A quantitative evaluation method and system for multidimensional point cloud structural similarity

Abstract

The present invention discloses a method and system for quantitatively evaluating the structural similarity of multidimensional point clouds. The method comprises: step 1) acquiring a three-dimensional point cloud image of a target object to obtain three-dimensional point cloud data; step 2) filtering the three-dimensional point cloud data; step 3) comparing the filtered point cloud data and the pre-filtered point cloud data with a standard three-dimensional point cloud set predetermined using a confidence level method, respectively, to obtain the three-dimensional structural similarity of the filtered data and the three-dimensional structural similarity of the pre-filtered data; and step 4) comparing the two three-dimensional structural similarities obtained in step 3) to evaluate the filtering effect of the three-dimensional point cloud data. The present invention utilizes a structural similarity method to objectively and quantitatively evaluate the filtering and sparseness effects of point cloud data, resolving the current situation in which there is no good objective evaluation of three-dimensional point cloud filtering.

Description

Technical Field

[0001] The present invention relates to the fields of scatter imaging and image processing, and in particular to a method and system for quantitatively evaluating multi-dimensional point cloud structural similarity. Background Art

[0002] Thanks to the rapid development of measuring equipment such as laser scanners, three-dimensional geometric models can better express the shape of objects. The point cloud data obtained by measuring three-dimensional models is becoming increasingly accurate and massive. Currently, general laser measuring equipment can obtain hundreds of thousands or even millions of point cloud data from an object's surface. This usually requires appropriate processing of the three-dimensional geometric model, such as lossy compression. Point cloud processing typically involves point cloud registration. As a core technology in point cloud data processing, this registration effect is affected by point cloud quality. Therefore, evaluating point cloud data quality has strong objective value, such as improving registration accuracy, enhancing spatial model integrity, and improving the performance of SLAM systems. Furthermore, the diverse sources of point cloud data, diverse preprocessing methods, and diverse point cloud noise types all increase the difficulty of defining and evaluating point cloud quality. Currently, the main evaluation methods for three-dimensional geometric model processing include subjective and objective evaluation methods.

[0003] Although many point cloud preprocessing methods are currently available, most rely on subjective evaluation to determine point cloud quality. There is no effective objective evaluation method for preprocessed point cloud quality. This paper proposes a structural similarity evaluation method for evaluating the quality of multidimensional point clouds—a quantitative evaluation method for structural similarity of multidimensional point clouds. This method is well-defined, accurate, and more systematic than HVS, closer to the human eye. Summary of the Invention

[0004] The purpose of the present invention is to evaluate the structural similarity of multidimensional point cloud quality, and a quantitative evaluation method and system for multidimensional point cloud structural similarity are proposed.

[0005] In order to achieve the above object, the present invention proposes a method for quantitatively evaluating the structural similarity of multidimensional point clouds, which includes:

[0006] Step 1) obtaining a three-dimensional point cloud image of the target object to obtain three-dimensional point cloud data;

[0007] Step 2) filtering the three-dimensional point cloud data;

[0008] Step 3) comparing the filtered point cloud data and the point cloud data before filtering with a standard three-dimensional point cloud set predetermined by a confidence method, respectively, to obtain the three-dimensional structural similarity after filtering and the three-dimensional structural similarity before filtering;

[0009] Step 4) comparing the similarities of the two 3D structures obtained in step 3) to evaluate the filtering effect of the 3D point cloud data.

[0010] As an improvement to the above method, the method further includes the step of predetermining a standard three-dimensional point cloud set using a confidence method; specifically, the step includes:

[0011] M point cloud images are superimposed at corresponding positions, and all point clouds contained in the M point cloud images are used as a point cloud set with 100% confidence, thereby obtaining a standard three-dimensional point cloud set, where M is greater than 10.

[0012] As an improvement to the above method, step 1) specifically includes: using a time-of-flight camera with spatial resolution capability to acquire three-dimensional point cloud data of the target object.

[0013] As an improvement to the above method, step 3) specifically includes:

[0014] Step 301) With the camera as the origin, obtain the 3D point cloud image after filtering, the 3D point cloud image before filtering, and the 3D coordinate information of different points in the standard 3D point cloud set;

[0015] Step 302) respectively calculates the mean, standard deviation and covariance of the three sets of three-dimensional coordinate information obtained in step 301) corresponding to the x-axis, y-axis and z-axis;

[0016] Step 303) Calculate the depth similarity index, depth contrast index and structural similarity index after and before filtering based on the mean, standard deviation and covariance obtained in step 302);

[0017] Step 304) Calculate the three-dimensional structural similarity after filtering and the three-dimensional structural similarity before filtering respectively based on the depth similarity index, depth contrast index and structural similarity index obtained in step 303).

[0018] As an improvement to the above method, step 302) specifically includes:

[0019] The x-axis mean μ is obtained according to the following formula Ipx,k , y-axis mean μ Ipy,k , z-axis mean μ Ipz,k for:

[0020]

[0021]

[0022]

[0023] Wherein, the subscript k∈[0,2] is an integer, where k=0 represents the standard 3D point cloud set, k=1 represents after filtering, k=2 represents before filtering, N represents the number of point clouds in the 3D point cloud image, Ip ix,a represents the i-th x-axis pixel in the k state; Ip iy,k Indicates the i-th y-axis pixel in state k, Ip iz,k represents the i-th z-axis pixel in state k;

[0024] According to the following formula, the x-axis standard deviation σ under the k state is obtained Ipix,k , y-axis standard deviation σ Ipiy,k , z-axis standard deviation σ Ipiz,k for:

[0025]

[0026]

[0027]

[0028] According to the following formula, the covariance σ of the filtered 3D point cloud image and the standard 3D point cloud set on the x-axis is obtained: Ipx,1,Ipx,0 , the covariance σ on the y-axis Ipy,1,Ipy,0 , the covariance σ on the z-axis Ipz,1,Ipz,0 for:

[0029]

[0030]

[0031]

[0032] As an improvement to the above method, step 303) specifically includes:

[0033] Step 303-1) The depth similarity index l(P) on the x-axis between the filtered 3D point cloud image and the standard 3D point cloud set is obtained according to the following formula: x,1 ,P x,0 ), depth contrast similarity index c(P x,1 ,P x,0 ), structural similarity index s(P x,1 ,P x,0 )for:

[0034]

[0035]

[0036]

[0037] Among them, C1, C2, and C3 are the corresponding index adjustment coefficients, and C1 = (K1N) 2 , C2=(K2N) 2 , K1=0.01, K2=0.03, N represents the number of points in the 3D point cloud image;

[0038] The structural similarity SSIM of the filtered 3D point cloud and the standard 3D point cloud on the x-axis is obtained according to the following formula: x,1,0 for:

[0039] SSIM x,1,0 =[l(P x,1 ,P x,0 )] α [c(P x,1 ,P x,0 )] β [s(P x,1 ,P x,0 )] γ

[0040] Among them, α, β, and γ are the weights of the corresponding depth similarity index, depth contrast similarity index, and structure similarity index respectively;

[0041] Step 303-2) uses the method of step 303-1) to obtain the depth similarity index l(P y,1 ,P y,0 ), depth contrast similarity index c(P y,1 ,P y,0 ), structural similarity index s(P y,1 ,P y,0 ), and the structural similarity SSIM between the filtered 3D point cloud and the standard 3D point cloud set on the y-axis y,1,0 ;

[0042] Step 303-3) uses the method of step 303-1) to obtain the depth similarity index l(P z,1 ,P z,0 ), depth contrast similarity index c(P z,1 ,P z,0 ), structural similarity index s(P z,1 ,P z,0 ), and the structural similarity SSIM between the filtered 3D point cloud and the standard 3D point cloud set on the z axis z,1,0 ;

[0043] Step 303-4) Obtain the filtered three-dimensional structural similarity 3D-SSIM according to the following formula:1,0 for:

[0044]

[0045] Among them, ρ, σ, ω are the weights corresponding to the z-axis, x-axis and y-axis respectively;

[0046] Step 303-5) uses the method of step 303-1) to step 303-4) to obtain the three-dimensional structural similarity 3D-SSIM before filtering 2,0 .

[0047] As an improvement to the above method, the step 304) specifically includes: performing the three-dimensional structural similarity 3D-SSIM after filtering. 1,0 3D-SSIM with 3D structure similarity before filtering 2,0 Compare and evaluate the filtering effect of 3D point cloud.

[0048] A multi-dimensional point cloud structure similarity quantitative evaluation system, the system includes: a point cloud data acquisition module, a filtering processing module, a three-dimensional structure similarity calculation module and a quantitative evaluation module; wherein,

[0049] The point cloud data acquisition module is used to acquire a three-dimensional point cloud image of the target object to obtain three-dimensional point cloud data;

[0050] The filtering processing module is used to perform filtering processing on the three-dimensional point cloud data;

[0051] The three-dimensional structure similarity calculation module is used to compare the filtered point cloud data and the point cloud data before filtering with a standard three-dimensional point cloud set predetermined by a confidence method, and obtain the three-dimensional structure similarity of the filtering process and the three-dimensional structure similarity before filtering respectively;

[0052] The quantitative evaluation module is used to compare the two three-dimensional structural similarities obtained by the three-dimensional structural similarity calculation module and evaluate the filtering effect of the three-dimensional point cloud data.

[0053] Compared with the existing technology, the advantages of this method are:

[0054] 1. The method of the present invention uses the structural similarity method to objectively and quantitatively evaluate the filtering and sparse effects of point cloud data, solving the current situation where there is no good objective evaluation of 3D point cloud filtering;

[0055] 2. The method of the present invention solves the difficulty in defining point cloud quality evaluation;

[0056] 3. The method of the present invention solves the problem that the previous evaluation filtering method can only rely on subjective evaluation without objective evaluation, and contributes to the objective evaluation method of point cloud preprocessing and provides a reference;

[0057] 4. The method of the present invention provides a reference for intuitive analysis of point cloud quality in the registration of multiple point clouds;

[0058] 5. The method of the present invention plays a certain role in the precise positioning of multiple point clouds and SLAM systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a flowchart of the method for quantitatively evaluating the structural similarity of multi-dimensional point clouds provided by the present invention. DETAILED DESCRIPTION

[0060] The present invention discloses a method and system for quantitatively evaluating the structural similarity of multidimensional point clouds. The method comprises: acquiring three-dimensional point cloud data of an object; using the point cloud data as distance information; respectively obtaining the mean, standard deviation and covariance of different axes in the point cloud data; obtaining a standard three-dimensional point cloud data set of the object using a confidence method; comparing the point cloud data before and after filtering with the standard three-dimensional point cloud data of the object obtained using the confidence method, and respectively comparing the mean, standard deviation and covariance of the coordinates of all points on the three-dimensional coordinate axis; performing weight distribution on the structural similarity values ​​on the three coordinate axes to finally obtain the three-dimensional structural similarity correlation, thereby evaluating the point cloud filtering effect.

[0061] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0062] Example 1

[0063] like Figure 1 As shown, embodiment 1 of the present invention discloses a method for quantitatively evaluating the similarity of multi-dimensional point cloud structures, comprising the following steps:

[0064] S1. Use the TOF camera to obtain the three-dimensional point cloud data of the target object;

[0065] S2. Analyze the point cloud data to obtain the three-dimensional coordinate information of different points in the point cloud with the camera as the origin;

[0066] S3. Calculate the mean, standard deviation and covariance of different axes in the point cloud data respectively;

[0067] S4, calculating depth similarity, depth contrast similarity and structural similarity indices;

[0068] S5. Obtaining a standard three-dimensional point cloud dataset of the object using a confidence method;

[0069] S6. Compare the point cloud data before and after filtering with the standard three-dimensional point cloud data of the object obtained using the confidence level, and compare the mean, standard deviation, and covariance of the coordinates of all points on the three-dimensional coordinate axis respectively;

[0070] S7. Weight the structural similarity values ​​on the three coordinate axes and finally obtain the three-dimensional structural similarity correlation to evaluate the point cloud filtering effect.

[0071] In the embodiment of the present invention, the HLT003S-001 time-of-flight camera is used to obtain point cloud data, which is used as distance information and three-dimensional coordinate information. Then, the three-dimensional coordinate information is obtained from the point cloud data obtained by the time-of-flight camera, and the mean, standard deviation and covariance of the point cloud data in different axes are calculated respectively. Using the confidence method

[0072] The above steps are further described below.

[0073] S1. Use the TOF camera to obtain the three-dimensional point cloud data of the target object;

[0074] S2. Analyze the point cloud data to obtain the three-dimensional coordinate information of different points in the point cloud with the camera as the origin;

[0075] S21, the different points in the point cloud are each point in the point cloud data;

[0076] S22. The three-dimensional coordinate information with the camera as the origin is the three-dimensional coordinates of each point in the point cloud measured by setting the camera coordinate information to (0, 0, 0).

