CN113592711B - Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data - Google Patents
Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data Download PDFInfo
- Publication number
- CN113592711B CN113592711B CN202110830434.4A CN202110830434A CN113592711B CN 113592711 B CN113592711 B CN 113592711B CN 202110830434 A CN202110830434 A CN 202110830434A CN 113592711 B CN113592711 B CN 113592711B
- Authority
- CN
- China
- Prior art keywords
- data
- point
- point cloud
- point set
- normal vector
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 239000013598 vector Substances 0.000 claims abstract description 72
- 239000011159 matrix material Substances 0.000 claims abstract description 45
- 238000005070 sampling Methods 0.000 claims abstract description 45
- 230000006870 function Effects 0.000 claims description 89
- 238000005457 optimization Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/005—Tree description, e.g. octree, quadtree
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Generation (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a three-dimensional reconstruction method for non-uniform point cloud data. Comprising the following steps: scanning the object to obtain complete point set data, and performing pretreatment such as normal vector normalization; manually adding brief contour lines for missing data to obtain a sparse point set; constructing an Hermite matrix by using the sparse point set and obtaining a normal vector; calculating a normal vector and point set data to obtain an interpolation function, and then up-sampling to obtain final point cloud; and performing triangularization operation to obtain a final three-dimensional reconstruction model. The invention also discloses a three-dimensional reconstruction system, computer equipment and a computer readable storage medium for the non-uniform point cloud data. The invention solves the object surface reconstruction problem of the missing part by using an interpolation method, combines the shape information complementation of the artificial sketch, can be freely modified at the missing part of the model, and can well combine the geometric characteristics of the sample points of the actual object and the characteristics of the artificial added sketch to obtain a complete reconstruction model.
Description
Technical Field
The invention relates to the technical field of graphic processing and three-dimensional reconstruction, in particular to a three-dimensional reconstruction method for non-uniform point cloud data, a three-dimensional reconstruction system for non-uniform point cloud data, computer equipment and a computer readable storage medium.
Background
In recent years, as the technology of scanning cameras is mature, more and more ways can be used for conveniently obtaining information on the surface of an object. More and more mobile phones or internet company platforms are now pushing portable scanning applications than have been possible in the past to obtain scans of object surface information from professional laboratories by means of infrared scanning only.
Reconstructing a three-dimensional object from the sampled point set refers to obtaining a model of the real object in space from the point cloud data through calculation in a computer. In past studies, reconstruction methods have been largely divided into two categories. One is a combination method, typically employing Destrian triangulation to obtain a triangular mesh model of the object. Although the grid obtained by this method has small errors on the input data, the grid obtained by the algorithm is unstable and often requires that the sampling points be sufficiently average and dense.
The method mainly comprises the steps of expressing a complex object surface by using linear combination of a group of basis functions, and obtaining an implicit function by solving a large sparse linear equation set and optimizing.
One of the existing technologies is a reconstruction method for point cloud data with normal vectors based on an implicit curved surface, which is proposed by Kazhdan et al (Poisson surface reconstruction) and Screened poisson surface reconstruction), and a system provided for point cloud reconstruction based on a Poisson equation, which comprises a function solving module, a surface extraction module, a self-adaptive space dividing module and the like. And solving the three-dimensional reconstruction problem by using an implicit function. The disadvantages of this method are: the point cloud data with normal vectors needs to be input, and meanwhile, the reconstruction effect is not good enough under the condition of processing the sparse point cloud data.
The second existing technology is a reconstruction method based on dynamic collision detection proposed by Bauchet et al, kinetic Shape Reconstruction, firstly, obtaining some patches of a scene through calculation, and diffusing the patches until the two patches collide. The process is processed through a priority queue, and finally a reconstruction model with a relatively smooth surface is obtained. The disadvantages of this method are: the time complexity required by the dynamic collision detection is high, and meanwhile, a model conforming to the original object cannot be well generated under the condition of uneven point cloud data.
Disclosure of Invention
The invention aims to overcome the defects of the existing method and provides a three-dimensional reconstruction method, a system, equipment and a storage medium for non-uniform point cloud data. The invention solves the main problems of uneven acquisition of point cloud data and high-efficiency reconstruction of the acquired point cloud data under the condition that some point cloud data lack normal vectors.
In order to solve the above problems, the present invention provides a three-dimensional reconstruction method for non-uniform point cloud data, the method comprising:
scanning the object at the full view angle as much as possible by using scanning equipment to obtain complete point set data in the space;
preprocessing the point set data, storing the point set data into PLY data format, and normalizing normal vector of each point in the data format;
manually adding brief contour lines to the part without scanned missing data to obtain a sparse point set of the missing data;
constructing an Hermite matrix by using the sparse point set of the missing data, and obtaining the normal vector of the sparse point set of the missing data by using the matrix through an optimization method;
calculating to obtain an interpolation function through the normal vector of the sparse point set and the data of the sparse point set, and then up-sampling and increasing the density of the point cloud according to the interpolation function to enable the density of the point cloud to be consistent with that of the sampling point of the complete data part, so as to obtain the final point cloud of the missing data;
and performing triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
Preferably, the preprocessing is performed on the point set data, and the point set data is stored in a PLY data format, specifically:
the point set data is represented as a point set P, where P ε P represents each point in space;
and adding a document head according to the PLY data format, writing the data type of each vertex, the three-dimensional coordinate information of the vertex and the normal vector information line by line, and marking correspondingly if the vertex lacks the normal vector information.
Preferably, the constructing an emmett matrix by using the sparse point set of the missing data, and then obtaining a normal vector of the sparse point set of the missing data by using the matrix through an optimization method, which specifically comprises:
converting the sparse point set of the missing data to obtain an interpolation matrix, calculating an inverse matrix of the interpolation matrix, and then performing transformation calculation on a submatrix in the inverse matrix to obtain an Hermite matrix;
obtaining the normal vector estimation of the sparse points of the missing data by constructing an optimization problem, wherein the optimization problem is described as follows:
Minimizes:g T Hg
wherein an H matrix is the emmet matrix, g represents a normal vector of the sparse point set of missing data;
and solving the optimization problem by using a Lagrangian method, and obtaining the normal vector of the sparse point set of the missing data.
Preferably, the interpolation function is obtained by calculating the normal vector of the sparse point set and the data of the sparse point set, and then the density of the point cloud is increased according to the up-sampling of the interpolation function so as to be consistent with the density of the sampling points of the complete data part, so that the final point cloud of the missing data is obtained, specifically:
the interpolation function represents an implicit function of the object surface of the missing part;
the interpolation function is expressed as:
the geometric meaning of the method is a distance function between any point x in space and a surface, wherein a, b, c and d are coefficients to be solved, and a kernel function phi is a radial base kernel;
the function value s is calculated by the Hermite matrix H and the normal vector g;
the linear equation set is accelerated and solved by using the equation solving method of the block matrix, coefficients a, b, c and d of the interpolation function are obtained, and therefore the complete interpolation function is obtained;
setting a radius r, calculating the number of other points in a sphere with each sampling point p as a center and the radius r in the point set data to obtain the number of neighbors of each sampling point, averaging the number of neighbors of each sampling point to obtain the average density of the sampling points, and finally sampling the complete interpolation function in space with the average density to obtain the final point cloud of the missing data.
Preferably, the triangulating operation is performed on the point cloud obtained by combining the scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model, which specifically includes:
combining the scanned point cloud and the final point cloud of the missing data to form a complete three-dimensional model space, dividing the space into a plurality of octree cube nodes by adopting an adaptive octree, and expressing a basis function F in each node by using a block B spline function o ：
Where q is any point in space, F is a cubic convolution of the three-dimensional box filter function, o.w represents the side length of the node cube, o.c represents the center coordinates of the node;
the surface of the three-dimensional model may be represented as an implicit function of a linear combination of basis functions:
wherein M is i For the coefficients to be solved, n is the total number of nodes, and i represents the ith node;
further, the vector at any point q in the vector field in space is expressed as:
wherein F (q) is a three-dimensional box filter function,normal vector s representing sampling point p The area size of the area where the sampling point is located is represented;
because the implicit function f (q), namely the derivative of the surface of the three-dimensional model, and the normal vector direction of the sampling point are the same, a poisson equation is obtained by combining the normal vector with the gradient of the implicit function;
converting the poisson equation into an optimization problem, and solving the coefficient M of each linear combination of the basic functions by minimizing the loss function i Thereby obtaining an implicit function f (q);
and (5) meshing the implicit function f (q) triangle by using an isosurface extraction algorithm to obtain a final triangle mesh model.
Correspondingly, the invention also provides a three-dimensional reconstruction system for non-uniform point cloud data, which comprises:
the point cloud data acquisition unit is used for scanning the object at the full view angle as far as possible by using scanning equipment to obtain complete point set data in the space, preprocessing the point set data, storing the point set data into PLY data format, and normalizing the normal vector of each point;
the interactive hole filling unit is used for artificially adding brief contour lines to the part of the missing data which is not scanned to obtain a sparse point set of the missing data;
the missing point cloud generation unit is used for constructing an Hermite matrix by using the sparse point set of the missing data, obtaining a normal vector of the sparse point set of the missing data by using the matrix through an optimization method, calculating to obtain an interpolation function by using the normal vector of the sparse point set and the data of the sparse point set, and then up-sampling and increasing the density of the point cloud according to the interpolation function to enable the density of the sampling points of the complete data part to be consistent, so as to obtain the final point cloud of the missing data;
and the three-dimensional reconstruction unit is used for carrying out triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
Correspondingly, the invention also provides computer equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the three-dimensional reconstruction method for the non-uniform point cloud data.
Correspondingly, the invention further provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, realizes the steps of the three-dimensional reconstruction method for the non-uniform point cloud data.
The implementation of the invention has the following beneficial effects:
the invention solves the object surface reconstruction problem of the missing part by using an interpolation method, combines the shape information complementation of the artificial sketch, can be freely modified at the missing part of the model, and can well combine the geometric characteristics of the sample points of the actual object and the characteristics of the artificial added sketch to obtain a complete reconstruction model.
Drawings
FIG. 1 is a general flow chart of a method for three-dimensional reconstruction of point cloud data non-uniformities in accordance with an embodiment of the present invention;
FIG. 2 is a schematic diagram of a process for computing a model surface from a point cloud according to an embodiment of the present invention;
fig. 3 is a block diagram of a three-dimensional reconstruction system of point cloud data non-uniformity according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
FIG. 1 is a general flow chart of a video super-resolution method based on data simulation according to an embodiment of the present invention, as shown in FIG. 1, the method includes:
s1, scanning an object at the full view angle as much as possible by using scanning equipment to obtain complete point set data in a space;
s2, preprocessing the point set data, storing the point set data into PLY data format, and normalizing normal vectors of each point in the data format;
s3, artificially adding brief contour lines to the part of the missing data which is not scanned to obtain a sparse point set of the missing data;
s4, constructing an Hermite matrix by using the sparse point set of the missing data, and obtaining the normal vector of the sparse point set of the missing data by using the matrix through an optimization method;
s5, calculating to obtain an interpolation function through the normal vector of the sparse point set and the data of the sparse point set, and then up-sampling and increasing the density of the point cloud according to the interpolation function to enable the density of the sampling points to be consistent with that of the whole data part, so as to obtain the final point cloud of the missing data;
and S6, performing triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
Step S1, specifically, the following steps are performed:
in this embodiment, a mobile phone application program or a professional infrared scanning instrument can be used to scan an object at an angle as full as possible, and the portable scanning of the mobile phone application program and the like may have a larger error in determining the spatial position of the point due to the immaturity of the existing scanning technology, so that the scanning with the professional instrument is more recommended.
Step S2, specifically, the following steps are performed:
the point set data is represented as a point set P, where P ε P represents each point in space;
PLY is a computer file format, and is named as a polygon file or a Stanford triangle file. And adding a document head according to the PLY data format, writing the data type of each vertex, the three-dimensional coordinate information of the vertex and the normal vector information line by line, and marking correspondingly if the vertex lacks the normal vector information.
The normal vector of the acquired points is then normalized, and each coordinate scalar of the normal vector is divided by the length of the normal vector itself.
Step S3, specifically, the following steps are performed:
because the acquired data has a certain deficiency, in order to obtain a complete model of the object, some point information is selected to be manually added as input. Specifically, program interaction is used to add the outline of the missing data portion in the file, i.e., manually add the sketch. The addition of part of the normal vector is missing because it is a point that is artificially added.
Step S4, specifically, the following steps are performed:
converting the sparse point set of the missing data to obtain an interpolation matrix, calculating an inverse matrix of the interpolation matrix, and then performing transformation calculation on a submatrix in the inverse matrix to obtain an Hermite matrix;
the normal vector estimation of the sparse points of the missing data is obtained by constructing an optimization problem, which is described as follows:
Minimizes:g T Hg
wherein an H matrix is the emmet matrix, g represents a normal vector of the sparse point set of missing data;
and solving the optimization problem by using a Lagrangian method, and obtaining the normal vector of the sparse point set of the missing data. Here a set of guesses is needed as an initial solution so that the method starts to iterate. In order to better and faster obtain the algorithm result, the characteristic that the included angles between the normal vectors of the similar points are smaller is used to make the initial solution uniform and smooth, so that the iteration times are reduced.
Step S5, specifically, the following steps are performed:
calculating to obtain an interpolation function through normal vectors of the sparse point sets and data of the sparse point sets, wherein the interpolation function represents an implicit function of the object surface of the missing part;
the interpolation function is expressed as:
the geometric meaning of the method is a distance function between any point x in space and a surface, wherein a, b, c and d are coefficients to be solved, and a kernel function phi is a radial base kernel;
the normal vector g comprises a normal vector obtained by optimizing a sparse point set lacking vertexes of the normal vector and lacking data, and the function value s is calculated by the Hermite matrix H and the normal vector g;
the linear equation set is accelerated and solved by using the equation solving method of the block matrix, coefficients a, b, c and d of the interpolation function are obtained, and therefore the complete interpolation function is obtained;
setting a radius r, and calculating the number of other points in the sphere with each sampling point p as the center and the radius r in the point set data to obtain the number of neighbors of each sampling point. For the value of the radius r, if a precise instrument is used for scanning, the radius can be set to be slightly smaller due to smaller noise; if a portable scanning tool is used, the detection radius needs to be set larger. This is due to the different decision of the noise of different instruments.
And then, averaging the neighbor number of each sampling point to obtain the average density of the sampling points, and finally sampling the complete interpolation function in space according to the average density, so that the density of the whole point cloud is not excessively different, and the final point cloud of the missing data is obtained.
Step S6, as shown in fig. 2, is specifically as follows:
and combining the scanned point cloud and the final point cloud of the missing data to form a complete three-dimensional model space, and dividing the space into a plurality of octree cube nodes by adopting the self-adaptive octree. The default adaptive octree depth in this embodiment is 8. The octree can be resized, the lower the octree depth, the smoother the reconstructed result, and the detail is missing. If the depth of the octree is high, the three-dimensional model can be well reconstructed when the data noise is small and the data points are dense. In addition, the influence of noise can be reduced by adjusting the number of sampling points in the minimum leaf node. If the noise is large, the number of sampling points in each node is larger, so that the noise point is prevented from occupying too large proportion in one node.
Then expressing the basis function F in each node by using the block B spline function o ：
Where q is any point in space, F is a three-dimensional convolution of the box filter function, o.w represents the side length of the node cube, and o.c represents the center coordinates of the node. The three-dimensional box filter function is used because it is shaped like a gaussian distribution, but is simple to calculate and can increase the calculation speed. The reason that three convolutions are adopted instead of four is that the function definition domain of 4 times is overlarge, so that each node has no good independence;
the surface of the three-dimensional model may be represented as an implicit function of a linear combination of basis functions:
wherein M is i For the coefficients to be solved, n is the total number of nodes, and i represents the ith node;
furthermore, since the implicit function is continuous, when representing a vector field, a continuous vector field is also used, and the vector at any point q in the vector field in space is represented as:
wherein F (q) is a three-dimensional box filter function,normal vector s representing sampling point p The area size of the area where the sampling point is located is represented;
because the implicit function f (q), namely the derivative of the surface of the three-dimensional model, and the normal vector direction of the sampling point are the same, a poisson equation is obtained by combining the normal vector with the gradient of the implicit function;
converting the poisson equation into an optimization problem, and solving the coefficient M of each linear combination of the basic functions by minimizing the loss function i Thereby obtaining an implicit function f (q). In order to accelerate the matrix calculation speed, a V model method or a W model method of multiple grids can be adopted, and a conjugate gradient method is adopted on each coarse grid to accelerate the solving process;
and (5) meshing the implicit function f (q) triangle by using an isosurface extraction algorithm to obtain a final triangle mesh model. Because the implicit function obtained by solving the equation is not necessarily close to the original sampling points, in order to well extract the surface grid, the function mean value on each sampling point is used as the input of a stepping method, and an equivalent surface is obtained. This triangular mesh is the reconstructed object.
Correspondingly, the invention also provides a video super-resolution system based on data simulation, as shown in fig. 3, comprising:
the point cloud data acquisition unit 1 is used for scanning an object at the full view angle as far as possible by using scanning equipment to obtain complete point set data in a space, preprocessing the point set data, storing the point set data into PLY data format, and normalizing normal vectors of each point;
the interactive hole supplementing unit 2 is used for artificially adding brief contour lines to the part of the missing data which is not scanned to obtain a sparse point set of the missing data;
the missing point cloud generating unit 3 is configured to construct an emmet matrix by using the sparse point set of the missing data, calculate a normal vector of the sparse point set of the missing data by using the matrix through an optimization method, calculate an interpolation function by using the normal vector of the sparse point set and the data of the sparse point set, and up-sample and increase the density of the point cloud according to the interpolation function so as to be consistent with the sampling point density of the complete data part, thereby obtaining the final point cloud of the missing data;
and the three-dimensional reconstruction unit 4 is used for carrying out triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
Therefore, the invention solves the object surface reconstruction problem of the missing part by using an interpolation method, combines the shape information complementation of the artificial sketch, can be freely modified at the missing part of the model, and can well combine the geometric characteristics of the sample points of the actual object with the characteristics of the artificial added sketch to obtain a complete reconstruction model.
Correspondingly, the invention also provides computer equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the three-dimensional reconstruction method for the non-uniform point cloud data when executing the computer program. Meanwhile, the invention also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, realizes the steps of the three-dimensional reconstruction method for the non-uniform point cloud data.
The three-dimensional reconstruction method, system, device and storage medium for non-uniform point cloud data provided by the embodiments of the present invention are described in detail, and specific examples are applied to illustrate the principles and embodiments of the present invention, and the description of the above embodiments is only used to help understand the method and core ideas of the present invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (8)
1. A method for three-dimensional reconstruction of point cloud data non-uniformity, the method comprising:
scanning the object at the full view angle as much as possible by using scanning equipment to obtain complete point set data in the space;
preprocessing the point set data, storing the point set data into PLY data format, and normalizing normal vector of each point in the data format;
manually adding brief contour lines to the part without scanned missing data to obtain a sparse point set of the missing data;
constructing an Hermite matrix by using the sparse point set of the missing data, and obtaining the normal vector of the sparse point set of the missing data by using the matrix through an optimization method;
calculating to obtain an interpolation function through the normal vector of the sparse point set and the data of the sparse point set, and then up-sampling and increasing the density of the point cloud according to the interpolation function to enable the density of the point cloud to be consistent with that of the sampling point of the complete data part, so as to obtain the final point cloud of the missing data;
and performing triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
2. The method for three-dimensional reconstruction of point cloud data non-uniformity according to claim 1, wherein said preprocessing said point set data and storing it in PLY data format is as follows:
the point set data is represented as a point set P, where P ε P represents each point in space;
and adding a document head according to the PLY data format, writing the data type of each vertex, the three-dimensional coordinate information of the vertex and the normal vector information line by line, and marking correspondingly if the vertex lacks the normal vector information.
3. The method for three-dimensional reconstruction of non-uniformity of point cloud data according to claim 1, wherein constructing an emmet matrix by using the sparse point set of the missing data, and obtaining a normal vector of the sparse point set of the missing data by using the matrix through an optimization method, specifically comprises:
converting the sparse point set of the missing data to obtain an interpolation matrix, calculating an inverse matrix of the interpolation matrix, and then performing transformation calculation on a submatrix in the inverse matrix to obtain an Hermite matrix;
obtaining the normal vector estimation of the sparse points of the missing data by constructing an optimization problem, wherein the optimization problem is described as follows:
Minimizes：g T Hg
wherein an H matrix is the emmet matrix, g represents a normal vector of the sparse point set of missing data;
and solving the optimization problem by using a Lagrangian method, and obtaining the normal vector of the sparse point set of the missing data.
4. The method for three-dimensional reconstruction of non-uniform point cloud data according to claim 2 or 3, wherein the interpolation function is obtained by calculating a normal vector of the sparse point set and data of the sparse point set, and the density of the point cloud is increased according to up-sampling of the interpolation function so as to be consistent with the sampling point density of the complete data part, so as to obtain the final point cloud of the missing data, specifically:
the interpolation function represents an implicit function of the object surface of the missing part;
the interpolation function is expressed as:
the geometric meaning of the method is a distance function between any point x in space and a surface, wherein a, b, c and d are coefficients to be solved, and a kernel function phi is a radial base kernel;
the function value s is calculated by the Hermite matrix H and the normal vector g;
the linear equation set is accelerated and solved by using the equation solving method of the block matrix, coefficients a, b, c and d of the interpolation function are obtained, and therefore the complete interpolation function is obtained;
setting a radius r, calculating the number of other points in a sphere with each sampling point p as a center and the radius r in the point set data to obtain the number of neighbors of each sampling point, averaging the number of neighbors of each sampling point to obtain the average density of the sampling points, and finally sampling the complete interpolation function in space with the average density to obtain the final point cloud of the missing data.
5. The method for three-dimensional reconstruction of non-uniform point cloud data according to claim 2, wherein the performing a triangularization operation on the point cloud obtained by combining the scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model comprises:
combining the scanned point cloud and the final point cloud of the missing data to form a complete three-dimensional model space, dividing the space into a plurality of octree cube nodes by adopting an adaptive octree, and expressing a basis function F in each node by using a block B spline function o ：
Where q is any point in space, F is a cubic convolution of the three-dimensional box filter function, o.w represents the side length of the node cube, o.c represents the center coordinates of the node;
the surface of the three-dimensional model may be represented as an implicit function of a linear combination of basis functions:
wherein M is i For the coefficients to be solved, n is the total number of nodes, and i represents the ith node;
further, the vector at any point q in the vector field in space is expressed as:
wherein F (q) is a three-dimensional box filter function,normal vector s representing sampling point p The area size of the area where the sampling point is located is represented;
because the implicit function f (q), namely the derivative of the surface of the three-dimensional model, and the normal vector direction of the sampling point are the same, a poisson equation is obtained by combining the normal vector with the gradient of the implicit function;
converting the poisson equation into an optimization problem, and solving the coefficient M of each linear combination of the basic functions by minimizing the loss function i Thereby obtaining an implicit function f (q);
and (5) meshing the implicit function f (q) triangle by using an isosurface extraction algorithm to obtain a final triangle mesh model.
6. A three-dimensional reconstruction system for non-uniformity of point cloud data, the system comprising:
the point cloud data acquisition unit is used for scanning the object at the full view angle as far as possible by using scanning equipment to obtain complete point set data in the space, preprocessing the point set data, storing the point set data into PLY data format, and normalizing the normal vector of each point;
the interactive hole filling unit is used for artificially adding brief contour lines to the part of the missing data which is not scanned to obtain a sparse point set of the missing data;
the missing point cloud generation unit is used for constructing an Hermite matrix by using the sparse point set of the missing data, obtaining a normal vector of the sparse point set of the missing data by using the matrix through an optimization method, calculating to obtain an interpolation function by using the normal vector of the sparse point set and the data of the sparse point set, and then up-sampling and increasing the density of the point cloud according to the interpolation function to enable the density of the sampling points of the complete data part to be consistent, so as to obtain the final point cloud of the missing data;
and the three-dimensional reconstruction unit is used for carrying out triangularization operation by combining the point cloud obtained by scanning and the final point cloud of the missing data to obtain a final three-dimensional reconstruction model.
7. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110830434.4A CN113592711B (en) | 2021-07-22 | 2021-07-22 | Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110830434.4A CN113592711B (en) | 2021-07-22 | 2021-07-22 | Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113592711A CN113592711A (en) | 2021-11-02 |
| CN113592711B true CN113592711B (en) | 2024-01-12 |
Family
ID=78249258
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110830434.4A Active CN113592711B (en) | 2021-07-22 | 2021-07-22 | Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113592711B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494611B (en) * | 2022-04-18 | 2022-09-02 | 清华大学 | Intelligent three-dimensional reconstruction method, device, equipment and medium based on nerve basis function |
| CN115619983B (en) * | 2022-12-02 | 2023-04-07 | 中南大学 | Average curvature controllable implicit surface generation method based on radial basis function |
| CN116468767B (en) * | 2023-03-28 | 2023-10-13 | 南京航空航天大学 | Airplane surface reconstruction method based on local geometric features and implicit distance field |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016001697A1 (en) * | 2014-07-03 | 2016-01-07 | Cgg Services Sa | Systems and methods for geologic surface reconstruction using implicit functions |
| CN106780751A (en) * | 2017-01-19 | 2017-05-31 | 桂林电子科技大学 | Three-dimensional point cloud method for reconstructing based on improved shielding Poisson algorithm |
| CN109285223A (en) * | 2018-09-07 | 2019-01-29 | 中山大学 | Image reconstruction quadrilateral mesh method and system based on voxel model |
| CN109584357A (en) * | 2018-11-27 | 2019-04-05 | 中南大学 | Three-dimensional modeling method, device, system and storage medium based on more contour lines |
| CN112150606A (en) * | 2020-08-24 | 2020-12-29 | 上海大学 | Thread surface three-dimensional reconstruction method based on point cloud data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103489218B (en) * | 2013-09-17 | 2016-06-29 | 中国科学院深圳先进技术研究院 | Point cloud data quality automatic optimization method and system |
-
2021
- 2021-07-22 CN CN202110830434.4A patent/CN113592711B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016001697A1 (en) * | 2014-07-03 | 2016-01-07 | Cgg Services Sa | Systems and methods for geologic surface reconstruction using implicit functions |
| CN106780751A (en) * | 2017-01-19 | 2017-05-31 | 桂林电子科技大学 | Three-dimensional point cloud method for reconstructing based on improved shielding Poisson algorithm |
| CN109285223A (en) * | 2018-09-07 | 2019-01-29 | 中山大学 | Image reconstruction quadrilateral mesh method and system based on voxel model |
| CN109584357A (en) * | 2018-11-27 | 2019-04-05 | 中南大学 | Three-dimensional modeling method, device, system and storage medium based on more contour lines |
| CN112150606A (en) * | 2020-08-24 | 2020-12-29 | 上海大学 | Thread surface three-dimensional reconstruction method based on point cloud data |
Non-Patent Citations (1)
| Title |
|---|
| 利用Voronoi协方差矩阵重建隐式曲面;李国俊 等;中国图象图形学报;第21卷(第3期);第323-330页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113592711A (en) | 2021-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113592711B (en) | Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data | |
| CN113096234B (en) | Method and device for generating three-dimensional grid model by using multiple color pictures | |
| CN110390638B (en) | High-resolution three-dimensional voxel model reconstruction method | |
| US8259110B1 (en) | Method and apparatus for computing direct lighting for global illumination rendering in 3-D computer graphics | |
| US11263356B2 (en) | Scalable and precise fitting of NURBS surfaces to large-size mesh representations | |
| CN111581776B (en) | Iso-geometric analysis method based on geometric reconstruction model | |
| EP3736776B1 (en) | Apparatus, system and method for the generation of polygonal meshes | |
| CN114049420B (en) | Model training method, image rendering method, device and electronic equipment | |
| JP2006031561A (en) | High-dimensional texture mapping device, method and program | |
| CN116958453B (en) | Three-dimensional model reconstruction method, device and medium based on nerve radiation field | |
| US20090309877A1 (en) | Soft shadow rendering | |
| JP2009521062A (en) | Modeling the 3D shape of an object by shading 2D images | |
| CN108986218A (en) | A kind of building point off density cloud fast reconstructing method based on PMVS | |
| CN115546371A (en) | Point cloud optimization method and system, electronic device and storage medium | |
| CN117274527B (en) | Method for constructing three-dimensional visualization model data set of generator equipment | |
| Jang et al. | Feature‐Preserving Displacement Mapping With Graphics Processing Unit (GPU) Tessellation | |
| CN110322548B (en) | Three-dimensional grid model generation method based on geometric image parameterization | |
| KR20030016477A (en) | Remeshing optimizing method and apparatus for polyhedral surface | |
| CN114418852B (en) | Point cloud arbitrary scale up-sampling method based on self-supervision deep learning | |
| CN117437362B (en) | Three-dimensional animation model generation method and system | |
| CN111445585B (en) | Three-dimensional shape corresponding method and device based on dual convex hull domain | |
| CN114241300A (en) | Remote sensing image target recognition training sample generation method based on grid parameterization | |
| CN111080764A (en) | Method, device and equipment for synthesizing body texture | |
| CN114693527A (en) | DEM super-resolution reconstruction method based on self-similarity characteristics and Transformer technology | |
| CN116266374A (en) | Image processing method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |