Method and device for evaluating acquisition quality of three-dimensional Gaussian splashing data

Abstract

The invention provides a three-dimensional Gaussian splash data acquisition quality evaluation method and device, and relates to the technical field of image data processing, and the method comprises the steps: carrying out the voxelization discrete processing of point cloud data corresponding to a target scene, so as to construct a voxel map based on a Hash structure, and the voxel map comprises a plurality of voxels; performing rasterization calculation on the plurality of voxels one by one to determine observability information of each voxel under different visual angles, and generating a visual angle observation set corresponding to each voxel; performing modeling on the voxel by adopting a regular icosahedron structure, and performing observation confidence calculation on the voxel based on the distance parameter corresponding to each view angle and the view angle observation set; and determining a quality evaluation result of the corresponding point in the voxel based on a calculation result. According to the method, the potential defect area is found in time in the data acquisition stage, later modeling failure or rendering quality degradation is avoided, and the overall efficiency and controllability of the three-dimensional data acquisition and three-dimensional reconstruction process are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and particularly relates to a three-dimensional Gaussian splatting data acquisition quality evaluation method and device. BACKGROUND

[0002] Three-dimensional Gaussian splatting (3DGS) is a new three-dimensional scene representation method, which can realize high-fidelity and free-view real-time rendering. In the construction process of 3DGS, point cloud data, image sequences and corresponding camera poses of the scene are usually required as input, and the final Gaussian representation is obtained by numerical optimization iteration. However, the rendering quality of 3DGS depends largely on the sufficiency and accuracy of data acquisition. In large-scale scenes, the optimization process of 3DGS usually takes a long time, and the rendering effect can only be evaluated after the optimization is completed, so as to find out which areas have insufficient data acquisition. This post-evaluation method often requires the acquisition personnel to return to the site for re-acquisition according to the optimized rendering result, but for some areas that are difficult to re-acquire (such as under a bridge, dangerous terrain or limited space), this process is almost impossible, which seriously affects the efficiency of three-dimensional data acquisition and modeling. SUMMARY

[0003] Therefore, the present application aims to provide at least a three-dimensional Gaussian splatting data acquisition quality evaluation method and device, which can find potential defect areas in the data acquisition stage in time, avoid late modeling failure or rendering quality degradation, and improve the overall efficiency and controllability of three-dimensional data acquisition and three-dimensional reconstruction process.

[0004] The present application mainly includes the following aspects: In a first aspect, the present application provides a three-dimensional Gaussian splatting data acquisition quality evaluation method, which comprises: performing voxelization and discrete processing on point cloud data corresponding to a target scene to construct a voxel map based on a hash structure, the voxel map comprising a plurality of voxels; performing rasterization calculation on each of the plurality of voxels to determine the observability information of each voxel under different viewing angles, and generating a viewing angle observation set corresponding to each voxel; modeling the voxels using an icosahedron structure, and performing observation confidence calculation on the voxels based on the distance parameters corresponding to each viewing angle and the viewing angle observation set; and determining the quality evaluation result of the corresponding points in the voxels based on the calculation result.

[0005] In one possible implementation, a voxel map is constructed as follows: based on the three-dimensional coordinates corresponding to each point in the point cloud data, multiple voxels and the voxel center corresponding to each voxel are determined by discretization; based on the three-dimensional coordinates corresponding to each voxel center, the hash value corresponding to each voxel center is calculated; and based on the multiple voxels, the voxel center corresponding to each voxel, the hash value, and the point index belonging to the hash value, a voxel map based on a hash structure is constructed.

[0006] In one possible implementation, the view observation set corresponding to each voxel is determined as follows: For each voxel: based on the voxel center and the extrinsic parameters of each camera, the spatial coordinates of the voxel in each camera coordinate system are determined; based on the spatial coordinates of the voxel center in the coordinate system corresponding to each camera and the camera intrinsic parameters, the projection of the voxel on the pixel plane corresponding to each camera is determined; based on the given voxel size, the projection radius of the voxel on the pixel plane corresponding to each camera is determined; based on the projection radius and projection of the voxel on the pixel plane corresponding to each camera, the voxel index set corresponding to each pixel in each camera plane and its corresponding depth cache are determined; based on the voxel index set and depth cache corresponding to each pixel in each camera plane, the view observation set corresponding to each voxel is determined, and the view observation set records the camera observation viewpoints that can observe the voxel.

[0007] In one possible implementation, the camera extrinsic parameters include a rotation matrix and a translation vector, wherein the target space coordinates of the voxel center in each camera coordinate system are determined by the following formula:

[0008] in, This represents the three-dimensional coordinates of the voxel center corresponding to the k-th voxel. , This represents the rotation matrix corresponding to the i-th camera. This represents the translation vector corresponding to the i-th camera. This represents the spatial coordinates of the voxel center corresponding to the k-th voxel in the coordinate system of the i-th camera; This represents the x-coordinate of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the ordinate of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the depth of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This indicates transpose.

[0009] In one possible implementation, the projection of a voxel onto the corresponding pixel plane of each camera is determined by the following formula:

[0010]

[0011]

[0012] in, Represents the center of the k-th voxel. The projection onto the pixel plane corresponding to the i-th camera. This represents the width coordinate of the center of the k-th voxel in the pixel plane corresponding to the i-th camera. This represents the height coordinate of the center of the k-th voxel in the pixel plane corresponding to the ith camera. This represents the pixel focal length of the i-th camera on the x-axis. This represents the pixel focal length of the i-th camera on the y-axis. This represents the x-coordinate of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the ordinate of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the coordinates of the principal point C corresponding to the i-th camera on the camera's x-axis. Let C represent the coordinates of the principal point C corresponding to the i-th camera on the camera's y-axis.

[0013] In one possible implementation, the projection radius of the voxel on the corresponding pixel plane of each camera is determined by the following formula:

[0014] In this formula, This represents the projection radius of the k-th voxel onto the pixel plane corresponding to the i-th camera. This represents the size of a given voxel corresponding to the k-th voxel. This represents the pixel focal length of the i-th camera along the x-axis. It represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera.

[0015] In one possible implementation, the voxel index set and depth cache corresponding to each pixel are determined by the following formula:

[0016]

[0017] In this formula, The coordinates of the i-th camera on the pixel plane are: The set of voxel indices corresponding to the pixels. This represents the projection of the k-th voxel onto the pixel plane corresponding to the i-th camera. This represents the projection radius of the k-th voxel on the pixel plane corresponding to the i-th camera; The coordinates of the i-th camera on the pixel plane are: The depth cache corresponding to the pixels, This represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera. Represents a set of voxel indices Corresponding multiple voxels The minimum value in.

[0018] In one possible implementation, the observation confidence level for each voxel is determined using the following formula:

[0019]

[0020] in, This represents the decay function based on Euclidean distance. This represents the camera translation corresponding to the p-th camera in the view observation set corresponding to the k-th voxel. This represents the center coordinates of the k-th voxel. express and The Euclidean norm between them Let represent the set of viewpoints corresponding to the m-th face in the icosahedral structure corresponding to the k-th voxel. This represents the observation confidence level of the m-th face in the icosahedral structure corresponding to the k-th voxel.

[0021] The overall observation confidence level corresponds to the k-th voxel, which is the sum of the observation confidence levels of the twenty faces in the icosahedral structure corresponding to the k-th voxel.

[0022] In one possible implementation, the calculation results describe the overall observation confidence of the voxel center corresponding to each voxel. The quality assessment results of corresponding points within a voxel are determined as follows: for each voxel, its corresponding voxel center is regarded as a control plotting point, and the observation confidence of each point within the voxel is obtained by linear interpolation of the overall observation confidence of the voxel center; for each point within the voxel, the point is visualized using the observation confidence of that point, and the quality of the point is determined based on the display results. The observation confidence of each point indicates the grayscale value corresponding to the point in the image display.

[0023] Secondly, embodiments of this application also provide a quality assessment device for the acquisition of three-dimensional Gaussian splash data. The device includes: a construction module, which performs voxelization discretization processing on the point cloud data corresponding to the target scene to construct a voxel map based on a hash structure, the voxel map including multiple voxels; a rasterization calculation module, which performs rasterization calculation on each of the multiple voxels to determine the observability information of each voxel under different viewpoints and generates a viewpoint observation set corresponding to each voxel; a modeling module, which models the voxels using a regular icosahedral structure and calculates the observation confidence of the voxels based on the distance parameters corresponding to each viewpoint and the viewpoint observation set; and a quality assessment module, which determines the quality assessment result of the corresponding points within the voxel based on the calculation results.

[0024] This application provides a method and apparatus for evaluating the acquisition quality of 3D Gaussian splash data, comprising: performing voxelization discretization processing on point cloud data corresponding to a target scene to construct a voxel map based on a hash structure, the voxel map including multiple voxels; performing rasterization calculations on each voxel to determine the observability information of each voxel under different viewpoints, and generating a viewpoint observation set corresponding to each voxel; modeling the voxels using a regular icosahedral structure, and calculating the observation confidence of the voxels based on the distance parameters corresponding to each viewpoint and the viewpoint observation set; and determining the quality evaluation result of the corresponding points within the voxel based on the calculation results. By analyzing the input data of 3DGS to determine the data acquisition quality, time-consuming optimization processes are eliminated, improving the overall efficiency of data acquisition and quality control.

[0025] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A flowchart of a method for evaluating the acquisition quality of three-dimensional Gaussian splash data provided in an embodiment of this application is shown; Figure 2 This document illustrates a flowchart of a method for determining the viewpoint observation set corresponding to a voxel, as provided in an embodiment of this application. Figure 3 This paper illustrates a functional block diagram of a three-dimensional Gaussian splash data acquisition quality assessment device provided in an embodiment of this application. Figure 4A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0029] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0030] 3D Gaussian Splatting (3DGS) is an emerging method for representing 3D scenes, enabling high-fidelity, free-viewpoint real-time rendering. The construction of 3DGS typically requires inputting point cloud data, image sequences, and their corresponding camera poses. Numerical optimization and iterative solutions are then used to obtain the final Gaussian representation. However, the rendering quality of 3DGS largely depends on the sufficiency and accuracy of data acquisition. In large-scale scenes, the optimization process of 3DGS is often time-consuming. Only after optimization can the rendering effect be evaluated, revealing areas with insufficient data acquisition. This retrospective evaluation method often requires data acquisition personnel to return to the site for supplementary data acquisition based on the optimized rendering results. For areas where repeated data acquisition is difficult (such as under bridges, dangerous terrain, or confined spaces), this process is almost impractical, severely impacting the efficiency of 3D data acquisition and modeling.

[0031] Based on this, this application provides a method and apparatus for evaluating the acquisition quality of 3D Gaussian splash data. Without requiring a time-consuming 3DGS optimization process, it directly evaluates the integrity and reliability of scene acquisition based on input data, thereby achieving rapid quality diagnosis of large-scale scenes. The method provided in this application can promptly identify potential defect areas during the data acquisition stage, avoiding later modeling failures or rendering quality degradation, and improving the overall efficiency and controllability of the 3D data acquisition and 3D reconstruction process. Specifically, as follows: Please see Figure 1 , Figure 1 A flowchart illustrating a method for evaluating the acquisition quality of three-dimensional Gaussian splash data provided in an embodiment of this application is shown. Figure 1 As shown, the method provided in this application embodiment includes the following steps: S100. Perform voxelization discretization on the point cloud data corresponding to the target scene to construct a voxel map based on a hash structure.

[0032] A voxel map consists of multiple voxels.

[0033] S200. Perform rasterization calculations on multiple voxels one by one to determine the observability information of each voxel under different viewpoints and generate a viewpoint observation set corresponding to each voxel.

[0034] S300. The voxels are modeled using a regular icosahedral structure, and the observation confidence of the voxels is calculated based on the distance parameters corresponding to each viewpoint and the viewpoint observation set.

[0035] S400. Based on the calculation results, determine the quality assessment results of the corresponding points within the voxel.

[0036] In a specific implementation, step S100 first involves collecting point cloud data within the target scene, and then constructing a voxel map based on a hash structure based on the point cloud data. In a preferred embodiment, step S100 includes: Based on the three-dimensional coordinates of each point in the point cloud data, multiple voxels and their corresponding voxel centers are determined by discretization. Based on the three-dimensional coordinates of each voxel center, the hash value of each voxel center is calculated. Based on multiple voxels, their corresponding voxel centers, hash values, and point indices belonging to the hash values, a voxel map based on a hash structure is constructed. In one example, the point cloud data P corresponding to the target scene is read, denoised, and the 3D coordinates of each point in the processed point cloud data P are determined. ,in, This represents the 3D coordinates of the q-th point in the point cloud data. This represents the x-coordinate of the q-th point. This represents the y-coordinate of the q-th point in the point cloud data. Let represent the z-axis coordinate of the q-th point in the point cloud data. Assuming a given voxel size Vs (for example, 0.2 meters), the voxel center corresponding to each point is determined by the following formula: .

[0037] Traverse the three-dimensional coordinates of each point in the point cloud data, determine the voxel center of each point in the point cloud data based on the above formula, and determine the voxel using the calculated voxel center.

[0038] In a preferred embodiment, the hash value corresponding to the voxel center is calculated using the following formula:

[0039] in, Represents the center of the k-th voxel. The corresponding hash value, Represents the center of the k-th voxel. The corresponding x-axis coordinates Represents the center of the k-th voxel. The corresponding y-axis coordinate, Represents the center of the k-th voxel. The corresponding z-axis coordinates are used to construct a voxel map V={ using the voxel center, its corresponding hash value, 3D coordinates, and all points belonging to the hash value. }, where N represents the total number of voxels from the point cloud data.

[0040] In a preferred embodiment, please refer to Figure 2 , Figure 2 A flowchart illustrating a method for determining the viewpoint observation set corresponding to a voxel, as provided in an embodiment of this application, is shown. Figure 2 As shown, step S200 includes: S2001. For each voxel, determine the spatial coordinates of the voxel in each camera coordinate system based on the voxel center and the external parameters of each camera.

[0041] S2002. For each voxel, determine the projection of the voxel onto the corresponding pixel plane of each camera based on the spatial coordinates of the voxel in the coordinate system corresponding to each camera and the camera intrinsic parameters.

[0042] S2003. For each voxel, determine the projection radius of the voxel on the corresponding pixel plane of each camera based on the given voxel size.

[0043] S2004. Based on the projection radius and projection of the voxel on the corresponding pixel plane of each camera, determine the voxel index set and its corresponding depth cache for each pixel of each camera.

[0044] S2005. Based on the voxel index set and depth cache corresponding to each pixel under each camera, determine the view observation set corresponding to each voxel.

[0045] The viewpoint observation set records the camera observation viewpoints that can be observed for voxels.

[0046] In one specific embodiment, in step S2001, the camera extrinsic parameters include a rotation matrix R and a translation vector t, wherein the spatial coordinates of the voxel center in each camera coordinate system are determined by the following formula:

[0047] In this formula, This represents the three-dimensional coordinates of the voxel center corresponding to the k-th voxel. , This represents the rotation matrix corresponding to the i-th camera. This represents the translation vector corresponding to the i-th camera. This represents the spatial coordinates of the center of the k-th voxel in the coordinate system corresponding to the i-th camera. This represents the x-coordinate of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the ordinate of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This represents the depth of the center of the k-th voxel in the camera coordinate system corresponding to the i-th camera. This indicates transpose.

[0048] In a preferred embodiment, in step S2002, the camera intrinsic parameters include the camera focal length and the camera principal point, wherein the projection of the voxel on the corresponding pixel plane of each camera is determined by the following formula:

[0049]

[0050]

[0051] In this formula, Represents the center of the k-th voxel. The projection onto the pixel plane corresponding to the i-th camera. Represents the center of the k-th voxel. The width coordinates of the pixel plane corresponding to the i-th camera. Represents the center of the k-th voxel. The height coordinates of the pixel plane corresponding to the i-th camera.

[0052] This represents the pixel focal length of the i-th camera on the x-axis. This represents the pixel focal length of the i-th camera on the y-axis. This represents the coordinates of the principal point C corresponding to the i-th camera on the camera's x-axis. Let C represent the coordinates of the principal point C corresponding to the i-th camera on the camera's y-axis.

[0053] In step S2003, based on the spatial scale corresponding to the voxel, the projection radius of the voxel on the pixel plane can be approximated as:

[0054] In this formula, This represents the projection radius of the center of the k-th voxel onto the pixel plane corresponding to the i-th camera. This represents the size of a given voxel corresponding to the k-th voxel. This represents the pixel focal length of the i-th camera on the x-axis.

[0055] In a preferred embodiment, in step S2004, the voxel index set and depth buffer corresponding to each pixel on the camera pixel plane are determined by the following formula:

[0056]

[0057] In this formula, The coordinates of the i-th camera on the pixel plane are: The set of voxel indices corresponding to the pixels. Represents the center of the k-th voxel. The projection onto the pixel plane corresponding to the i-th camera. This represents the projection radius of the k-th voxel onto the pixel plane corresponding to the i-th camera. Represents pixels on the pixel plane With the kth voxel center The distance between the projections on the corresponding pixel plane of the i-th camera.

[0058] The coordinates of the i-th camera on the pixel plane are: The depth cache corresponding to the pixels, This represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera. Represents a set of voxel indices Multiple voxel centers correspond The minimum value in.

[0059] In step S2005, for each voxel The set of perspective observations that it can observe is defined as follows:

[0060] For the i-th camera, that is, if and only if the voxel center When the depth at the projection position of the i-th camera is equal to the depth buffer of the corresponding pixel at that projection position, it is considered that the voxel to which the center of the corresponding voxel belongs can be observed by camera i.

[0061] In a preferred embodiment, in step S300, for each voxel, the observation confidence is calculated based on its corresponding view observation set to measure the reliability of the 3D reconstruction of the region near the voxel. The higher the observation confidence, the better the geometric reconstruction quality of the 3DGS model in the corresponding region.

[0062] In one specific embodiment, the voxel center is modeled as a regular icosahedron with a radius r = 0.1 meters, and each face is used as the basic unit for confidence accumulation. It is assumed that... Let represent the set of observations corresponding to the k-th voxel. Then, the observation confidence level for each face of the icosahedron is determined by the following formula:

[0063] in, This represents the decay function based on Euclidean distance. This represents the camera translation corresponding to the p-th camera in the view observation set corresponding to the k-th voxel. This represents the center coordinates of the k-th voxel. express and The Euclidean norm between them Let represent the set of viewpoints corresponding to the m-th face in the icosahedral structure corresponding to the k-th voxel. This represents the observation confidence level of the m-th face in the icosahedral structure corresponding to the k-th voxel.

[0064] In one specific embodiment, (For example, a=-0.005, b=20), finally, the overall confidence score corresponding to the voxel is obtained by summing the observation confidence scores of all faces of the modeled icosahedron:

[0065] The overall observation confidence level corresponds to the k-th voxel, which is the sum of the observation confidence levels of the twenty faces in the icosahedral structure corresponding to the k-th voxel.

[0066] In this application, if the calculated If it is less than 0, then let .

[0067] In a preferred embodiment, step S400 includes: For each voxel, its corresponding voxel center is regarded as a control plotting point. The observation confidence of each point within the voxel is obtained by linear interpolation of the overall observation confidence of the voxel center. For each point within the voxel, the observation confidence of the point is used to visualize the point. The quality of the point is determined based on the display results. The observation confidence of each point indicates the gray value of the point in the image display.

[0068] In this application, the observation confidence level of each point in the point cloud corresponds to the gray value of the pixel, that is, the observation confidence level of each point is mapped to the gray value space to visualize the point cloud. In this way, it is possible to show the user which points have passed the quality assessment and which points have not passed the quality assessment. Specifically, points with darker colors (with a gray value threshold set) are shown to have passed the quality assessment, while points with lighter colors are shown to have not passed the quality assessment.

[0069] Specifically, for a point located within a voxel, its confidence level is obtained by linear interpolation of the overall confidence level corresponding to the voxel center, so that points closer to the voxel center have higher confidence consistency.

[0070] The technical solution provided in this application allows for the direct evaluation of the completeness and reliability of scene acquisition based on input data without requiring a time-consuming 3DGS optimization process, thereby enabling rapid quality diagnosis of large-scale scenes. This method can promptly identify potential defective areas during the data acquisition phase, preventing later modeling failures or rendering quality degradation, and improving the overall efficiency and controllability of the 3D data acquisition and 3D reconstruction process.

[0071] Based on the same application concept, this application also provides a three-dimensional Gaussian splash data acquisition quality assessment device corresponding to the three-dimensional Gaussian splash data acquisition quality assessment method provided in the above embodiments. Since the principle of the device in this application embodiment is similar to the three-dimensional Gaussian splash data acquisition quality assessment method in the above embodiments of this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0072] Please see Figure 3 , Figure 3 This diagram illustrates the functional modules of a three-dimensional Gaussian splash data acquisition quality assessment device provided in an embodiment of this application. Figure 3 As shown, the three-dimensional Gaussian splash data acquisition quality assessment device provided in this application includes: Module 500 is used to perform voxelization discretization on the point cloud data corresponding to the target scene in order to construct a voxel map based on a hash structure. The voxel map includes multiple voxels. The rasterization calculation module 510 performs rasterization calculations on multiple voxels one by one to determine the observability information of each voxel under different viewpoints and generate a viewpoint observation set corresponding to each voxel. Modeling module 520 models voxels using a regular icosahedral structure and calculates the observation confidence of voxels based on the distance parameters corresponding to each viewpoint and the viewpoint observation set. The quality assessment module 530 determines the quality assessment results of corresponding points within a voxel based on the calculation results.

[0073] Based on the same application concept, please refer to Figure 4 , Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Figure 4 As shown, the electronic device 60 includes a processor 601, a memory 602, and a bus 603. The memory 602 stores machine-readable instructions that can be executed by the processor 601. When the electronic device 60 is running, the processor 601 and the memory 602 communicate through the bus 603. The machine-readable instructions are executed by the processor 601 to perform the steps of the three-dimensional Gaussian splash data acquisition quality assessment method provided in any of the above embodiments.

[0074] Based on the same concept, this application also provides a computer-readable storage medium storing a computer program, which, when run by a processor, executes the steps of the three-dimensional Gaussian splash data acquisition quality assessment method provided in the above embodiments.

[0075] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0078] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method of evaluating the quality of acquisition of three-dimensional Gaussian sputtering data, characterized in that, The method comprises: perform voxelization discrete processing on point cloud data corresponding to a target scene to construct a voxel map based on a hash structure, the voxel map comprising a plurality of voxels; perform rasterization calculation on the plurality of voxels one by one to determine the observability information of each voxel under different viewing angles and generate a viewing angle observation set corresponding to each voxel; model the voxels using a regular icosahedron structure and perform observation confidence calculation on the voxels based on the distance parameters corresponding to each viewing angle and the viewing angle observation set; determine the quality evaluation result of the corresponding points in the voxel based on the calculation result.

2. The method of claim 1, wherein, The voxel map is constructed in the following manner: determine a plurality of voxels and the voxel center corresponding to each voxel in a discrete manner according to the three-dimensional coordinates corresponding to each point in the point cloud data; calculate the hash value corresponding to each voxel center according to the three-dimensional coordinates corresponding to each voxel center; construct a voxel map based on the hash structure based on the plurality of voxels, the voxel center corresponding to each voxel, the hash value and the point index belonging to the hash value.

3. The method of claim 1, wherein, The viewing angle observation set corresponding to each voxel is determined in the following manner: for each voxel: determine the spatial coordinates of the voxel in each camera coordinate system according to the voxel center and the camera extrinsic parameters; determine the projection of the voxel on the pixel plane corresponding to each camera according to the spatial coordinates of the voxel center in the coordinate system corresponding to each camera and the camera intrinsic parameters; determine the projection radius of the voxel on the pixel plane corresponding to each camera according to the given voxel size of the voxel; determine the voxel index set corresponding to each pixel under each camera plane and the corresponding depth buffer according to the projection radius and the projection of the voxel on the pixel plane corresponding to each camera; determine the viewing angle observation set corresponding to each voxel according to the voxel index set corresponding to each pixel under each camera and the depth buffer, the viewing angle observation set recording the camera viewing angle that can observe the voxel.

4. The method of claim 1, wherein, The camera extrinsic parameters comprise a rotation matrix and a translation vector, wherein the target spatial coordinates of the voxel center in each camera coordinate system are determined by the following formula: wherein, represents the three-dimensional coordinate of the voxel center corresponding to the kth voxel, , represents the rotation matrix corresponding to the ith camera, represents the translation vector corresponding to the ith camera, represents the spatial coordinate of the voxel center corresponding to the kth voxel in the camera coordinate system corresponding to the ith camera; denotes the horizontal coordinate of the kth voxel center in the camera coordinate system corresponding to the i th camera, denotes the vertical coordinate of the kth voxel center in the camera coordinate system corresponding to the i th camera, denotes the depth of the kth voxel center in the camera coordinate system corresponding to the i th camera, denotes the transpose.

5. The method of claim 3, wherein, the projection of the voxel on the pixel plane corresponding to each camera is determined by the following formula: wherein, represents the kth voxel center represents the projection of the kth voxel center on the i th camera corresponding pixel plane, represents the kth voxel center in the width direction coordinate of the i th camera corresponding pixel plane, represents the kth voxel center in the height direction coordinate of the i th camera corresponding pixel plane; represents the pixel focal length of the i-th camera in the x-axis, represents the pixel focal length of the i-th camera in the y-axis, represents the horizontal coordinate of the k-th voxel in the camera coordinate system corresponding to the i-th camera, represents the vertical coordinate of the k-th voxel in the camera coordinate system corresponding to the i-th camera, represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera, represents the coordinate of the main point C corresponding to the i-th camera in the x-axis of the camera, represents the coordinate of the main point C corresponding to the i-th camera in the y-axis of the camera.

6. The method of claim 3, wherein, the projection radius of the voxel on the pixel plane corresponding to each camera is determined by the following formula: In this formula, represents the projection radius of the kth voxel on the pixel plane of the ith camera, represents the given voxel size corresponding to the kth voxel, represents the pixel focal length of the x-axis of the ith camera, represents the depth of the kth voxel in the camera coordinate system corresponding to the ith camera.

7. The method of claim 3, wherein, the voxel index set corresponding to each pixel and the depth buffer are determined by the following formula: In this formula, represents the coordinates on the pixel plane of the i-th camera corresponding to the voxel index set represents the projection of the k-th voxel on the pixel plane corresponding to the i-th camera, represents the projection radius of the k-th voxel on the pixel plane corresponding to the i-th camera;​ represents the coordinates on the pixel plane of the i-th camera the depth buffer corresponding to the pixel pair represents the depth of the k-th voxel in the camera coordinate system corresponding to the i-th camera represents the set of voxel indices corresponding to the minimum value in the set of corresponding voxels.

8. The method of claim 3, wherein, the observation confidence corresponding to each voxel is determined by the following formula: wherein, represents an attenuation function based on the Euclidean distance, represents the camera translation corresponding to the pth camera in the perspective observation set corresponding to the kth voxel, represents the center coordinate corresponding to the kth voxel, represents the Euclidean norm between represents the perspective observation set corresponding to the mth face in the icosahedron structure corresponding to the kth voxel, represents the observation confidence of the mth face in the icosahedron structure corresponding to the kth voxel,​ represents the overall observation confidence corresponding to the kth voxel, which is the sum of the observation confidence corresponding to the twenty faces in the icosahedron structure corresponding to the kth voxel.

9. The method of claim 1, wherein, the calculation result describes the overall observation confidence of the voxel center corresponding to each voxel, wherein the quality evaluation result of the corresponding points in the voxel is determined in the following manner: for each voxel, the observation confidence of each point in the voxel is obtained by linearly interpolating the overall observation confidence corresponding to the voxel center of the voxel. The device comprises:

10. An apparatus for evaluating the quality of acquisition of three-dimensional Gaussian sputtering data, characterized in that it comprises: ​ A construction module voxelizes and discretizes point cloud data corresponding to a target scene to construct a voxel map based on a hash structure, the voxel map including a plurality of voxels; A rasterization calculation module performs rasterization calculation on each of the plurality of voxels to determine observability information of each voxel under different viewing angles, and generates a viewing angle observation set corresponding to each voxel; A modeling module models the voxels in a regular icosahedron structure, and calculates observation confidence of the voxels based on distance parameters corresponding to each viewing angle and the viewing angle observation set; A quality evaluation module determines a quality evaluation result of corresponding points in the voxels based on the calculation result.

Images