[0077] S3. Calculate the mean, standard deviation and covariance of different axes in the point cloud data respectively;

[0078] S31, the mean, standard deviation, and covariance of different axes, namely the mean, standard deviation, and covariance of the x-axis, y-axis, and z-axis;

[0079] S32, the mean formula is:

[0080]

[0081] S33. The standard deviation formula is:

[0082]

[0083] S34. The covariance formula is:

[0084]

[0085] Among them, the above three formulas are the mean, standard deviation and covariance values ​​on one coordinate axis.

[0086] In the above formula, it is assumed that the number of point clouds in a point cloud data is N, and the pixels are recorded as Ip i , Ip 1i 、Ip 2i , represents the pixels in point cloud data 1 and point cloud data 2, μIp1 、μ Ip2 Represents the mean of point cloud data 1 and point cloud data 2.

[0087] This gives the x-axis mean μ Ipx,k , y-axis mean μ Ipy,k , z-axis mean μ Ipz,k , where the subscript k∈[0,2] is an integer, where k=0 represents the standard 3D point cloud set, k=1 represents after filtering, k=2 represents before filtering, N represents the number of point clouds in the 3D point cloud image, Ip ix,a represents the i-th x-axis pixel in the k state; Ip iy,a Indicates the i-th y-axis pixel in state k, Ip iz,a Represents the i-th z-axis pixel in state k.

[0088] Get the x-axis standard deviation σ Ipix,k , y-axis standard deviation σ Ipiy,k , z-axis standard deviation σ Ipiz,k .

[0089] Then we get the covariance σ between the filtered 3D point cloud and the standard 3D point cloud on the x-axis. Ipx,1,Ipx,0 , the covariance σ on the y-axis Ipy,1,Ipy,0 , the covariance σ on the z-axis Ipz,1,Ipz,0 ;

[0090] S4, calculating depth similarity, depth contrast similarity and structural similarity indices;

[0091] S41. The depth similarity index is:

[0092]

[0093] S42, depth contrast similarity index:

[0094]

[0095] S43, the structural similarity value is:

[0096]

[0097] Among them, C1, C2, and C3 are the corresponding index adjustment coefficients, and C1 = (K1N) 2 , C2=(K2N) 2 , Usually K1=0.01, K2=0.03, N represents the number of points in the 3D point cloud image;

[0098] S44, similarity is the cosine similarity, which is calculated by subtracting the mean of the two images:

[0099]

[0100] The depth similarity index l(P x,1 ,P x,0 ), depth contrast similarity index c(P x,1 ,P x,0 ), structural similarity index s(P x,1 ,P x,0 );

[0101] The depth similarity index l(P y,1 ,P y,0 ), depth contrast similarity index c(P y,1 ,P y,0 ), structural similarity index s(P y,1 ,P y,0 ),

[0102] The depth similarity index l(P z,1 ,P z,0 ), depth contrast similarity index c(P z,1 ,P z,0 ), structural similarity index s(P z,1 ,P z,0 ).

[0103] Similarly, the corresponding indicators of the three-dimensional point cloud before filtering and the standard three-dimensional point cloud set on different coordinate axes are obtained.

[0104] S5. Obtaining a standard three-dimensional point cloud dataset of the object using a confidence method;

[0105] The confidence method involves superimposing ten (or more) point cloud images at corresponding locations. For example, the point clouds included in all ten point cloud images are considered a 100% confidence point cloud set, while the point clouds included in all nine point cloud images are considered a 90% confidence point cloud set. As the number of images in the same point cloud set decreases, the confidence level gradually decreases until it reaches 10%. This method can obtain absolutely correct point cloud data points, thereby filtering out erroneous point cloud points.

[0106] S6. Compare the point cloud data before and after filtering with the standard three-dimensional point cloud data of the object obtained using the confidence level, and compare the mean, standard deviation, and covariance of the coordinates of all points on the three-dimensional coordinate axis respectively;

[0107] S61. The comparison includes the mean, standard deviation and covariance of the coordinates of the points in the point cloud data before and after filtering. The mean, standard deviation and covariance of all the coordinates of the points in the standard 3D point cloud data of the object obtained by using the confidence level.

[0108] S62,

[0109] SSIM=[l(P1,P2)] α [c(P1,P2)] β [s(P1,P2)] γ

[0110] The structural similarity SSIM between the filtered 3D point cloud and the standard 3D point cloud set on the x-axis is obtained. x,1,0 for:

[0111] SSIM x,1,0 =[l(P x,1 ,P x,0 )] α [c(P x,1 ,P x,0 )] β [s(P x,1 ,P x,0 )] γ

[0112] Among them, α, β, and γ are the weights of the corresponding depth similarity index, depth contrast similarity index, and structure similarity index respectively;

[0113] Similarly, the structural similarity SSIM of the filtered 3D point cloud and the standard 3D point cloud on the y-axis is obtained. y,1,0 , SSIM structural similarity in the z-axis z,1,0 .

