Three-dimensional digital asset generation method and system based on gaussian light field and blockchain

Abstract

The application relates to the technical field of three-dimensional modeling, and relates to a three-dimensional digital asset generation method and system based on a Gaussian light field and a blockchain, which comprises the following steps: collecting a target object from multiple perspectives to obtain a multi-view image set and a point coordinate set; obtaining a camera pose parameter set according to the multi-view image set and the point coordinate set; reconstructing a Gaussian light field based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model; performing encapsulation processing on the target three-dimensional model to obtain a three-dimensional digital asset file; performing blockchain notarization on the three-dimensional digital asset file to obtain a notarization identifier; and obtaining a three-dimensional digital asset based on the notarization identifier and the three-dimensional digital asset file. The application can solve the problems of low model precision, low generation efficiency and low asset right confirmation credibility in the current three-dimensional digital asset generation process.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of three-dimensional modeling, in particular to a three-dimensional digital asset generation method and system based on a Gaussian light field and a blockchain. BACKGROUND

[0002] With the rapid development of virtual reality, digital twin and metaverse industries, the application demand of three-dimensional digital assets in cultural entertainment, industrial design, cultural relic protection, education and training and other fields is increasing, and three-dimensional modeling technology, as a core link of three-dimensional digital asset production, has great application value.

[0003] Traditional three-dimensional digital asset generation methods usually rely on laser scanning, photogrammetry or multi-view stereo vision technology to obtain three-dimensional models, although three-dimensional digital assets can be obtained, but there are deficiencies in model quality, modeling efficiency and other aspects, which cannot meet the demand of three-dimensional digital assets for high-precision models. In addition, the traditional method lacks the ability of credible traceability and tamper-proofing of the generated three-dimensional digital assets, and it is difficult to effectively prove the source and integrity in the process of digital asset circulation, therefore, there are problems of low model precision, low generation efficiency and low asset right confirmation credibility in the current three-dimensional digital asset generation process. SUMMARY

[0004] The application provides a three-dimensional digital asset generation method and system based on a Gaussian light field and a blockchain, which mainly aims to solve the problems of low model precision, low generation efficiency and low asset right confirmation credibility in the current three-dimensional digital asset generation process.

[0005] To achieve the above-mentioned purpose, the application provides a three-dimensional digital asset generation method based on a Gaussian light field and a blockchain, which comprises:

[0006] obtaining a target object and an image acquisition unit;

[0007] using the image acquisition unit to collect the target object from multiple perspectives to obtain a multi-view image set and a point coordinate set;

[0008] obtaining a camera pose parameter set according to the multi-view image set and the point coordinate set;

[0009] performing Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model;

[0010] performing encapsulation processing on the target three-dimensional model to obtain a three-dimensional digital asset file;

[0011] performing blockchain storage on the three-dimensional digital asset file to obtain a storage identification;

[0012] Obtain the three-dimensional digital asset based on the storage evidence identifier and the three-dimensional digital asset file.

[0013] Optionally, the multi-view image set and the point coordinate set are obtained by using the image acquisition unit to collect the target object from multiple perspectives.

[0014] Determine the three-dimensional bounding box data of the target object.

[0015] Obtain a spatial collection path point set according to the three-dimensional bounding box data and a preset collection accuracy, wherein the spatial collection path point set includes a plurality of spatial collection path points.

[0016] For each spatial collection path point in the spatial collection path point set, the following operations are performed:

[0017] Collect an original image frame based on the spatial collection path point using the image acquisition unit.

[0018] Preprocess the original image frame to obtain a standard image frame.

[0019] Quality check the standard image frame to obtain a qualified image frame.

[0020] Determine the three-dimensional point coordinates of the spatial collection path point.

[0021] Summarize the qualified image frame and the three-dimensional point coordinates, respectively, to obtain a multi-view image set and a point coordinate set, wherein the qualified image frame in the multi-view image set and the three-dimensional point coordinates in the point coordinate set correspond one-to-one.

[0022] Optionally, the quality check of the standard image frame to obtain a qualified image frame includes:

[0023] Detect the features of the standard image frame to obtain a plurality of image feature points.

[0024] Determine a plurality of feature coordinates based on the plurality of image feature points, wherein an image feature point and a feature coordinate correspond one-to-one.

[0025] Divide the standard image frame into a plurality of regular grids.

[0026] For each regular grid in the plurality of regular grids, the following operations are performed:

[0027] Determine the number of image feature points in the regular grid based on the plurality of feature coordinates.

[0028] Summarize the number of image feature points to obtain a set of image feature point numbers.

[0029] calculate a distribution dispersion degree based on the set of image feature point quantities;

[0030] calculate a definition evaluation value of the standard image frame;

[0031] determine a total number of image feature points of the plurality of image feature points;

[0032] if the total number of image feature points is greater than or equal to a preset image feature point quantity threshold, the distribution dispersion degree is less than or equal to a preset dispersion degree threshold, and the definition evaluation value is greater than a preset definition threshold, the standard image frame is confirmed as a qualified image frame;

[0033] otherwise, determine an unqualified type, and adjust parameters of the image acquisition unit according to the unqualified type to obtain an adjusted image acquisition unit, take the adjusted image acquisition unit as the image acquisition unit, and return to the step of acquiring an image based on the spatial acquisition path point and using the image acquisition unit to acquire an image until a qualified image frame is obtained.

[0034] Optionally, the obtaining of the camera pose parameter set based on the set of multi-view images and the set of point coordinates comprises:

[0035] the following operations are performed on each three-dimensional point coordinate in the set of point coordinates:

[0036] based on the qualified image frame in the set of multi-view images corresponding to the three-dimensional point coordinate, determine a plurality of qualified image feature points and a plurality of qualified feature coordinates, wherein the qualified image feature points and the qualified feature coordinates are in one-to-one correspondence;

[0037] determine a plurality of real feature coordinates based on the plurality of qualified image feature points;

[0038] perform pose solving using a preset pose algorithm based on the three-dimensional point coordinate, the plurality of real feature coordinates, and the plurality of qualified feature coordinates to obtain a camera pose parameter, wherein the camera pose parameter comprises a rotation matrix and a translation vector;

[0039] aggregate the camera pose parameters to obtain a set of camera pose parameters.

[0040] Optionally, the Gaussian light field reconstruction based on the set of multi-view images and the set of camera pose parameters to obtain a target three-dimensional model comprises:

[0041] obtain a three-dimensional sparse point cloud based on the set of multi-view images and the set of camera pose parameters, wherein the three-dimensional sparse point cloud comprises a plurality of three-dimensional points;

[0042] the following operations are performed on each three-dimensional point in the three-dimensional sparse point cloud:

[0043] performing Gaussian attribute initialization on the three-dimensional points to obtain Gaussian attribute parameters;

[0044]

[0045] constructing a Gaussian point cloud model based on the Gaussian attribute parameter set;

[0046] sampling the camera pose parameter set to obtain a verification camera pose parameter set;

[0047] performing differentiable rendering on the Gaussian point cloud model according to the verification camera pose parameter set to obtain a synthesized image set, wherein the synthesized image set includes a plurality of synthesized images, and each synthesized image in the synthesized image set corresponds to a verification camera pose parameter in the verification camera pose parameter set;

[0048] calculating a photometric reconstruction loss based on the synthesized image set;

[0049] when the photometric reconstruction loss is greater than a preset convergence threshold, performing parameter updating on the Gaussian attribute parameter set to obtain an updated Gaussian parameter set, taking the updated Gaussian parameter set as the Gaussian attribute parameter set, and returning to the step of constructing the Gaussian point cloud model based on the Gaussian attribute parameter set until the photometric reconstruction loss is less than or equal to the convergence threshold;

[0050] when the photometric reconstruction loss is less than or equal to the convergence threshold, confirming the Gaussian point cloud model as a target three-dimensional model.

[0051] Optionally, the obtaining of the three-dimensional sparse point cloud based on the multi-view image set and the camera pose parameter set includes:

[0052] performing the following operations on each qualified image frame in the multi-view image set:

[0053] extracting feature descriptor sets from the plurality of qualified image feature points of the qualified image frame;

[0054]

[0055] performing cross-image feature matching according to the plurality of feature descriptor sets to obtain a matching feature point pair set;

[0056] performing the following operations on each matching feature point pair in the matching feature point pair set:

[0057] determining a first qualified image frame and a second qualified image frame from the multi-view image set based on the matching feature point pair;

[0058] ​​determine a first feature coordinate and a second feature coordinate based on the matched feature point pair, the first qualified image frame, and the second qualified image frame;

[0059] determine a first camera pose parameter and a second camera pose parameter from the camera pose parameter set according to the first qualified image frame and the second qualified image frame;

[0060] obtain a three-dimensional point by using a preset triangulation algorithm according to the first feature coordinate, the second feature coordinate, the first camera pose parameter, and the second camera pose parameter;

[0061] obtain a three-dimensional sparse point cloud by aggregating the three-dimensional points.

