Full-plane projection three-dimensional point cloud up-sampling method and system

Abstract

The invention belongs to but not limited to the technical field of point cloud acquisition, and particularly relates to a three-dimensional point cloud up-sampling method and system for full-plane projection, and the method comprises the steps: projecting all scanned point clouds to a two-dimensional plane; in order to conveniently operate points projected on a two-dimensional plane, a quadtree data structure is established, coordinate transformation is performed on the points on a projection plane, namely, three-dimensional coordinates are converted into two-dimensional coordinates, the projection plane is rotated to be parallel to an xoy plane by using a Rodrigue rotation formula, and then x and y coordinates are directly taken; establishing a quadtree data structure, and performing quick Poisson disk sampling; the points are restored to the three-dimensional space; and reforming and smoothing the point cloud: if the point cloud restored to the three-dimensional space is not completely fit with the original point cloud, smoothing the point cloud and reforming the point cloud.

Description

Technical Field

[0001] The present invention belongs to but is not limited to the field of point cloud acquisition technology, and in particular relates to a three-dimensional point cloud upsampling method and system for full plane projection. Background Art

[0002] Point cloud upsampling aims to convert sparse, noisy, and incomplete point clouds into dense, clean, and complete point clouds, which is beneficial to improve the performance of downstream tasks.

[0003] Existing related technologies, such as Voxel Structure Based Mesh Reconstruction From a 3D Point Cloud, mention that the input point cloud is denoised using MLS, the density of the point cloud is adjusted through the octree, and then the voxel structure is resampled. The effect is very good, but the computational efficiency is not high.

[0004] Point cloud is the most basic way to represent 3D data. It is widely used in target detection, object recognition, 3D reconstruction and virtual reality in the field of computer vision and machine learning. With the popularity of commercial 3D scanners, it is becoming more and more convenient to capture point clouds from real-world scenes. However, the original point clouds generated by 3D scanners are usually sparse and noisy, and sometimes have holes, which affects the performance of downstream tasks. Therefore, it is crucial to upsample the point cloud into a dense, clean, complete point cloud with a variety of geometric details.

[0005] Existing point cloud upsampling technologies are divided into traditional algorithms and learning-based algorithms. Learning-based algorithms do perform well in point cloud upsampling, but due to the poor interpretability of deep learning networks, designing an ideal network often requires fine-tuning parameters and network architecture, and in order to ensure the generalization ability of the network after training, it is often necessary to use a data set that includes point clouds of various shapes. The larger the data set, the longer the training time and the higher the computing power requirements. In reality, some devices have limited computing power and often do not support these networks. At this time, a lightweight network model is needed, or a traditional algorithm is used.

[0006] Not only that, some point cloud acquisition devices often collect a large number of points from the real environment. Due to the huge amount of data, such point clouds must be downsampled first. Many details will be lost in the downsampling process, so upsampling is needed to supplement the details.

[0007] In view of the above analysis, the technical problems that need to be solved urgently in the existing technology are: due to the poor interpretability of deep learning networks, it is often necessary to fine-tune parameters and network architecture to design an ideal network, and in order to ensure the generalization ability of the network after training, it is often necessary to select a data set that includes point clouds of various shapes. The larger the data set, the longer the training time and the higher the computing power required. However, in reality, some devices have limited computing power and often do not support these networks. Summary of the invention

[0008] In view of the problems existing in the prior art, the present invention provides a three-dimensional point cloud upsampling method and system for full plane projection.

[0009] The present invention is implemented as follows: a three-dimensional point cloud upsampling method for full plane projection, comprising:

[0010] S1. Project all the scanned point clouds onto a two-dimensional plane;

[0011] S2. In order to facilitate the operation of the points projected onto the two-dimensional plane, a quadtree data structure is established to transform the coordinates of the points on the projection plane, that is, to convert the three-dimensional coordinates into two-dimensional coordinates, and use the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly take the x and y coordinates;

[0012] S3. Establish a quadtree data structure and perform fast Poisson disk sampling;

[0013] S4. restore the point back to the three-dimensional space;

[0014] S5. Reshape and smooth the point cloud: The point cloud restored to the three-dimensional space does not completely match the original point cloud, so it is smoothed and reshaped.

[0015] Further, the coordinates of the intersection point in S1 are:

[0016]

[0017]

[0018]

[0019] In order to restore the point cloud back to three-dimensional space, in this step, it is necessary to calculate the distance from each point to the projection plane and record it. The formula is as follows:

[0020]

[0021] Furthermore, in S2, v rot =Rv, where v rotis the rotated vector, v is the vector to be rotated, and R is the rotation matrix to be obtained;

[0022]

[0023] v rot That is, the coordinates of the projection plane rotated to the xoy plane.

[0024] Furthermore, in S3, the tree building rule is: ensure that each leaf node contains at most 5 points.

[0025] Sampling rule: Sampling is performed on each leaf node containing 35 points.

[0026] Furthermore, the restored three-dimensional coordinates in S4 are:

[0027]

[0028]

[0029]

[0030] Another object of the present invention is to provide a three-dimensional point cloud upsampling system of full plane projection that implements the three-dimensional point cloud upsampling method of full plane projection, comprising:

[0031] Point cloud two-dimensional plane projection module projects all scanned point clouds onto a two-dimensional plane;

[0032] The projection plane coordinate transformation module, in order to facilitate the operation of points projected onto the two-dimensional plane, establishes a quadtree data structure to transform the coordinates of the points on the projection plane, that is, converts the three-dimensional coordinates into two-dimensional coordinates, uses the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly takes the x and y coordinates;

[0033] Quadtree data structure establishment module, establishes quadtree data structure and performs fast Poisson disk sampling;

[0034] The three-dimensional space restoration module restores the points back to the three-dimensional space;

[0035] Smooth point cloud reorganization module: The point cloud restored to the three-dimensional space does not completely fit the original point cloud, so it is smoothed and reorganized.

[0036] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the three-dimensional point cloud upsampling method of full plane projection.

[0037] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the three-dimensional point cloud upsampling method of full plane projection.

[0038] Another object of the present invention is to provide an information data processing terminal, which includes the three-dimensional point cloud upsampling system of full-plane projection.

[0039] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0040] First, in similar sampling algorithms of the prior art or in model reconstruction, efficiency and accuracy are often not achieved at the same time. Some algorithms have low computational efficiency while ensuring high accuracy of the results; some algorithms have high efficiency, but the accuracy of the results is often unsatisfactory. The present invention designs an algorithm that takes both computational efficiency and accuracy into account. The results show that the present invention can indeed make the downstream point cloud processing effect better.

[0041] The present invention designs a method that can project the point cloud scanned by the real device onto a two-dimensional plane, and perform upsampling operations on the movie on the two-dimensional plane. When one dimension is reduced, the amount of calculation is reduced, and the requirements for the computing power of the device are reduced. The algorithm provided by the present invention is efficient and the result is accurate. It can be seen from the test results that the present invention can make the downstream point cloud processing effect better.

[0042] The present invention proposes a three-dimensional point cloud upsampling method for full plane projection, which solves multiple technical problems in the prior art by establishing a quadtree data structure for fast Poisson disk sampling and realizing the reorganization and smoothing of three-dimensional point clouds. First, when projecting a three-dimensional point cloud onto a two-dimensional plane, the method uses the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, which simplifies the complexity of coordinate transformation and makes subsequent operations more convenient and efficient. Through this method, point cloud data can be conveniently processed and operated on a two-dimensional plane, significantly improving computational efficiency.

[0043] Secondly, the present invention manages the points on the projection plane by establishing a quadtree data structure and ensures that each leaf node contains at most 5 points, thereby effectively reducing data redundancy and computational complexity. At the same time, the fast Poisson disk sampling technology is used to sample each leaf node containing 35 points, further improving the uniformity and sparsity of the point cloud data. This method not only improves the quality of point cloud data, but also reduces the amount of data, thereby reducing storage and computing costs.

[0044] In addition, in the process of restoring the point cloud back to three-dimensional space, the present invention ensures accurate restoration of the point cloud data by calculating the distance from each point to the projection plane and recording it. This method makes the restored three-dimensional point cloud more consistent with the original point cloud, reducing data loss and distortion. At the same time, the restored point cloud data is reorganized and smoothed, further improving the accuracy and smoothness of the point cloud data, thereby improving the overall quality and usability of the point cloud.

[0045] In summary, the present invention significantly improves the efficiency and quality of point cloud data processing by optimizing the projection, sampling and restoration process of point cloud data. By using quadtree data structure and fast Poisson disk sampling technology, not only the problems of point cloud data redundancy and high computational complexity in the prior art are solved, but also the uniformity and sparsity of point cloud data are improved. In the process of restoring and reorganizing point cloud data, the method of the present invention effectively reduces data loss and distortion, ensures the accuracy and smoothness of point cloud data, significantly improves the overall quality and availability of point cloud data, and has important application value and technological progress.

[0046] Second, the technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0047] Among the point cloud related technologies that can be retrieved at present, the technical solutions using point cloud projection methods are mostly local projections, among which local projections are mostly related to differential geometry, that is, cutting plane projection sampling is performed on local areas of 3D point clouds and then restored. No one has proposed a technology for full plane projection, processing point clouds on a 2D plane and then restoring the entire point cloud back to 3D space.

[0048] Most of the local projections are related to differential geometry, that is, the tangent plane projection sampling is performed on the local area of ​​the 3D point cloud and then restored. Reducing the information of the 3D object to 2D will inevitably cause information loss. Because it is a local projection, the information loss to the 3D object is very small, and the error of restoration is also very small. However, the global projection loses a lot of 3D information, and the error of restoration is also very large. How to restore the sampling points back to the 3D space after sampling and minimize this loss is the key.

[0049] Third, this paper proposes a 3D point cloud upsampling method for full-plane projection, which aims to optimize the processing and enhancement of 3D point clouds through a series of carefully designed steps. The following are the technical problems solved by this method and the significant technical progress achieved:

[0050] Technical issues solved:

[0051] 1. Point cloud processing efficiency: Traditional 3D point cloud processing methods are often performed directly in 3D space, which has high computational complexity and low processing efficiency. The present invention reduces the processing dimension and improves the processing efficiency by projecting the point cloud onto a 2D plane.

[0052] 2. Point cloud sampling quality: In the process of sampling 3D point clouds, it is a challenge to maintain the geometric characteristics and distribution uniformity of point clouds. This paper achieves high-quality point cloud sampling through quadtree data structure and fast Poisson disk sampling, effectively maintaining the geometric characteristics and distribution uniformity of point clouds.

[0053] 3. Point cloud restoration accuracy: When restoring a point cloud on a two-dimensional plane back to three-dimensional space, the restoration accuracy is often not high due to information loss during the projection process. The present invention significantly improves the restoration accuracy of the point cloud by recording the distance from each point to the projection plane and using the distance information during the restoration process.

[0054] Significant technological advancements:

[0055] 1. Dimensionality reduction and efficiency improvement: Through full plane projection, the three-dimensional point cloud problem is converted into a two-dimensional problem, which greatly reduces the computational complexity and improves the processing efficiency.

[0056] 2. High-quality sampling and geometric feature preservation: By using the quadtree data structure and fast Poisson disk sampling, high-quality upsampling of 3D point clouds is achieved, while effectively maintaining the geometric features and distribution uniformity of the point clouds.

[0057] 3. High-precision restoration and detail recovery: By recording the distance from the point to the projection plane and using this information in the restoration process, high-precision restoration of the three-dimensional point cloud is achieved, and more detail information is restored.

[0058] 4. Innovation of mathematical models and algorithms: This invention proposes a series of innovative mathematical models and algorithms, such as intersection coordinate calculation, rotation matrix solution, quadtree construction and sampling rules, etc., which provide new ideas and methods for three-dimensional point cloud processing.

[0059] In summary, the present invention solves key problems such as efficiency, sampling quality and restoration accuracy in three-dimensional point cloud processing through a series of innovative technical means, and has achieved significant technical progress. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a flow chart of a three-dimensional point cloud upsampling method for full plane projection provided by an embodiment of the present invention;

[0061] Figure 2 It is a structural diagram of a three-dimensional point cloud upsampling system for full plane projection provided by an embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of a projection process provided by an embodiment of the present invention;

[0063] Figure 4 is a schematic diagram of point cloud projection provided by an embodiment of the present invention;

[0064] Figure 5 It is a schematic diagram of rotating the projection plane to the xoy plane provided by an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of the rotation of the projection plane provided by an embodiment of the present invention;

[0066] Figure 7 is a schematic diagram of rapid Poisson disk sampling provided by an embodiment of the present invention;

[0067] Figure 8 is a schematic diagram of a geometric model provided by an embodiment of the present invention;

[0068] Fig. 9 Schematic diagram of restoring and smoothing point clouds provided by an embodiment of the present invention

[0069] Fig.10 is a schematic diagram of the model morphology after the reconstruction of the point cloud of a part of the workpiece provided by an embodiment of the present invention;

[0070] Fig.11 It is a schematic diagram of the reconstructed model morphology after upsampling of the point cloud provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] The following is a specific example from a manufacturing application:

[0073] Manufacturing Examples: Additive Manufacturing and Laser Manufacturing

[0074] This sub-project of the National Key R&D Program is aimed at meeting the urgent needs of major national projects for new-generation launch vehicles, fighter jets and other high-end equipment. It provides technical support such as optomechanical coordinated control and information processing for the development of the national large-scale structural and functional integrated shape and control manufacturing platform.

[0075] Solution:

[0076] Relying on the national key R&D plan projects, focusing on the development of shape-controllable manufacturing platforms with integrated structure and functions in the field of additive manufacturing, we develop optomechanical collaborative control systems and dedicated verification platforms for key algorithms, and based on this, we carry out research in signal processing, data processing, and automatic control, providing algorithm support for the development of high-precision laser manufacturing platforms.

[0077] Implementation effect:

[0078] Development of a control platform that integrates multi-sensor information data fusion, optical-mechanical-electrical-computing interaction, and high-precision collaboration of multi-functional modules; algorithm optimization based on point cloud data reconstruction, simplification, splicing, and registration, high-efficiency and high-precision selected area topography measurement and precise positioning technology; laser focus intelligent tracking and instantaneous deviation control algorithm based on deep learning, etc.

[0079] Manufacturing Example: Research and Development of Parameter-Controlled Femtosecond Laser Precision Manufacturing Equipment

[0080] Through the research on optomechanical coordinated control and signal processing in the field of additive manufacturing, we provide technical support such as optomechanical coordinated control and information processing for the development of the national large-scale structural and functional integrated shape-controllable manufacturing platform.

[0081] like Figure 1 As shown, the three-dimensional point cloud upsampling method of full plane projection provided by an embodiment of the present invention includes:

[0082] S1. Project all the scanned point clouds onto a two-dimensional plane;

[0083] S2. In order to facilitate the operation of the points projected onto the two-dimensional plane, a quadtree data structure is established to transform the coordinates of the points on the projection plane, that is, to convert the three-dimensional coordinates into two-dimensional coordinates, and use the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly take the x and y coordinates;

[0084] S3. Establish a quadtree data structure and perform fast Poisson disk sampling;

[0085] S4. restore the point back to the three-dimensional space;

[0086] S5. Reshape and smooth the point cloud: The point cloud restored to the three-dimensional space does not completely match the original point cloud, so it is smoothed and reshaped.

[0087] like Figure 2 As shown, the full-plane projection 3D point cloud upsampling system provided by the embodiment of the present invention includes:

[0088] Point cloud two-dimensional plane projection module projects all scanned point clouds onto a two-dimensional plane;

[0089] The projection plane coordinate transformation module, in order to facilitate the operation of points projected onto the two-dimensional plane, establishes a quadtree data structure to transform the coordinates of the points on the projection plane, that is, converts the three-dimensional coordinates into two-dimensional coordinates, uses the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly takes the x and y coordinates;

[0090] Quadtree data structure establishment module, establishes quadtree data structure and performs fast Poisson disk sampling;

[0091] The three-dimensional space restoration module restores the points back to the three-dimensional space;

[0092] Smooth point cloud reorganization module: The point cloud restored to the three-dimensional space does not completely fit the original point cloud, so it is smoothed and reorganized.

[0093] like Figure 3 , Figure 4 As shown, all the point clouds obtained by scanning are projected onto a two-dimensional plane. The following is the specific projection process

[0094] Here is the parametric equation of a line:

[0095] General equation of a plane: Ax+By+Cz+D=0

[0096] Assume the coordinates of the point cloud are: (x n ,y n , z n )

[0097] The coordinates of the camera are: (x0, y0, z0)

[0098] The direction vector of the camera is (normal vector of the projection plane): (n, m, l)

[0099] From the knowledge of spatial geometry, we know that a plane can be uniquely determined by knowing a point on a plane and the normal vector of the plane; a straight line can be uniquely determined by knowing a point on a straight line and the direction vector of the straight line.

[0100] So the problem is transformed into finding the coordinates of the intersection of the line and the projection plane

[0101] Substitute the point cloud coordinates and normal vector coordinates into the parametric equation of the line:

[0102]

[0103] Substituting the camera and normal coordinates into the general equation for a plane:

[0104] nx0+my0+lz0+D=0

[0105] Solution: D = -(nx0 + my0 + lz0)

[0106] The plane equation can be determined:

[0107] nx+my+lz-(nx0+my0+lz0)=0

[0108] Substitute the parametric equation of the line and solve for t:

[0109]

[0110] Substituting t into the parametric equation, we can find the coordinates of the intersection point:

[0111]

[0112]

[0113]

[0114] In order to restore the point cloud back to three-dimensional space, in this step, it is necessary to calculate the distance from each point to the projection plane and record it. The formula is as follows:

[0115]

[0116] like Figure 5 As shown, in order to facilitate the operation of points projected onto the two-dimensional plane, a quadtree data structure needs to be established. In order to establish the quadtree structure, the coordinates of the points on the projection plane need to be transformed, that is, a three-dimensional coordinate is converted into a two-dimensional coordinate. Here, the Rodriguez rotation formula is used to rotate the projection plane to be parallel to the xoy plane, and then the x and y coordinates can be directly taken.

[0117] like Figure 6 Shown is a schematic diagram of the projection plane rotation.

[0118] v rot =Rv, where v rot is the rotated vector, v is the vector to be rotated, and R is the rotation matrix to be calculated.

[0119]

[0120] Where E is a third-order unit matrix, θ is the angle between the projection plane and the xoy plane, which can be obtained by the angle between the normal vectors of the two planes, kx, ky, kz are the unit vectors of the rotation axis, that is, the unit vectors of the direction vector of the intersection line of the projection plane and the xoy plane,

[0121] Let the normal vector of the projection plane be (n, m, l)

[0122] It can be calculated

[0123]

[0124]

[0125] The direction vector of the projection plane can be obtained by the cross product of the normal vectors of the two planes and then normalized, that is,

[0126]

[0127] At the same time, let the point on the rotation axis be (x r ,y r , z r ), since the projection plane and the xoy surface must have an intersection line, the equation of the intersection line is determined by the projection plane equation and the xoy surface equation, that is:

[0128]

[0129] From the first step, we can see that the plane equation can be uniquely determined by any point on it and the normal vector, thus we can get the axis equation:

[0130] nx+my-(nx0+my0+lz0)=0

[0131] Next, determine the coordinates of any point on the axis. For convenience, take any point on the xoy plane and the coordinate origin to form a straight line. Connect the equation of this line with the axis equation to get the coordinates of any point on the axis:

[0132] Assume that the coordinates of any point on the xoy plane are (p,q,0), then:

[0133]

[0134] The coordinates of any point on the axis of rotation are:

[0135]

[0136] Substituting the above derivation results into v rot =Rv, we get:

[0137]

[0138] After sorting, we get:

[0139]

[0140] v rot That is, the coordinates of the projection plane rotated to the xoy plane.

[0141] like Figure 7 As shown, a quadtree data structure is established and fast Poisson disk sampling is performed

[0142] Tree building rules: Ensure that each leaf node contains at most 5 points.

[0143] Sampling rule: Sampling is performed on each leaf node containing 35 points.

[0144] Restore the point back to three-dimensional space. First, the rotation formula v mentioned in the second step rot =Rv, we only need to multiply both sides by R-1 The inverse transformation can be completed, that is, v = R -1 v rot , you can rotate the plane back and then create Figure 8 The geometry shown is:

[0145] Depend on Figure 8 It can be seen that a sphere is established with the point of the projection plane as the center and the distance d as the radius. The required three-dimensional space coordinates are the coordinates of the intersection of the straight line passing through the center of the sphere and in the direction of the normal vector of the projection plane and the spherical surface:

[0146] The coordinates of the center of the sphere (x n ,y n , z n ), given the radius d, we can get the equation of the sphere as:

[0147] (xx n ) 2 +(yy n ) 2 +(zz n ) 2 =d 2

[0148] Substituting the parametric equation of the straight line into the above equation of the sphere, we get:

[0149]

[0150] Substituting into the parametric equation of the line, we get the restored three-dimensional coordinates:

[0151]

[0152]

[0153]

[0154] Reshape and smooth the point cloud. The point cloud restored to the three-dimensional space after the fourth step does not completely match the original point cloud. Fig. 9 It is shown that the point cloud needs to be smoothed and reshaped.

[0155] Changes in the number of points before and after sampling: pointCloud with 3895 points before encryption, pointCloudwith4684 points after encryption.

[0156] The application implementation example of the present invention provides a structural and functional integrated shape-controllable manufacturing platform, develops an optomechanical and electrical collaborative control system and a dedicated verification platform for key algorithms, and conducts research on signal processing, data processing, and automatic control.

[0157] Relevant evidence of the technical effects achieved by the embodiments of the present invention.

[0158] After sampling the upstream point cloud and reconstructing the downstream point cloud model, the model morphology of some workpiece point clouds after reconstruction is as follows: Fig.10 Shown

[0159] Fig.11 After upsampling the point cloud, the reconstructed model morphology can be seen intuitively from the image. The upper surface is obviously flatter and the side becomes smoother.

[0160] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. A person of ordinary skill in the art will appreciate that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVDROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0161] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A three-dimensional point cloud upsampling method for full plane projection, characterized in that: include: S1. Project all the scanned point clouds onto a two-dimensional plane; S2. Establish a quadtree data structure, transform the coordinates of the points on the projection plane, that is, convert the three-dimensional coordinates into two-dimensional coordinates, use the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly obtain the x and y coordinates; S3. Establish a quadtree data structure and perform fast Poisson disk sampling; S4. restore the point back to the three-dimensional space; S5. Reshape and smooth the point cloud: The point cloud restored to the three-dimensional space does not completely match the original point cloud, so it is smoothed and reshaped.

2. The three-dimensional point cloud upsampling method of full plane projection according to claim 1, characterized in that: Coordinates of the intersection point in S1: In the formula, the coordinates of the point cloud are: (x n ,y n , z n ), the coordinates of the camera are: (x0, y0, z0), the normal vector of the projection plane is: (n, m, l); Restore the point cloud back to three-dimensional space, calculate the distance from each point to the projection plane and record it. The formula is as follows:

3. The three-dimensional point cloud upsampling method of full plane projection according to claim 1, characterized in that: In S2, v rot =Rv, where v rot is the rotated vector, v is the vector to be rotated, and R is the rotation matrix to be obtained; v rot That is, the coordinates of the projection plane rotated to the xoy plane. The normal vector of the projection plane is (n, m, l), and the coordinates of the point cloud are: (x n ,y n , z n ), the coordinates of the camera are: (x0, y0, z0), and the coordinates of any point on the xoy plane are (p, q, 0).

4. The three-dimensional point cloud upsampling method of full plane projection according to claim 1, characterized in that: In S3, the tree building rule is: ensure that each leaf node contains at most 5 points. Sampling rule: Sampling is performed on each leaf node containing 35 points.

5. The three-dimensional point cloud upsampling method of full plane projection according to claim 1, characterized in that: The restored three-dimensional coordinates in S4: The normal vector of the projection plane is (n, m, l), and the coordinates of the point cloud are: (x n ,y n , z n ), d is the distance from each point to the projection plane.

6. A system for upsampling a 3D point cloud by full plane projection that implements the method for upsampling a 3D point cloud by full plane projection as claimed in any one of claims 1 to 5, characterized in that: include: Point cloud two-dimensional plane projection module projects all scanned point clouds onto a two-dimensional plane; The projection plane coordinate transformation module, in order to facilitate the operation of points projected onto the two-dimensional plane, establishes a quadtree data structure to transform the coordinates of the points on the projection plane, that is, converts the three-dimensional coordinates into two-dimensional coordinates, uses the Rodriguez rotation formula to rotate the projection plane to be parallel to the xoy plane, and then directly takes the x and y coordinates; Quadtree data structure establishment module, establishes quadtree data structure and performs fast Poisson disk sampling; The three-dimensional space restoration module restores the points back to the three-dimensional space; Smooth point cloud reorganization module: The point cloud restored to the three-dimensional space does not completely fit the original point cloud, so it is smoothed and reorganized.

7. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the full-plane projection three-dimensional point cloud upsampling method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the full-plane projection three-dimensional point cloud upsampling method as described in any one of claims 1 to 5.

9. An information data processing terminal, characterized in that: The information data processing terminal includes the full-plane projection three-dimensional point cloud upsampling system as described in claim 6.

Images