CN117011299A - Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics - Google Patents
Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics Download PDFInfo
- Publication number
- CN117011299A CN117011299A CN202311280774.XA CN202311280774A CN117011299A CN 117011299 A CN117011299 A CN 117011299A CN 202311280774 A CN202311280774 A CN 202311280774A CN 117011299 A CN117011299 A CN 117011299A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- dimensional gradient
- reference point
- points
- gradient
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012952 Resampling Methods 0.000 title claims abstract description 33
- 238000001303 quality assessment method Methods 0.000 title claims abstract description 11
- 238000000034 method Methods 0.000 claims abstract description 30
- 230000004044 response Effects 0.000 claims abstract description 25
- 238000011176 pooling Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims description 29
- 238000013441 quality evaluation Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000012546 transfer Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000010845 search algorithm Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 4
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 230000000007 visual effect Effects 0.000 description 6
- 238000013461 design Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H17/00—Networks using digital techniques
- H03H17/02—Frequency selective networks
- H03H17/0248—Filters characterised by a particular frequency response or filtering method
- H03H17/028—Polynomial filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Algebra (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Hardware Design (AREA)
- Processing Or Creating Images (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a reference point cloud quality assessment method and a system for merging graph resampling and gradient characteristics, which relate to the field of image processing, and comprise the following steps: extracting key points from the input reference point cloud by using a graph-based key point resampling method; dividing a local neighborhood group of a reference point cloud and a distortion point cloud by taking the key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space; respectively extracting three-dimensional gradient amplitude characteristics and three-dimensional gradient pattern characteristics of the reference point cloud and the distortion point cloud, calculating three-dimensional gradient amplitude similarity based on the three-dimensional gradient amplitude characteristics, calculating three-dimensional gradient pattern similarity based on the three-dimensional gradient pattern characteristics, and calculating joint three-dimensional gradient characteristic similarity; and based on the similarity of the combined three-dimensional gradient characteristics, carrying out weighted pooling by using the response intensity values to obtain the objective quality score of the distorted point cloud. The application has higher identification accuracy, sensitivity and robustness.
Description
Technical Field
The application relates to the field of image processing, in particular to a reference point cloud quality evaluation method and system integrating graph resampling and gradient characteristics.
Background
With the continuous development of imaging technology, the demand for three-dimensional data in the field of computer vision is increasing. The 3D point cloud has a wide application prospect in a modern communication system due to the flexible expression form, and becomes one of three-dimensional data formats commonly used in immersive media application. The three-dimensional data contains depth information, brings more visual space sense and stereoscopic sense on visual expression, and accords with the perception and understanding of a human visual system on objects during observation.
However, the point cloud is affected by noise and distortion in the data processing process, so that the visual quality of the display content watched by human eyes is reduced while the distortion is caused, and the satisfaction degree of the end user on visual experience is affected. Therefore, the point cloud quality evaluation method which simultaneously considers the special structural characteristics of the point cloud and accords with the human visual characteristics has important application value in the aspects of acquisition, processing and application of the point cloud data, and the point cloud quality evaluation has higher theoretical research significance and practical application value at present.
Disclosure of Invention
The application mainly aims to overcome the defects in the prior art, and provides a reference point cloud quality evaluation method and a system for merging graph resampling and gradient characteristics, which effectively describe gradient direction characteristics and response intensity changes caused by point cloud distortion, accord with subjective perceptibility of human eyes to distorted point clouds, and have higher recognition accuracy, sensitivity and robustness.
The application adopts the following technical scheme:
in one aspect, a reference point cloud quality assessment method fusing graph resampling and gradient features includes:
s101, extracting key points from an input reference point cloud by using a key point resampling method based on a graph;
s102, dividing a local neighborhood group of a reference point cloud and a distortion point cloud by taking a key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space;
s103, respectively extracting three-dimensional gradient amplitude characteristics and three-dimensional gradient pattern characteristics of the reference point cloud and the distortion point cloud in the local area, calculating three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient amplitude characteristics, calculating three-dimensional gradient pattern similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient pattern characteristics, and calculating joint three-dimensional gradient characteristic similarity based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity;
and S104, carrying out weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity to obtain the objective quality score of the distorted point cloud.
