A method and system for photon point cloud filtering based on adaptive resolution voxels

Abstract

The application belongs to the technical field of laser detection, and discloses a photon point cloud filtering method and system based on adaptive resolution voxels, which comprises the following steps: S1, determining the resolution and boundary of voxels according to the elevation distribution characteristics and spatial distribution range of an initial photon point cloud, and voxelizing the original photon point cloud; S2, screening voxels greater than a determined threshold by using the density attribute of photons in the voxels, and retaining the photons in the voxels as a rough denoising photon point cloud; S3, determining the resolution and boundary of refined voxels according to the nearest neighbor distance and spatial distribution range of the rough denoising photon point cloud, and voxelizing the rough denoising photon point cloud; and S4, removing noise and extracting signal photons by using the topological relationship of voxels and the number of photons in the voxels. The application has the advantages of fast, automatic identification and extraction of signal photons on the surface of ground objects according to the distribution characteristics of photon point clouds, good accuracy, strong robustness, easy operation and the like.

Description

Technical Field

[0001] This invention belongs to the field of laser detection technology, and in particular relates to a photon point cloud filtering method and system based on adaptive resolution voxels. Background Technology

[0002] In space exploration, photon-counting lidar utilizes highly sensitive single-photon detectors to record the weak photon signals scattered or reflected by all targets during laser propagation, obtaining high-resolution photon point cloud data. Influenced by environmental conditions (atmospheric scattering and solar radiation), target characteristics (surface reflectivity and land cover type), and instrument performance (transmission energy, detector efficiency, dead time, and afterpulse effect), the photon point cloud data recorded by this technology contains a significant amount of noise in addition to the high-density target signals. The accuracy and effectiveness of photon point cloud filtering directly affect the application scope and data quality of subsequent data.

[0003] Currently, image filtering methods based on two-dimensional grids, probability density or histogram statistical methods, and density clustering analysis methods are commonly used to denoise photon point cloud data, but all of them have the following drawbacks:

[0004] (1) Image filtering method based on two-dimensional grid: The photon point cloud is rasterized into a two-dimensional image along the track according to the photon density, and then image processing technology is used to identify signal photons, such as edge and region detection, probability distribution function, median filter, etc. This method loses some useful information in the process of rasterizing the discrete photon point cloud into a two-dimensional image, which affects the integrity of the data and the denoising accuracy.

[0005] (2) Based on probability density or histogram statistical methods, signal photons are identified by using local statistical analysis, local angle mapping, local anomaly factor, K nearest neighbor, Bayesian decision and other analysis methods according to the local density information and relative adjacency relationship of photon point cloud. However, this method lacks consideration of the influence of terrain slope and the problem of uneven distribution of photon point cloud data density.

[0006] (3) Based on density clustering analysis, signal photons are identified by local clustering criteria through the spatial distribution characteristics of photon point clouds, such as traditional and improved DBSCAN (Density Based Spatial Clustering of Applications with Noise), OPTICS (ordering points to identify the clustering structure), crisp clustering, etc. This method is currently the most popular and effective method. However, it requires traversing all photon points, which makes it less efficient, and it is difficult to effectively filter noisy photons for complex terrain with steep terrain. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a photon point cloud filtering method and system based on adaptive resolution voxels.

[0008] This invention is implemented as follows: a photon point cloud filtering method based on adaptive resolution voxels, comprising:

[0009] S1. Based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, determine the resolution and boundary of the voxels and voxelize the original photon point cloud.

[0010] S2, using the density property of photons within voxels, selects voxels with a density greater than a certain threshold and retains the photons within voxels as photon point clouds after coarse denoising.

[0011] S3. Based on the nearest neighbor distance and spatial distribution range of the photon point cloud after coarse denoising, determine the resolution and boundary of the refined voxels and voxelize the photon point cloud after coarse denoising.

[0012] S4 utilizes the topological relationship of voxels and the number of photons within voxels to remove noise and extract signal photons.

[0013] Furthermore, S1 specifically includes:

[0014] S101, the voxel resolution is determined based on the elevation information distribution characteristics of the photon point cloud. Histogram statistical analysis is performed on the elevation information of the original photon point cloud. The elevation and frequency distribution curves are fitted using a Gaussian function to obtain the mean μ and standard deviation σ of the elevation distribution. The equation is as follows:

[0015]

[0016] In the formula, x is the elevation bin of the photon point cloud in histogram statistics, and f(x) is the distribution curve of elevation frequency in histogram statistics; in order for the voxel to be able to enclose most of the dense signal photons, the side length of the voxel lattice is determined according to σ, that is, the voxel resolution.

[0017] S102, determine the voxel grid boundary based on the spatial distribution range of the photon point cloud, analyze the three-dimensional coordinates composed of the planar position and elevation information of the original photon point cloud, obtain its distribution range in the X, Y, and Z axes, and obtain the minimum value in each of the three directions (X... min Y min Z min ) and maximum value (X) max Y max Z max This allows for the determination of the voxel bounding boxes of the initial photon point cloud;

[0018] S103. Based on the bounding box and grid size of the voxels, the original photon point cloud is voxelized, and a voxel grid with a resolution of 2.5 times σ is created to cover the entire space. At the same time, all photon point clouds are mapped to the corresponding voxel grids, and voxel grids with no photon distribution are deleted.

[0019] Furthermore, S2 specifically includes: using the density characteristics of voxels to remove noise photons, calculating the photon density within each voxel as the density attribute of that voxel, and the expression for the voxel density ρ is:

[0020]

[0021] In the formula, n is the number of photon points in a voxel, and D = 2.5σ is the resolution of the voxel. The average value a of the density attributes of all voxels is obtained as the threshold. Voxels greater than or equal to the threshold are regarded as signal voxels, and voxels less than the threshold are regarded as noise voxels. Photons in the signal voxels are extracted and retained as the coarsely denoised photon point cloud.

[0022] Furthermore, S3 specifically includes:

[0023] S301, iterate through and calculate the average and standard deviation of the nearest neighbor distances of all photon points in the photon point cloud, and find the i-th photon point P. i With the j-th photon point P j The nearest distance between (P) i ,P j The distance is expressed using Euclidean distance, and its expression is:

[0024]

[0025] In the formula, (x i ,y i ,z i ) and (x j ,y j ,z j () are points P i and point P j The three-dimensional coordinates are compared; the mean and standard deviation are compared to determine the resolution of the fine voxels. If the standard deviation is less than the mean, it indicates that the photon point cloud is densely distributed, and the voxel grid size is the sum of the mean and the standard deviation. Conversely, if the standard deviation is greater than the mean, it indicates that the photon point cloud is discretely distributed, and the voxel grid size is the standard deviation.

[0026] S302, determine the grid boundary of fine voxels based on the spatial distribution range of photon point clouds;

[0027] S303 voxels the coarsely denoised photon point cloud based on the bounding box and grid size of the fine voxels.

[0028] Furthermore, S4 specifically includes:

[0029] S401 uses the topological structure of voxels to determine the connectivity coefficients of voxels. Based on the adjacency relationships of each voxel in three-dimensional space, including 26 types between faces, edges, and vertices, and because signal photons are densely distributed on the surface of a continuous object, the connectivity between voxels is transitive, that is, two voxels that can be connected through a core voxel are also considered connected. The total number of all connected voxels can be used to represent the connectivity coefficient of a voxel. A depth-optimized search tree is used to traverse the connectivity coefficients of all voxels and retain voxel clusters with a connectivity coefficient greater than or equal to 27.

[0030] S402, further eliminate isolated voxels in voxel clusters by utilizing the topological relationship between voxels and the number of photons within voxels, that is, eliminate isolated voxels with a topological adjacency relationship between voxels less than or equal to 2 and a number of photons within voxels less than or equal to 1.

[0031] S403, the retained voxel photons are used as the signal photon point cloud after fine denoising.

[0032] Another object of the present invention is to provide a photon point cloud filtering system based on adaptive resolution voxels to implement the aforementioned photon point cloud filtering method, comprising:

[0033] The initial photon point cloud voxelization module is used to determine the resolution and boundaries of voxels based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, and to voxelize the original photon point cloud.

[0034] The photon point cloud coarse denoising module is used to utilize the density properties of photons within voxels to filter voxels that are greater than a certain threshold, and retain the photons within the voxels as the photon point cloud after coarse denoising.

[0035] The denoised photon point cloud voxelization module is used to determine the resolution and boundary of the fine voxels based on the nearest neighbor distance and spatial distribution range of the coarsely denoised photon point cloud, and to voxelize the coarsely denoised photon point cloud.

[0036] The signal photon extraction module is used to extract signal photons by eliminating noise using the topological relationship of voxels and the number of photons within voxels.

[0037] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the photon point cloud filtering method based on adaptive resolution voxels.

[0038] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the photon point cloud filtering method based on adaptive resolution voxels.

[0039] Another objective of this invention is to provide an information data processing terminal for implementing the aforementioned photon point cloud filtering system based on adaptive resolution voxels.

[0040] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0041] This invention can quickly and automatically identify and extract signal photons from the surface of ground objects based on the distribution characteristics of photon point clouds, and has the advantages of high accuracy, strong robustness and ease of operation.

[0042] The photon point cloud filtering method based on adaptive resolution voxels provided by this invention proposes a novel processing flow for denoising photon point cloud data. Specific technical advancements include:

[0043] Two-stage voxelization: This invention employs a two-stage voxelization process (coarse denoising and fine voxelization) to adaptively adjust the voxel resolution based on the different characteristics of the photon point cloud, making the filtering process more refined and efficient. The first stage of voxelization aims to quickly remove a large number of obvious noise points, while the second stage performs more detailed processing on the denoised point cloud, improving the extraction accuracy of signal photons.

[0044] Density-based noise removal: In the first stage of voxelization, a threshold is set to filter out voxels containing the true signal by utilizing the density properties of photons within the voxel, effectively removing sparsely distributed noise points. This step reduces the amount of data in subsequent processing and improves filtering efficiency.

[0045] Refined noise removal: In the second stage of voxelization, the resolution of voxels is further refined based on the features of the photon point cloud after coarse denoising, and even smaller noises are removed. This step allows the filtering method to adapt to the subtle changes in the photon point cloud data, and to more accurately preserve the signal photons.

[0046] By utilizing the topological relationships of voxels and analyzing the number of photons within each voxel, this invention can more accurately eliminate isolated noise points while retaining the signal photons that constitute the actual ground surface. This processing takes into account the spatial continuity of the photon point cloud, enhancing the reliability of the filtering results.

[0047] Improving filtering accuracy: By using adaptive resolution voxelization and density- and topology-based filtering, this invention can effectively improve the accuracy of photon point cloud filtering and reduce signal loss.

[0048] Enhanced data processing efficiency: Two-stage voxelization and density-based initial denoising significantly reduce the amount of data and improve the efficiency of subsequent processing.

[0049] High adaptability: This invention is applicable to photon point cloud data with different characteristics and distribution ranges, and has strong adaptability and versatility.

[0050] High degree of automation: The entire filtering process can be automated, reducing manual intervention and improving processing efficiency and reliability.

[0051] In summary, the photonic point cloud filtering method based on adaptive resolution voxels provided by this invention demonstrates significant technological advancements and application value in the field of point cloud data processing. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of the photon point cloud filtering method based on adaptive resolution voxels provided in an embodiment of the present invention;

[0054] Figure 2 This is a structural diagram of the photon point cloud filtering system based on adaptive resolution voxels provided in an embodiment of the present invention;

[0055] Figure 3 These are schematic diagrams of photon point cloud voxels and fine denoising voxels provided in the embodiments of the present invention; wherein, (a) is a photon point cloud with coarse resolution voxels in the embodiment, and (b) is a photon point cloud with fine resolution voxels in the embodiment.

[0056] Figure 4 These are schematic diagrams of photon point clouds before and after denoising provided in the embodiments of the present invention; wherein, (a) is the original photon point cloud before denoising in the embodiment, and (b) is the signal photon point cloud after denoising in the embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] To address the problems existing in the prior art, this invention provides a photon point cloud filtering method and system based on adaptive resolution voxels. The invention will be described in detail below with reference to the accompanying drawings.

[0059] Example 1: Photon point cloud filtering in urban environments

[0060] Photon point cloud data in urban environments contains a large amount of information about ground features such as buildings, vehicles, and trees, as well as noise points caused by reflection and scattering. The filtering method of this invention can be effectively applied to photon point cloud filtering in such complex environments. The specific steps include:

[0061] 1. Voxelization: First, based on the elevation distribution and spatial range of the urban environmental photon point cloud, a suitable voxel resolution and boundary are determined, and preliminary voxelization is performed to convert the original photon point cloud into a voxel grid representation.

[0062] 2. Coarse denoising: Based on voxelization, voxels with a density greater than a set threshold are selected according to the density properties of photons within the voxels, and the photon points within these voxels are retained to form a coarsely denoised photon point cloud.

[0063] 3. Refined Voxelization: Based on the photon point cloud after coarse denoising, the resolution and boundaries of the voxels are redefined, and more refined voxelization is performed to more accurately remove the remaining noise.

[0064] 4. Noise Removal and Signal Extraction: By utilizing the number of photons within refined voxels and the topological relationships between voxels, isolated noise voxels are removed and signal photons are extracted, thereby obtaining a filtered photon point cloud that more accurately represents the features of ground objects in the urban environment.

[0065] Example 2: Photon point cloud filtering in forest-covered areas

[0066] When filtering photon point clouds in forested areas, it is necessary to process photon point clouds generated by reflections from tree canopies and the ground, while removing noise introduced by factors such as atmospheric scattering. The method of this invention is also applicable to such natural environments, and the specific steps include:

[0067] 1. Preliminary Voxelization: Based on the characteristics of photon point clouds in forest areas, determine appropriate voxel resolution and boundaries, and perform preliminary voxelization to simplify the complex forest photon point clouds into a voxel grid representation.

[0068] 2. Density-based preliminary denoising: By analyzing the density of photons within voxels, voxels with lower density are eliminated. These low-density voxels usually correspond to noise points caused by scattering or reflection.

[0069] 3. Voxelization again: Based on the photon point cloud after initial denoising, the resolution of the voxels is further refined in order to perform more refined filtering on the remaining photon point cloud.

[0070] 4. Detailed noise removal: By utilizing the topological relationships and photon counts of refined voxels, isolated noise voxels are further removed, thereby extracting signal photons representing the forest floor and canopy.

[0071] These two embodiments demonstrate the application capability and advantages of the filtering method of the present invention in different environments, including effectively removing noise and extracting signal photons in complex urban environments and natural forest areas. Through adaptive resolution voxelization and density- and topology-based filtering, the present invention not only improves the accuracy of photon point cloud filtering but also enhances the efficiency and adaptability of data processing, providing a high-quality data foundation for subsequent photon point cloud analysis and applications.

[0072] like Figure 1 As shown, the photon point cloud filtering method based on adaptive resolution voxels provided in this embodiment of the invention includes:

[0073] S1. Based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, determine the resolution and boundary of the voxels and voxelize the original photon point cloud.

[0074] S2, using the density property of photons within voxels, selects voxels with a density greater than a certain threshold and retains the photons within voxels as photon point clouds after coarse denoising.

[0075] S3. Based on the nearest neighbor distance and spatial distribution range of the photon point cloud after coarse denoising, determine the resolution and boundary of the refined voxels and voxelize the photon point cloud after coarse denoising.

[0076] S4 utilizes the topological relationship of voxels and the number of photons within voxels to remove noise and extract signal photons.

[0077] S1 specifically includes:

[0078] S101, the voxel resolution is determined based on the elevation information distribution characteristics of the photon point cloud. Histogram statistical analysis is performed on the elevation information of the original photon point cloud. The elevation and frequency distribution curves are fitted using a Gaussian function to obtain the mean μ and standard deviation σ of the elevation distribution. The equation is as follows:

[0079]

[0080] In the formula, x is the elevation bin of the photon point cloud in histogram statistics, and f(x) is the distribution curve of elevation frequency in histogram statistics; in order for the voxel to be able to enclose most of the dense signal photons, the side length of the voxel lattice is determined according to σ, that is, the voxel resolution.

[0081] S102, determine the voxel grid boundary based on the spatial distribution range of the photon point cloud, analyze the three-dimensional coordinates composed of the planar position and elevation information of the original photon point cloud, obtain its distribution range in the X, Y, and Z axes, and obtain the minimum value in each of the three directions (X... min Y min Z min ) and maximum value (X) max Ymax Z max This allows for the determination of the voxel bounding boxes of the initial photon point cloud;

[0082] S103. Based on the bounding box and grid size of the voxels, the original photon point cloud is voxelized, and a voxel grid with a resolution of 2.5 times σ is created to cover the entire space. At the same time, all photon point clouds are mapped to the corresponding voxel grids, and voxel grids with no photon distribution are deleted.

[0083] S2 specifically includes: using the density characteristics of voxels to remove noise photons, calculating the photon density within each voxel as the density attribute of that voxel, and the expression for the voxel density ρ is:

[0084]

[0085] In the formula, n is the number of photon points in a voxel, and D = 2.5σ is the resolution of the voxel. The average value a of the density attributes of all voxels is obtained as the threshold. Voxels greater than or equal to the threshold are regarded as signal voxels, and voxels less than the threshold are regarded as noise voxels. Photons in the signal voxels are extracted and retained as the coarsely denoised photon point cloud.

[0086] S3 specifically includes:

[0087] S301, iterate through and calculate the average and standard deviation of the nearest neighbor distances of all photon points in the photon point cloud, and find the i-th photon point P. i With the j-th photon point P j The nearest distance between (P) i ,P j The distance is expressed using Euclidean distance, and its expression is:

[0088]

[0089] In the formula, (x i ,y i ,z i ) and (x j ,y j ,z j () are points P i and point P j The three-dimensional coordinates are compared; the mean and standard deviation are compared to determine the resolution of the fine voxels. If the standard deviation is less than the mean, it indicates that the photon point cloud is densely distributed, and the voxel grid size is the sum of the mean and the standard deviation. Conversely, if the standard deviation is greater than the mean, it indicates that the photon point cloud is discretely distributed, and the voxel grid size is the standard deviation.

[0090] S302, determine the grid boundary of fine voxels based on the spatial distribution range of photon point clouds;

[0091] S303 voxels the coarsely denoised photon point cloud based on the bounding box and grid size of the fine voxels.

[0092] S4 specifically includes:

[0093] S401 uses the topological structure of voxels to determine the connectivity coefficients of voxels. Based on the adjacency relationships of each voxel in three-dimensional space, including 26 types between faces, edges, and vertices, and because signal photons are densely distributed on the surface of a continuous object, the connectivity between voxels is transitive, that is, two voxels that can be connected through a core voxel are also considered connected. The total number of all connected voxels can be used to represent the connectivity coefficient of a voxel. A depth-optimized search tree is used to traverse the connectivity coefficients of all voxels and retain voxel clusters with a connectivity coefficient greater than or equal to 27.

[0094] S402, further eliminate isolated voxels in voxel clusters by utilizing the topological relationship between voxels and the number of photons within voxels, that is, eliminate isolated voxels with a topological adjacency relationship between voxels less than or equal to 2 and a number of photons within voxels less than or equal to 1.

[0095] S403, the retained voxel photons are used as the signal photon point cloud after fine denoising.

[0096] like Figure 2 As shown, the photon point cloud filtering system based on adaptive resolution voxels provided in this embodiment of the invention includes:

[0097] The initial photon point cloud voxelization module is used to determine the resolution and boundaries of voxels based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, and to voxelize the original photon point cloud.

[0098] The photon point cloud coarse denoising module is used to utilize the density properties of photons within voxels to filter voxels that are greater than a certain threshold, and retain the photons within the voxels as the photon point cloud after coarse denoising.

[0099] The denoised photon point cloud voxelization module is used to determine the resolution and boundary of the fine voxels based on the nearest neighbor distance and spatial distribution range of the coarsely denoised photon point cloud, and to voxelize the coarsely denoised photon point cloud.

[0100] The signal photon extraction module is used to extract signal photons by eliminating noise using the topological relationship of voxels and the number of photons within voxels.

[0101] like Figure 1 As shown, an adaptive resolution voxel photon point cloud filtering method includes the following steps:

[0102] (1) Using the ATL03 level photon point cloud of the ICESat2 satellite ATLAS laser altimetry system as input data, such as Figure 4As shown in (a), the elevation information of the photon point cloud is first statistically analyzed using a histogram with a bin size of 10 meters. A Gaussian function is then used to fit the mean and standard deviation of the elevation and frequency distribution curves. To ensure that voxels can encapsulate most of the dense photons, σ is set to a resolution of 2.5 times the coarsely denoised voxels. Then, considering the spatial distribution range of the photon point cloud, the photon point cloud is voxelized, as shown... Figure 3 As shown in (a).

[0103] (2) Divide the number of photon point clouds contained in the voxel by the voxel volume to obtain the density attribute of the voxel, count and obtain the average value of the density of all voxels, and use it as the coarse denoising threshold. Voxels with a density less than the threshold are regarded as noise voxels, and those with a density greater than the threshold are regarded as signal voxels. The photons in all signal voxels are retained as signal photon point clouds after coarse denoising.

[0104] (3) Calculate the mean and standard deviation of the nearest neighbor distances of the photon point cloud after coarse denoising. Compare the mean and standard deviation to determine the voxel resolution. If the standard deviation is less than the mean, the voxel resolution is the sum of the standard deviation and the mean; otherwise, it is the mean. Simultaneously, based on the spatial distribution range of the coarse denoising photon point cloud, voxelize the coarse denoising photon point cloud, such as... Figure 3 As shown in (b).

[0105] (4) Noise is removed and signal photons are extracted using the topological relationships and the number of photons within voxels. The connectivity coefficients of voxels are determined based on their topological structure. A depth-optimized search tree is used to traverse the connectivity coefficients of all voxels based on their adjacency relationships in 3D space and the transitivity of connectivity between voxels, retaining voxel clusters with connectivity coefficients greater than or equal to 27. Then, based on the topological relationships between voxels and the number of photons within voxels, isolated voxels within voxel clusters are further removed, specifically those with a topological adjacency relationship less than or equal to 2 and a number of photons less than or equal to 1. Finally, the retained voxel photon point cloud is used as the finely denoised signal photon point cloud, such as... Figure 4 As shown in (b).

[0106] The denoised signal photon point cloud was compared with the manually labeled signal photon, as shown in Table 1. The denoising accuracy was quantitatively evaluated using precision (P), recall (R), and F1 score. The method proposed in this invention achieved precision of 0.9870, recall of 0.9721, and F1 score of 0.9844.

[0107] Table 1. Denoising accuracy of manually annotated photon point clouds

[0108]

[0109] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a photon point cloud filtering method based on adaptive resolution voxels.

[0110] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform steps of a photon point cloud filtering method based on adaptive resolution voxels.

[0111] An application embodiment of the present invention provides an information data processing terminal, which is used to implement a photon point cloud filtering system based on adaptive resolution voxels.

[0112] It should be noted that embodiments of the present invention can be implemented using hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented using hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or using software executed by various types of processors, or using a combination of the above-described hardware circuitry and software, such as firmware.

[0113] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A photon point cloud filtering method based on adaptive resolution voxels, characterized in that, include: S1. Based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, determine the resolution and boundary of the voxels and voxelize the original photon point cloud. S2, using the density property of photons within voxels, selects voxels with a density greater than a certain threshold and retains the photons within voxels as photon point clouds after coarse denoising. S3. Based on the nearest neighbor distance and spatial distribution range of the photon point cloud after coarse denoising, determine the resolution and boundary of the refined voxels and voxelize the photon point cloud after coarse denoising. S4 utilizes the topological relationship of voxels and the number of photons within voxels to remove noise and extract signal photons; S4 specifically includes: S401 uses the topological structure of voxels to determine the connectivity coefficients of voxels. Based on the adjacency relationships of each voxel in three-dimensional space, including 26 types between faces, edges, and vertices, and because signal photons are densely distributed on the surface of a continuous object, the connectivity between voxels is transitive, that is, two voxels that can be connected through a core voxel are also considered connected. The total number of all connected voxels is used to represent the connectivity coefficient of a voxel. A depth-optimized search tree is used to traverse the connectivity coefficients of all voxels and retain voxel clusters with a connectivity coefficient greater than or equal to 27. S402, further eliminate isolated voxels in voxel clusters by utilizing the topological relationship between voxels and the number of photons within voxels, that is, eliminate isolated voxels with a topological adjacency relationship between voxels less than or equal to 2 and a number of photons within voxels less than or equal to 1. S403, the retained voxel photons are used as the signal photon point cloud after fine denoising.

2. The photon point cloud filtering method based on adaptive resolution voxels as described in claim 1, characterized in that, S1 specifically includes: S101, the voxel resolution is determined based on the elevation information distribution characteristics of the photon point cloud. Histogram statistical analysis is performed on the elevation information of the original photon point cloud, and the elevation-frequency distribution curve is fitted using a Gaussian function to obtain the average value of the elevation distribution. with standard deviation The equation is expressed as: ； In the formula, It is the elevation bin of the photon point cloud in histogram statistics. It is the distribution curve of elevation frequency in histogram statistics; in order for the voxel to encapsulate most of the dense signal photons, according to Determine the side length of the voxel lattice, i.e., the voxel resolution; S102, determine the voxel grid boundary based on the spatial distribution range of the photon point cloud, analyze the three-dimensional coordinates composed of the planar position and elevation information of the original photon point cloud, obtain its distribution range in the X, Y, and Z axes, and obtain the minimum value in each of the three directions (X... min Y min Z min ) and maximum value (X) max Y max Z max ), thereby determining the voxel bounding box of the initial photon point cloud; S103 voxels the original photon point cloud based on the voxel bounding boxes and grid size, creating a voxel with a resolution 2.5 times higher. The voxel grid covers the entire space, and all photon point clouds are mapped to the corresponding voxel grids, while voxel grids with no photon distribution are deleted.

3. The photon point cloud filtering method based on adaptive resolution voxels as described in claim 1, characterized in that, S2 specifically includes: using the density characteristics of voxels to remove noise photons, calculating the photon density within each voxel as the density attribute of that voxel, and voxel density. The expression is: ； In the formula, This represents the number of photon points within a voxel. The resolution of the voxels is determined; the average value 'a' of the density attributes of all voxels is obtained as the threshold; voxels greater than or equal to the threshold are taken as signal voxels, and voxels less than the threshold are taken as noise voxels; photons in the signal voxels are extracted and retained as coarsely denoised photon point clouds.

4. The photon point cloud filtering method based on adaptive resolution voxels as described in claim 1, characterized in that, S3 specifically includes: S301, iterate through and calculate the average and standard deviation of the nearest neighbor distances of all photon points in the photon point cloud, and find the i-th photon point P. i With the j-th photon point P j closest distance between Using Euclidean distance, its expression is: ； In the formula, (x i , y i , z i ) and (x j , y j , z j () are points P i and point P j The three-dimensional coordinates are compared; the mean and standard deviation are compared to determine the resolution of the fine voxels. If the standard deviation is less than the mean, it indicates that the photon point cloud is densely distributed, and the voxel grid size is the sum of the mean and the standard deviation. Conversely, if the standard deviation is greater than the mean, it indicates that the photon point cloud is discretely distributed, and the voxel grid size is the standard deviation. S302, determine the grid boundary of fine voxels based on the spatial distribution range of photon point clouds; S303 voxels the coarsely denoised photon point cloud based on the bounding box and grid size of the fine voxels.

5. A photon point cloud filtering system based on adaptive resolution voxels, implementing the photon point cloud filtering method based on adaptive resolution voxels as described in any one of claims 1 to 4, characterized in that, include: The initial photon point cloud voxelization module is used to determine the resolution and boundaries of voxels based on the elevation distribution characteristics and spatial distribution range of the initial photon point cloud, and to voxelize the original photon point cloud. The photon point cloud coarse denoising module is used to utilize the density properties of photons within voxels to filter voxels that are greater than a certain threshold, and retain the photons within the voxels as the photon point cloud after coarse denoising. The denoised photon point cloud voxelization module is used to determine the resolution and boundary of the fine voxels based on the nearest neighbor distance and spatial distribution range of the coarsely denoised photon point cloud, and to voxelize the coarsely denoised photon point cloud. The signal photon extraction module is used to extract signal photons by eliminating noise using the topological relationship of voxels and the number of photons within voxels.

6. A computer device comprising a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the photon point cloud filtering method based on adaptive resolution voxels as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the photon point cloud filtering method based on adaptive resolution voxels as described in any one of claims 1 to 4.

8. An information data processing terminal, the information data processing terminal being used to implement the photon point cloud filtering system based on adaptive resolution voxels as described in claim 5.

Images