Ship outfitting piece multi-angle RGB graph and standard coordinate graph data set construction method

Abstract

The invention relates to the field of three-dimensional modeling and data set construction of ship outfitting parts, and particularly provides a ship outfitting part multi-angle RGB graph and standard coordinate graph data set construction method. The method comprises the steps of performing direction label analysis and normalization processing on a fitting-out part three-dimensional model based on a deep learning model, optimizing a view angle configuration file and generating multi-view angle configuration; adjusting the intensity and distribution of a light source according to the material characteristics of the fitting-out part, and rendering an RGB image and a standard coordinate graph; analyzing geometric consistency and direction matching of rendering results through multi-modal consistency verification, and generating an abnormal data list; the diversity of the data set is expanded by combining a data enhancement module; and finally, evaluating the multi-view coverage rate and consistency of the data set, and feeding back and optimizing a direction label generation model and a view configuration file to form a closed-loop improved mechanism. According to the method, a high-quality multi-angle RGB image and standard coordinate graph data set can be automatically generated, the consistency and geometric alignment degree of images are ensured, a three-dimensional reconstruction task and deep learning model training are effectively supported, and the quality and adaptability of the data set are improved.

Description

Technical Field

[0001] The present invention relates to the fields of computer graphics, three-dimensional modeling and deep learning, and in particular to a method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts. Background Art

[0002] In ship design and manufacturing, outfitting parts, as an important part of ships, have complex geometric shapes and diverse material properties. Traditional 3D modeling methods usually rely on manual modeling and rendering. Especially in the details and geometric structures of outfitting parts, manual rendering is difficult to ensure image consistency and high-precision 3D reconstruction effects. In addition, in the process of generating multi-view images and standard coordinate diagrams of ship outfitting parts, accurate rendering and verification are usually required for different viewpoints, lighting and materials to ensure the geometric consistency between multi-angle images and accurate matching of direction labels.

[0003] With the application of deep learning technology in 3D reconstruction, automated and efficient data generation has become particularly important. In particular, it is necessary to generate multi-angle images of ship outfitting parts to ensure the high quality and geometric consistency of the images to meet the training data requirements of deep learning models. However, existing technologies still face great challenges in multi-angle dataset generation, image consistency verification, geometric alignment, and directional label matching. How to automatically generate accurate multi-angle RGB images and standard coordinate map datasets, and ensure the high quality and consistency of the datasets, is still a key issue that needs to be solved in the current technical field.

[0004] The present invention introduces an automated rendering and verification mechanism and combines it with deep learning technology to propose a new method that can efficiently generate high-quality multi-angle RGB images and standard coordinate map data sets of outfitting parts, providing high-quality data support for three-dimensional reconstruction tasks and the training of deep learning models. Summary of the invention

[0005] The present invention provides a method for constructing a multi-angle RGB image and a standard coordinate image dataset of ship outfitting parts, so as to solve the problem of how to automatically generate multi-angle RGB images and a standard coordinate image dataset of ship outfitting parts to ensure image consistency, precise geometric alignment and matching of direction labels to support high-quality three-dimensional reconstruction tasks and deep learning model training.

[0006] In order to solve the above technical problems, the present invention provides a method for constructing a multi-angle RGB image and a standard coordinate image data set of ship outfitting parts, comprising:

[0007] Based on the deep learning model, the directional label analysis is performed on the three-dimensional model of the outfitting parts, and the three-dimensional model is normalized to generate the directional label and the default position parameter;

[0008] Based on the direction label and the default position parameter, the perspective configuration file is optimized, the perspective, focal length and position parameters of the camera are set, and a multi-perspective configuration of the outfitting part is generated;

[0009] Based on the viewing angle configuration file and the material characteristics of the outfitting parts, the intensity and distribution of the light source are adjusted, the material of the high-reflective area is optimized, and a uniform illumination distribution scheme is generated;

[0010] Based on the view configuration file and the illumination distribution scheme, cyclically render the RGB image and the standard coordinate image of the outfitting part, and verify the illumination consistency and geometric alignment in real time during the rendering process to generate preliminary verification data;

[0011] Performing a multimodal consistency check on the preliminary verification data and the rendered image, analyzing the geometric consistency and directional matching of the rendering results using a deep learning model, and generating a list of abnormal data;

[0012] Based on the rendered image and the orientation label, a standardized output file is generated, including an RGB image, a standard coordinate image, and an orientation label; and the diversity of the outfitting data set is expanded through a data enhancement module;

[0013] Comprehensive evaluation of the multi-view coverage and canonical coordinates of the dataset Figure 1 The evaluation results are used to optimize the direction label generation model and the view configuration file, forming a closed-loop improvement mechanism for data generation.

[0014] Furthermore, before the step of performing direction label analysis on the outfitting three-dimensional model based on the deep learning model, it also includes:

[0015] The three-dimensional model of the outfitting part is imported from a variety of standard three-dimensional model formats, and the geometric structure and material information of the model are extracted.

[0016] Furthermore, the step of performing direction label analysis based on the deep learning model specifically includes:

[0017] The geometric features of the three-dimensional model are analyzed to generate direction labels corresponding to the front, top and side of the model, and a normalized scale factor is generated according to the size of the model.

[0018] Furthermore, the step of optimizing the viewing angle configuration file specifically includes:

[0019] The six fixed viewing angles of the outfitting parts are determined by the direction tags, and the viewing angle, focal length and position of the camera are adjusted based on the size and geometric characteristics of the outfitting parts.

[0020] Furthermore, the step of adjusting the intensity and distribution of the light source specifically includes:

[0021] Based on the material properties of the outfitting parts, the position, intensity and angle of the light source are optimized to reduce overexposure problems in reflective areas and adjust the light distribution.

[0022] Furthermore, the step of cyclically rendering the RGB image and the standard coordinate image of the outfitting part specifically includes:

[0023] According to the view configuration file and the illumination distribution scheme, the RGB map and the standard coordinate map are rendered for each view using the automatic rendering control, and the illumination consistency and the geometric alignment are verified.

[0024] Furthermore, the step of multimodal consistency verification specifically includes:

[0025] A deep learning model is used to analyze the geometric consistency between images and the matching degree of orientation labels, generating abnormal data about perspective deviation, geometric inconsistency and orientation error.

[0026] Furthermore, the steps of the data enhancement module specifically include:

[0027] Lighting simulation, material transformation and geometry enhancement are performed on the rendered RGB map and the canonical coordinate map to generate a new dataset variant.

[0028] Furthermore, the multi-view coverage and standard coordinates of the comprehensive evaluation dataset Figure 1 The steps to ensure consistency include:

[0029] Calculate and evaluate the image coverage area at each viewpoint, verify the geometric alignment between views and the consistency of the canonical coordinate map, generate an evaluation report and feed it back to the orientation label generation model and viewpoint profile optimization.

[0030] Furthermore, a system for constructing a multi-angle RGB image and a standard coordinate image data set of ship outfitting parts includes:

[0031] A data acquisition module, which imports the three-dimensional model of the outfitting part from a variety of standard three-dimensional model formats, extracts the geometric structure and material information of the three-dimensional model, performs normalization processing, and generates direction labels and default position parameters;

[0032] A view configuration and parameter optimization module optimizes the view configuration file based on the direction label and the default position parameter, sets the view, focal length and position parameters of the camera, and generates a multi-view configuration file for the outfitting part;

[0033] The illumination and material optimization module adjusts the intensity and distribution of the light source according to the material properties and geometric features of the outfitting parts, optimizes the material of the high-reflective area, and generates a uniform illumination distribution plan;

[0034] A rendering and verification module, based on the viewing angle configuration file and the illumination distribution scheme, cyclically renders the RGB image and the standard coordinate image of the outfitting part, verifies the illumination consistency and geometric alignment in real time, and generates preliminary verification data;

[0035] A multimodal consistency check module performs a multimodal consistency check on the preliminary check data and the rendered image, uses a deep learning model to analyze the geometric consistency and directional matching of the rendering result, and generates an abnormal data list;

[0036] A data set generation and enhancement module generates an RGB image, a standard coordinate image and an orientation label based on the rendered image and the orientation label, and expands the diversity of the outfitting data set through a data enhancement module;

[0037] Closed-loop optimization and feedback module to comprehensively evaluate the multi-view coverage and standard coordinates of the dataset Figure 1 The evaluation results are used to optimize the directional label generation model and the perspective configuration file, forming a closed-loop improvement mechanism for data generation, promoting the continuous optimization of the system and the improvement of the data set quality.

[0038] The key innovative features of the present invention include:

[0039] (1) Automatic generation of multi-angle RGB images and standard coordinate images: Through automated viewing angle configuration and lighting optimization, the uniformity and consistency of multi-angle images under different conditions are ensured.

[0040] (2) Deep learning-assisted directional label analysis and geometric verification: The deep learning model is used to verify the geometric consistency and directional label matching of the rendering results, which improves the accuracy of the data.

[0041] (3) Closed-loop improvement mechanism: Through the evaluation and feedback optimization of the data set, a self-improving data generation closed loop is formed to ensure that data quality is continuously improved to meet the high standards required for 3D modeling and deep learning.

[0042] The following are its main beneficial effects: The present invention provides an automated method for constructing a multi-angle RGB image and a standard coordinate map data set of ship outfitting parts, which can efficiently generate a high-quality three-dimensional reconstruction data set. Compared with traditional manual modeling and rendering methods, the present invention introduces a deep learning model to analyze direction labels, automatically optimize the viewing angle configuration and camera parameters, and accurately control the lighting and materials, thereby greatly improving the consistency of the image, the geometric alignment, and the precise matching of direction labels, and reducing errors and inconsistencies in manual operations. In addition, through automated multi-view rendering and real-time verification mechanisms, the present invention can monitor the rendering quality in real time to ensure accurate matching of images from different viewing angles. The introduction of the data enhancement module further expands the diversity of the data set and improves the training effect and adaptability of the deep learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a process for constructing a multi-angle RGB image and a standard coordinate image data set of a ship outfitting part provided in an embodiment of the present application;

[0044] Figure 2 A structural block diagram of a method for constructing a multi-angle RGB image and a standard coordinate image data set of ship outfitting parts provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0046] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0047] Example 1: Reference Figure 1 , is a flow chart of an end-to-end low-latency image transmission method provided by an embodiment of the present invention, and the flow may at least include steps S100-S700:

[0048] S100: Based on the deep learning model, perform directional label analysis on the three-dimensional model of the outfitting parts, normalize the three-dimensional model, and generate directional labels and default position parameters.

[0049] S200, based on the direction label and the default position parameters, optimizing the viewing angle configuration file, setting the viewing angle, focal length and position parameters of the camera, and generating a multi-view configuration of the outfitting part.

[0050] S300: Based on the viewing angle configuration file and the material characteristics of the outfitting parts, the intensity and distribution of the light source are adjusted, the material of the high-reflective area is optimized, and a uniform illumination distribution scheme is generated.

[0051] S400, based on the viewing angle configuration file and the illumination distribution scheme, cyclically render the RGB image and the standard coordinate image of the outfitting parts, and verify the illumination consistency and geometric alignment in real time during the rendering process to generate preliminary verification data.

[0052] S500, performing a multimodal consistency check on the preliminary verification data and the rendered image, using a deep learning model to analyze the geometric consistency and directional matching of the rendering results, and generating an abnormal data list.

[0053] S600, based on the rendered image and the direction label, generate a standardized output file, including an RGB image, a standard coordinate image and a direction label; and expand the diversity of the outfitting data set through a data enhancement module.

[0054] S700: Comprehensively evaluate the multi-view coverage and standard coordinates of the data set Figure 1 The evaluation results are used to optimize the direction label generation model and the view configuration file, forming a closed-loop improvement mechanism for data generation.

[0055] Step S100 at least includes steps S110-S130:

[0056] S110: Obtaining an input three-dimensional model from an outfitting part three-dimensional model import module, performing coordinate normalization processing on the model, and obtaining a unitized outfitting part model.

[0057] First, the 3D model input data is obtained from the outfitting 3D model import module. The input data is a .obj or .gl b file that supports the standard 3D model format and contains the geometric structure and material information of the outfitting.

[0058] Specifically, the input 3D model is loaded into a default rendering environment, and its vertex coordinate data is parsed to extract the boundary range of the model, including the minimum boundary point (min_x, min_y, min_z) and the maximum boundary point (max_x, max_y, max_z).

[0059] The size ratio of the outfitting model is calculated based on the above boundary range, and the maximum boundary length is normalized to unit length. At the same time, the center position of the model is adjusted to the coordinate origin (0,0,0) to form a unitized outfitting model. During the unitization process, the generated scale factor and center offset value are recorded for subsequent steps.

[0060] S120: Using a deep learning model to perform directional analysis on the normalized geometric features of the outfitting model to generate a directional label for the outfitting.

[0061] The vertex coordinates and center offset values ​​of the unitized outfitting model are obtained from the normalized output of S110 as input data for the deep learning model.

[0062] The pre-trained geometric feature direction classification model is used to analyze the structural features of the normalized outfitting model. The model determines the front direction, top direction and side direction of the model by analyzing the geometric shape, vertex distribution and edge characteristics of the outfitting.

[0063] Furthermore, the direction information is marked as a direction label (front, top, side), and is bound to the normalization model and output in the form of a label to provide a direction correction basis for subsequent modules.

[0064] S130: Based on the direction label, the outfitting model is adjusted to a default position and direction, and the processed three-dimensional model and the direction label are output.

[0065] The orientation label of the outfitting part is obtained from S120, and the orientation and position of the outfitting part model are adjusted in combination with the center offset value of the normalized output.

[0066] Specifically, according to the direction label, the front direction of the outfitting model is corrected to the positive direction of the x-axis, the upper direction is corrected to the positive direction of the z-axis, and the model is adjusted to the origin (0,0,0) through the offset value.

[0067] The processed outfitting model and its orientation label are output as input data of the S200 module. The output includes a corrected three-dimensional geometric model, a normalized scale factor, an orientation label, and a position offset parameter.

[0068] Description of the connection between modules: S110→S120: The unitized outfitting model and normalized scale factor generated by S110 are directly input into S120 for geometric feature analysis and direction label generation.

[0069] S120→S130: The direction label generated by S120 is used as the key input of S130, and the direction label determines the rotation correction direction of the model in the three-dimensional space.

[0070] S130→S200: The correction model and direction label outputted from S130 are used as basic data for optimizing the viewing angle configuration in S200, and the direction information is used to determine specific settings of the six fixed viewing angles.

[0071] Through the above process, each step of the S100 module has a clear input and output relationship, forming a complete logical closed loop.

[0072] Step S200 at least includes steps S210-S230:

[0073] S210: Obtain direction labels and normalized models from the S100 module, and set six fixed viewing angles (front, back, left, right, top, and bottom) of the outfitting model.

[0074] The normalized data and direction labels of the outfitting model are obtained from the S100 module, wherein the normalized data includes the unitized model size and its center position, and the direction labels include the front, top and side directions.

[0075] Specifically, the positions of six fixed viewing angles are determined based on the direction labels, wherein the front view corresponds to the front side of the model, the rear view corresponds to the back side of the model, the left view and the right view correspond to the side directions of the model respectively, and the top view and the bottom view correspond to the top and the bottom of the model respectively.

[0076] Furthermore, according to the size of the normalized model, the coordinate value and orientation of the camera in the three-dimensional space are preliminarily set to ensure that the camera's viewing angle can cover the entire model, providing basic position parameters for subsequent viewing angle optimization.

[0077] S220: Based on the size characteristics and orientation labels of the outfitting parts, the focal length, position and viewing angle parameters of the camera are optimized to ensure uniformity of multi-view coverage.

[0078] The preliminary parameters of six fixed view angles are obtained from the S210 module, including the initial position and orientation of the camera and the size range of the normalized model.

[0079] Specifically, according to the maximum boundary length of the outfitting model, the camera focal length is optimized to ensure that the complete outline of the model is included in the viewing angle, and the distance between the camera and the model is appropriately adjusted to avoid occlusion or viewing angle deformation.

[0080] Furthermore, the camera’s orientation angle is corrected based on the direction label so that the line of sight of each perspective is perpendicular to the model surface, ensuring that the projection of the model in the six perspectives is evenly distributed, providing parameter support for subsequent rendering consistency.

[0081] S230: Generate a multi-view configuration file, including view name, camera position, focal length and angle parameters, as input for subsequent modules.

[0082] Get the optimized camera parameters from the S220 module, including the names of the six viewing angles, the camera's 3D spatial position, focal length, orientation angle, and viewing angle range.

