The weighted guidance point cloud model denoising method based on curvature and neighborhood reconstruction of the present invention includes the following steps:

Step 1: Calculate the curvature information of each point in the point cloud model, and extract the feature points in the model according to the set threshold;

Step 2, according to the extracted feature points, reconstruct the neighborhood points on the basis of the neighborhood obtained by the K-nearest neighbor method, and make the reconstructed neighborhood on one surface;

Step 3: According to the reconstructed neighborhood, use the three-dimensional position information of each point as a guide signal, and add the curvature information as a weighted signal to the position guide signal, so as to perform linear transformation on each point in the point cloud model. ;

Step 4: Perform linear transformation on each point according to the linear transformation coefficients calculated in Step 3 to realize denoising of the point cloud model.

As a specific example, the curvature information of each point in the point cloud model is calculated in step 1, and the feature points in the model are extracted according to the set threshold, as follows:

Step 1.1. For each point p _i , calculate its curvature value σ(N _i ) corresponding to the neighborhood N _i as

Among them, λ ₀ , λ ₁ , λ ₂ are the singular _values of the covariance matrix of Ni, and λ ₀ <λ ₁ <λ ₂ , reflecting the distribution of three orthogonal _singular vectors of Ni;

Step 1.2. Set the threshold t, the points with the curvature value greater than the threshold t are the feature points, and the points with the curvature value less than the threshold t are the non-feature points.

Further, according to the extracted feature points described in step 2, the neighborhood points are reconstructed on the basis of the neighborhood obtained by the K-nearest neighbor method, and the reconstructed neighborhood is on one plane, as follows:

Step 2.1. Use the K-nearest neighbor method to assign an initial neighborhood N to each feature point p _i ;

初始化为只包含其自身p_ij和特征点p_i两个点；Step 2.2. For each neighborhood point p _ij in the initial neighborhood N, also use the K-nearest neighbor method to obtain the neighborhood N _ij of the neighborhood point p _ij , and at the same time select the candidate neighbor of the feature point p _i at p _ij area

Initialized to include only two points of its own p _ij and feature point p _i ;

中，根据当前邻域的曲率值以及点的位置关系构建下式评判标准：Step 2.3. Scan each point in N _ij to determine whether the point can be added to the candidate neighborhood

, the following evaluation criteria are constructed according to the curvature value of the current neighborhood and the positional relationship of the points:

表示衡量当前邻域的曲率值以及邻域中点的位置关系的标准值，

表示p_ij在邻域

下的曲率值，K代表邻域

中所有点的个数，α和β为用户自定义的控制系数；p_ijk表示邻域

中的第k个点，k＝1,2,…,K，K表示邻域

中点的总数；in,

Represents the standard value for measuring the curvature value of the current neighborhood and the positional relationship of the points in the neighborhood,

Indicates that p _ij is in the neighborhood

The curvature value under , K represents the neighborhood

The number of all points in , α and β are user-defined control coefficients; p _ijk represents the neighborhood

The kth point in , k=1,2,...,K, K represents the neighborhood

the total number of midpoints;

的值减小，则代表该点能够加入到候选邻域

中，并将距离该点最近的5个点也加入到候选邻域

中；Step 2.4. If this point is added, the formula (2) is

The value of is reduced, it means that the point can be added to the candidate neighborhood

, and the 5 points closest to this point are also added to the candidate neighborhood

middle;

的值，其中最小的一个值对应的邻域即为该特征点p_i重构出的邻域N'。Step 2.5. For the remaining points in N, a candidate neighborhood including the feature point p _i is obtained. For each candidate neighborhood, calculate according to formula (2).

, and the neighborhood corresponding to the smallest value is the neighborhood N' reconstructed by the feature point p _i .

As a specific example, according to the reconstructed neighborhood described in step 3, the three-dimensional position information of each point is used as the guiding signal, and the curvature information is added to the position guiding signal as a weighted signal, so as to improve the accuracy of the point cloud model. Each point in the linear transformation is performed as follows:

The cost function E of the weighted guided filtering algorithm is:

γ(i)=(σ-t) ^s(i) +χ (4)

s(i)=-sgn(σ-t)×μ×σ (5)

动态决定。Among them, N(pi ) represents the neighborhood of the current point _pi , p _ij is the point _in the neighborhood, a _i and _bi are the linear transformation coefficients to be obtained, ε is the parameter controlling the filtering effect; σ is the neighborhood The curvature value of the point calculated before the domain reconstruction, t is the threshold for judging the feature point; χ is a positive number, used to prevent the weight γ(i) from being 0; μ is the magnification, determined by

Dynamic decision.

