A parameter automatic identification and tuning system for a gaussian sputtering model reconstruction

Abstract

The application discloses a kind of parameter automatic identification tuning systems of Gaussian sputtering model reconstruction, it is related to image analysis field, including: data receiving management module is used to obtain data source information and corresponding resource monitoring data;Data source analysis module is used to obtain the type identification result and scene demand information corresponding to data source information;Data preprocessing module is used to analyze and process data source information according to type identification result, obtain corresponding preprocessed image set, carry out comprehensive analysis and processing to preprocessed image set, obtain corresponding priority mode based on scene demand information and analysis processing result;Parameter hierarchical identification module carries out parameter identification analysis according to corresponding priority mode and resource monitoring data, obtains initial data source parameter;Parameter model generation module is used to construct Gaussian model file;Parameter adjustment optimization module is used to optimize the Gaussian model file constructed;The application improves the accuracy and efficiency of image analysis.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and in particular to an automatic parameter identification and optimization system for Gaussian splash model reconstruction. Background Technology

[0002] With the development of computer graphics, the demand for realistic graphics rendering effects is constantly increasing, especially in fields such as film, games, advertising, animation production, and navigation. Gaussian splash model has a promising future in the field of 3D reconstruction due to its efficient training and rendering speed and realistic detail restoration capabilities. However, its reconstruction quality is highly dependent on parameter settings, and manual optimization is inefficient and difficult to adapt to complex scenes.

[0003] A search revealed Chinese patent CN119068123A, which discloses a method, system, device, and storage medium for reconstructing a real-world 3D scene. Belonging to the field of computer vision modeling technology, the method includes: acquiring multi-angle 2D images of a target scene; converting the 2D images into a Gaussian splash model of the target scene; assigning semantic labels to each Gaussian ellipsoid in the Gaussian splash model to obtain a semantic Gaussian splash model; and editing the semantic Gaussian splash model to reconstruct the 3D target scene. This invention can significantly improve model editing accuracy and efficiency, providing technical support for film and television special effects, game development, virtual reality, and other fields.

[0004] Compared with existing technologies, the Chinese patent with patent number CN119068123A can obtain a semantic Gaussian splash model by assigning semantic labels to each Gaussian ellipsoid in the Gaussian splash model, and reconstruct a three-dimensional target scene based on the semantic Gaussian splash model, thereby improving the accuracy of model editing. However, in actual use, the above method assigns corresponding semantic labels to each Gaussian ellipsoid, which increases the workload in the model construction process to a certain extent, and it is difficult to ensure the accuracy of each semantic label, thus affecting the efficiency and accuracy of reconstructing the target scene to a certain extent. To this end, this invention proposes an automatic parameter identification and optimization system for Gaussian splash model reconstruction. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of insufficient accuracy in existing technologies by proposing an automatic parameter identification and optimization system for Gaussian splash model reconstruction.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An automatic parameter identification and tuning system for Gaussian splash model reconstruction includes: The data receiving and management module is used to obtain the corresponding input account information. Users upload the corresponding data source information through the input account information, mark the data source information, and obtain the corresponding resource monitoring data. The data source analysis module is used to perform type identification and requirement feature extraction on the data source information and the corresponding marked input account information, respectively, to obtain the corresponding type identification results and scenario requirement information, and to set the preprocessing process path information based on the obtained information; The data preprocessing module is used to analyze and process the data according to the preprocessing process path information, obtain the preprocessed image set corresponding to the data source information, extract features from each preprocessed image in the preprocessed image set, obtain the corresponding image scene elements, and perform comprehensive analysis based on the image scene elements and the corresponding scene requirement information to obtain the priority mode of the corresponding preprocessed image. The parameter hierarchical recognition module is used to perform initial parameter analysis based on the priority patterns in the preprocessed image set corresponding to the data source information, obtain initial pattern parameter recognition information, optimize and adjust the initial pattern parameter recognition information based on the corresponding resource monitoring data, and obtain the initial data source parameters. The parametric model generation module is used to construct a Gaussian model file based on the initial data source parameters; The parameter adjustment and optimization module is used to analyze and process the constructed Gaussian model file, obtain feedback evaluation data, optimize the corresponding initial data source parameters based on the feedback evaluation data until the feedback evaluation data meets the standard, and output the corresponding Gaussian model file.

[0007] The above technical solution further includes: the data receiving and management module includes: The data receiving unit is used to set up a data receiving port, obtain the corresponding input account information through the data receiving port, and receive the uploaded data source information according to the corresponding input account information. The resource management unit is used to mark the received data source information according to the corresponding input account information and set up a resource processing space, which is used to perform image analysis processing on the corresponding data source information. The resource monitoring unit is used to monitor and process the resources of each corresponding resource processing space within the system and obtain the corresponding resource monitoring data.

[0008] Furthermore, the data source analysis module includes: The type identification unit is used to identify the type of data source information in the corresponding resource processing space. The data receiving port includes the adaptation identifier type, dimension type and format type. The corresponding identification node is set according to the corresponding type, and the type identification architecture model is set according to the identification node. The data source information is analyzed and processed through the type identification architecture model. The identification results corresponding to the three dimensions are mutually verified and analyzed, and the corresponding type identification results are output. The requirement identification unit is used to parse and process the type identification results corresponding to the data source information to obtain the corresponding data source features; obtain the historical account behavior information corresponding to the entered account information to obtain the corresponding account attribute features; construct a scenario requirement mapping library, input the data source features and account attribute features into the scenario requirement mapping library for comparative analysis to obtain feature requirement mapping information, and perform weighted fusion processing and conflict resolution processing on the obtained feature requirement mapping information to obtain the corresponding scenario requirement information. The process analysis unit is used to analyze and process the preprocessing processes involved in the system and the corresponding type identification results and scenario requirement information, and to construct a process mapping table; based on the type identification results and scenario requirement information corresponding to the relevant data source information, the preprocessing process path information is obtained by inputting them into the process mapping table.

[0009] Furthermore, the data preprocessing module includes: The process processing unit is used to obtain the preprocessing process path information corresponding to the data source information, analyze and process the corresponding data source information sequentially according to the preprocessing process path information, obtain the corresponding preprocessed images sequentially according to the spatiotemporal sequence, and generate a preprocessed image set. Based on the preprocessing results of each preprocessed image in each preprocessed image set through the preprocessing process path information, the corresponding resolution parameters, texture complexity parameters, and image scene elements are obtained. A multi-level priority mode analysis model is set, and the resolution parameters, texture complexity parameters, and image scene elements in the corresponding preprocessed images are superimposed sequentially through the multi-level priority mode analysis model to obtain the demand marking results at the corresponding positions in the corresponding preprocessed images. The demand marking results include quality markings and velocity markings. The processing and analysis unit is used to perform logical association analysis on the labeling results corresponding to each preprocessed image in the preprocessed image set according to the scene requirement information corresponding to the data source information, and to determine the priority mode of the corresponding preprocessed image based on the obtained logical association analysis results. The priority mode includes quality priority mode and speed priority mode.

[0010] Furthermore, the parameter hierarchical identification module includes: The pattern recognition and analysis unit is used to obtain the priority pattern corresponding to each preprocessed image in the preprocessed image set corresponding to the data source information, perform initial parameter recognition according to the corresponding priority pattern, and obtain the corresponding initial pattern parameter recognition information, which includes the corresponding COLMAP core parameters and Gaussian-Splatting core parameters. The resource identification and analysis unit is used to acquire the resource monitoring data and the initial mode parameter identification information corresponding to the corresponding data source information obtained in the corresponding resource processing space, acquire the corresponding resource demand information based on the corresponding initial mode parameter identification information, compare and analyze the acquired resource demand data and resource monitoring data to acquire the corresponding space load data, and perform scheduling analysis based on the space load data of each resource processing space to acquire resource scheduling data. The initial optimization analysis unit is used to dynamically adjust the corresponding initial mode parameter identification information based on the resource scheduling data corresponding to the corresponding resource processing space, and to obtain the initial data source parameters.

[0011] Furthermore, the process by which the initial optimization analysis unit obtains the initial data source parameters includes: Based on the priority mode corresponding to the data source information, the importance of the corresponding initial mode parameter identification information is evaluated using the single-parameter variable method. The corresponding importance evaluation data is obtained, and the corresponding initial mode parameter identification information is sorted from low to high. According to the sorting results, the corresponding initial mode parameter identification information is optimized and adjusted according to the resource scheduling data. The optimization and adjustment results are integrated until the corresponding resource monitoring data all meet the requirements, and the initial data source parameters are generated.

[0012] Furthermore, the parameter model generation module includes: Obtain the initial data source parameters corresponding to the data source information within the corresponding resource processing space, call the COLMAP core parameters for calculation and analysis to generate the corresponding sparse point cloud, and then call the Gaussian-Splatting core parameters for iterative optimization to generate the corresponding Gaussian model file.

[0013] Furthermore, the parameter adjustment and optimization module includes: The feedback evaluation unit is used to feed back the Gaussian model file to the corresponding input account information. Users can use the corresponding input account information to evaluate the Gaussian model file corresponding to the corresponding data source information and obtain feedback evaluation data. The adjustment and optimization unit is used to obtain the corresponding feedback evaluation data, perform deviation processing on the corresponding positions in the Gaussian model file based on the feedback evaluation data, obtain the corresponding preprocessed image and initial data source parameters in the preprocessed image set corresponding to the corresponding data source information based on the deviation processing results, optimize the initial data source parameters of the corresponding preprocessed image based on the corresponding deviation processing results until the deviation processing results reach the minimum value, and output the corresponding Gaussian model file.

[0014] The present invention has the following beneficial effects: 1. In this invention, by performing scene recognition on the corresponding data source information, the corresponding scene requirement information is obtained. Based on the recognized scene requirement information, the labeling results corresponding to the pixel value resolution analysis processing results are logically verified and analyzed to obtain the corresponding priority mode. Thus, the corresponding parameter adjustment method is set according to different priority modes, which can improve the accuracy of the image analysis process to a certain extent.

[0015] 2. In this invention, the initial parameters are analyzed and processed by the priority mode under the scene requirement information to obtain the corresponding initial mode parameter identification information. Based on the obtained resource monitoring data, the initial mode parameter identification information under the corresponding priority mode is optimized and adjusted, so as to maximize resource utilization under the corresponding resource conditions and improve the efficiency of the image analysis and processing process. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the automatic parameter identification and optimization system for Gaussian splash model reconstruction proposed in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 like Figure 1 As shown, the present invention proposes an automatic parameter identification and tuning system for Gaussian splash model reconstruction, comprising: The data receiving and management module is used to obtain the corresponding input account information. Users upload the corresponding data source information through the input account information, mark the data source information, and obtain the corresponding resource monitoring data. The data source analysis module is used to perform type identification and requirement feature extraction on the data source information and the corresponding marked input account information, respectively, to obtain the corresponding type identification results and scenario requirement information, and to set the preprocessing process path information based on the obtained information; The data preprocessing module is used to analyze and process the data according to the preprocessing process path information, obtain the preprocessed image set corresponding to the data source information, extract features from each preprocessed image in the preprocessed image set, obtain the corresponding image scene elements, and perform comprehensive analysis based on the image scene elements and the corresponding scene requirement information to obtain the priority mode of the corresponding preprocessed image. The parameter hierarchical recognition module is used to perform initial parameter analysis based on the priority patterns in the preprocessed image set corresponding to the data source information, obtain initial pattern parameter recognition information, optimize and adjust the initial pattern parameter recognition information based on the corresponding resource monitoring data, and obtain the initial data source parameters. The parametric model generation module is used to construct a Gaussian model file based on the initial data source parameters; The parameter adjustment and optimization module is used to analyze and process the constructed Gaussian model file, obtain feedback evaluation data, optimize the corresponding initial data source parameters based on the feedback evaluation data until the feedback evaluation data meets the standard, and output the corresponding Gaussian model file. As shown above, this invention improves the efficiency and accuracy of image rendering analysis to a certain extent by identifying the types of information from different data sources, obtaining the corresponding priority mode based on the identification results, and combining the corresponding priority mode with scene features, resource constraints, and quantitative indicators.

[0019] In practical implementation, the data receiving and management module includes: A data receiving unit is used to set up a data receiving port and obtain the corresponding input account information through the data receiving port. The data receiving port receives the uploaded data source information according to the corresponding input account information. The data source information includes, but is not limited to, videos, panoramic videos, image sets, and panoramic image sets. The obtained data source information is marked according to the corresponding input account information. The resource management unit is used to set up corresponding resource processing spaces for the data source information received by the data receiving unit in the system according to the corresponding labeling processing results, and to perform image analysis processing on the corresponding data source information through the corresponding resource processing spaces. The resource monitoring unit is used to analyze and process the corresponding resource processing spaces within the system and obtain the corresponding resource monitoring data. It should be further explained that, in the specific implementation process, the data receiving unit involves input account information such as device user information and scene type. The corresponding input account information provides a reference for image analysis and processing of subsequent data source information, thereby improving the accuracy and analysis efficiency in the Gaussian splash model reconstruction process.