[0083] Specifically, the parameter information of each perspective is stored in the form of structured data as a multi-perspective configuration file, which includes the perspective name (such as front view, rear view, etc.), camera position coordinates (x, y, z), focal length (unit length) and rotation angle value.

[0084] The multi-view configuration file is output to provide an input basis for the lighting optimization and rendering modules, ensuring the consistency of the multi-view data and the reproducibility of the parameters.

[0085] Description of the connection between modules: S210→S220: The six preliminary parameters of fixed viewing angles generated by S210, including the camera position and direction, are directly input into S220 for further optimization of the viewing angle range and focal length.

[0086] S220→S230: The camera focal length, position and viewing angle parameters optimized by S220 are used as the core data for generating a configuration file by S230.

[0087] S230→S300: The multi-view configuration file generated by S230 is directly input into the S300 module to determine the illumination distribution and rendering viewing angle.

[0088] Through the above steps, the S200 module forms a close connection from the acquisition of direction labels to the generation of view configuration files, and provides clear data input support for subsequent modules.

[0089] Step S300 at least includes steps S310-S330:

[0090] S310: Obtain the view configuration file from the S200 module, analyze the material properties of the outfitting model, and determine the initial value of the light source distribution and intensity.

[0091] The camera position, orientation and focal length parameters of six fixed viewing angles are parsed from the viewing angle configuration file obtained from the S200 module as input conditions for light source distribution analysis.

[0092] From the normalized model of the outfitting parts, its material property data is further analyzed, including the surface material type (such as metal, composite material) and its reflection characteristic parameters.

[0093] Based on the geometric relationship between the material properties and the viewing angle configuration file, the distribution position and intensity of the light source are preliminarily determined to ensure that the model is evenly illuminated at six fixed viewing angles, providing basic parameters for subsequent optimization steps.

[0094] S320: Based on the geometric characteristics of the outfitting model and the material reflective parameters, the light source angle and intensity are optimized to reduce overexposure and shadow problems.

[0095] The initial distribution parameters of the light source and the geometric characteristic data of the outfitting model are obtained from the S310 module, and the shading relationship between the complex geometric structure of the model (such as hollow design and protruding parts) and the light source is analyzed.

[0096] Specifically, the angle and intensity distribution of the light source are adjusted according to the reflective characteristic parameters of the model surface to ensure that the light can evenly cover the model surface and avoid excessive light intensity in the overexposed area and insufficient light intensity in the shadow area.

[0097] Furthermore, combined with the reflectivity of the material, special lighting compensation is performed on the metal area to reduce the impact of metal highlight reflection on the consistency of RGB images, while optimizing the projection effect of light in the hollow area.

[0098] S330: Generate a lighting distribution scheme file, including light source position, intensity and angle parameters, and output it to the rendering module

[0099] The optimized light source position, intensity and angle parameters are obtained from the S320 module and integrated into a structured lighting distribution solution file, which contains light source arrangement data corresponding to six fixed viewing angles.

[0100] Specifically, the lighting distribution scheme file contains the following contents:

[0101] The light source position coordinates (x,y,z) for each fixed view angle.

[0102] The intensity value of the light source is adapted to the material characteristics and geometric distribution of the model.

[0103] The irradiation angle and light diffusion range parameters of the light source.

[0104] Output the illumination distribution scheme file to the rendering module to ensure that subsequent rendering can generate high-quality RGB images and standard coordinate images based on the optimized lighting environment.

[0105] Description of the connection between modules: S310→S320: The initial distribution and intensity parameters of the light source determined by S310 are directly used as the input conditions for optimizing the angle and intensity of the light source in S320.

[0106] S320→S330: The light source position, angle and intensity parameters optimized by S320 are used as the core data for generating the illumination distribution scheme file in S330.

[0107] S330→S400: The illumination distribution scheme file generated by S330 is directly input into the S400 module to guide the rendering of multi-view RGB images and standard coordinate maps.

[0108] Through the above steps, the S300 module realizes the complete process from initial light source distribution analysis to light source parameter optimization and then to lighting plan file generation, and forms a close connection with the previous and next modules.

[0109] Step S400 at least includes steps S410-S430:

[0110] S410: Obtain the view configuration file and the lighting distribution plan from the S200 and S300 modules respectively, load the outfitting model and initialize the rendering environment.

[0111] The viewing angle configuration file is obtained from the S200 module, wherein the configuration file includes the camera position, focal length and rotation angle information of six fixed viewing angles.

[0112] The lighting distribution scheme is obtained from the S300 module, and the scheme includes light source position, intensity and angle parameters as a basis for lighting arrangement in the rendering environment.

[0113] Specifically, the normalized 3D model of the outfitting is loaded into the rendering environment, the camera parameters are set based on the perspective configuration file, and the light source is initialized according to the lighting distribution scheme to ensure that the perspective and lighting environment of the model are consistent with the previous steps.

[0114] S420: Utilize automatic loop rendering control to render the RGB image and the standard coordinate image for each viewing angle, and verify the lighting consistency and geometric alignment in real time.

[0115] Call the automatic rendering control script, loop through the six fixed perspectives in the perspective configuration file, and load the corresponding camera parameters and lighting solutions one by one.

[0116] At each viewing angle, the RGB image of the outfitting is rendered first, and the surface color, light and shadow, and texture information of the outfitting are specifically mapped to the image output; then the standard coordinate map is rendered through coordinate mapping technology to ensure that the three-dimensional position of the outfitting is accurately converted into unit cube coordinates.

[0117] Furthermore, the consistency of the lighting distribution and the geometric alignment are checked in real time during each rendering process, including:

[0118] Verify the intensity distribution of the light source at six viewing angles to determine whether there are overexposed or underlit areas.