As a specific example, step 4 performs linear transformation on each point according to the linear transformation coefficient calculated in step 3 to realize denoising of the point cloud model, as follows:

Step 4.1. The coefficients a _i and b _i of the linear transformation obtained from step 3 are:

in:

为邻域的中心点，ε为控制滤波效果的参数；Among them, |N(pi )| represents the number of points contained in the neighborhood of point _pi , p _ij is a point in the neighborhood N( _{pi )} of _pi ,

is the center point of the neighborhood, ε is the parameter that controls the filtering effect;

Step 4.2. According to the obtained linear transformation coefficients a _i and b _i , perform linear transformation on each feature point to obtain the position of the denoised point, and obtain the denoised point cloud model after all points are updated.

The present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

Example

1, the weighted guidance point cloud model denoising method based on curvature and neighborhood reconstruction of the present invention includes the following steps:

Step 1: Calculate the curvature information of each point in the point cloud model, and extract the feature points in the model according to the set threshold, as follows:

Step 1.1. For each point p _i , calculate its curvature value σ(N _i ) corresponding to the neighborhood N _i as:

Among them, λ ₀ , λ ₁ , λ ₂ are the singular _values of the covariance matrix of Ni, and λ ₀ <λ ₁ <λ ₂ , reflecting the distribution of three orthogonal _singular vectors of Ni;

Step 1.2. Set the threshold t, the points with the curvature value greater than the threshold t are the feature points, and the points with the curvature value less than the threshold t are the non-feature points.

Step 2, according to the extracted feature points, reconstruct the neighborhood points on the basis of the neighborhood obtained by the K-nearest neighbor method, and make the reconstructed neighborhood on one plane, combined with Figure 2, as follows:

Step 2.1. Use the K-nearest neighbor method to assign an initial neighborhood N to each feature point p _i ;

初始化为只包含其自身p_ij和特征点p_i两个点；Step 2.2. For each neighborhood point p _ij in the initial neighborhood N, also use the K-nearest neighbor method to obtain the neighborhood N _ij of the neighborhood point p _ij , and at the same time select the candidate neighbor of the feature point p _i at p _ij area

Initialized to include only two points of its own p _ij and feature point p _i ;

中，根据当前邻域的曲率值以及点的位置关系构建下式评判标准：Step 2.3. Scan each point in N _ij to determine whether the point can be added to the candidate neighborhood

, the following evaluation criteria are constructed according to the curvature value of the current neighborhood and the positional relationship of the points:

表示衡量当前邻域的曲率值以及邻域中点的位置关系的标准值，

表示p_ij在邻域

下的曲率值，K代表邻域

中所有点的个数，α和β为用户自定义的控制系数；p_ijk表示邻域

中的第k个点，k＝1,2,…,K，K表示邻域

中点的总数；in,

Represents the standard value for measuring the curvature value of the current neighborhood and the positional relationship of the points in the neighborhood,

Indicates that p _ij is in the neighborhood

The curvature value under , K represents the neighborhood

The number of all points in , α and β are user-defined control coefficients; p _ijk represents the neighborhood

The kth point in , k=1,2,...,K, K represents the neighborhood

the total number of midpoints;

的值减小，则代表该点能够加入到候选邻域

中，并将距离该点最近的5个点也加入到候选邻域

中；Step 2.4. If this point is added, the formula (2) is

The value of is reduced, it means that the point can be added to the candidate neighborhood

, and the 5 points closest to this point are also added to the candidate neighborhood

middle;

的值，其中最小的一个值对应的邻域即为该特征点p_i重构出的邻域N'。Step 2.5. For the remaining points in N, a candidate neighborhood including the feature point p _i is obtained. For each candidate neighborhood, calculate according to formula (2).

, and the neighborhood corresponding to the smallest value is the neighborhood N' reconstructed by the feature point p _i .

Step 3: According to the reconstructed neighborhood, use the three-dimensional position information of each point as a guide signal, and add the curvature information as a weighted signal to the position guide signal, so as to perform linear transformation on each point in the point cloud model. , combined with Figure 3, as follows:

The traditional guided filtering algorithm uses a unified linear model and the same regularization parameter for each part of the point cloud model, so it will smooth out the feature area, and use a weighted guided point cloud denoising method that fuses curvature information. An effective guided filtering weight model needs to solve two problems: first, the point cloud model processing method used by the weights should be able to accurately identify the feature area of the model; second, in the feature area, the superposition should be smaller The smoothing multiple of , that is, the final regular term should be smaller, the weight at the denominator should be larger; in the flat area of the model, a slightly larger smoothing multiple should be superimposed, that is, the final regular term should be slightly larger, then the denominator should be slightly larger. The weights should be small. In the field of point cloud processing, the feature points of most point clouds can be extracted by calculating the curvature information of each point.