Preferably, the S101 specifically includes:
reference point cloud for inputA high-pass diagram filtering method is used for the +.>Resampling is carried out to extract key points; in particular using the graph transfer operator +.>To describe the relative positional relationship between nodes, +.>Representing a graph transfer operator +.>Is a dimension of (2); use of adjacency matrix->And transfer matrix->To express +.>I.e. +.>Each of the linear shift invariant map filters is +.>Polynomial function of (c):
wherein,representing a linear shift invariant graph filter; />The +.>Personal factor (F)>Representing +.>Power of the order; />Indicate->A coefficient; />Representing the length of the graphics filter; />A representation matrix; />Representation->An inverse matrix of (a); />Representing the identity matrix;
implementation using a haar-like graph filterThe high pass filtering of (2) is as follows:
wherein,representing a high pass filtering of the haar-like graph filter; />Representation->Is->A feature vector; />Representation->Is a characteristic value of (2); in the graph vertex field, +.>Point->The frequency response formula of (2) is:
wherein,indicate->Point->Is a frequency response formula for (1); />Representing the total number of points in the spatial frequency domain; />Indicate->Point->Is a neighboring point of (2);
for input reference point cloudExtracting key points to generate skeleton point cloud->。
Preferably, the step S102 specifically includes:
the resampled key point skeleton point cloudIs used as a reference point cloud +.>And distortion point cloud->For +.>The%>Key points->Use +.>And->Euclidean distance pair of corresponding geometric parts in +.>The neighbors of the points are clustered as follows:
wherein,expressed as key points->Is the center of sphere>A neighborhood group of reference point clouds for a radius; />Expressed as key points->Is the center of sphere>A neighborhood group of the distorted point cloud for a radius; />Representing a set of all points in a keypoint skeleton of a reference point cloud; />Key point skeleton for representing distortion point cloudA set of all points in (a); />Representing the square of the two norms.
Preferably, the step S103 specifically includes:
from a reference point cloud Three-dimensional gradient amplitude characteristic>And three-dimensional gradient pattern->And distortion point cloud->Three-dimensional gradient amplitude characteristic>And gradient pattern->Calculating to obtain three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud>Three-dimensional gradient direction similarity ++>The calculation formula is as follows:
wherein,and->Is a constant for ensuring the stability of the numerical value;
points for a reference point cloud local areaThree-dimensional gradient amplitude feature->The calculation formula is as follows:
wherein,、/>、/>respectively representing gradient amplitude differences of points in the reference point cloud local neighborhood group in the x, y and z directions; />Representing a gradient calculation method on a point xyz axis in a reference point cloud local neighborhood group;
introducing three-dimensional gradient directions to describe change information on a point cloud space so as to obtain key pointsIs the center of sphere, is->Constructing a local area for a radius forA point falling on the local area +.>The ball coordinate value formula for calculating this point is as follows:
wherein,representation dot->Polar angle of>Representation dot->Carrying out direction clustering according to the ball coordinate values of all points in the local area; setting eight directions of the three-dimensional gradient, wherein the closer the ball coordinate values of points in the local area are to which direction, the clustering is carried out in the direction; the three-dimensional gradient eight directions are calculated according to polar angles and azimuth angles and distributed at equal intervals in space; the specific numerical calculation formula is as follows:
wherein the method comprises the steps of；/>Representing polar angle/>Is (are) direction of->Representing azimuth angleThe two directions are mutually combined to form a three-dimensional gradient eight direction;
three-dimensional gradient patternThe calculation formula is as follows:
wherein,for->A convolution kernel in the three-dimensional eight direction, the direction of the convolution kernel depending on the direction of which three-dimensional gradient the point clusters in; />Representing a convolution operation;
for points of a distorted point cloud local areaThree-dimensional gradient amplitude feature->The calculation formula is as follows:
wherein,、/>、/>gradient amplitude differences of points in the distortion point cloud local neighborhood group in the x, y and z directions are respectively represented; />Representing a gradient calculation method on a point xyz axis in a local neighborhood group of the distorted point cloud;
introducing three-dimensional gradient directions to describe change information on a point cloud space so as to obtain key pointsIs the center of sphere, is->Constructing a local region for a radius, for a point lying in the local region +.>The ball coordinate value formula for calculating this point is as follows:
wherein,representation dot->Polar angle of>Representation dot->Carrying out direction clustering according to the ball coordinate values of all points in the local area; setting eight directions of the three-dimensional gradient, wherein the closer the ball coordinate values of points in the local area are to which direction, the clustering is carried out in the direction; the three-dimensional gradient eight directions are calculated according to polar angles and azimuth angles and distributed at equal intervals in space; the specific numerical calculation formula is as follows:
wherein the method comprises the steps of；/>Representing polar angle->Is (are) direction of->Representing azimuth angleThe two directions are mutually combined to form a three-dimensional gradient eight direction;
three-dimensional gradient patternThe calculation formula is as follows:
wherein,for->A convolution kernel in the three-dimensional eight direction, the direction of the convolution kernel depending on the direction of which three-dimensional gradient the point clusters in;
and calculating the feature similarity of the combined three-dimensional gradient based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity, wherein the feature similarity is as follows:
wherein the parameters areIs used for adjusting->And->Positive numbers of relative importance between them.
Preferably, the step S104 is specifically as follows:
wherein,response intensity values for points within the reference point cloud local area; />Distorting response intensity values of points within the point cloud local area; />Representing the number of local areas; />Representing a local range of the reference point cloud and the distorted point cloud, i.e. the xyz three-dimensional point falls within the local range; />Representing the objective quality score of the distorted point cloud.
Preferably, the calculation method of the response intensity value is a K neighbor search algorithm.
In another aspect, a reference point cloud quality assessment system that fuses graph resampling and gradient features includes:
the key point extraction module is used for extracting key points from the input reference point cloud by using a graph-based key point resampling method;
the local area point clustering module is used for dividing a local neighborhood group of the reference point cloud and the distortion point cloud by taking the key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space;
the gradient feature similarity calculation module is used for respectively extracting three-dimensional gradient amplitude features and three-dimensional gradient pattern features of the reference point cloud and the distortion point cloud in the local area, calculating the three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient amplitude features, calculating the three-dimensional gradient pattern similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient pattern features, and calculating the joint three-dimensional gradient feature similarity based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity;
and the quality evaluation module is used for carrying out weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity to obtain the objective quality score of the distorted point cloud.
Compared with the prior art, the application has the following beneficial effects:
(1) According to the application, geometric information of a reference point cloud is extracted by a graph-based resampling method to generate a key point skeleton, a local region parameterized model taking key points as sphere centers is constructed for the reference point cloud and the distorted point cloud, the similarity of gradient amplitude values and gradient directions is measured respectively, and finally the similarity of the two characteristics is weighted and pooled by using a weighted combination calculation method of response intensity values, so that a point cloud quality evaluation score is calculated, the gradient direction characteristics and response intensity changes caused by the point cloud distortion are effectively described, the subjective perceptibility of human eyes on the distorted point cloud is met, and the method has high recognition accuracy, sensitivity and robustness, and good point cloud quality evaluation performance;
(2) According to the application, the geometric information of the reference point cloud is extracted by a graph-based resampling method to generate the key point skeleton, so that the problem that feature mapping of the reference and distorted point cloud is difficult in quality evaluation is solved, the sampled key points not only accord with human perception, but also obviously reduce calculation complexity.
Drawings
FIG. 1 is a flowchart of a reference point cloud quality assessment method for merging graph resampling and gradient features according to an embodiment of the present application;
FIG. 2 is a structural flow diagram of a reference point cloud quality assessment method that fuses graph resampling and gradient features in accordance with an embodiment of the present application;
fig. 3 is a block diagram of a reference point cloud quality assessment system that fuses graph resampling and gradient features in accordance with an embodiment of the present application.