[0114] Similarly, the structural similarity SSIM of the 3D point cloud before filtering and the standard 3D point cloud set in different coordinate axes is obtained. x,2,0 , SSIM y,2,0 , SSIM z,2,0 .

[0115] S7. Weight the structural similarity values ​​on the three coordinate axes and finally obtain the three-dimensional structural similarity correlation to evaluate the point cloud filtering effect.

[0116] The structural similarity algorithms of the three separate axes in the three-dimensional coordinates are multiplied together according to the weights. The three-dimensional structural similarity formula is:

[0117] 3D-SSIM=[SSIM z ] ρ [SSIM x ] σ [SSIM y ] ω

[0118] Among them, ρ, σ, ω are the weights corresponding to the z-axis, x-axis and y-axis respectively;

[0119] The filtered three-dimensional structural similarity 3D-SSIM is obtained. 1,0 for:

[0120]

[0121] Similarly, the three-dimensional structural similarity 3D-SSIM before filtering is obtained 2,0 , according to 3D-SSIM 1,0 and 3D-SSIM 2,0 Compare the results to evaluate the filtering effect.

[0122] Example 2

[0123] Embodiment 2 of the present invention proposes a multi-dimensional point cloud structural similarity quantitative evaluation system, which is implemented based on the method of embodiment 1. The system includes: a point cloud data acquisition module, a filtering processing module, a three-dimensional structure similarity calculation module and a quantitative evaluation module; wherein,

[0124] The point cloud data acquisition module is used to acquire a three-dimensional point cloud image of the target object to obtain three-dimensional point cloud data;

[0125] The filtering processing module is used to perform filtering processing on the three-dimensional point cloud data;

[0126] The three-dimensional structure similarity calculation module is used to compare the filtered point cloud data and the point cloud data before filtering with a standard three-dimensional point cloud set predetermined by a confidence method, and obtain the three-dimensional structure similarity of the filtering process and the three-dimensional structure similarity before filtering respectively;

[0127] The quantitative evaluation module is used to compare the two three-dimensional structural similarities obtained by the three-dimensional structural similarity calculation module and evaluate the filtering effect of the three-dimensional point cloud data.

[0128] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and are intended to be encompassed by the claims of the present invention.

Claims

1. A method for quantitatively evaluating the structural similarity of a multidimensional point cloud, the method comprising: Step 1) obtaining a three-dimensional point cloud image of the target object to obtain three-dimensional point cloud data; Step 2) filtering the three-dimensional point cloud data; Step 3) comparing the filtered point cloud data and the point cloud data before filtering with a standard three-dimensional point cloud set predetermined by a confidence method, respectively, to obtain the three-dimensional structural similarity after filtering and the three-dimensional structural similarity before filtering; Step 4) comparing the similarities of the two three-dimensional structures obtained in step 3) to evaluate the filtering effect of the three-dimensional point cloud data; The method further includes the step of predetermining a standard three-dimensional point cloud set using a confidence method; specifically, the step includes: Superimpose M point cloud images at corresponding positions, and take all the point clouds contained in the M point cloud images as the point cloud set with 100% confidence, thereby obtaining a standard three-dimensional point cloud set, where M is greater than 10; The step 3) specifically includes: Step 301) With the camera as the origin, obtain the 3D point cloud image after filtering, the 3D point cloud image before filtering, and the 3D coordinate information of different points in the standard 3D point cloud set; Step 302) respectively calculates the mean, standard deviation and covariance of the three sets of three-dimensional coordinate information obtained in step 301) corresponding to the x-axis, y-axis and z-axis; Step 303) Calculate the depth similarity index, depth contrast index and structural similarity index after and before filtering based on the mean, standard deviation and covariance obtained in step 302); Step 304) calculating the three-dimensional structural similarity after filtering and the three-dimensional structural similarity before filtering based on the depth similarity index, depth contrast index, and structural similarity index obtained in step 303); The step 303) specifically includes: Step 303-1) The depth similarity index l(P) on the x-axis between the filtered 3D point cloud image and the standard 3D point cloud set is obtained according to the following formula: x,1 ,P x,0 ), depth contrast similarity index c(P x,1 ,P x,0 ), structural similarity index s(P x,1 ,P x,0 )for: Among them, C1, C2, and C3 are the corresponding index adjustment coefficients, and C1 = (K1N) 2 , C2=(K2N) 2 , K1=0.01, K2=0.03, N represents the number of points in the 3D point cloud image; The structural similarity SSIM of the filtered 3D point cloud and the standard 3D point cloud on the x-axis is obtained according to the following formula: x,1,0 for: SSIM x,1,0 =[l(P x,1 ,P x,0 )] α [c(P x,1 ,P x,0 )] β [s(P x,1 ,P x,0 )] γ Among them, α, β, and γ are the weights of the corresponding depth similarity index, depth contrast similarity index, and structure similarity index respectively; Step 303-2) uses the method of step 303-1) to obtain the depth similarity index l(P y,1 ,P y,0 ), depth contrast similarity index c(P y,1 ,P y,0 ), structural similarity index s(P y,1 ,P y,0 ), and the structural similarity SSIM between the filtered 3D point cloud and the standard 3D point cloud set on the y-axis y,1,0 ; Step 303-3) uses the method of step 303-1) to obtain the depth similarity index l(P z,1 ,P z,0 ), depth contrast similarity index c(P z,1 ,P z,0 ), structural similarity index s(P z,1 ,P z,0 ), and the structural similarity SSIM between the filtered 3D point cloud and the standard 3D point cloud set on the z axis z,1,0 ; Step 303-4) Obtain the filtered three-dimensional structural similarity 3D-SSIM according to the following formula: 1,0 for: Among them, ρ, σ, ω are the weights corresponding to the z-axis, x-axis and y-axis respectively; Step 303-5) uses the method of step 303-1) to step 303-4) to obtain the three-dimensional structural similarity 3D-SSIM before filtering 2,0 .

2. The method for quantitatively evaluating the similarity of multidimensional point clouds according to claim 1, wherein: The step 1) specifically includes: using a time-of-flight camera with spatial resolution capability to acquire three-dimensional point cloud data of the target object.

3. The method for quantitatively evaluating the structural similarity of multidimensional point clouds according to claim 1, wherein: The step 302) specifically includes: The x-axis mean μ is obtained according to the following formula Ipx,k , y-axis mean μ Ipy,k , z-axis mean μ Ipz,k for: Wherein, the subscript k∈[0,2] is an integer, k=0 represents the standard 3D point cloud set, k=1 represents after filtering, k=2 represents before filtering, N represents the number of point clouds in the 3D point cloud image, Ip ix,k represents the i-th x-axis pixel in the k state; Ip iy,k Indicates the i-th y-axis pixel in state k, Ip iz,k represents the i-th z-axis pixel in state k; According to the following formula, the x-axis standard deviation σ under the k state is obtained Ipix,k , y-axis standard deviation σ Ipiy,k , z-axis standard deviation σ Ipiz,k for: According to the following formula, the covariance σ of the filtered 3D point cloud image and the standard 3D point cloud set on the x-axis is obtained: Ipx,1,Ipx,0 , the covariance σ on the y-axis Ipy,1,Ipy,0 , the covariance σ on the z-axis Ipz,1,Ipz,0 for:

4. The method for quantitatively evaluating the structural similarity of multidimensional point clouds according to claim 1, wherein: The step 304) specifically includes: performing the three-dimensional structural similarity 3D-SSIM after filtering. 1,0 3D-SSIM with 3D structure similarity before filtering 2,0 Compare and evaluate the filtering effect of 3D point cloud.

5. A system based on the multi-dimensional point cloud structural similarity quantitative evaluation method according to claim 1, characterized in that: The system includes: a point cloud data acquisition module, a filtering processing module, a three-dimensional structure similarity calculation module and a quantitative evaluation module; wherein, The point cloud data acquisition module is used to acquire a three-dimensional point cloud image of the target object to obtain three-dimensional point cloud data; The filtering processing module is used to perform filtering processing on the three-dimensional point cloud data; The three-dimensional structure similarity calculation module is used to compare the filtered point cloud data and the point cloud data before filtering with a standard three-dimensional point cloud set predetermined by a confidence method, and obtain the three-dimensional structure similarity of the filtering process and the three-dimensional structure similarity before filtering respectively; The quantitative evaluation module is used to compare the two three-dimensional structural similarities obtained by the three-dimensional structural similarity calculation module and evaluate the filtering effect of the three-dimensional point cloud data.

Images