[0062] Optionally, the calculating the photometric reconstruction loss based on the synthesized image set comprises:

[0063] perform the following operations on each synthesized image in the synthesized image set:

[0064] determine a real image from the multi-view image set based on the synthesized image;

[0065] obtain a synthesized image pixel set and a real image pixel set based on the synthesized image and the real image, wherein the synthesized image pixel set comprises a plurality of synthesized image pixels, the real image pixel set comprises a plurality of real image pixels, and the synthesized image pixels and the real image pixels correspond to each other in a one-to-one manner;

[0066] perform the following operations on each synthesized image pixel in the synthesized image pixel set:

[0067] obtain an RGB three-channel pixel value of the synthesized image pixel, wherein the RGB three-channel pixel value comprises a red channel pixel value, a green channel pixel value, and a blue channel pixel value;

[0068] determine a real RGB three-channel pixel value of the corresponding real image pixel based on the synthesized image pixel, wherein the real RGB three-channel pixel value comprises a real red channel pixel value, a real green channel pixel value, and a real blue channel pixel value;

[0069] calculate a photometric difference value based on the RGB three-channel pixel value of the synthesized image pixel and the real RGB three-channel pixel value of the real image pixel, and the calculation formula is as follows:

[0070] ,

[0071] wherein, represents the photometric difference value, represents the red channel pixel value of the synthesized image pixel, represents the green channel pixel value of the synthesized image pixel, a blue channel pixel value representing a pixel of the synthetic image, a real red channel pixel value representing a pixel of the real image, a real green channel pixel value representing a pixel of the real image, a real blue channel pixel value representing a pixel of the real image;

[0072] aggregating the luminosity difference values to obtain a plurality of luminosity difference values;

[0073] calculating an image luminosity loss value according to the plurality of luminosity difference values;

[0074] performing structural similarity analysis on the synthetic image and the real image to obtain a structural similarity index;

[0075] calculating a luminosity reconstruction loss component according to the image luminosity loss value and the structural similarity index;

[0076] aggregating the luminosity reconstruction loss components to obtain a plurality of luminosity reconstruction loss components;

[0077] calculating a luminosity reconstruction loss based on the plurality of luminosity reconstruction loss components, wherein the luminosity reconstruction loss is an arithmetic mean of the plurality of luminosity reconstruction loss components.

[0078] Optionally, the blockchain notarization of the three-dimensional digital asset file to obtain a notarization identifier comprises:

[0079] performing file reading on the three-dimensional digital asset file to obtain a file byte sequence;

[0080] performing hash operation on the file byte sequence by using a preset one-way hash algorithm to obtain a three-dimensional digital asset fingerprint;

[0081] constructing blockchain notarization data based on the three-dimensional digital asset fingerprint and preset asset metadata;

[0082] performing on-chain processing on the blockchain notarization data to obtain blockchain feedback data;

[0083] performing analysis on the blockchain feedback data to obtain the notarization identifier.

[0084] Optionally, the on-chain processing of the blockchain notarization data to obtain the blockchain feedback data comprises:

[0085] obtaining a blockchain notarization node, wherein the blockchain notarization node comprises a master node and a plurality of distributed nodes;

[0086] performing digital signature processing on the blockchain notarization data by using the master node to obtain to-be-consensus notarization data;

[0087] extracting a distributed node from the plurality of distributed nodes in turn, and performing the following operation on each distributed node:

[0088] performing validity verification on the to-be-consensus stored evidence data by using the distributed node to obtain a node verification result, wherein the node verification result is pass or fail;

[0089] aggregating the node verification results to obtain a verification result set;

[0090] determining a number of verification passes according to the verification result set;

[0091] when the number of verification passes is less than a preset number threshold, performing stored evidence parameter debugging on the master node to obtain a debugged master node, taking the debugged master node as the master node, and returning to the step of performing digital signature processing on the blockchain stored evidence data by using the master node until the number of verification passes is greater than or equal to the number threshold;

[0092] when the number of verification passes is greater than or equal to the number threshold, performing block packaging on the to-be-consensus stored evidence data to obtain blockchain feedback data.

[0093] To achieve the above object, the application further provides a three-dimensional digital asset generation system based on a Gaussian light field and a blockchain, comprising:

[0094] an equipment acquisition module configured to acquire a target object and an image acquisition unit;

[0095] an image acquisition and pose solution module configured to acquire the target object by using the image acquisition unit to obtain a multi-view image set and a point coordinate set, and acquire a camera pose parameter set according to the multi-view image set and the point coordinate set;

[0096] a Gaussian light field reconstruction module configured to reconstruct a Gaussian light field based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model;

[0097] a blockchain stored evidence module configured to encapsulate the target three-dimensional model to obtain a three-dimensional digital asset file, store the three-dimensional digital asset file in a blockchain to obtain a stored evidence identifier, and acquire a three-dimensional digital asset based on the stored evidence identifier and the three-dimensional digital asset file.

[0098] To solve the above problems, the application further provides an electronic device, comprising:

[0099] a memory configured to store at least one instruction; and a processor configured to execute the instruction stored in the memory to implement the three-dimensional digital asset generation method based on a Gaussian light field and a blockchain.

[0100] To solve the above problems, the application further provides a computer readable storage medium, wherein at least one instruction is stored in the computer readable storage medium, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned three-dimensional digital asset generation method based on a Gaussian light field and a blockchain.

[0101] To solve the problems in the background art, the application obtains a target object and an image acquisition unit, and the physical entity of three-dimensional reconstruction and the hardware basis of image acquisition are determined through this step. The image acquisition unit is used to collect the target object from multiple perspectives to obtain a multi-view image set and a point coordinate set. According to the multi-view image set and the point coordinate set, a camera pose parameter set is obtained. The position and orientation of the image acquisition unit in the three-dimensional space when each qualified image frame is taken are determined through feature extraction and pose solving, which provides accurate geometric constraints for subsequent three-dimensional reconstruction. Gaussian light field reconstruction is performed based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model. The sparse point cloud is extracted from the multi-view image, and the Gaussian point cloud model is optimized based on the differentiable rendering, so that the target three-dimensional model is closer to the target object, and the accuracy and authenticity of the model are improved. The target three-dimensional model is encapsulated to obtain a three-dimensional digital asset file, and the three-dimensional digital asset file is stored in a blockchain to obtain a storage identification. Based on the storage identification and the three-dimensional digital asset file, a three-dimensional digital asset is obtained. Therefore, the application can solve the problems of low model accuracy, low generation efficiency and low asset right confirmation credibility in the current three-dimensional digital asset generation process. BRIEF DESCRIPTION OF DRAWINGS

[0102] Figure 1 A flowchart of a three-dimensional digital asset generation method based on a Gaussian light field and a blockchain is provided for an embodiment of the application.

[0103] Figure 2 A functional module diagram of a three-dimensional digital asset generation system based on a Gaussian light field and a blockchain is provided for an embodiment of the application.

[0104] Figure 3 A structural diagram of an electronic device for implementing the three-dimensional digital asset generation method based on a Gaussian light field and a blockchain is provided for an embodiment of the application.

[0105] REFERENCE SIGNS:

[0106] 1, electronic device; 10, processor; 11, memory; 12, bus.

[0107] The implementation, functional characteristics and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0108] It should be understood that the specific embodiments described herein are merely exemplary and are not intended to limit the application.

[0109] The embodiment of the present application provides a three-dimensional digital asset generation method based on a Gaussian light field and a blockchain. The execution subject of the three-dimensional digital asset generation method based on the Gaussian light field and the blockchain includes but is not limited to at least one of electronic devices such as a server, a terminal and the like which can be configured to execute the method provided by the embodiment of the present application. In other words, the three-dimensional digital asset generation method based on the Gaussian light field and the blockchain can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster and the like.

[0110] Referring to Figure 1 Fig. 1 shows a flowchart of a three-dimensional digital asset generation method based on a Gaussian light field and a blockchain provided by an embodiment of the present application. In the embodiment, the three-dimensional digital asset generation method based on the Gaussian light field and the blockchain includes the following steps.

[0111] S1, obtaining a target object and an image acquisition unit.

[0112] It should be explained that the target object refers to a physical entity or a scene waiting to generate a three-dimensional digital asset, for example: cultural relics, buildings, natural scenes and the like. The image acquisition unit refers to a device for capturing images of the target object from different spatial perspectives, for example: a monocular camera, a multi-view camera, a laser scanning camera and the like.

[0113] S2, performing multi-perspective acquisition on the target object by using the image acquisition unit to obtain a multi-perspective image set and a point coordinate set.

[0114] In detail, the multi-perspective acquisition on the target object by using the image acquisition unit to obtain the multi-perspective image set and the point coordinate set includes the following steps.