[0020] In the specific implementation process, the data source analysis module includes: The type identification unit is used to identify the type of the obtained data source information. According to the adaptation identifier type, dimension type and format type corresponding to the data receiving port, the corresponding identification nodes are set according to the corresponding adaptation type, dimension type and format type, and the corresponding identification nodes are set into the type identification architecture model. The obtained data source information is input into the type recognition architecture model. The corresponding recognition nodes are traversed in turn. The corresponding recognition nodes perform matching analysis on the corresponding data source information for the corresponding adaptation identifier type, dimension type and format type in turn. The data source information type recognition results corresponding to the three dimensions are obtained. The type recognition results corresponding to each of the three dimensions are verified and analyzed until the verification is passed. The type recognition results corresponding to the corresponding data source information are then output. The demand identification unit is used to acquire relevant data source information and its corresponding input account information, analyze and process the type identification results corresponding to the input account information and data source information, and acquire the scene identification results corresponding to the relevant data source information. Its specific implementation process includes: Obtain the type recognition result corresponding to the corresponding data source information, and perform parsing and processing based on the corresponding type recognition result to obtain the corresponding spatiotemporal features, acquisition features and content features. Among them, the spatiotemporal features are the time series information and three-dimensional spatial information corresponding to the corresponding data source information, the acquisition features are the device identification features corresponding to the corresponding data source information, and the content features are the scene elements corresponding to the corresponding data source information. Obtain historical account behavior information corresponding to the entered account information, analyze and process the historical account behavior information, and obtain the corresponding account attribute characteristics; Set up a scenario requirement mapping library based on the feature data corresponding to the historical data source information and the account information entered. Input the corresponding data source features and account features into the scenario requirement mapping library to obtain the feature requirement mapping information corresponding to the feature elements. The obtained feature requirement mapping information is subjected to weighted fusion processing and conflict resolution processing according to its corresponding data source features and account features in sequence to obtain the scenario requirement information corresponding to the data source information; The process analysis unit is used to match the type identification results corresponding to the relevant data source information with the corresponding preprocessing process based on the relevant scenario requirements, wherein: The type recognition results corresponding to the data source information include video type and image type, and the scene requirement information includes accuracy requirements, efficiency requirements and interaction requirements; The preprocessing process involved in the system and the corresponding type identification results and scene requirement information are obtained and analyzed to obtain the corresponding process mapping table. The process mapping table includes whether each preprocessing process node needs to be processed under the conditions of video type, image type, accuracy requirement, efficiency requirement or interaction requirement. Input the type identification results and scenario requirements corresponding to the data source information into the process mapping table to obtain the preprocessing process path information corresponding to the data source information.