Detailed Description
The application will be further illustrated with reference to specific examples. It is to be understood that these examples are illustrative of the present application and are not intended to limit the scope of the present application. Furthermore, it should be understood that various changes and modifications can be made by one skilled in the art after reading the teachings of the present application, and such equivalents are intended to fall within the scope of the application as defined in the appended claims.
Referring to fig. 1 and fig. 2, the reference point cloud quality assessment method integrating graph resampling and gradient features of the present application specifically includes the following steps:
s101, extracting key points from the input reference point cloud by using a key point resampling method based on the graph.
Concrete embodimentsIn the form of a reference point cloud for receiving inputA high-pass diagram filtering method is used for +.>Resampling is performed to extract keypoints. Use of the graph transfer operator->To describe the relative positional relationship between nodes, +.>Representing a graph transfer operator +.>Is a dimension of (2); />Adjacency matrix can be used +.>And transfer matrix->To express, i.e.)>. Each linear shift invariant graph filter is +.>Polynomial function of (c):
wherein,representing a linear shift invariant graph filter; />Representing a graph transfer operatorIs->Power of the order; />Indicate->A coefficient; />Representing the length of the graphics filter; />A representation matrix; />Representation->An inverse matrix of (a); />Representing the identity matrix. In this embodiment a class Ha Ertu filter is used to implement +.>Is a high pass filter of (2):
wherein,representing a high pass filtering of the haar-like graph filter; />Representation->Is->Features ofVector; />Representation->Is a characteristic value of (2); in the graph vertex field, +.>Point->The frequency response formula (of a point in the reference point cloud spatial frequency domain) is:
wherein,indicate->Point->Is a frequency response formula for (1); />Representing the total number of points in the spatial frequency domain; />Indicate->Point->Is a neighboring point of (2);
is used for reference point cloud in spatial frequency domain>Ordering for sampling, which reflects the +.>How much information about a point can be collected from neighboring points, the more information points are more likely to be chosen as keypoints. Finally, filter length is taken>And the number of valid keypoints after resampling +.>To achieve a balance between complexity and efficiency.
S102, dividing a local neighborhood group of the reference point cloud and the distortion point cloud by taking the key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space.
Specifically, the resampled key point skeleton point cloudIs used as a reference point cloud +.>And distortion point cloud->For +.>The%>Key points->The method uses->And->Euclidean distance pair of corresponding geometric parts in +.>The neighbors of the points are clustered as follows:
wherein,and->Represents the key point->Is the center of sphere>A neighborhood group for a reference and distorted point cloud of radius; />Representing a set of all points in a keypoint skeleton of a reference point cloud; />Representing a set of all points in a key point skeleton of the distorted point cloud; />Representing the square of the two norms.
And S103, respectively extracting three-dimensional gradient amplitude characteristics and three-dimensional gradient pattern characteristics of the reference point cloud and the distortion point cloud in the local area, calculating three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient amplitude characteristics, calculating three-dimensional gradient pattern similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient pattern characteristics, and calculating joint three-dimensional gradient characteristic similarity based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity.
Specifically, on the basis of extracting the three-dimensional gradient amplitude characteristics, a three-dimensional gradient eight-direction calculation method is expanded, a K neighbor search algorithm is adopted to calculate the response intensity value of each point, and a local parameterized model of three-dimensional gradient information is constructed as follows.
Points for a three-dimensional point cloud local areaThe three-dimensional gradient magnitude calculation formula is as follows:
wherein,、/>、/>gradient magnitude differences in three directions for points within each local neighborhood group are shown.
The three-dimensional gradient eight-direction division angle formula is as follows:
the calculation formula of the gradient pattern is as follows:
wherein,point calculated for eight-directional division angle of three-dimensional gradient +.>A convolution kernel of the direction; />Representing a convolution operation.
And extracting three-dimensional gradient features of the reference point cloud and the distorted point cloud by using a local parameterized model, and calculating feature similarity of the three-dimensional gradient features, including three-dimensional gradient amplitude features and gradient pattern features.
Specifically, according to the reference point cloud Three-dimensional gradient amplitude characteristic>And gradient patternAnd distortion point cloud->Three-dimensional gradient amplitude characteristic>And gradient pattern->Calculating to obtain three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud>And three-dimensional gradient direction similarityThe calculation formula is as follows:
wherein,and->Is a constant for ensuring numerical stability.
In the embodiment, three-dimensional gradient amplitude and direction similarity are jointly considered as comprehensive indexes for evaluating the quality of the point cloud, and the calculation mode can be expressed as a combination of the three-dimensional gradient amplitude and the direction similarity:
wherein the parameters areIs used for adjusting->And->Positive numbers of relative importance between them.
And S104, carrying out weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity to obtain the objective quality score of the distorted point cloud.
Specifically, a three-dimensional response intensity value weighted pooling method is used for calculating to obtain objective quality scores of point cloudsThe following are provided:
wherein,and->For response intensity values of the reference point cloud and the distorted point cloud, a larger value in the reference point cloud and the distorted point cloud is selected as a weight in a three-dimensional edge intensity pooling strategy>；/>Representing a local range of the reference point cloud and the distorted point cloud, i.e. the xyz three-dimensional point falls within the local range; />Representing the objective quality score of the distorted point cloud.
Referring to fig. 3, the embodiment further discloses a reference point cloud quality evaluation system fusing graph resampling and gradient features, which includes:
the key point extraction module 301 is configured to extract key points from an input reference point cloud by using a graph-based key point resampling method;
the local area point clustering module 302 is configured to divide a local neighborhood group of the reference point cloud and the distorted point cloud with the key point as a center, and cluster the points in each local area according to euclidean distances between the key point and other points in the coordinate space;
the gradient feature similarity calculation module 303 is configured to extract three-dimensional gradient amplitude features and three-dimensional gradient pattern features of the reference point cloud and the distorted point cloud in the local area, calculate three-dimensional gradient amplitude similarities of the reference point cloud and the distorted point cloud based on the three-dimensional gradient amplitude features, calculate three-dimensional gradient pattern similarities of the reference point cloud and the distorted point cloud based on the three-dimensional gradient pattern features, and calculate joint three-dimensional gradient feature similarities based on the three-dimensional gradient amplitude similarities and the three-dimensional gradient pattern similarities;
the quality evaluation module 304 is configured to perform weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity, so as to obtain an objective quality score of the distorted point cloud.
The embodiment of the reference point cloud quality evaluation method for integrating the image resampling and gradient characteristics is not repeated.
The foregoing is merely illustrative of specific embodiments of the present application, but the design concept of the present application is not limited thereto, and any insubstantial modification of the present application by using the design concept shall fall within the scope of the present application.
Claims (7)
1. The reference point cloud quality assessment method integrating graph resampling and gradient characteristics is characterized by comprising the following steps of:
s101, extracting key points from an input reference point cloud by using a key point resampling method based on a graph;
s102, dividing a local neighborhood group of a reference point cloud and a distortion point cloud by taking a key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space;
s103, respectively extracting three-dimensional gradient amplitude characteristics and three-dimensional gradient pattern characteristics of the reference point cloud and the distortion point cloud in the local area, calculating three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient amplitude characteristics, calculating three-dimensional gradient pattern similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient pattern characteristics, and calculating joint three-dimensional gradient characteristic similarity based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity;
and S104, carrying out weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity to obtain the objective quality score of the distorted point cloud.