[0115] determining three-dimensional bounding box data of the target object;

[0116] acquiring a spatial acquisition path point set according to the three-dimensional bounding box data and a preset acquisition accuracy, wherein the spatial acquisition path point set includes a plurality of spatial acquisition path points;

[0117] performing the following operations on each spatial acquisition path point in the spatial acquisition path point set:

[0118] performing image acquisition by using the image acquisition unit based on the spatial acquisition path point to obtain an original image frame;

[0119] performing a preprocessing operation on the original image frame to obtain a standard image frame;

[0120] performing a quality check on the standard image frame to obtain a qualified image frame;

[0121] determining three-dimensional point coordinates of the spatial collection path points;

[0122] respectively collecting the qualified image frames and the three-dimensional point coordinates to obtain a multi-view image set and a point coordinate set, wherein the qualified image frames in the multi-view image set correspond to the three-dimensional point coordinates in the point coordinate set one by one.

[0123] It can be understood that the three-dimensional bounding box data refers to the spatial range of the smallest cuboid region capable of completely containing the target object in the world coordinate system, used to define the spatial distribution range of the target object. For example, a target object is a statue, the X-axis range of which in the world coordinate system is 200mm to 800mm, the Y-axis range is 300mm to 900mm, and the Z-axis range is 0mm to 1200mm. The three-dimensional bounding box data of the statue can be represented as the minimum point coordinates (200, 300, 0) and the maximum point coordinates (800, 900, 1200). The world coordinate system refers to the three-dimensional rectangular coordinate system uniformly used in the entire three-dimensional digital asset generation process. Optionally, the geometric center point of the target object is taken as the origin of the world coordinate system, the positive north direction is taken as the positive direction of the Y-axis, the positive east direction is taken as the positive direction of the X-axis, and the direction perpendicular to the ground upward is taken as the positive direction of the Z-axis. The collection accuracy refers to a parameter artificially set in advance to determine the distance between the spatial collection path points, for example, 100mm. The spatial collection path point set refers to a set of three-dimensional coordinate points generated based on the three-dimensional bounding box data and the collection accuracy through a path planning algorithm (such as a spiral path planning method), used to guide the shooting position of the image collection unit. The spatial collection path point refers to a single path point in the spatial collection path point set.

[0124] It should be noted that based on the spatial collection path points, the image collection unit is used to perform image collection to obtain the original image frame, which refers to controlling the image collection unit to move to the position of the spatial collection path point, adjusting the orientation of the image collection unit to align with the target object, and then shooting the target object to obtain the image of the target object, which is the original image frame. The preprocessing operation refers to a processing operation performed to eliminate interference noise in the original image frame and improve the image quality standard, specifically including denoising, image distortion correction, size normalization, etc. The standard image frame refers to the original image frame after the preprocessing operation. The three-dimensional point coordinates refer to the coordinates of a certain spatial collection path point in the world coordinate system. The multi-view image set refers to a set composed of all qualified image frames. The point coordinate set refers to a set composed of all three-dimensional point coordinates.

[0125] Specifically, the quality checking on the standard image frame to obtain a qualified image frame comprises:

[0126] detecting features of the standard image frame to obtain a plurality of image feature points;

[0127] determining a plurality of feature coordinates based on the plurality of image feature points, wherein an image feature point corresponds to a feature coordinate one by one;

[0128] dividing the standard image frame into a plurality of regular grids;

[0129] performing the following operations on each of the plurality of regular grids:

[0130] determining the number of image feature points in the regular grid based on the plurality of feature coordinates;

[0131] summarizing the number of image feature points to obtain a set of image feature point numbers;

[0132] calculating a distribution dispersion based on the set of image feature point numbers;

[0133] calculating a definition evaluation value of the standard image frame;

[0134] determining the total number of image feature points of the plurality of image feature points;

[0135] if the total number of image feature points is greater than or equal to a preset image feature point number threshold, the distribution dispersion is less than or equal to a preset dispersion threshold, and the definition evaluation value is greater than a preset definition threshold, the standard image frame is confirmed as a qualified image frame;

[0136] otherwise, determining an unqualified type, and adjusting parameters of the image acquisition unit according to the unqualified type to obtain an adjusted image acquisition unit, taking the adjusted image acquisition unit as the image acquisition unit, and returning to the step of acquiring an image based on the spatial acquisition path point and using the image acquisition unit until a qualified image frame is obtained.

[0137] It should be understood that the feature detection refers to a process of identifying pixels with obvious local structure or gray level change (such as corner, edge intersection, spot, etc.) in a standard image frame by a feature recognition algorithm (such as Harris corner detection method). The image feature point refers to a pixel identified after feature detection. The feature coordinate refers to a specific coordinate of the image feature point in a two-dimensional pixel coordinate system, which is specifically represented as (u, v), wherein u represents a horizontal pixel position, and v represents a vertical pixel position. The two-dimensional pixel coordinate system refers to a plane rectangular coordinate system with the upper left corner of the standard image frame as the origin, the u-axis positive direction to the right, and the v-axis positive direction downward. The grid division of the standard image frame to obtain a plurality of regular grids refers to uniformly dividing the standard image frame along the horizontal and vertical directions to divide a plurality of unit cells, each unit cell being a regular grid, for example, dividing the standard image frame into M rows and N columns of unit cells with equal areas. The image feature point quantity refers to the number of image feature points falling into a regular grid. The image feature point quantity set refers to a set composed of image feature point quantities of all regular grids. The distribution dispersion degree refers to a quantitative index for indicating the uniformity of the distribution of image feature points on the standard image frame, and the smaller the value is, the more uniform the distribution is. Optionally, the coefficient of variation of the image feature point quantity set is used as the distribution dispersion degree, and the calculation formula of the coefficient of variation is: wherein, the coefficient of variation of the image feature point quantity set, the standard deviation of the image feature point quantity set, the average value of the image feature point quantity set, the calculation methods of the standard deviation and the average value are both prior art, and will not be described here. The definition value refers to a quantitative index for indicating the definition degree of the standard image frame, and the larger the value is, the more clear the standard image frame is. Optionally, the definition value is calculated by using the Laplacian variance method (the Laplacian variance calculated by the Laplacian variance method is used as the definition value). The total number of image feature points refers to the total number of all image feature points.

[0138] It needs to be explained that the image feature point quantity threshold refers to a critical value of the number of image feature points set artificially in advance, which is used to determine whether the image features of the standard image frame contain sufficient number of image feature points. Optionally, the image feature point quantity threshold can be set according to the target object, for example: for objects with rich surface texture and complex structure (such as sculptures, precision mechanical parts, etc.), the threshold can be set in the range of 500-1000, and for objects with smooth surface, single texture or simple structure (such as smooth pottery, smooth metal ball, etc.), the threshold can be set in the range of 100-300. If the target object is a complex sculpture, the image feature point can be set to 1000. The dispersion threshold refers to a critical value of the distribution dispersion set artificially in advance. When the distribution dispersion is greater than the dispersion threshold, it indicates that the image feature point distribution of the standard image frame is uneven. The value of the dispersion threshold can be set according to the structural complexity of the target object and the accuracy requirement of subsequent three-dimensional modeling, for example, for objects with complex surface structure and high-precision reconstruction requirements (such as precision parts, sculptures, etc.), the dispersion threshold can be set smaller (such as 0.5) to ensure that the image feature points are uniformly distributed in the image; for objects with simple surface structure and low reconstruction accuracy requirements (such as smooth pottery, smooth metal ball, etc.), the dispersion threshold can be set larger (such as 1.2) to avoid insufficient number of qualified image frames due to excessive requirements. The clarity threshold refers to a critical value of the clarity evaluation value set artificially in advance. When the clarity evaluation value is less than or equal to the clarity threshold, it indicates that the standard image frame is blurred. The value of the clarity threshold can be set according to the performance of the image acquisition unit and the environmental lighting conditions, for example: if the image acquisition unit used is a high-resolution camera and is located in a well-lit indoor scene, a higher clarity threshold (such as 100) can be set to select clear images with rich details. The qualified image frame refers to a standard image frame whose total number of image feature points is greater than or equal to the image feature point quantity threshold, the distribution dispersion is less than or equal to the dispersion threshold, and the clarity evaluation value is greater than the clarity threshold. The unqualified type refers to the classification type of the standard image frame that is not confirmed as a qualified image frame, including: insufficient feature point type (total number of image feature points is less than image feature point quantity threshold), uneven feature point distribution type (distribution dispersion is greater than dispersion threshold), insufficient clarity type (clarity evaluation value is less than or equal to clarity threshold). The parameter adjustment refers to modifying the parameter configuration of the image acquisition unit according to the unqualified type to improve the quality of the subsequent acquired images. The specific method includes: for the insufficient feature point type, adjusting the focal length and exposure time of the image acquisition unit; for the uneven feature point distribution type, adjusting the pitch angle of the image acquisition unit; for the insufficient clarity type, adjusting the aperture size and sensitivity of the image acquisition unit. If the standard image frame meets multiple unqualified types at the same time, the corresponding parameters will be adjusted. The adjusted image acquisition unit refers to the image acquisition unit after parameter adjustment.

[0139] S3, obtaining a camera pose parameter set according to the multi-view image set and the point position coordinate set.

[0140] Further, the obtaining of the camera pose parameter set according to the multi-view image set and the point position coordinate set comprises:

[0141] The following operations are performed on each three-dimensional point position coordinate in the point position coordinate set:

[0142] Based on the qualified image frames in the multi-view image set corresponding to the three-dimensional point position coordinate, a plurality of qualified image feature points and a plurality of qualified feature coordinates are determined, wherein the qualified image feature points and the qualified feature coordinates are one-to-one corresponding;

[0143] Based on the plurality of qualified image feature points, a plurality of real feature coordinates are determined;

[0144] According to the three-dimensional point position coordinate, the plurality of real feature coordinates and the plurality of qualified feature coordinates, a pose algorithm is used for pose solving to obtain a camera pose parameter, wherein the camera pose parameter comprises a rotation matrix and a translation vector;

[0145] The camera pose parameters are summarized to obtain a camera pose parameter set.

[0146] It needs to be explained that the qualified image feature point refers to the image feature point of the qualified image frame. The qualified feature coordinates refer to the feature coordinates of the qualified image feature point. The real feature coordinates refer to the actual position coordinates of the qualified image feature point in the three-dimensional physical space (world coordinate system), which can be represented as (x, y, z), and optionally, the real feature coordinates can be obtained by a cross-image feature matching method and a triangulation method. The pose algorithm refers to a mathematical method for solving the accurate position and orientation of the image acquisition unit according to the known real feature coordinates and their qualified feature coordinates in the image, and optionally, the PnP algorithm is used as the pose algorithm. The pose solution refers to a process of taking the three-dimensional point coordinates as the initial value of the image acquisition unit position, using the pose algorithm to obtain the camera pose parameters based on multiple real feature coordinates and multiple qualified feature coordinates. The camera pose parameters refer to a set of mathematical parameters describing the rigid transformation relationship between the camera coordinate system and the world coordinate system, including a rotation matrix and a translation vector. The camera coordinate system refers to a three-dimensional rectangular coordinate system with the optical center of the image acquisition unit as the origin, the Z-axis along the camera optical axis pointing to the shooting direction, the X-axis along the horizontal direction parallel to the shooting image (qualified image frame) plane, and the Y-axis along the vertical direction parallel to the shooting image (qualified image frame) plane, used to describe the orientation and position of the image acquisition unit itself. The rotation matrix refers to a 3x3 orthogonal matrix used to represent the directional relationship of the camera coordinate system relative to the world coordinate system, which can describe the spatial orientation of the image acquisition unit. The translation vector refers to a vector representing the spatial offset of the origin of the camera coordinate system relative to the origin of the world coordinate system, which can describe the accurate position of the image acquisition unit. The camera pose parameter set refers to a set composed of camera pose parameters corresponding to all three-dimensional point coordinates.

[0147] S4, Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set is performed to obtain a target three-dimensional model.

[0148] In detail, the Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set to obtain the target three-dimensional model includes:

[0149] Based on the multi-view image set and the camera pose parameter set, a three-dimensional sparse point cloud is obtained, wherein the three-dimensional sparse point cloud includes a plurality of three-dimensional points;

[0150] For each three-dimensional point in the three-dimensional sparse point cloud, the following operations are performed:

[0151] The three-dimensional point is initialized with Gaussian attributes to obtain Gaussian attribute parameters;

[0152] The Gaussian attribute parameters are summarized to obtain a Gaussian attribute parameter set;

[0153] Based on the Gaussian attribute parameter set, a Gaussian point cloud model is constructed;

[0154] sampling the set of camera pose parameters to obtain a set of verification camera pose parameters;

[0155] performing differentiable rendering on the Gaussian point cloud model according to the set of verification camera pose parameters to obtain a set of synthesized images, wherein the set of synthesized images comprises a plurality of synthesized images, and each synthesized image in the set of synthesized images corresponds to a verification camera pose parameter in the set of verification camera pose parameters;

[0156] calculating a photometric reconstruction loss based on the set of synthesized images;

[0157] when the photometric reconstruction loss is greater than a preset convergence threshold, performing parameter updating on the set of Gaussian attribute parameters to obtain an updated set of Gaussian parameters, taking the updated set of Gaussian parameters as the set of Gaussian attribute parameters, and returning to the step of constructing the Gaussian point cloud model based on the set of Gaussian attribute parameters until the photometric reconstruction loss is less than or equal to the convergence threshold;

[0158] when the photometric reconstruction loss is less than or equal to the convergence threshold, confirming the Gaussian point cloud model as the target three-dimensional model.

[0159] It should be explained that the Gaussian attribute initialization refers to a process of assigning a set of Gaussian attribute parameters for defining a Gaussian ellipsoid to a three-dimensional point. The Gaussian ellipsoid refers to a representation form of a three-dimensional Gaussian distribution with the three-dimensional point as the center, which is represented as an ellipsoid-shaped volume region, and is used to approximate the influence range, directionality and diffusion characteristics of surface details of the three-dimensional point in the three-dimensional space. The Gaussian attribute parameters refer to parameters for describing the Gaussian ellipsoid, specifically including: a center position, a covariance matrix, an opacity and a spherical harmonic function coefficient. The center position refers to the position of the Gaussian ellipsoid in the world coordinate system, i.e. the coordinates of the three-dimensional point. The covariance matrix refers to a 3x3 symmetric matrix (such as ), and the diagonal elements (x, y, z) in the symmetric matrix respectively correspond to the stretching degree of the Gaussian ellipsoid in the X, Y and Z axis directions of the world coordinate system (the larger the value, the greater the stretching degree of the Gaussian ellipsoid in the corresponding axis direction); the non-diagonal elements (xy, yz, zx) correspond to the correlation between the X, Y and Z axis directions of the Gaussian ellipsoid. 、 、 、 、 ​The direction of the Gaussian ellipsoid corresponds to the regulation of the Gaussian ellipsoid; by adjusting the values of the elements of the covariance matrix, the shape and direction of the Gaussian ellipsoid can be controlled. The opacity refers to a quantitative parameter artificially set in advance to represent the blocking ability of the Gaussian ellipsoid to light, for example, when the target object is a metal material, it is set to 1, and when the target object is a translucent plastic material, it is set to 0.4. The spherical harmonic function coefficient refers to a coefficient for describing the color change of the Gaussian ellipsoid with the observation direction, and its calculation method is a prior art, which will not be described here. The Gaussian attribute parameter set refers to a set composed of Gaussian attribute parameters of all three-dimensional points. The Gaussian point cloud model is constructed based on the Gaussian attribute parameter set, which means that all Gaussian ellipsoids are combined into a whole model in the world coordinate system according to the Gaussian attribute parameter set, and the model is the Gaussian point cloud model.

[0160] It should be understood that sampling the camera pose parameter set to obtain the verification camera pose parameter set means randomly sampling the camera pose parameter set to obtain a plurality of camera pose parameters, and the set composed of the obtained camera pose parameters is the verification camera pose parameter set. The differentiable rendering refers to a process of simulating a plurality of images according to the verification camera pose parameter set and the Gaussian point cloud model, which is a prior art and will not be described in detail. The synthetic image refers to an image simulated by differentiable rendering. The synthetic image set refers to a set composed of all synthetic images. The convergence threshold refers to a critical value of the photometric reconstruction loss artificially set in advance. When the photometric reconstruction loss is less than or equal to the convergence threshold, it indicates that the accuracy of the Gaussian point cloud model meets the requirements and can be used as the target three-dimensional model. The parameter update refers to a process of adjusting the Gaussian attribute parameters according to the photometric reconstruction loss. The adjustment logic is to correct the parameters that cause the loss to increase (such as adjusting the center position for the center position deviation and correcting the spherical harmonic function coefficient for the color deviation), and the purpose is to reduce the photometric reconstruction loss and make the Gaussian point cloud model closer to the real image. The updated Gaussian parameter refers to the Gaussian attribute parameter after the parameter update. The target three-dimensional model refers to the Gaussian point cloud model whose photometric reconstruction loss is less than or equal to the convergence threshold, and the restoration accuracy of the target object meets the user's requirements.

[0161] Specifically, the three-dimensional sparse point cloud is obtained based on the multi-view image set and the camera pose parameter set, which includes:

[0162] The following operations are performed on each qualified image frame in the multi-view image set:

[0163] Feature descriptor extraction is performed on the plurality of qualified image feature points of the qualified image frame to obtain a feature descriptor set;

[0164] The feature descriptor set is summarized to obtain a plurality of feature descriptor sets;

[0165] According to the feature descriptor subset, cross-image feature matching is performed to obtain a matched feature point pair set;

[0166] For each matched feature point pair in the matched feature point pair set, the following operation is performed:

[0167] Based on the matched feature point pair, a first qualified image frame and a second qualified image frame are determined from a multi-view image set;

[0168] Based on the matched feature point pair, the first qualified image frame and the second qualified image frame, a first feature coordinate and a second feature coordinate are determined;

[0169] According to the first qualified image frame and the second qualified image frame, a first camera pose parameter and a second camera pose parameter are confirmed from a camera pose parameter set;

[0170] According to the first feature coordinate, the second feature coordinate, the first camera pose parameter and the second camera pose parameter, a three-dimensional point is obtained by using a preset triangulation algorithm;

[0171] The three-dimensional points are summarized to obtain a three-dimensional sparse point cloud.

[0172] It should be noted that the feature descriptor set is obtained by extracting feature descriptors from the plurality of qualified image feature points of the qualified image frame, that is, for each qualified image feature point in the plurality of qualified image feature points, a mathematical vector (i.e., a feature descriptor) describing the appearance, texture and structure information of the surrounding local image region of the qualified image feature point is obtained by using a feature descriptor extraction algorithm (such as SIFT algorithm), and the set composed of all feature descriptors is the feature descriptor set. The cross-image feature matching refers to a cross-image feature matching technology, which is used to find a set of qualified image feature points corresponding to the actual positions of feature points of the same target object between feature descriptor sets of different qualified image frames. The cross-image feature matching technology is a prior art and will not be described here. The matched feature point pair set refers to a set composed of all matched feature point pairs obtained after cross-image feature matching. The matched feature point pair is composed of a pair of qualified image feature points corresponding to the actual positions of feature points of the same target object in two qualified image frames.

[0173] It can be understood that the first qualified image frame refers to a qualified image frame in which a first qualified image feature point in a matching feature point pair is located. The second qualified image frame refers to a qualified image frame in which a second qualified image feature point in the matching feature point pair is located. The first feature coordinate refers to a feature coordinate of the first qualified image feature point in the matching feature point pair. The second feature coordinate refers to a feature coordinate of the second qualified image feature point in the matching feature point pair. The first camera pose parameter refers to a camera pose parameter corresponding to the image acquisition unit when the first qualified image frame is captured. The second camera pose parameter refers to a camera pose parameter corresponding to the image acquisition unit when the second qualified image frame is captured. The triangulation algorithm refers to a mathematical method for determining the accurate position of an actual physical position in a world coordinate system by solving the geometric projection relationship of two two-dimensional image point coordinates corresponding to the actual physical position. The triangulation algorithm is prior art and will not be described here. The three-dimensional point obtained by using the preset triangulation algorithm according to the first feature coordinate, the second feature coordinate, the rotation matrix and the translation vector in the first camera pose parameter, and the rotation matrix and the translation vector in the second camera pose parameter refers to: substituting the known first feature coordinate, the second feature coordinate, and the first camera pose parameter and the second camera pose parameter into the triangulation algorithm, calculating the three-dimensional coordinates of the intersection point in the world coordinate system by solving the intersection point of two rays with the image acquisition unit as the starting point in the world coordinate system, and the intersection point is the three-dimensional point. The three-dimensional sparse point cloud refers to a set composed of all three-dimensional points.

[0174] Further, the photometric reconstruction loss is calculated based on the synthesized image set, including:

[0175] The following operations are performed on each synthesized image in the synthesized image set:

[0176] Based on the synthesized image, a real image is identified from the multi-view image set;

[0177] Based on the synthesized image and the real image, a synthesized image pixel set and a real image pixel set are obtained, wherein the synthesized image pixel set includes a plurality of synthesized image pixels, the real image pixel set includes a plurality of real image pixels, and the synthesized image pixels and the real image pixels correspond one-to-one;

[0178] The following operations are performed on each synthesized image pixel in the synthesized image pixel set:

[0179] The RGB three-channel pixel value of the synthesized image pixel is obtained, wherein the RGB three-channel pixel value includes a red channel pixel value, a green channel pixel value, and a blue channel pixel value;

[0180] determine a real RGB three-channel pixel value of a corresponding real image pixel based on the synthetic image pixel, wherein the real RGB three-channel pixel value comprises a real red channel pixel value, a real green channel pixel value and a real blue channel pixel value;

[0181] calculate a luminosity difference value based on the RGB three-channel pixel value of the synthetic image pixel and the real RGB three-channel pixel value of the real image pixel, and the calculation formula is as follows:

[0182]

[0183] wherein, the luminosity difference value is represented by L, the red channel pixel value of the synthetic image pixel is represented by R, the green channel pixel value of the synthetic image pixel is represented by G, the blue channel pixel value of the synthetic image pixel is represented by B, the real red channel pixel value of the real image pixel is represented by Rr, the real green channel pixel value of the real image pixel is represented by Gr, the real blue channel pixel value of the real image pixel is represented by Br;

[0184] aggregate the luminosity difference values to obtain a plurality of luminosity difference values;

[0185] calculate an image luminosity loss value according to the plurality of luminosity difference values;

[0186] perform structural similarity analysis on the synthetic image and the real image to obtain a structural similarity index;

[0187] calculate a luminosity reconstruction loss component according to the image luminosity loss value and the structural similarity index;

[0188] aggregate the luminosity reconstruction loss components to obtain a plurality of luminosity reconstruction loss components;

[0189] calculate a luminosity reconstruction loss based on the plurality of luminosity reconstruction loss components, wherein the luminosity reconstruction loss is an arithmetic mean of the plurality of luminosity reconstruction loss components.

[0190] ​It can be understood that the real image refers to a synthetic image frame corresponding to the same camera pose parameter in the multi-view image set. The synthetic image pixel set refers to a set composed of all pixels in the synthetic image. The real image pixel set refers to a set composed of all pixels in the real image. The synthetic image pixel refers to a certain pixel in the synthetic image pixel set. The real image pixel refers to a certain pixel in the real image pixel set. The RGB three-channel pixel value refers to a set of parameters for describing the pixel color of the synthetic image pixel, including the red channel pixel value, the green channel pixel value and the blue channel pixel value. The red channel pixel value refers to the value of the pixel in the red color channel. The green channel pixel value refers to the value of the pixel in the green color channel. The blue channel pixel value refers to the value of the pixel in the blue color channel. The real RGB three-channel pixel value refers to the RGB three-channel pixel value of the real image pixel, the real red channel pixel value refers to the red channel pixel value of the real image pixel. The real green channel pixel value refers to the green channel pixel value of the real image pixel. The real blue channel pixel value refers to the blue channel pixel value of the real image pixel.

[0191] It should be understood that the luminosity difference value refers to a quantitative parameter for representing the color difference between the synthetic image pixel and the corresponding real image pixel. The smaller the luminosity difference value, the closer the color between the synthetic image pixel and the real image pixel. The image luminosity loss value refers to the color difference between the entire synthetic image and the real image. The smaller the image luminosity loss value, the closer the color between the synthetic image and the real image. The calculation method is: wherein, represents the image luminosity loss value, represents the sum of all luminosity difference values, represents the total number of pixels of the synthetic image. The structural similarity analysis refers to a process of analyzing and evaluating the similarity between the synthetic image and the real image in contrast, brightness and structure by using structural similarity evaluation methods (such as MS-SSIM, etc.). The structural similarity index refers to a quantitative index obtained by structural similarity analysis for representing the similarity between the synthetic image and the real image. The larger the value, the more similar the structure between the synthetic image and the real image. The luminosity reconstruction loss component refers to a quantitative index obtained by comprehensively considering the image luminosity loss value and the structural similarity index for representing the overall difference between the synthetic image and the real image. The calculation method is: wherein, represents the luminosity reconstruction loss component, represents the structural similarity index, represents the weight coefficient of the pre-set image luminosity loss value, A weight coefficient representing a pre-set structure similarity index. The smaller the photometric reconstruction loss component, the more similar the synthetic image is to the real image in color and structure.

[0192] S5, encapsulating the target three-dimensional model to obtain a three-dimensional digital asset file.

[0193] It needs to be explained that the encapsulation process refers to the process of organizing, packaging and storing the target three-dimensional model and its related metadata according to a certain file format and structure, which specifically includes: model data packaging, metadata addition and format standardization. The model data packaging refers to converting the data of the target three-dimensional model into a specified three-dimensional model storage format (such as glTF, OBJ, etc.), and the method of converting the data of the target three-dimensional model into a specified three-dimensional model storage format is a prior art, which will not be repeated here. The metadata addition refers to embedding the data information describing the target three-dimensional model into the file together with the target three-dimensional model after model data packaging, and the data information describing the target three-dimensional model includes: a set of camera pose parameters, a set of multi-view images, a set of matching feature point pairs, etc. The format standardization refers to the process of standardizing the file format of the packaged target three-dimensional model and metadata, unifying the coding and structure layout, so as to meet the syntax and semantic requirements of a specific file format (such as glTF, OBJ, etc.). The method of standardizing the file format of the packaged target three-dimensional model and metadata is a prior art, which will not be repeated here. The three-dimensional digital asset file refers to the digital file obtained after encapsulation, which contains the data of the target three-dimensional model and related metadata (divided into target three-dimensional model data and related metadata), which is the digital carrier of the target three-dimensional model.

[0194] S6, storing the three-dimensional digital asset file on the blockchain to obtain a storage identification.

[0195] In detail, the storing the three-dimensional digital asset file on the blockchain to obtain a storage identification includes:

[0196] reading the three-dimensional digital asset file to obtain a file byte sequence;

[0197] performing a hash operation on the file byte sequence using a preset one-way hash algorithm to obtain a three-dimensional digital asset fingerprint;

[0198] constructing a blockchain storage data based on the three-dimensional digital asset fingerprint and a preset asset metadata;

[0199] performing a chain processing on the blockchain storage data to obtain a blockchain feedback data;

[0200] analyzing the blockchain feedback data to obtain a storage identification.

[0201] It can be explained that the file reading refers to the process of opening the three-dimensional digital asset file by binary mode and sequentially reading all the contents thereof. The file byte sequence refers to the byte sequence presented in binary form after file reading. The one-way hash algorithm refers to an algorithm for mapping a file byte sequence to a fixed-length output value (such as the SHA-256 algorithm). The three-dimensional digital asset fingerprint obtained by hashing the file byte sequence using the preset one-way hash algorithm refers to performing mathematical operations on the file byte sequence using a one-way hash algorithm to generate a fixed-length hash value, which is the three-dimensional digital asset fingerprint. The asset metadata refers to descriptive information associated with the three-dimensional digital asset file, including: a unique identifier, a file generation timestamp, an owner identity of the three-dimensional digital asset file, etc. The blockchain storage data based on the three-dimensional digital asset fingerprint and the preset asset metadata refers to integrating the three-dimensional digital asset fingerprint and the asset metadata to form a complete data package, which is the blockchain storage data. The blockchain storage data can be used to establish an unalterable binding relationship between the three-dimensional digital asset and the asset metadata, thereby determining the ownership of the three-dimensional digital asset file. The blockchain feedback data refers to the result information returned by the blockchain about the on-chain operation after the completion of the on-chain operation of the blockchain storage data, including: transaction hash, transaction execution status (success or failure), block hash of the blockchain storage data, etc. The storage identification obtained by analyzing the blockchain feedback data refers to analyzing the blockchain return data to extract the transaction hash (i.e., obtaining an identification that can uniquely point to the on-chain record), which is the storage identification.

[0202] Specifically, the on-chain processing of the blockchain storage data to obtain the blockchain feedback data includes:

[0203] Obtaining a blockchain storage node, wherein the blockchain storage node includes: a master node and a plurality of distributed nodes;

[0204] Performing digital signature processing on the blockchain storage data using the master node to obtain consensus storage data;

[0205] Sequentially extracting distributed nodes from the plurality of distributed nodes and performing the following operations on the distributed nodes:

[0206] Performing validity verification on the consensus storage data using the distributed node to obtain a node verification result, wherein the node verification result is pass or fail;

[0207] Summarizing the node verification results to obtain a verification result set;

[0208] Determining the number of passed verifications according to the verification result set;

[0209] When the number of passed verifications is less than the preset number threshold, the main node is subjected to parameter debugging for evidence storage, a debugged main node is obtained, the debugged main node is taken as the main node, and the step of performing digital signature processing on the blockchain evidence storage data by using the main node is returned until the number of passed verifications is greater than or equal to the number threshold.

[0210] When the number of passed verifications is greater than or equal to the number threshold, the to-be-consensus evidence storage data is subjected to block packaging to obtain blockchain feedback data.

[0211] It should be understood that the blockchain evidence storage node refers to a blockchain node in a blockchain network, which is divided into a main node and a distributed node. The main node refers to a node responsible for initiating blockchain evidence storage in the process of on-chain processing. The distributed node refers to other participating nodes in the blockchain network except the main node, which is responsible for validity verification of the to-be-consensus evidence storage data submitted by the main node, to ensure the integrity and consistency of data in the blockchain network. The data signature refers to a process in which the main node uses a private key held by itself to perform an encryption operation on the blockchain evidence storage data. Optionally, an elliptic curve digital signature algorithm is used for the encryption operation. The to-be-consensus evidence storage data refers to a verifiable digital signature generated after the blockchain evidence storage data is subjected to data signature, which can be used to prove the source and integrity of the blockchain evidence storage data. The validity verification refers to a process in which the distributed node decrypts the digital signature through a public key, and compares whether the decrypted result is consistent with the blockchain evidence storage data. The node verification result refers to a conclusion obtained after the validity verification by the distributed node, which is divided into passed and not passed. The passed refers to a result when the content obtained after the distributed node decrypts the digital signature is consistent with the blockchain evidence storage data, indicating that the format of the to-be-consensus evidence storage data is standard and the data is not tampered with, and the to-be-consensus evidence storage data is in compliance and effective. The not passed refers to a result when the content obtained after the distributed node decrypts the digital signature is inconsistent with the blockchain evidence storage data, indicating that the format of the to-be-consensus evidence storage data does not conform to the standard or the data is tampered with.

[0212] It should be noted that the set of verification results refers to a set composed of all node verification results. The number of verification passes refers to the number of node verification results in the verification result set that pass. The number threshold refers to a critical number of nodes set by a person for judging the validity of the to-be-consensus evidence data. When the number of verification passes is greater than or equal to the number threshold, the to-be-consensus evidence data is considered valid. Optionally, the number threshold can be set to 2 / 3 of the total number of node verification results. The evidence parameter debugging refers to a process of adjusting and optimizing related parameters or configurations of the master node in the digital signature process when the number of verification passes is less than the number threshold, specifically including: replacing the private key, adjusting the algorithm used in the encryption operation, modifying the data transmission format, etc. The debug master node refers to the master node after the evidence parameter debugging. The block packaging refers to the operation of chaining the blockchain evidence data when the number of verification passes is greater than or equal to the number threshold.

[0213] S7, based on the evidence identifier and the three-dimensional digital asset file, obtaining a three-dimensional digital asset.

[0214] It should be explained that the three-dimensional digital asset based on the evidence identifier and the three-dimensional digital asset file refers to writing the evidence identifier into the related metadata of the three-dimensional digital asset file to form a new data file, which is the three-dimensional digital asset. The three-dimensional digital asset binds the related data content of the target three-dimensional model with the on-chain evidence record on the blockchain, ensuring the legality and authenticity of the digital asset, and guaranteeing the trusted right of the digital asset.

[0215] The present application is to solve the problems in the background art. The present application acquires a target object and an image acquisition unit. The present application determines the physical entity of three-dimensional reconstruction and the hardware basis of image acquisition through this step. The present application acquires a multi-view image set and a point coordinate set by using the image acquisition unit to collect the target object from multiple perspectives. The present application acquires a camera pose parameter set according to the multi-view image set and the point coordinate set. The present application determines the position and orientation of the image acquisition unit in three-dimensional space when each qualified image frame is taken by feature extraction and pose solving, which provides accurate geometric constraints for subsequent three-dimensional reconstruction. The present application performs Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model. The present application extracts sparse point clouds from multi-view images and optimizes the Gaussian point cloud model based on differentiable rendering, so that the obtained target three-dimensional model is closer to the target object, improving the accuracy and authenticity of the model. The target three-dimensional model is encapsulated to obtain a three-dimensional digital asset file. The three-dimensional digital asset file is stored in a blockchain to obtain a storage identification. The three-dimensional digital asset is obtained based on the storage identification and the three-dimensional digital asset file. Therefore, the present application can solve the problems of low model accuracy, low generation efficiency and low asset right confirmation credibility in the current three-dimensional digital asset generation process.

[0216] As shown in Figure 2 It is a functional module diagram of a three-dimensional digital asset generation system based on Gaussian light field and blockchain provided by an embodiment of the present application.

[0217] The three-dimensional digital asset generation system based on Gaussian light field and blockchain 100 can be installed in an electronic device. According to the functions to be realized, the three-dimensional digital asset generation system based on Gaussian light field and blockchain 100 can include a device acquisition module 101, an image acquisition and pose solving module 102, a Gaussian light field reconstruction module 103 and a blockchain storage module 104. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0218] The device acquisition module 101 is used to acquire a target object and an image acquisition unit.

[0219] The image acquisition and pose solving module 102 is used to collect the target object from multiple perspectives by using the image acquisition unit to obtain a multi-view image set and a point coordinate set, and acquire a camera pose parameter set according to the multi-view image set and the point coordinate set.

[0220] The Gaussian light field reconstruction module 103 is used to perform Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model.

[0221] The blockchain storage module 104 is configured to encapsulate the target three-dimensional model to obtain a three-dimensional digital asset file, store the three-dimensional digital asset file in a blockchain to obtain a storage identification, and acquire a three-dimensional digital asset based on the storage identification and the three-dimensional digital asset file.

[0222] In detail, the modules in the three-dimensional digital asset generation system 100 based on the Gaussian light field and the blockchain in the embodiments of the present application adopt the same technical means as the three-dimensional digital asset generation method based on the Gaussian light field and the blockchain in the above Figure 1 , and can produce the same technical effects, which will not be described here in detail.

[0223] As shown in Figure 3 , it is a structural schematic diagram of an electronic device for implementing the three-dimensional digital asset generation method based on the Gaussian light field and the blockchain according to an embodiment of the present application.

[0224] The electronic device 1 can include a processor 10, a memory 11 and a bus 12, and can further include a computer program stored in the memory 11 and executable on the processor 10, such as a three-dimensional digital asset generation method program based on the Gaussian light field and the blockchain.

[0225] The memory 11 includes at least one type of readable storage medium, including a flash memory, a mobile hard disk, a multimedia card, a card-type memory (such as an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 11 includes an internal storage unit of the electronic device 1 and also includes an external storage device. The memory 11 can be used not only to store application software and various data installed in the electronic device 1, such as the code of the three-dimensional digital asset generation method program based on the Gaussian light field and the blockchain, but also to temporarily store data that has been output or will be output.

[0226] The processor 10 can be composed of integrated circuits in some embodiments, for example, can be composed of a single packaged integrated circuit, or can be composed of multiple packaged integrated circuits with the same function or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors, combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, which connects various components of the entire electronic device through various interfaces and lines, executes programs or modules stored in the memory 11 (such as a three-dimensional digital asset generation method based on Gaussian light field and blockchain), and calls data stored in the memory 11, to perform various functions and process data of the electronic device 1.

[0227] The bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize the connection and communication between the memory 11, the at least one processor 10, etc.

[0228] Figure 3 Only the electronic device with components is shown, and those skilled in the art can understand that, Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0229] For example, although not shown, the electronic device 1 can also include a power supply (such as a battery) to power various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, so as to realize functions such as charge management, discharge management, and power consumption management through the power management device. The power supply can also include one or more direct current or alternating current power supplies, recharging devices, power supply fault detection circuits, power supply converters or inverters, power supply status indicators, etc. The electronic device 1 can also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which are not described here.

[0230] Further, the electronic device 1 can also include a network interface, which can optionally include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), and is usually used to establish a communication connection between the electronic device 1 and other electronic devices.

[0231] Optionally, the electronic device 1 can also include a user interface, which can be a display, an input unit such as a keyboard, and optionally a standard wired interface, a wireless interface. Optionally, in some embodiments, the display can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. The display can also be appropriately referred to as a display screen or a display unit, for displaying information processed in the electronic device 1 and for displaying a visualized user interface.

[0232] The memory 11 in the electronic device 1 stores a Gaussian light field and blockchain-based three-dimensional digital asset generation method program, which is a combination of multiple instructions and can achieve the following when running in the processor 10:

[0233] Obtaining a target object and an image acquisition unit;

[0234] Using the image acquisition unit to collect the target object from multiple perspectives to obtain a multi-view image set and a point coordinate set;

[0235] Obtaining a camera pose parameter set according to the multi-view image set and the point coordinate set;

[0236] Reconstructing a Gaussian light field based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model;

[0237] Packaging the target three-dimensional model to obtain a three-dimensional digital asset file;

[0238] Storing the three-dimensional digital asset file on a blockchain to obtain a storage identification;

[0239] Obtaining a three-dimensional digital asset based on the storage identification and the three-dimensional digital asset file.

[0240] Specifically, the processor 10 can refer to the description of the specific implementation method of the above instructions in the corresponding embodiments, which will not be repeated here. Figures 1 to 3

[0241] ​Further, the modules / units integrated in the electronic device 1 are stored in a computer readable storage medium if they are realized in the form of software function units and sold or used as independent products. The computer readable storage medium can be volatile or non-volatile. For example, the computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM).

[0242] The application further provides a computer readable storage medium, which stores a computer program, and the computer program can realize the following when executed by a processor of an electronic device:

[0243] obtaining a target object and an image acquisition unit;

[0244] collecting the target object by using the image acquisition unit to obtain a multi-view image set and a point coordinate set;

[0245] obtaining a camera pose parameter set according to the multi-view image set and the point coordinate set;

[0246] performing Gaussian light field reconstruction based on the multi-view image set and the camera pose parameter set to obtain a target three-dimensional model;

[0247] performing encapsulation processing on the target three-dimensional model to obtain a three-dimensional digital asset file;

[0248] performing blockchain notarization on the three-dimensional digital asset file to obtain a notarization identifier;

[0249] obtaining a three-dimensional digital asset based on the notarization identifier and the three-dimensional digital asset file.

[0250] In the several embodiments provided in the present application, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and actual implementation can have another division manner.

[0251] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. According to actual needs, part or all of the modules can be selected to achieve the purpose of the present embodiment.

[0252] In addition, each function module in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software function module.

[0253] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be realized in other specific forms without departing from the spirit or essential characteristics of the present application.

[0254] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for generating three-dimensional digital assets based on Gaussian light fields and blockchain, characterized in that, The method includes: Acquire the target object and image acquisition unit; The image acquisition unit is used to acquire the target object from multiple perspectives, resulting in a multi-view image set and a set of point coordinates. Based on the multi-view image set and the point coordinate set, obtain the camera pose parameter set; Gaussian light field reconstruction is performed based on the multi-view image set and camera pose parameter set to obtain the target 3D model; The step of reconstructing the target 3D model using Gaussian light field based on the multi-view image set and camera pose parameter set includes: Based on a multi-view image set and a camera pose parameter set, a three-dimensional sparse point cloud is obtained, wherein the three-dimensional sparse point cloud includes multiple three-dimensional points. For each 3D point in the 3D sparse point cloud, the following operation is performed: The three-dimensional points are initialized with Gaussian properties to obtain Gaussian property parameters; By summarizing the Gaussian attribute parameters, a Gaussian attribute parameter set is obtained; Construct a Gaussian point cloud model based on the Gaussian attribute parameter set; The camera pose parameter set is sampled to obtain the verification camera pose parameter set; Based on the set of verification camera pose parameters, the Gaussian point cloud model is rendered in a differentiable manner to obtain a set of synthetic images. The set of synthetic images includes multiple synthetic images, and the synthetic images in the set of synthetic images correspond one-to-one with the verification camera pose parameters in the set of verification camera pose parameters. Based on the synthesized image set, the photometric reconstruction loss is calculated; When the photometric reconstruction loss is greater than the preset convergence threshold, the Gaussian attribute parameter set is updated to obtain the updated Gaussian parameter set. The updated Gaussian parameter set is used as the Gaussian attribute parameter set, and the above steps of constructing a Gaussian point cloud model based on the Gaussian attribute parameter set are returned until the photometric reconstruction loss is less than or equal to the convergence threshold. When the photometric reconstruction loss is less than or equal to the convergence threshold, the Gaussian point cloud model is confirmed as the target 3D model; The target 3D model is encapsulated to obtain a 3D digital asset file; The three-dimensional digital asset file is stored on the blockchain to obtain a storage identifier; Based on the evidence identification and the 3D digital asset file, the 3D digital asset is obtained.

2. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 1, characterized in that, The step of using the image acquisition unit to acquire multi-view images of the target object, obtaining a multi-view image set and a set of point coordinates, includes: Determine the three-dimensional bounding box data of the target object; Based on the 3D bounding box data and the preset acquisition precision, a spatial acquisition path point set is obtained, wherein the spatial acquisition path point set includes: multiple spatial acquisition path points; For each spatial acquisition path point in the set of spatial acquisition path points, the following operation is performed: Based on the spatial acquisition path points, image acquisition is performed using the image acquisition unit to obtain the original image frame; The original image frame is preprocessed to obtain a standard image frame; The standard image frames are subjected to quality verification to obtain qualified image frames; Determine the three-dimensional coordinates of the spatial acquisition path points; The qualified image frames and three-dimensional point coordinates are summarized to obtain a multi-view image set and a point coordinate set, wherein the qualified image frames in the multi-view image set correspond one-to-one with the three-dimensional point coordinates in the point coordinate set.

3. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 2, characterized in that, The process of performing quality verification on the standard image frame to obtain a qualified image frame includes: Feature detection is performed on the standard image frame to obtain multiple image feature points; Based on the multiple image feature points, multiple feature coordinates are determined, wherein the image feature points and feature coordinates correspond one-to-one; The standard image frame is divided into grids to obtain multiple regular grids; For each of the plurality of rule grids, the following operation is performed: The number of image feature points in a regular grid is determined based on multiple feature coordinates; Summarize the number of image feature points to obtain the image feature point count set; Based on the set of image feature points, the distribution dispersion is calculated, wherein the coefficient of variation of the set of image feature points is used as the distribution dispersion, and the formula for calculating the coefficient of variation is as follows: ， in, The coefficient of variation represents the set of feature points in an image. The standard deviation of the set of feature points in an image. It represents the average value of the set of feature points in an image; Calculate the sharpness evaluation value of the standard image frame, wherein the Laplacian variance is used as the sharpness evaluation value; Determine the total number of image feature points from multiple image feature points; If the total number of image feature points is greater than or equal to a preset threshold for the number of image feature points, the distribution dispersion is less than or equal to a preset dispersion threshold, and the sharpness evaluation value is greater than a preset sharpness threshold, then the standard image frame is confirmed as a qualified image frame. Otherwise, determine the type of non-compliance, adjust the parameters of the image acquisition unit according to the type of non-compliance, obtain the adjusted image acquisition unit, use the adjusted image acquisition unit as the image acquisition unit, and return to the above steps of using the image acquisition unit to acquire images based on the spatial acquisition path point until a qualified image frame is obtained.

4. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 3, characterized in that, The step of obtaining the camera pose parameter set based on the multi-view image set and the point coordinate set includes: For each three-dimensional point coordinate in the aforementioned point coordinate set, the following operation is performed: Based on qualified image frames in a multi-view image set corresponding to three-dimensional point coordinates, multiple qualified image feature points and multiple qualified feature coordinates are determined, wherein the qualified image feature points and qualified feature coordinates correspond one-to-one. Based on the multiple qualified image feature points, multiple true feature coordinates are determined; Based on the three-dimensional point coordinates, multiple real feature coordinates, and multiple qualified feature coordinates, the pose is calculated using a preset pose algorithm to obtain camera pose parameters, wherein the camera pose parameters include: rotation matrix and translation vector. The camera pose parameters are summarized to obtain the camera pose parameter set.

5. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 4, characterized in that, The process of acquiring a 3D sparse point cloud based on a multi-view image set and a camera pose parameter set includes: For each qualified image frame in the multi-view image set, the following operation is performed: Feature transformation descriptors are extracted from multiple qualified image feature points of the qualified image frame to obtain a feature descriptor subset; By summing up the aforementioned feature description subsets, multiple feature description subsets are obtained; Based on the multiple feature description subsets, cross-image feature matching is performed to obtain a set of matching feature point pairs; For each matching feature point pair in the set of matching feature point pairs, the following operation is performed: Based on the matching feature point pairs, the first qualified image frame and the second qualified image frame are determined from the multi-view image set. Based on the matching feature point pairs, the first qualified image frame, and the second qualified image frame, the first feature coordinates and the second feature coordinates are determined. Based on the first qualified image frame and the second qualified image frame, the first camera pose parameters and the second camera pose parameters are determined from the camera pose parameter set; Based on the first feature coordinates, the second feature coordinates, the first camera pose parameters, and the second camera pose parameters, a preset triangulation algorithm is used to obtain three-dimensional points. By summing the three-dimensional points, a three-dimensional sparse point cloud is obtained.

6. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 5, characterized in that, The calculation of photometric reconstruction loss based on the synthesized image set includes: For each synthesized image in the synthesized image set, the following operation is performed: Based on synthetic images, real images are identified from a multi-view image set; Based on the synthesized image and the real image, obtain the pixel set of the synthesized image and the pixel set of the real image, wherein the pixel set of the synthesized image includes multiple synthesized image pixels, the pixel set of the real image includes multiple real image pixels, and the synthesized image pixels correspond one-to-one with the real image pixels; For each pixel in the composite image pixel set, the following operation is performed: Obtain the RGB three-channel pixel values ​​of the composite image pixels, wherein the RGB three-channel pixel values ​​include: red channel pixel values, green channel pixel values ​​and blue channel pixel values; Based on the synthesized image pixels, the true RGB three-channel pixel values ​​of the corresponding real image pixels are determined, wherein the true RGB three-channel pixel values ​​include: true red channel pixel values, true green channel pixel values, and true blue channel pixel values; Based on the RGB three-channel pixel values ​​of the synthesized image pixels and the real RGB three-channel pixel values ​​of the real image pixels, the luminance difference value is calculated using the following formula: ， in, Indicates the difference in photometric intensity. This represents the red channel pixel value of the composite image. The green channel pixel value represents the pixel value of the composite image. This represents the blue channel pixel value of the composite image. This represents the true red channel pixel value of a real image pixel. This represents the true green channel pixel value of a real image pixel. This represents the true blue channel pixel value of the actual image pixel; By summing up the aforementioned photometric difference values, multiple photometric difference values ​​are obtained; Based on multiple photometric difference values, the image photometric loss value is calculated. This value refers to the color difference between the synthesized image and the original image. The formula for calculating the image photometric loss value is shown below: ， in, Indicates the image luminance loss value. This represents the sum of all photometric differences. This represents the total number of pixels in the composite image; Structural similarity analysis is performed on the synthetic image and the real image to obtain a structural similarity index, wherein the structural similarity index refers to a quantitative indicator used to represent the similarity between the synthetic image and the real image, obtained by using the MS-SSIM algorithm. Based on the image photometric loss value and structural similarity index, the photometric reconstruction loss component is calculated. The photometric reconstruction loss component is a quantitative indicator representing the overall difference between the synthesized image and the real image. The formula for calculating the photometric reconstruction loss component is as follows: ， in, This represents the loss component in photometric reconstruction. Represents the structural similarity index. This represents a weighting coefficient that is pre-defined by the user to indicate the image photometric loss value. The weighting coefficients of the pre-defined structural similarity index; By summing the photometric reconstruction loss components, multiple photometric reconstruction loss components are obtained; Based on the multiple photometric reconstruction loss components, the photometric reconstruction loss is calculated, wherein the photometric reconstruction loss refers to the arithmetic mean of the multiple photometric reconstruction loss components.

7. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 6, characterized in that, The step of performing blockchain-based notarization on the three-dimensional digital asset file to obtain a notarization identifier includes: The three-dimensional digital asset file is read to obtain the file byte sequence; A three-dimensional digital asset fingerprint is obtained by performing a hash operation on the file byte sequence using a preset one-way hash algorithm. Based on the three-dimensional digital asset fingerprint and the preset asset metadata, blockchain-based evidence storage data is constructed. The blockchain-stored data is processed on the blockchain to obtain blockchain feedback data; The blockchain feedback data is parsed to obtain the evidence storage identifier.

8. The method for generating three-dimensional digital assets based on Gaussian light fields and blockchain as described in claim 7, characterized in that, The process of uploading the blockchain-stored evidence data to the blockchain to obtain blockchain feedback data includes: Obtain blockchain evidence storage nodes, wherein the blockchain evidence storage nodes include: a master node and multiple distributed nodes; The master node is used to perform digital signature processing on the blockchain evidence storage data to obtain evidence storage data to be reached through consensus. Extract distribution nodes sequentially from the plurality of distribution nodes, and perform the following operations on each distribution node: The distributed nodes are used to verify the validity of the data to be stored in consensus, and the node verification result is obtained, wherein the node verification result is either passed or failed. The node verification results are summarized to obtain a verification result set; Based on the set of verification results, determine the number of verifications that passed. When the number of successful verifications is less than a preset threshold, the master node is debugged to obtain a debug master node. The debug master node is then used as the master node, and the steps described above for digitally signing the blockchain evidence data using the master node are returned until the number of successful verifications is greater than or equal to the threshold. When the number of verifications that pass is greater than or equal to the threshold, the data to be stored for consensus is packaged into blocks to obtain blockchain feedback data.

9. A system using the three-dimensional digital asset generation method based on Gaussian light field and blockchain as described in any one of claims 1 to 8, characterized in that, The system includes: The device acquisition module is used to acquire the target object and the image acquisition unit; The image acquisition and pose calculation module is used to acquire the target object from multiple perspectives using the image acquisition unit, obtain a set of multi-view images and a set of point coordinates, and obtain a set of camera pose parameters based on the set of multi-view images and the set of point coordinates. The Gaussian light field reconstruction module is used to reconstruct the Gaussian light field based on the multi-view image set and the camera pose parameter set to obtain the target three-dimensional model. The blockchain notarization module is used to encapsulate the target 3D model to obtain a 3D digital asset file, perform blockchain notarization on the 3D digital asset file to obtain a notarization identifier, and obtain the 3D digital asset based on the notarization identifier and the 3D digital asset file.

Images