[0021] In specific implementation, the data preprocessing module includes: The process processing unit is used to obtain the preprocessing process path information corresponding to the corresponding data source information, and to perform preprocessing analysis on the corresponding preprocessing images sequentially according to the preprocessing process path information. This process obtains the preprocessing image set corresponding to the corresponding data source information, as well as the resolution parameters, texture complexity parameters, and image scene elements corresponding to each preprocessing image within the preprocessing image set. Specifically: Pixel value resolution analysis is performed on each preprocessed image. The width (WD) and height (HD) pixel values ​​of each preprocessed image are read, and the corresponding resolution parameter FR is calculated, where: ; Texture parameter analysis is performed on each preprocessed image. Pixels in the preprocessed image are sequentially traversed and marked as target pixels. Texture feature data is extracted from each target pixel, and the corresponding texture complexity parameter WR is calculated based on the texture feature data. , is the grayscale value of the corresponding pixel, and k is the number of pixels in the edge region of the corresponding target pixel; Each preprocessed image is analyzed and processed by image scene elements. Each preprocessed image is classified, detected and segmented based on machine vision to obtain the corresponding scene type, core elements and element attributes. The obtained data information is integrated to obtain the corresponding image scene elements. The image scene elements include scene type elements, scene location elements and scene attribute elements in sequence. Based on the analysis process of historical data sources, a multi-level priority mode analysis model is set up according to the corresponding resolution parameters, texture complexity parameters, and image scene elements. This multi-level priority mode analysis model is used to sequentially mark requirements based on the resolution parameters, texture complexity parameters, and image scene elements, setting resolution analysis levels, texture analysis levels, and image scene analysis levels respectively. The analysis process of the texture analysis level is processed based on the analysis results of the resolution analysis level, and the analysis process of the image scene analysis level is processed based on the analysis results of the texture analysis level, proceeding in a progressive analysis manner. The process includes: The resolution analysis level is equipped with a resolution parameter threshold FR0, and the corresponding resolution parameter FR is compared and analyzed with the resolution parameter threshold FR0: If FR≥FR0, then quality marking is performed; If FR < FR0, then velocity marking is performed; The texture analysis layer is equipped with a texture parameter threshold WR0, and the corresponding texture complexity parameter WR is compared and analyzed with the texture parameter threshold WR0: If WR≥WR0, then quality marking is performed; If WR < WR0, then speed marking is performed; The system determines whether the labeling results at the corresponding location in the preprocessed image at the resolution analysis level and the texture analysis level are consistent. If they are consistent, they are directly input into the image scene analysis level. If they are inconsistent, the texture complexity parameters at the corresponding location in the preprocessed image are corrected to obtain the effective texture complexity parameter YWR. , The adaptation coefficient is the corresponding labeling result for the analysis level at the corresponding resolution. When the label is a velocity label, the adaptation coefficient is greater than 1, and when the label is a mass label, the adaptation coefficient is less than or equal to 1. Forced comparative analysis is performed based on effective texture complexity parameters and texture parameter threshold WR0 to obtain the corresponding labeling results within the texture analysis level; The image scene analysis layer is equipped with an image scene element reference library. Based on the historical priority marking of the image scene elements corresponding to the corresponding historical data source information, the obtained image scene elements are sequentially traversed in the corresponding image scene element reference library, and the marking results of the corresponding image scene elements are obtained sequentially according to the scene structure of the image scene elements. The textural complexity parameters of the labeling results corresponding to the corresponding image scene elements and their location information are integrated and processed. According to the type and texture complexity parameters of the image scene elements, the corresponding weight coefficients are set respectively. The labeling results of the corresponding image scene elements are statistically analyzed according to the corresponding weight coefficients. The statistical values ​​of the labeling results of the image scene elements at the corresponding locations are obtained, and the labeling results with larger statistical values ​​are obtained. Based on the corresponding labeling results, labeling processing is performed in the corresponding preprocessed image, and the labeling processing results include quality labels and velocity labels; It should be further explained that, in the specific implementation process, when the process processing unit acquires each preprocessed image in the preprocessed image set, it transfers and marks each preprocessed image according to the distribution results of the corresponding data source information in the data source analysis module, and obtains the type recognition results and scene requirement information corresponding to each preprocessed image. The processing and analysis unit performs verification processing and analysis based on the labeling results of each preprocessed image in the corresponding preprocessed image set and the corresponding transfer labels. The process includes: The scenario requirement type corresponding to the corresponding transfer mark is analyzed and processed with the quality mark and speed mark respectively to obtain the logical association relationship between the corresponding scenario requirement type and the quality mark and speed mark, and a logical association verification table is generated. Based on the established logical association verification table, logical analysis is performed on the corresponding transfer markers with the quality markers and the velocity markers respectively, and the corresponding logical analysis results are obtained. The logical analysis results include verification consistency, verification crossover, and verification separation. Based on the corresponding logical analysis results, the labeling results corresponding to the preprocessed images are integrated to obtain the corresponding priority modes. These priority modes include a quality-priority mode and a speed-priority mode, wherein: If the corresponding logical analysis result is consistent with the verification, then set the priority mode according to the corresponding marking result; If the corresponding logical analysis result is a cross-validation, then a mapping comparison analysis is performed based on the deviation between the corresponding resolution parameter FR and the resolution parameter threshold FR0 and the mapping quantity of the corresponding scene elements. The corresponding deviation value and mapping quantity are normalized, and a comparison analysis is performed based on the normalization result. If the normalization result corresponding to the deviation value is greater than or equal to the normalization result corresponding to the mapping quantity, then the corresponding priority mode is set according to the marking result; otherwise, the corresponding priority mode is set according to the scene requirements. If the corresponding logical analysis result is verification separation, then perform image anomaly reporting analysis; The obtained priority pattern is sent to the parameter processing and identification module; It should be further explained that, in the specific implementation process, within the workflow processing unit, if the corresponding data source information is a video, the corresponding preprocessing process also includes extracting keyframes from the corresponding video or panoramic video to obtain the corresponding image sequence; if the corresponding data source information is an image set or panoramic image set, the corresponding preprocessing process does not include keyframe extraction, and proceeds directly to the subsequent preprocessing process; in addition, the corresponding data source information is pre-classified, and the corresponding preprocessing process is set according to the classification results, thereby improving the efficiency of data analysis in the Gaussian splash model reconstruction process.