[0119] Verify the alignment of geometric feature points of outfitting parts in rendered images to ensure consistent image projection from different viewing angles.

[0120] S430: Generate a rendered RGB image and a standard coordinate image, as well as preliminary verification data (such as illumination distribution and alignment information)

[0121] The six rendered view RGB images and standard coordinate images are obtained from the S420 module and stored in the specified path according to the view naming rules to ensure the correspondence between the image and the view.

[0122] Specifically, preliminary verification data is generated according to the real-time verification result, and the data includes illumination distribution information (such as intensity value distribution) of each viewing angle, geometric alignment measurement values, and integrity check records of the rendered image.

[0123] The RGB image, standard coordinate image and preliminary verification data are output to the next module for subsequent verification and data management.

[0124] Description of the connection between modules: S410→S420: The model loaded by S410 and the rendering environment initialization parameters (view configuration file and lighting distribution plan) are directly used as the input conditions for the automatic loop rendering of S420.

[0125] S420→S430: The RGB image, the standard coordinate image and the real-time verification result rendered by S420 are directly used in S430 to generate preliminary verification data.

[0126] S430→S500: The preliminary verification data and image files generated in S430 are used as input for further analysis by the subsequent multi-modal consistency verification module.

[0127] Through the above steps, the S400 module completes the rendering environment initialization, automatic loop rendering and preliminary verification data generation based on the perspective configuration file and lighting scheme, and realizes data flow and logical connection with subsequent modules.

[0128] Step S500 at least includes steps S510-S530:

[0129] S510: Obtain preliminary verification data and rendered image sequence from S400 module, perform multi-view geometric feature matching, and verify consistency between images.

[0130] Preliminary calibration data is obtained from the S400 module, including illumination distribution information for each viewing angle, geometric alignment measurement values, and RGB images and standard coordinate images generated by rendering.

[0131] The RGB image and the standard coordinate image are loaded, and based on the multi-view geometric feature matching technology, the spatial consistency of the corresponding geometric feature points in the six view images is compared one by one. Specifically, the feature point distribution of the model at different view angles is extracted through the feature point matching algorithm.

[0132] Furthermore, the geometric feature alignment between the six views is verified, and a preliminary consistency verification report is generated, including the feature offset value and image matching quality score between each pair of views.

[0133] S520: Analyze the matching degree between the direction label and the image sequence using a deep learning model to determine whether there is an abnormal viewing angle or direction deviation.

[0134] The feature offset value and image matching score are obtained from the consistency verification report and preliminary verification data output by the S510 module, and combined with the direction label generated by the S400 module.

[0135] Load the pre-trained deep learning model, input the image sequences of six perspectives and their direction labels, analyze the matching degree between each perspective image and the direction label, and specifically detect whether there is a direction error or perspective offset problem.

[0136] Furthermore, based on the analysis results of the deep learning model, abnormal viewing angles and potential directional deviations are marked, including the size of the offset angle, the type of error (such as viewing angle offset, lighting anomaly, etc.), and the degree of inconsistency with the directional label.

[0137] S530: Output an abnormal data list, including abnormal viewing angle, direction deviation and geometric consistency problem description.

[0138] The verification results of the S510 and S520 modules are integrated to generate an abnormal data list, which includes the following contents:

[0139] Abnormal view angle name (such as front view, left view, etc.).

[0140] Description of the geometric consistency problem, including the offset of feature points between views and the matching score.

[0141] Direction deviation description, including the offset angle and deviation type between the viewing angle and direction label.

[0142] Specifically, the above abnormal data is formatted into a structured file, including the image file name, matching score and offset details of the abnormal perspective, and output to the subsequent module for further verification and repair.

[0143] Synchronize the abnormal data list to the data management module and record the verification log to provide a traceability basis for data set quality management.

[0144] Description of the connection between modules: S510→S520: The geometric feature matching verification report and consistency score generated by S510 are directly used as the input conditions for the S520 deep learning model to analyze the direction deviation.

[0145] S520→S530: The abnormal viewing angle and direction deviation analysis results marked in S520 are used to generate an abnormal data list in S530.

[0146] S530→S600: The abnormal data list and verification log outputted from S530 are used as input for optimization and processing of subsequent data management and enhancement modules.

[0147] Through the above steps, the S500 module fully realizes the closed-loop verification process from preliminary verification data loading, multi-view feature matching, deep learning direction analysis to abnormal data output, and maintains a close connection with the previous and next modules.

[0148] Step S600 at least includes steps S610-S630:

[0149] S610: Integrate the rendering image generated by S400 and the direction label generated by S100 to generate a standardized output file (including an RGB image, a standard coordinate image and direction information).

[0150] Obtain rendered image sequences of six perspectives from the S400 module, including RGB images and canonical coordinate images; and obtain corresponding direction labels from the S100 module.

[0151] Specifically, the RGB image, the standard coordinate image and the direction label are integrated into a standardized file, which uses the unique identifier (UID) of the outfitting part as an index and records the image file name and direction information corresponding to each viewing angle.

[0152] Furthermore, the rendered images are stored according to the naming rules, where the RGB image is named rgb_<viewing angle number>.png, the standard coordinate image is named nocs_<viewing angle number>.png, and the direction label and outfitting classification information are added to the associated file.

[0153] S620: Use the data enhancement module to perform diversified processing on the RGB image and the canonical coordinate image, including lighting simulation and material transformation, to expand the diversity of the data set.

[0154] Load the RGB image and the canonical coordinate map from the standardized output file generated by the S610 module and input them into the data enhancement module for diversification processing.

[0155] Specifically, based on lighting simulation technology, the intensity and distribution of the light source in the RGB image are adjusted to generate image variants under different lighting conditions; based on material transformation technology, the surface material properties of the outfitting parts (such as metallic gloss and surface roughness) are modified to generate images in multi-material scenes.

[0156] Furthermore, the geometric shape enhancement algorithm is used to perform random scaling and rotation transformations on the canonical coordinate map to generate a variant canonical coordinate map, ensuring that the enhanced dataset can cover more training scenario requirements.

[0157] S630: Output the final data set file, including the enhanced multi-view image, the canonical coordinate map, and the direction label.

[0158] Obtain the enhanced RGB image and canonical coordinate map from the S620 module, as well as the direction label obtained from the S610 module, and integrate them into the final dataset file.

[0159] Specifically, the enhanced image is stored in a specified path according to the standardized naming rules, and an associated file is generated to record the UI of each outfitting part, the enhanced variant file name, the orientation label, and the enhancement type (such as lighting conditions, material transformation).

[0160] Output the final dataset file, including multi-view enhanced images, standard coordinate maps, and orientation labels. At the same time, upload the dataset file to the data management system for subsequent deep learning model calls and verification.

[0161] Description of the connection between modules: S610→S620: The standardized file (RGB image, standard coordinate map and direction label) generated by S610 is used as the input condition of the S620 data enhancement module.

[0162] S620→S630: The enhanced image and variant information of S620 are directly input into the S630 module to generate the final dataset file.

[0163] S630→S700: The final data set file and its description information output by S630 serve as the core input for the quality evaluation of the S700 module.

[0164] Through the above steps, the S600 module realizes the complete process from the integration of rendered images and orientation labels, data enhancement processing to the final data set output, forming a strict logical connection with the previous and next modules.

[0165] Step S700 at least includes steps S710-S730:

[0166] S710: Obtain the generated data set from the S600 module, and analyze the consistency between the multi-view coverage and the standard coordinate map.

[0167] The generated dataset, including the enhanced multi-view RGB image, canonical coordinate map, and orientation labels, is obtained from the S600 module and loaded into the evaluation module for coverage and consistency analysis.

[0168] Specifically, based on the view coverage analysis algorithm, the RGB images of six fixed view angles are evaluated to detect whether there are any missing geometric features between view angles, and further analyze whether the requirements for full view coverage are met.

[0169] At the same time, the geometric feature points of the outfitting parts are extracted from the specification coordinate diagram, their consistency in multi-view projection is verified, and potential inconsistent areas are recorded, such as feature point alignment errors caused by lighting or rendering deviations.

[0170] S720: Based on the evaluation results, feedback optimization is performed on the direction label generation model and the view configuration file, and the model and configuration parameters are updated.

[0171] Obtain the view coverage analysis results and the consistency deviation data of the specification coordinate diagram from the evaluation report output by the S710 module.

[0172] Based on the deviation data, the directional label generation model is further optimized, specifically including adjusting the geometric feature extraction strategy of the model to reduce the problem of insufficient multi-view coverage caused by inaccurate directional label generation.

[0173] At the same time, the perspective configuration file is optimized based on the evaluation results, and the camera focal length, rotation angle and position parameters are updated to ensure more uniform perspective coverage in subsequent rendering and reduce the omission of geometric features caused by improper camera configuration.

[0174] S730: Complete the closed-loop optimization and apply the optimized parameters and models to the next round of data set construction process.

[0175] The optimized parameters and updated direction label generation model output from the S720 module are loaded into the data generation process for the next round of data set construction.

[0176] Specifically, the optimized viewing angle configuration file is input into the S200 module to reset the viewing angle, focal length and position parameters of the camera; the optimized direction label generation model is applied to the S100 module to improve the accuracy of direction label generation.

[0177] Output the model and parameter version records after closed-loop optimization to ensure that the next round of data set generation process is based on the latest optimization results, and record feedback logs to track the optimization effect.

[0178] Inter-module connection description: S710 → S720: S710’s visual angle coverage evaluation results and standard coordinates Figure 1 The consistency deviation data is directly used as the input condition for the S720 model and configuration file feedback optimization.

[0179] S720→S730: The model and parameters optimized in S720 are used as input for the closed-loop optimization of S730 to ensure the update and application of the next round of data generation process.

[0180] S730→S100 / S200: The optimization results of S730 react to the S100 direction label generation module and the S200 view configuration module, forming a closed-loop improvement of data generation.

[0181] Through the above steps, the S700 module realizes a complete logical chain from data set evaluation, optimization feedback to closed-loop process update, and effectively connects data flow and improvement mechanisms with the previous and next modules.

[0182] The key innovative features of the present invention include:

[0183] (1) Automatic generation of multi-angle RGB images and standard coordinate images: Through automated viewing angle configuration and lighting optimization, the uniformity and consistency of multi-angle images under different conditions are ensured.

[0184] (2) Deep learning-assisted directional label analysis and geometric verification: The deep learning model is used to verify the geometric consistency and directional label matching of the rendering results, which improves the accuracy of the data.

[0185] (3) Closed-loop improvement mechanism: Through the evaluation and feedback optimization of the data set, a self-improving data generation closed loop is formed to ensure that data quality is continuously improved to meet the high standards required for 3D modeling and deep learning.

[0186] The following are its main beneficial effects: The present invention provides an automated method for constructing a multi-angle RGB image and a standard coordinate map data set of ship outfitting parts, which can efficiently generate a high-quality three-dimensional reconstruction data set. Compared with traditional manual modeling and rendering methods, the present invention introduces a deep learning model to analyze direction labels, automatically optimize the viewing angle configuration and camera parameters, and accurately control the lighting and materials, thereby greatly improving the consistency of the image, the geometric alignment, and the precise matching of direction labels, and reducing errors and inconsistencies in manual operations. In addition, through automated multi-view rendering and real-time verification mechanisms, the present invention can monitor the rendering quality in real time to ensure accurate matching of images from different viewing angles. The introduction of the data enhancement module further expands the diversity of the data set and improves the training effect and adaptability of the deep learning model.