In addition, an ideal weight model requires a smaller weight in the flat area of the model and a larger weight in the feature area. Since the change law of the weight value is similar to that of the exponential function, the curvature information of the point can be used as the base, and the weight model based on the exponential function can be used to amplify or suppress the curvature information of the point. In order to obtain the feature area of the model, the feature point curvature threshold t is set. When the curvature value of a point is greater than t, the point is considered as a feature point, otherwise it is a non-feature point. In order to make the curvature information more reasonable, a constraint factor s(i) is introduced into the weight γ(i) as the exponential term of the weight model. The constraint factor can set a constraint boundary for the weight, so that it can Accurately determine the different weighting behaviors that should be taken for different regions, so that for feature regions, it is sensitive to feature information and amplifies the information at feature points; for flat regions, it is insensitive to feature information and inhibits the growth of feature information. , so as to achieve the purpose of retaining features.

Based on the above analysis, the weights are defined as:

γ(i)=(σ-t) ^s(i) +χ

s(i)=-sgn(σ-t)×μ×σ

动态决定，α为一个常数项，用于防止γ(i)作为分母为零的情况。where σ is the curvature value of the point calculated before neighborhood reconstruction, t is the threshold for judging feature points, and μ is the magnification, which is defined by

Dynamically determined, α is a constant term that prevents γ(i) as a zero denominator.

Obviously, when the point belongs to the feature point, the sign of s(i) is negative. At this time, the greater the curvature, the more significant the feature information, the greater the weight, which is sensitive to the feature information; when the point does not belong to the feature point, s( i) The sign is positive, and the weight at this time is very small, which is insensitive to feature information.

Finally, the combined weights are rewritten on the basis of guided filtering, and the cost function is:

where N(pi ) represents the neighborhood of the current point pi, p _ij is the point _in the neighborhood, a _i and _bi are the linear transformation coefficients to be obtained _, and ε is a parameter that controls the filtering effect.

Step 4: Perform linear transformation on each point according to the linear transformation coefficient calculated in Step 3 to realize denoising of the point cloud model, as follows:

Step 4.1, the coefficient of linear transformation obtained from step 3 is:

in:

为邻域的中心点，ε为控制滤波效果的参数；Among them, |N(pi )| represents the number of points contained in the neighborhood of point _pi , p _ij is a point in the neighborhood N( _{pi )} of _pi ,

is the center point of the neighborhood, ε is the parameter that controls the filtering effect;

Step 4.2. According to the obtained linear transformation coefficients a _i and b _i , perform linear transformation on each feature point to obtain the position of the denoised point, and obtain the denoised point cloud model after all points are updated.

4 is a denoising effect diagram in an embodiment of the present invention, wherein FIG. 4(a) is a schematic diagram of an input point cloud model with noise, and FIG. 4(b) is a schematic diagram of an output point cloud model after denoising. 5 is a denoising effect diagram in an embodiment of the present invention, wherein FIG. 5(a) is a schematic diagram of an input point cloud model with noise, and FIG. 5(b) is a schematic diagram of an output point cloud model after denoising. It can be seen from Fig. 4 and Fig. 5 that the denoising method of the point cloud model based on the weighted guide point cloud model based on curvature and neighborhood reconstruction of the present invention is used to denoise the point cloud model, which can remove the noise while maintaining the sharp features of the model, which proves that the present invention can Widely used in various models.

Claims (5)

1. A curvature and neighborhood reconstruction-based weighted guide point cloud model denoising method is characterized by comprising the following steps:

step 1, calculating curvature information of each point in a point cloud model, and extracting characteristic points in the model according to a set threshold value;

step 2, reconstructing neighborhood points on the basis of neighborhoods acquired by a K neighbor method according to the extracted feature points, and enabling the reconstructed neighborhoods to be on one surface;

step 3, according to the reconstructed neighborhood, using the three-dimensional position information of each point as a guide signal, and simultaneously adding curvature information as a weighting signal into the position guide signal, thereby performing linear transformation on each point in the point cloud model;

and 4, performing linear transformation on each point according to the linear transformation coefficient calculated in the step 3, and denoising the point cloud model.

2. The curvature and neighborhood reconstruction-based weighted guided point cloud model denoising method according to claim 1, wherein the curvature information of each point in the point cloud model is calculated in step 1, and feature points in the model are extracted according to a set threshold, specifically as follows:

step 1.1, for each point p_iCalculate that it corresponds to neighborhood N_iCurvature value of (a) (N)_i) Is composed of

Wherein λ is₀、λ₁、λ₂Is N_iOf the covariance matrix of, and λ₀＜λ₁＜λ₂Reflect N_iDistribution of three orthogonal singular vectors;

step 1.2, setting a threshold value t, wherein points with curvature values larger than the threshold value t are characteristic points, and points with curvature values smaller than the threshold value t are non-characteristic points.

3. The curvature and neighborhood reconstruction-based weighted guided point cloud model denoising method of claim 1, wherein the step 2 reconstructs neighborhood points based on the neighborhood obtained by the K-nearest neighbor method according to the extracted feature points, and makes the reconstructed neighborhood be on one surface, specifically as follows:

step 2.1, using K nearest neighbor method to each feature point p_iAssigning an initial neighborhood N;

step 2.2, for each neighborhood point p in the initial neighborhood N_ijObtaining the neighborhood point p by using the K nearest neighbor method_ijNeighborhood N of_ijWhile simultaneously applying the feature points p_iAt p_ijCandidate neighborhood of (c)

Initialised to contain only its own p_ijAnd a feature point p_iTwo points;

step 2.3, scan N_ijEach point in (1) determines whether the point can join the candidate neighborhood

In the method, the following judgment standard is constructed according to the curvature value of the current neighborhood and the position relation of the points:

wherein,

a standard value for measuring the curvature value of the current neighborhood and the position relation of the middle point of the neighborhood is expressed,

represents p_ijIn the neighborhood

Curvature value of the lower, K representing the neighborhood

The number of all the points, alpha and beta, are the user-defined control coefficients; p is a radical of_ijkRepresenting a neighborhood

K-th point in (1), 2, …, K representing the neighborhood

The total number of midpoints;

step 2.4, if the addition of this point is made in formula (2)

Is decreased, it means that the point canJoining to a candidate neighborhood

And 5 points nearest to the point are also added into the candidate neighborhood

Performing the following steps;

step 2.5, obtaining a characteristic point p for the rest points in the N_iThe candidate neighborhoods within, for each candidate neighborhood, are calculated according to equation (2)

Wherein the neighborhood corresponding to the minimum value is the feature point p_iReconstructed neighborhood N'.

4. The curvature-and-neighborhood reconstruction-based weighted guided point cloud model denoising method of claim 1,2 or 3, wherein the step 3 utilizes the three-dimensional position information of each point as a guiding signal according to the reconstructed neighborhood, and adds the curvature information as a weighted signal to the position guiding signal, thereby performing linear transformation on each point in the point cloud model, specifically as follows:

the cost function E of the weighted guided filtering algorithm is:

γ(i)＝(σ-t)^s(i)+χ (4)

s(i)＝-sgn(σ-t)×μ×σ (5)

wherein, N (p)_i) Indicates the current point p_iNeighborhood of p_ijIs a point in the neighborhood, a_iAnd b_iIs a linear transformation coefficient to be solved, and epsilon is a parameter for controlling the filtering effect; sigma is a curvature value of a point calculated before neighborhood reconstruction, and t is a threshold value for judging a characteristic point; χ is a positive number for preventing the weight γ (i) from being 0(ii) a Mu is a magnification factor of

And (4) dynamically determining.

5. The curvature and neighborhood reconstruction based weighted guided point cloud model denoising method according to claim 4, wherein the linear transformation coefficient calculated in step 4 is used for performing linear transformation on each point to realize point cloud model denoising, and the method comprises the following steps:

step 4.1, coefficient a of linear transformation obtained in step 3_i、b_iComprises the following steps:

wherein:

wherein, | N (p)_i) I represents a point p_iP, the number of points contained in the neighborhood of_ijIs p_iNeighborhood of N (p)_i) At one point of the inner side of the body,

is the central point of the neighborhood, and epsilon is a parameter for controlling the filtering effect;

step 4.2, according to the obtained linear transformation coefficient a_iAnd b_iAnd performing linear transformation on each characteristic point to obtain the position of the denoised point, and obtaining the denoised point cloud model after all the points are updated.

Images