2. The method for evaluating quality of reference point cloud of fusion map resampling and gradient characteristics according to claim 1, wherein S101 specifically comprises:
reference point cloud for inputA high-pass diagram filtering method is used for the +.>Resampling is carried out to extract key points; in particular using the graph transfer operator +.>To describe the relative positional relationship between nodes, +.>Representing a graph transfer operator +.>Is a dimension of (2); use of adjacency matrix->And transfer matrix->To express +.>I.e. +.>Each of the linear shift invariant map filters is +.>Polynomial function of (c):
；
wherein,representing a linear shift invariant graph filter; />The +.>The number of factors that are used to determine the quality of the product,representing +.>Power of the order; />Indicate->A coefficient; />Representing the length of the graphics filter; />A representation matrix; />Representation->An inverse matrix of (a); />Representing the identity matrix;
implementation using a haar-like graph filterThe high pass filtering of (2) is as follows:
；
wherein,representing a high pass filtering of the haar-like graph filter; />Representation->Is->A feature vector; />Representation->Is a characteristic value of (2); in the graph vertex field, +.>Point->The frequency response formula of (2) is:
；
wherein,indicate->Point->Is a frequency response formula for (1); />Representing the total number of points in the spatial frequency domain; />Represent the firstPoint->Is a neighboring point of (2);
for input reference point cloudExtracting key points to generate skeleton point cloud->。
3. The method for evaluating quality of reference point cloud for merging graph resampling and gradient features according to claim 2, wherein S102 specifically comprises:
the resampled key point skeleton point cloudIs used as a reference point cloud +.>And distortion point cloud->For +.>The%>Key points->Use +.>And->Euclidean distance pair of corresponding geometric parts in +.>The neighbors of the points are clustered as follows:
；
wherein,expressed as key points->Is the center of sphere>A neighborhood group of reference point clouds for a radius; />Expressed as key points->Is the center of sphere>A neighborhood group of the distorted point cloud for a radius; />Representing referencesA set of all points in a key point skeleton of the point cloud; />Representing a set of all points in a key point skeleton of the distorted point cloud; />Representing the square of the two norms.
4. The method for evaluating quality of reference point cloud for merging graph resampling and gradient features according to claim 3, wherein S103 specifically comprises:
from a reference point cloudThree-dimensional gradient amplitude characteristic>And three-dimensional gradient pattern->And distortion point cloud->Three-dimensional gradient amplitude characteristic>And gradient pattern->Calculating to obtain three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud>Three-dimensional gradient direction similarity ++>The calculation formula is as follows:
；
；
wherein,and->Is a constant for ensuring the stability of the numerical value;
points for a reference point cloud local areaThree-dimensional gradient amplitude feature->The calculation formula is as follows:
；
；
wherein,、/>、/>respectively representing gradient amplitude differences of points in the reference point cloud local neighborhood group in the x, y and z directions; />Representing a gradient calculation method on a point xyz axis in a reference point cloud local neighborhood group;
introducing three-dimensional gradient directions to describe change information on a point cloud space so as to obtain key pointsIs the center of sphere, is->Constructing a local region for a radius, for a point lying in the local region +.>The ball coordinate value formula for calculating this point is as follows:
；
；
wherein,representation dot->Polar angle of>Representation dot->Carrying out direction clustering according to the ball coordinate values of all points in the local area; setting eight directions of the three-dimensional gradient, wherein the closer the ball coordinate values of points in the local area are to which direction, the clustering is carried out in the direction; the three-dimensional gradient eight directions are calculated according to polar angles and azimuth angles and distributed at equal intervals in space; calculation formula for specific valueThe formula is:
；
；
wherein the method comprises the steps of；/>Representing polar angle->Is (are) direction of->Representing azimuth angleThe two directions are mutually combined to form a three-dimensional gradient eight direction;
three-dimensional gradient patternThe calculation formula is as follows:
；
wherein,for->A convolution kernel in the three-dimensional eight direction, the direction of the convolution kernel depending on the direction of which three-dimensional gradient the point clusters in; />Representing a convolution operation;
for points of a distorted point cloud local areaThree-dimensional gradient amplitude feature->The calculation formula is as follows:
；
；
wherein,、/>、/>gradient amplitude differences of points in the distortion point cloud local neighborhood group in the x, y and z directions are respectively represented; />Representing a gradient calculation method on a point xyz axis in a local neighborhood group of the distorted point cloud;
introducing three-dimensional gradient directions to describe change information on a point cloud space so as to obtain key pointsIs the center of sphere, is->Constructing local areas for radii to fall onPoint ∈of local region>The ball coordinate value formula for calculating this point is as follows:
；
；
wherein,representation dot->Polar angle of>Representation dot->Carrying out direction clustering according to the ball coordinate values of all points in the local area; setting eight directions of the three-dimensional gradient, wherein the closer the ball coordinate values of points in the local area are to which direction, the clustering is carried out in the direction; the three-dimensional gradient eight directions are calculated according to polar angles and azimuth angles and distributed at equal intervals in space; the specific numerical calculation formula is as follows:
；
；
wherein the method comprises the steps of；/>Representing polar angle->Is (are) direction of->Representing azimuth angleThe two directions are mutually combined to form a three-dimensional gradient eight direction;
three-dimensional gradient patternThe calculation formula is as follows:
；
wherein,for->A convolution kernel in the three-dimensional eight direction, the direction of the convolution kernel depending on the direction of which three-dimensional gradient the point clusters in;
and calculating the feature similarity of the combined three-dimensional gradient based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity, wherein the feature similarity is as follows:
；
wherein the parameters areIs used for adjusting->And->Positive numbers of relative importance between them.