[0022] In specific implementation, the parameter initial identification module includes: The pattern recognition and analysis unit is used to set up a corresponding pattern analysis layer within the resource processing space corresponding to the data source information. Based on the set pattern analysis layer, it performs initial parameter recognition on the pattern analysis results corresponding to the data source information, obtaining the corresponding initial pattern parameter recognition information. This initial pattern parameter recognition information includes corresponding COLMAP core parameters and Gaussian-Splatting core parameters. The COLMAP core parameters include the feature extraction threshold TC, matching ratio RC, sparse reconstruction iteration count IC, and point cloud density factor DC. The Gaussian-Splatting core parameters include the iteration count IG, Gaussian number NG, resolution scaling factor SG, and regularization coefficient RG. The corresponding parameter formulas are as follows: The feature extraction threshold TC in the quality-first mode is ; The feature extraction threshold TC in speed-first mode is ,in, =0.04~0.06, =0.12~0.15, = =0.2~0.3, The baseline resolution is recommended (1920×1080=2073600). The matching ratio RC in the quality-first mode is ; The matching ratio RC in speed priority mode is The constraints are as follows: ; The IC of sparse reconstruction iterations in the quality-first mode is: ; The IC of sparse reconstruction iterations in the speed-first mode is: The constraint is that the number of sparse reconstruction iterations is greater than or equal to 30 and less than or equal to 200. Cloud density factor DC under quality-first mode is ; Cloud density factor DC in speed-priority mode is The constraints are as follows: ; The number of iterations IG in the quality-first mode is ; The number of iterations IG in speed-first mode is The constraints are as follows: The number of iterations IG is greater than or equal to 8000 and less than or equal to 80000; The number of Gaussians NG in the quality-first mode is ; Gaussian quantity in speed-priority mode (NG) The constraints are as follows: ; The resolution scaling factor SG in quality-first mode is ; The resolution scaling factor SG in speed-priority mode is: The constraints are as follows: ; The regularization coefficient RG in the quality-first mode is ; The regularization coefficient RG in speed-priority mode is The constraints are as follows: ; The initial identification results of the obtained parameters are labeled according to the corresponding pattern analysis layer; A resource identification and analysis unit is used to set up a corresponding resource analysis layer within the resource processing space corresponding to the relevant data source information. The resource analysis layer is used to adaptively adjust itself based on the corresponding resource monitoring data and the identification information of each initial mode parameter corresponding to the mode analysis layer. The process includes: Obtain the resource requirement data for each time series within the preprocessed image set corresponding to the initial pattern parameter recognition information corresponding to the relevant data source information; The resource monitoring data obtained from each resource processing space corresponding to each time series are integrated and processed. The resource demand data corresponding to the corresponding time series is compared and analyzed with the corresponding resource monitoring data to obtain spatial load data, which includes spatial overload, spatial full load, and spatial load. Perform resource scheduling analysis on the space load data corresponding to each resource processing space, obtain resource scheduling data, and feed the corresponding resource scheduling data back to each resource processing space. The initial optimization analysis unit is used to perform optimization processing based on the initial pattern parameter identification information and resource scheduling data corresponding to the pattern analysis layer and the resource analysis layer. Based on the corresponding resource scheduling data, perform resource reset analysis on the corresponding resource processing space, and determine the resource monitoring data within the corresponding resource processing space based on the results of the corresponding resource reset analysis. If the resource demand data corresponding to the initial pattern parameter identification information in the corresponding resource processing space is still considered to be space overloaded, then the initial pattern parameter identification information in the corresponding resource processing space will be optimized and analyzed. The importance of the initial mode parameter identification information for each type is assessed using the single-parameter variable method based on the quality and velocity impact dimensions, respectively, to obtain the importance assessment data for the corresponding impact dimensions; Based on the obtained importance assessment data, sort the data from low to high, and set up a corresponding resource-parameter elastic dynamic adjustment model based on the sorting results: Set the elastic adjustment parameters based on the corresponding importance assessment data. Where n is the corresponding parameter type, the corresponding optimization adjustment amount is obtained. Mark the current parameter value corresponding to the corresponding parameter type as ,in: ,in, and Set the corresponding parameter type constraints for the current resource rate and target resource rate, respectively, based on the priority mode corresponding to the corresponding data source information; The optimization adjustment amounts obtained by analyzing and processing the identification information of the corresponding initial mode parameters based on the resource-parameter elastic dynamic adjustment model are integrated until the corresponding resource monitoring data meet the requirements, and the initial data source parameters are generated.

[0023] In specific implementation, the parameter model generation module includes: Obtain the initial data source parameters corresponding to the data source information within the corresponding resource processing space, call the COLMAP core parameters for calculation and analysis to generate the corresponding sparse point cloud, and then call the Gaussian-Splatting core parameters for iterative optimization to generate the corresponding Gaussian model file.

[0024] In specific implementation, the parameter adjustment and optimization module includes: The feedback evaluation unit is used to feed back the Gaussian model file to the corresponding input account information. The user uses the corresponding input account information to perform feedback evaluation on the Gaussian model file corresponding to the corresponding data source information and obtain feedback evaluation data. The feedback evaluation data is the corresponding quantization deviation data, which corresponds to pixel-level deviation, set deviation and texture deviation respectively. The adjustment and optimization unit is used to acquire corresponding feedback evaluation data, perform deviation processing on corresponding positions within the Gaussian model file based on the feedback evaluation data, map the corresponding feedback evaluation data to acquire a spatial deviation heatmap, confirm the position based on the spatial deviation heatmap, acquire the corresponding preprocessed image and initial data source parameters within the preprocessed image set corresponding to the corresponding data source information, and optimize the initial data source parameters of the corresponding preprocessed image based on the corresponding deviation processing results. The optimization process involves determining the corresponding adjustment range based on the corresponding deviation magnitude, iteratively verifying and analyzing the corresponding adjustment range until the deviation processing result reaches the minimum value, and outputting the corresponding Gaussian model file.

[0025] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automatic parameter identification and optimization system for Gaussian splash model reconstruction, characterized in that, include: The data receiving and management module is used to obtain the corresponding input account information. Users upload the corresponding data source information through the input account information, mark the data source information, and obtain the corresponding resource monitoring data. The data source analysis module is used to perform type identification and requirement feature extraction on the data source information and the corresponding marked input account information, respectively, to obtain the corresponding type identification results and scenario requirement information, and to set the preprocessing process path information based on the obtained information; The data preprocessing module is used to analyze and process the data according to the preprocessing process path information, obtain the preprocessed image set corresponding to the data source information, extract features from each preprocessed image in the preprocessed image set, obtain the corresponding image scene elements, and perform comprehensive analysis based on the image scene elements and the corresponding scene requirement information to obtain the priority mode of the corresponding preprocessed image. The parameter hierarchical recognition module is used to perform initial parameter analysis based on the priority patterns in the preprocessed image set corresponding to the data source information, obtain initial pattern parameter recognition information, optimize and adjust the initial pattern parameter recognition information based on the corresponding resource monitoring data, and obtain the initial data source parameters. The parametric model generation module is used to construct a Gaussian model file based on the initial data source parameters; The parameter adjustment and optimization module is used to analyze and process the constructed Gaussian model file, obtain feedback evaluation data, optimize the corresponding initial data source parameters based on the feedback evaluation data until the feedback evaluation data meets the standard, and output the corresponding Gaussian model file. The data receiving and management module includes: The data receiving unit is used to set up a data receiving port, obtain the corresponding input account information through the data receiving port, and receive the uploaded data source information according to the corresponding input account information. The resource management unit is used to mark the received data source information according to the corresponding input account information and set up a resource processing space, which is used to perform image analysis processing on the corresponding data source information. The resource monitoring unit is used to monitor and process the resources of each corresponding resource processing space within the system and obtain the corresponding resource monitoring data; The data source analysis module includes: The type identification unit is used to identify the type of data source information in the corresponding resource processing space. The data receiving port includes the adaptation identifier type, dimension type and format type. The corresponding identification node is set according to the corresponding type data receiving port, and the type identification architecture model is set according to the identification node. The data source information is analyzed and processed through the type identification architecture model. The identification results corresponding to the three dimensions are mutually verified and analyzed, and the corresponding type identification results are output. The requirement identification unit is used to parse and process the type identification results corresponding to the data source information to obtain the corresponding data source features; obtain the historical account behavior information corresponding to the entered account information to obtain the corresponding account attribute features; construct a scenario requirement mapping library, input the data source features and account attribute features into the scenario requirement mapping library for comparative analysis to obtain feature requirement mapping information, and perform weighted fusion processing and conflict resolution processing on the obtained feature requirement mapping information to obtain the corresponding scenario requirement information. The process analysis unit is used to analyze and process the type identification results and scenario requirement information corresponding to the preprocessing processes involved in the system, and construct a process mapping table; based on the type identification results and scenario requirement information corresponding to the relevant data source information, the preprocessing process path information is obtained by inputting them into the process mapping table. The data preprocessing module includes: The process processing unit is used to obtain the preprocessing process path information corresponding to the data source information, analyze and process the corresponding data source information sequentially according to the preprocessing process path information, obtain the corresponding preprocessed images sequentially according to the spatiotemporal sequence, and generate a preprocessed image set. Based on the preprocessing results of each preprocessed image in each preprocessed image set through the preprocessing process path information, the corresponding resolution parameters, texture complexity parameters, and image scene elements are obtained. A multi-level priority mode analysis model is set, and the resolution parameters, texture complexity parameters, and image scene elements in the corresponding preprocessed images are superimposed sequentially through the multi-level priority mode analysis model to obtain the demand marking results at the corresponding positions in the corresponding preprocessed images. The demand marking results include quality markings and velocity markings. The processing and analysis unit is used to perform logical association analysis on the labeling results corresponding to each preprocessed image in the preprocessed image set according to the scene requirement information corresponding to the data source information, and to determine the priority mode of the corresponding preprocessed image based on the obtained logical association analysis results. The priority mode includes quality priority mode and speed priority mode.

2. The automatic parameter identification and optimization system for Gaussian splash model reconstruction according to claim 1, characterized in that, The parameter hierarchical identification module includes: The pattern recognition and analysis unit is used to obtain the priority pattern corresponding to each preprocessed image in the preprocessed image set corresponding to the data source information, perform initial parameter recognition according to the corresponding priority pattern, and obtain the corresponding initial pattern parameter recognition information, which includes the corresponding COLMAP core parameters and Gaussian-Splatting core parameters. The resource identification and analysis unit is used to acquire the resource monitoring data and the initial mode parameter identification information corresponding to the corresponding data source information obtained in the corresponding resource processing space, acquire the corresponding resource demand information based on the corresponding initial mode parameter identification information, compare and analyze the acquired resource demand data and resource monitoring data to acquire the corresponding space load data, and perform scheduling analysis based on the space load data of each resource processing space to acquire resource scheduling data. The initial optimization analysis unit is used to dynamically adjust the corresponding initial mode parameter identification information based on the resource scheduling data corresponding to the corresponding resource processing space, and to obtain the initial data source parameters.

3. The automatic parameter identification and optimization system for Gaussian splash model reconstruction according to claim 2, characterized in that, The process by which the initial optimization analysis unit obtains the initial data source parameters includes: Based on the priority mode corresponding to the data source information, the importance of the corresponding initial mode parameter identification information is evaluated using the single-parameter variable method. The corresponding importance evaluation data is obtained, and the corresponding initial mode parameter identification information is sorted from low to high. According to the sorting results, the corresponding initial mode parameter identification information is optimized and adjusted according to the resource scheduling data. The optimization and adjustment results are integrated until the corresponding resource monitoring data all meet the requirements, and the initial data source parameters are generated.

4. The automatic parameter identification and optimization system for Gaussian splash model reconstruction according to claim 3, characterized in that, The parameter model generation module includes: Obtain the initial data source parameters corresponding to the data source information within the corresponding resource processing space, call the COLMAP core parameters for calculation and analysis to generate the corresponding sparse point cloud, and then call the Gaussian-Splatting core parameters for iterative optimization to generate the corresponding Gaussian model file.

5. The automatic parameter identification and optimization system for Gaussian splash model reconstruction according to claim 4, characterized in that, The parameter adjustment and optimization module includes: The feedback evaluation unit is used to feed back the Gaussian model file to the corresponding input account information. Users can use the corresponding input account information to evaluate the Gaussian model file corresponding to the corresponding data source information and obtain feedback evaluation data. The adjustment and optimization unit is used to obtain the corresponding feedback evaluation data, perform deviation processing on the corresponding positions in the Gaussian model file based on the feedback evaluation data, obtain the corresponding preprocessed image and initial data source parameters in the preprocessed image set corresponding to the corresponding data source information based on the deviation processing results, optimize the initial data source parameters of the corresponding preprocessed image based on the corresponding deviation processing results until the deviation processing results reach the minimum value, and output the corresponding Gaussian model file.

Images