[0187] Embodiment 2: Figure 2 FIG. 1 is a structural block diagram of a method for constructing a multi-angle RGB image and a standard coordinate image data set of a ship outfitting according to an embodiment of the present invention. Figure 2 As shown, the structure may include:

[0188] The data acquisition module 10 is responsible for extracting raw data from the three-dimensional model of the outfitting parts and converting it into a format that can be used for subsequent processing. Specific functions include:

[0189] Import outfitting 3D models from a variety of standard 3D model formats (such as .obj and .glb) to ensure data diversity and extensiveness.

[0190] The position of the three-dimensional model is normalized, the outfitting model is adjusted to a unified coordinate system, and the direction label of the model (such as front, top, side, etc.) is determined.

[0191] Input the processed model and material information into subsequent modules to ensure data consistency and accuracy.

[0192] The view configuration and parameter optimization module 20 is responsible for setting appropriate view angles and camera parameters according to the three-dimensional model of the outfitting parts and its direction labels obtained from the data acquisition module 10 to ensure uniform coverage of the model under different view angles. The specific functions of this module include:

[0193] Six fixed viewing angles (front, back, left, right, top, and bottom) are set based on the direction tag, and the corresponding camera position, focal length, and rotation angle are configured for each viewing angle.

[0194] The distribution of viewing angles is optimized based on the geometric characteristics and model size of the outfitting parts to ensure that all viewing angles can fully cover the geometric characteristics of the outfitting parts.

[0195] Generate an optimized view configuration file for input into subsequent rendering modules.

[0196] The lighting and material optimization module 30 is responsible for adjusting the distribution and intensity of the light source according to the material properties and geometric features of the outfitting parts, optimizing the rendering quality, and ensuring the consistency of the image. Specific functions include:

[0197] Analyze the material properties of outfitting parts (such as metal surface reflection, surface roughness, etc.), and optimize the light intensity and light source position according to the material properties.

[0198] Special treatment is given to highly reflective areas and complex geometric structures (such as hollow structures) to avoid overexposure or insufficient shadows.

[0199] Generate a lighting distribution plan and output it to the subsequent rendering module to ensure the uniformity and consistency of the light source during the rendering process.

[0200] The rendering and verification module 40 completes the rendering of the RGB image and the standard coordinate image of the outfitting parts according to the aforementioned view configuration file and the illumination distribution scheme, and verifies the rendering results in real time to ensure the image quality and geometric consistency. The specific functions of this module include:

[0201] Call the automated rendering control system to loop through rendering the RGB map and the standard coordinate map based on the optimized viewing angle profile and lighting distribution scheme.

[0202] During the rendering process, lighting consistency and geometric alignment are checked in real time to ensure that images from different perspectives have consistent lighting distribution and geometric features.

[0203] Output rendering results and preliminary verification data for further verification by subsequent verification modules.

[0204] The multimodal consistency check module 50 is responsible for performing consistency check between the rendered image and the standard coordinate map, and uses a deep learning algorithm to analyze the geometric alignment between the images and the matching degree of the direction labels. The functions of this module include:

[0205] Feature points are extracted from the image sequence obtained by the rendering and verification module 40 to perform geometric consistency verification.

[0206] A deep learning model is used to analyze the matching between the orientation label and the image sequence to determine whether there is an abnormal viewing angle or orientation deviation.

[0207] Output a list of abnormal data, including abnormal viewing angles, directional deviations, and geometric consistency issues, for subsequent data correction and optimization.

[0208] The dataset generation and enhancement module 60 is responsible for integrating rendered images, canonical coordinate maps, and orientation labels, and expanding the diversity of the dataset through data enhancement. Functions include:

[0209] Integrate the rendered images with the corresponding orientation labels and standard coordinate maps into a standardized dataset file to ensure the consistency of the data format.

[0210] Diversified training data is generated through data enhancement methods such as lighting simulation, material transformation, and geometric transformation to improve the diversity and adaptability of the data set.

[0211] Output the final dataset file for deep learning model training and further use.

[0212] The closed-loop optimization and feedback module 70 ensures continuous optimization of the system by continuously monitoring and evaluating each module in the process of building the data set. The functions of this module include:

[0213] The final dataset is obtained from the dataset generation and enhancement module 60, and the consistency of the multi-view coverage and the canonical coordinate map is evaluated.

[0214] Based on the evaluation results, feedback is provided to optimize the orientation label generation model and view profile, ensuring that the next round of dataset generation can provide higher quality images and annotations.

[0215] Complete closed-loop optimization and apply the optimized parameters and models to the next round of data set construction process to achieve continuous improvement of the system.

[0216] Through the ship outfitting multi-angle RGB image and standard coordinate map data set construction method system provided by the present invention, a high-quality multi-view data set that meets the needs of deep learning training can be generated efficiently and accurately.

[0217] The system has significant beneficial effects in the following aspects:

[0218] 1. Automated data generation: Through automated perspective configuration, lighting optimization, material rendering and other steps, manual intervention is reduced and the efficiency and consistency of data generation are improved.

[0219] 2. High-precision verification: The multimodal consistency verification module uses deep learning technology to accurately verify the rendering results to ensure the geometric alignment of the image and the consistency of the direction labels.

[0220] 3. Data diversity enhancement: Through the data enhancement module, a diverse data set containing different lighting, materials, and perspective changes is generated, which improves the robustness of the model.

[0221] 4. Closed-loop optimization mechanism: The directional label generation model and perspective profile are optimized based on the feedback of the dataset evaluation results, realizing a closed-loop improvement mechanism for data generation and continuously improving data quality.

[0222] The system not only supports deep learning tasks for 3D reconstruction of ship outfitting parts, but also has strong scalability and can be applied to 3D reconstruction and model training tasks in other fields.

[0223] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for constructing a multi-angle RGB image and a standard coordinate image dataset of ship outfitting parts, characterized in that: The steps include: Based on the deep learning model, the directional label analysis is performed on the three-dimensional model of the outfitting parts, and the three-dimensional model is normalized to generate the directional label and the default position parameter; Based on the direction label and the default position parameter, the perspective configuration file is optimized, the perspective, focal length and position parameters of the camera are set, and a multi-perspective configuration of the outfitting part is generated; Based on the viewing angle configuration file and the material characteristics of the outfitting parts, the intensity and distribution of the light source are adjusted, the material of the high-reflective area is optimized, and a uniform illumination distribution scheme is generated; Based on the view configuration file and the illumination distribution scheme, cyclically render the RGB image and the standard coordinate image of the outfitting part, and verify the illumination consistency and geometric alignment in real time during the rendering process to generate preliminary verification data; Performing a multimodal consistency check on the preliminary verification data and the rendered image, analyzing the geometric consistency and directional matching of the rendering results using a deep learning model, and generating a list of abnormal data; Based on the rendered image and the orientation label, a standardized output file is generated, including an RGB image, a standard coordinate image, and an orientation label; And expand the diversity of outfitting data sets through data enhancement modules; The multi-view coverage and the consistency of the canonical coordinate map of the dataset are comprehensively evaluated, and the evaluation results are used to optimize the direction label generation model and the view configuration file to form a closed-loop improvement mechanism for data generation.

2. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: Before the step of performing direction label analysis on the three-dimensional model of the outfitting parts based on the deep learning model, the method further includes: The three-dimensional model of the outfitting part is imported from a variety of standard three-dimensional model formats, and the geometric structure and material information of the model are extracted.

3. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The step of performing direction label analysis based on the deep learning model specifically includes: The geometric features of the three-dimensional model are analyzed to generate direction labels corresponding to the front, top and side of the model, and a normalized scale factor is generated according to the size of the model.

4. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The step of optimizing the viewing angle configuration file specifically includes: The six fixed viewing angles of the outfitting parts are determined by the direction tags, and the viewing angle, focal length and position of the camera are adjusted based on the size and geometric characteristics of the outfitting parts.

5. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The step of adjusting the intensity and distribution of the light source specifically includes: Based on the material properties of the outfitting parts, the position, intensity and angle of the light source are optimized to reduce overexposure problems in reflective areas and adjust the light distribution.

6. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The step of cyclically rendering the RGB image and the standard coordinate image of the outfitting part specifically includes: According to the view configuration file and the illumination distribution scheme, the RGB map and the standard coordinate map are rendered for each view using the automatic rendering control, and the illumination consistency and the geometric alignment are verified.

7. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The multi-modal consistency check step specifically includes: A deep learning model is used to analyze the geometric consistency between images and the matching degree of orientation labels, generating abnormal data about perspective deviation, geometric inconsistency and orientation error.

8. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The steps of the data enhancement module specifically include: Lighting simulation, material transformation and geometry enhancement are performed on the rendered RGB map and the canonical coordinate map to generate a new dataset variant.

9. The method for constructing a data set of multi-angle RGB images and standard coordinate images of ship outfitting parts according to claim 1, characterized in that: The step of comprehensively evaluating the consistency between the multi-view coverage of the data set and the standard coordinate map specifically includes: Calculate and evaluate the image coverage area at each viewpoint, verify the geometric alignment between views and the consistency of the canonical coordinate map, generate an evaluation report and feed it back to the orientation label generation model and viewpoint profile optimization.

10. A system for constructing a multi-angle RGB image and a standard coordinate image data set of ship outfitting parts, characterized in that: include: A data acquisition module, which imports the three-dimensional model of the outfitting part from a variety of standard three-dimensional model formats, extracts the geometric structure and material information of the three-dimensional model, performs normalization processing, and generates direction labels and default position parameters; A view configuration and parameter optimization module optimizes the view configuration file based on the direction label and the default position parameter, sets the view, focal length and position parameters of the camera, and generates a multi-view configuration file for the outfitting part; The illumination and material optimization module adjusts the intensity and distribution of the light source according to the material properties and geometric features of the outfitting parts, optimizes the material of the high-reflective area, and generates a uniform illumination distribution plan; A rendering and verification module, based on the viewing angle configuration file and the illumination distribution scheme, cyclically renders the RGB image and the standard coordinate image of the outfitting part, verifies the illumination consistency and geometric alignment in real time, and generates preliminary verification data; A multimodal consistency check module performs a multimodal consistency check on the preliminary check data and the rendered image, uses a deep learning model to analyze the geometric consistency and directional matching of the rendering result, and generates an abnormal data list; A data set generation and enhancement module generates an RGB image, a standard coordinate image and an orientation label based on the rendered image and the orientation label, and expands the diversity of the outfitting data set through a data enhancement module; The closed-loop optimization and feedback module comprehensively evaluates the multi-view coverage and consistency of the standard coordinate map of the dataset, and uses the evaluation results to optimize the direction label generation model and view configuration file, forming a closed-loop improvement mechanism for data generation, promoting continuous optimization of the system and improvement of dataset quality.

Images