5. The method for evaluating quality of reference point cloud for merging graph resampling and gradient features according to claim 4, wherein S104 is specifically as follows:
；
；
wherein,response intensity values for points within the reference point cloud local area; />Distorting response intensity values of points within the point cloud local area; />Representing the number of local areas; />Representing a local range of the reference point cloud and the distorted point cloud, i.e. the xyz three-dimensional point falls within the local range; />Representing the objective quality score of the distorted point cloud.
6. The method for evaluating the quality of the reference point cloud fusing map resampling and gradient features of claim 5, wherein the method for calculating the response intensity value is a K-nearest neighbor search algorithm.
7. A reference point cloud quality assessment system that fuses graph resampling and gradient features, comprising:
the key point extraction module is used for extracting key points from the input reference point cloud by using a graph-based key point resampling method;
the local area point clustering module is used for dividing a local neighborhood group of the reference point cloud and the distortion point cloud by taking the key point as a center, and clustering the points in each local area according to Euclidean distance between the key point and other points in a coordinate space;
the gradient feature similarity calculation module is used for respectively extracting three-dimensional gradient amplitude features and three-dimensional gradient pattern features of the reference point cloud and the distortion point cloud in the local area, calculating the three-dimensional gradient amplitude similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient amplitude features, calculating the three-dimensional gradient pattern similarity of the reference point cloud and the distortion point cloud based on the three-dimensional gradient pattern features, and calculating the joint three-dimensional gradient feature similarity based on the three-dimensional gradient amplitude similarity and the three-dimensional gradient pattern similarity;
and the quality evaluation module is used for carrying out weighted pooling by using the response intensity value based on the joint three-dimensional gradient feature similarity to obtain the objective quality score of the distorted point cloud.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311280774.XA CN117011299B (en) | 2023-10-07 | 2023-10-07 | Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311280774.XA CN117011299B (en) | 2023-10-07 | 2023-10-07 | Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117011299A true CN117011299A (en) | 2023-11-07 |
| CN117011299B CN117011299B (en) | 2024-02-20 |
Family
ID=88562197
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311280774.XA Active CN117011299B (en) | 2023-10-07 | 2023-10-07 | Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117011299B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117872389A (en) * | 2024-01-19 | 2024-04-12 | 武汉万曦智能科技有限公司 | Wireless speed measuring method and system for field vehicle |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113850748A (en) * | 2020-06-09 | 2021-12-28 | 上海交通大学 | Point cloud quality evaluation system and method |
| CN114331989A (en) * | 2021-12-16 | 2022-04-12 | 重庆邮电大学 | Full-reference 3D point cloud quality evaluation method based on point feature histogram geodesic distance |
| WO2023288231A1 (en) * | 2021-07-13 | 2023-01-19 | Tencent America LLC | Image based sampling metric for quality assessment |
| WO2023093824A1 (en) * | 2021-11-26 | 2023-06-01 | 中兴通讯股份有限公司 | Point cloud quality evaluation method, and device and storage medium |
| CN116205886A (en) * | 2023-03-07 | 2023-06-02 | 重庆邮电大学 | Point cloud quality assessment method based on relative entropy |
-
2023
- 2023-10-07 CN CN202311280774.XA patent/CN117011299B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113850748A (en) * | 2020-06-09 | 2021-12-28 | 上海交通大学 | Point cloud quality evaluation system and method |
| WO2023288231A1 (en) * | 2021-07-13 | 2023-01-19 | Tencent America LLC | Image based sampling metric for quality assessment |
| WO2023093824A1 (en) * | 2021-11-26 | 2023-06-01 | 中兴通讯股份有限公司 | Point cloud quality evaluation method, and device and storage medium |
| CN114331989A (en) * | 2021-12-16 | 2022-04-12 | 重庆邮电大学 | Full-reference 3D point cloud quality evaluation method based on point feature histogram geodesic distance |
| CN116205886A (en) * | 2023-03-07 | 2023-06-02 | 重庆邮电大学 | Point cloud quality assessment method based on relative entropy |
Non-Patent Citations (2)
| Title |
|---|
| HUANQIANG ZENG: "Point Cloud Quality Assessment via 3D Edge Similarity Measurement", 《IEEE SIGNAL PROCESSING LETTERS ( VOLUME: 29)》, pages 1804 * |
| 郭小敏: "基于引导调制的彩色点云无参考质量评价方法", 《光电子·激光》, vol. 34, no. 7, pages 713 - 722 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117872389A (en) * | 2024-01-19 | 2024-04-12 | 武汉万曦智能科技有限公司 | Wireless speed measuring method and system for field vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117011299B (en) | 2024-02-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101236602B (en) | Image processing apparatus, image processing method and computer program | |
| CN107833181B (en) | Three-dimensional panoramic image generation method based on zoom stereo vision | |
| CN102236794B (en) | Recognition and pose determination of 3D objects in 3D scenes | |
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| CN100559398C (en) | Automatic deepness image registration method | |
| CN104574347B (en) | Satellite in orbit image geometry positioning accuracy evaluation method based on multi- source Remote Sensing Data data | |
| CN108876749A (en) | A kind of lens distortion calibration method of robust | |
| CN108509848A (en) | The real-time detection method and system of three-dimension object | |
| CN106447601B (en) | Unmanned aerial vehicle remote sensing image splicing method based on projection-similarity transformation | |
| CN103400384A (en) | Large viewing angle image matching method capable of combining region matching and point matching | |
| CN105930382A (en) | Method for searching for 3D model with 2D pictures | |
| CN107481274B (en) | Robust reconstruction method of three-dimensional crop point cloud | |
| CN106204557A (en) | A kind of extracting method of the non-complete data symmetrical feature estimated with M based on extension Gaussian sphere | |
| CN111340862B (en) | Point cloud registration method and device based on multi-feature fusion and storage medium | |
| CN110111375B (en) | Image matching gross error elimination method and device under Delaunay triangulation network constraint | |
| CN110363179B (en) | Map acquisition method, map acquisition device, electronic equipment and storage medium | |
| CN110009745B (en) | Method for extracting plane from point cloud according to plane element and model drive | |
| CN117011299B (en) | Reference point cloud quality assessment method and system integrating graph resampling and gradient characteristics | |
| CN101794459A (en) | Seamless integration method of stereoscopic vision image and three-dimensional virtual object | |
| CN110211163A (en) | A kind of point cloud matching algorithm based on EPFH feature | |
| CN113963174B (en) | Bogie identification image feature extraction method for multi-view intensity domain and frequency domain fusion | |
| CN113688842B (en) | Local image feature extraction method based on decoupling | |
| CN116935013B (en) | Circuit board point cloud large-scale splicing method and system based on three-dimensional reconstruction | |
| CN117745780A (en) | Outdoor large scene 3D point cloud registration method based on isolated cluster removal | |
| CN110647925A (en) | Rigid object identification method and device based on improved LINE-MOD template matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zeng Huanqiang Inventor after: Lu Zian Inventor after: Hou Junhui Inventor after: Chen Jing Inventor after: Zhu Jianqing Inventor after: Chen Longtao Inventor before: Zeng Huanqiang Inventor before: Lu Zian Inventor before: Hou Junhui Inventor before: Chen Jing Inventor before: Zhu Jianqing Inventor before: Chen Longtao |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |