Three-dimensional surface flattening methods, devices, electronic equipment, media and products

Abstract

This invention relates to the field of image processing technology, and particularly to a three-dimensional surface flattening method, apparatus, electronic device, medium, and product. The method includes the following steps: acquiring multi-view surface images of a target object and reconstructing a three-dimensional dense point cloud; after preprocessing, flattening a reference plane by plane fitting and aligning the point cloud to its coordinate system to obtain a unified coordinate point cloud; constructing a discrete point cloud mesh model and compressing it to the reference plane to obtain a particle displacement field; and finally generating a two-dimensional flattened image of the target object based on this displacement field and the target three-dimensional dense point cloud. This invention solves the problems of incomplete display of two-dimensional flattened images and low flattening efficiency when processing surfaces with severe wrinkles, tears, or complex curvatures, thereby improving the accuracy of two-dimensional images of three-dimensional surface flattening and increasing the efficiency of three-dimensional flattening.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a three-dimensional surface flattening method, apparatus, electronic device, medium, and product. Background Technology

[0002] In fields such as document digitization, cultural relic restoration, and medical image analysis, it is often necessary to flatten the texture information on a three-dimensional curved surface onto a two-dimensional plane without loss.

[0003] In related technologies, the main methods for flattening include acquiring existing 3D models of the target object or using models based on the generalized cylindrical surface assumption, deep learning methods, and traditional mesh-based flattening based on a single viewpoint of the target object.

[0004] However, when dealing with surfaces that are severely wrinkled, torn, or have complex curvature, there are often problems such as overlapping, broken, loss of detail, or low computational efficiency after flattening, which urgently need to be solved. Summary of the Invention

[0005] This invention provides a three-dimensional surface flattening method, apparatus, electronic device, medium, and product to solve the problems of incomplete two-dimensional flattening image display and low flattening efficiency when dealing with surfaces with severe wrinkles, tears, or complex curvatures, thereby improving the accuracy of two-dimensional images for three-dimensional surface flattening and increasing three-dimensional flattening efficiency.

[0006] To achieve the above objectives, a first aspect of the present invention provides a three-dimensional surface flattening method, comprising the following steps: Surface images of the target object from multiple perspectives are acquired, and a three-dimensional dense point cloud is reconstructed based on the surface images from multiple perspectives. The three-dimensional dense point cloud is preprocessed to obtain a preprocessed point cloud, and a flattening reference plane is obtained by performing plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm. The three-dimensional dense point cloud is aligned to the plane coordinate system of the flattening reference plane to obtain a three-dimensional point cloud with unified coordinates. Based on the three-dimensional point cloud with unified coordinates, a discrete point cloud mesh model is constructed according to a preset discretization algorithm, and the point cloud mesh model is compressed to the flattening reference plane according to a preset surface mesh compression method to obtain the particle displacement field during the compression process. Based on the particle displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated.

[0007] Furthermore, in some embodiments, the construction of a discrete point cloud mesh model based on the three-dimensional point cloud with the unified coordinates according to a preset discretization algorithm includes: constructing a triangular mesh model based on the three-dimensional point cloud with the unified coordinates according to the preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the three-dimensional point cloud with the unified coordinates; obtaining virtual charge particles corresponding to each vertex in the triangular mesh model, constructing a Coulomb-Hooke particle system based on the virtual charge particles corresponding to each vertex, and obtaining the discrete point cloud mesh model based on the Coulomb-Hooke particle system.

[0008] Furthermore, in some embodiments, each particle interacts with the others through Coulomb repulsion.

[0009] Furthermore, in some embodiments, generating a two-dimensional flattened image of the target object based on the particle displacement field and the target three-dimensional dense point cloud includes: interpolating the particle displacement field and the target three-dimensional dense point cloud according to a preset interpolation algorithm to obtain a flattened two-dimensional point cloud; projecting the flattened two-dimensional point cloud onto a flattened reference plane to obtain an initial two-dimensional image; and repairing the initial two-dimensional image based on a pre-constructed image inpainting neural network to generate a two-dimensional flattened image of the target object.

[0010] Furthermore, in some embodiments, before compressing the point cloud mesh model to the flattened reference plane based on the preset curved mesh compression method, the method further includes: eliminating folded point clouds and overlapping point clouds in the discrete point cloud mesh model based on Coulomb repulsion force and Hooke support force until the discrete point cloud mesh model meets the preset point cloud mesh model flattening condition, thereby obtaining the point cloud mesh model.

[0011] The three-dimensional surface flattening method provided by the present invention acquires multi-view surface images of a target object and reconstructs a three-dimensional dense point cloud. After preprocessing, planar fitting to a flattening reference plane, and aligning the point cloud to its coordinate system to obtain a unified coordinate point cloud, a discrete point cloud mesh model is constructed and compressed to the reference plane to obtain a particle displacement field. Finally, a two-dimensional flattened image of the target object is generated based on this displacement field and the target three-dimensional dense point cloud. This method solves the problems of incomplete display of two-dimensional flattened images and low flattening efficiency when dealing with surfaces with severe wrinkles, tears, or complex curvatures in related technologies, improves the accuracy of two-dimensional images of three-dimensional surface flattening, and increases the efficiency of three-dimensional flattening.

[0012] To achieve the above objectives, a second aspect of the present invention provides a three-dimensional surface flattening device, comprising: an acquisition module, configured to acquire surface images of a target object from multiple perspectives, and reconstruct a three-dimensional dense point cloud based on the surface images from the multiple perspectives; a processing module, configured to preprocess the three-dimensional dense point cloud to obtain a preprocessed point cloud, and perform plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm to obtain a flattening reference plane, aligning the three-dimensional dense point cloud to the plane coordinate system of the flattening reference plane to obtain a three-dimensional point cloud with unified coordinates; and a generation module, configured to construct a discrete point cloud mesh model based on the three-dimensional point cloud with unified coordinates according to a preset discretization algorithm, and compress the point cloud mesh model to the flattening reference plane based on a preset surface mesh compression method to obtain a particle displacement field during the compression process, and generate a two-dimensional flattened image of the target object based on the particle displacement field and the target three-dimensional dense point cloud.

[0013] Further, in some embodiments, the generation module is specifically used for: constructing a triangular mesh model based on the three-dimensional point cloud of the unified coordinates according to the preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the three-dimensional point cloud of the unified coordinates; obtaining virtual charge particles corresponding to each vertex in the triangular mesh model, constructing a Coulomb-Hooke particle system based on the virtual charge particles corresponding to each vertex, and obtaining the discrete point cloud mesh model based on the Coulomb-Hooke particle system.

[0014] Furthermore, in some embodiments, each particle interacts with the others via Coulomb repulsion. Furthermore, in some embodiments, the generation module is also used to: interpolate based on the particle displacement field and the target three-dimensional dense point cloud according to a preset interpolation algorithm to obtain a flattened two-dimensional point cloud; project the flattened two-dimensional point cloud onto a flattened reference plane to obtain an initial two-dimensional image; and repair the initial two-dimensional image based on a pre-constructed image inpainting neural network to generate a two-dimensional flattened image of the target object.

[0015] Furthermore, in some embodiments, before the point cloud mesh model is compressed to the flattened reference plane using the preset curved mesh compression method, the generation module is further configured to: eliminate folded point clouds and overlapping point clouds in the discrete point cloud mesh model based on Coulomb repulsion force and Hooke support force until the discrete point cloud mesh model meets the preset point cloud mesh model flattening conditions, thereby obtaining the point cloud mesh model.

[0016] The three-dimensional surface flattening device provided in this embodiment of the invention acquires multi-view surface images of a target object and reconstructs a three-dimensional dense point cloud. After preprocessing, plane fitting to a flattening reference plane, and aligning the point cloud to its coordinate system to obtain a unified coordinate point cloud, a discrete point cloud mesh model is constructed and compressed to the reference plane to obtain a particle displacement field. Finally, based on this displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated. This solves the problems of incomplete display of two-dimensional flattened images and low flattening efficiency when dealing with surfaces with severe wrinkles, tears, or complex curvatures in related technologies. It improves the accuracy of two-dimensional images of three-dimensional surface flattening and increases the efficiency of three-dimensional flattening.

[0017] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the three-dimensional surface flattening method as described in the above embodiments.

[0018] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the three-dimensional surface flattening method as described in the above embodiments.

[0019] A fifth aspect of the present invention provides a computer program product, including a computer program that is executed to implement the three-dimensional surface flattening method as described in the above embodiments.

[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a three-dimensional surface flattening method provided according to an embodiment of the present invention; Figure 2 A schematic diagram of a three-dimensional curved surface flattening structure for multi-angle imaging and virtual compression according to a specific embodiment of the present invention; Figure 3 This is a schematic diagram of a three-dimensional surface flattening method according to a specific embodiment of the present invention; Figure 4 A schematic diagram of a three-dimensional curved surface flattening device provided according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0022] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0023] The following describes a three-dimensional surface flattening method, apparatus, electronic device, medium, and product according to embodiments of the present invention with reference to the accompanying drawings. First, the three-dimensional surface flattening method according to embodiments of the present invention will be described with reference to the accompanying drawings.

[0024] Figure 1 This is a flowchart of a three-dimensional surface flattening method provided according to an embodiment of the present invention.

[0025] like Figure 1 As shown, the three-dimensional surface flattening method includes the following steps: In step S101, surface images of the target object from multiple perspectives are acquired, and a three-dimensional dense point cloud is reconstructed based on the surface images from multiple perspectives.

[0026] Specifically, embodiments of the present invention may employ a single camera, using a multi-angle fixed bracket or manually moving the viewing angle to capture images from different angles, or may use multiple cameras to form a camera array to acquire surface images of the target object from multiple perspectives. The layout of the camera array may be circular, spherical, or linear. Simultaneously, structured light or laser scanning is used to assist in 3D reconstruction, and open-source tools such as OpenMVG and OpenMVS are used to reconstruct the 3D dense point cloud.

[0027] For example, Figure 2 This is a schematic diagram of a three-dimensional curved surface flattening structure for multi-angle imaging and virtual compression according to a specific embodiment of the present invention, as shown below. Figure 2As shown, the adjustable guide rail 201 is an arc-shaped slide rail. This guide rail not only has its own slide rail but can also rotate 360 ​​degrees around the central axis of the device, thus achieving two degrees of freedom of movement. The rotational movement of the adjustable guide rail 201 is achieved by the drive motor 203 driving the guide rail steering gear 202. A 20-megapixel industrial camera 204 is fixedly mounted on the slide rail of the adjustable guide rail 201, and can move along the slide rail on a circle with a radius of 50cm for precise positioning. The camera lens always faces the center of the device. The light source 205 moves in coordination with the camera 204. In this embodiment of the invention, a ring-shaped LED light is used to provide uniform, shadowless illumination for the object to be scanned. The control center (301) integrates a high-performance CPU (e.g., Intel Xeon W-3365), GPU (e.g., NVIDIA GeForce RTX 4090), memory, and storage media, forming the computing processing core of the device. The upper surface of the outer shell is designed as a flat platform. The upper surface of the control center 206, made of matte black material, is used to place the target object 207. The wrinkled document to be flattened is placed at the center of the platform as the target object 207. The control center (301) issues a command to start the drive motor 203 and the camera 204. The drive motor 203 drives the guide rail steering gear 202, causing the entire adjustable guide rail 201 to rotate around the document. At the same time, the camera 204 and the light source 205 slide along the guide rail. Through the synthesis of these two movements, the camera can stop and take pictures at any preset point in the spherical coordinate system, thereby automatically and efficiently acquiring a multi-angle image sequence covering the entire surface of the document, ensuring the integrity of the three-dimensional reconstruction data. The control center 206 receives the acquired multi-angle images and calls its internal three-dimensional reconstruction module (which can be based on open source libraries such as OpenMVS) to generate a high-precision dense three-dimensional point cloud of the document surface using motion reconstruction technology.

[0028] In step S102, the three-dimensional dense point cloud is preprocessed to obtain a preprocessed point cloud, and a flattened reference plane is obtained by performing plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm. The three-dimensional dense point cloud is aligned to the plane coordinate system of the flattened reference plane to obtain a three-dimensional point cloud with unified coordinates.

[0029] Among them, the flattening reference plane refers to the two-dimensional reference plane of the three-dimensional curved surface flattening target, and the preset plane fitting algorithm refers to the algorithm of calculating the two-dimensional plane of the overall spatial distribution trend of the point cloud from the discrete three-dimensional point cloud data.

[0030] As one possible implementation, embodiments of the present invention first preprocess the 3D dense point cloud using statistical filtering, radius filtering, or machine learning-based segmentation algorithms to achieve denoising and purification of the 3D dense point cloud. Then, based on the denoised discrete 3D point cloud data, a flattening reference plane is calculated using a plane fitting algorithm, such as least squares method, RANSAC algorithm, or principal component analysis, to obtain the principal component directions and geometric centers of the point cloud, eliminating point cloud attitude and position deviations, and aligning the 3D dense point cloud with the planar coordinate system of the flattening reference plane.

[0031] In step S103, a discrete point cloud mesh model is constructed based on a three-dimensional point cloud with unified coordinates according to a preset discretization algorithm. The point cloud mesh model is then compressed to a flattened reference plane based on a preset curved surface mesh compression method to obtain the particle displacement field during the compression process. Based on the particle displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated.

[0032] Among them, the discrete point cloud mesh model refers to the structured geometric model that is transformed from the preprocessed three-dimensional dense point cloud through the discretization algorithm. The particle displacement field in the compression process refers to the set of three-dimensional displacement vectors of each particle in the discrete point cloud mesh model from the initial position after pre-flattening to the final position after compression to the flattened reference plane.

[0033] In some embodiments, a discrete point cloud mesh model is constructed based on a three-dimensional point cloud with unified coordinates according to a preset discretization algorithm. This includes: constructing a triangular mesh model based on the three-dimensional point cloud with unified coordinates according to a preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the three-dimensional point cloud with unified coordinates; obtaining virtual charge particles corresponding to each vertex in the triangular mesh model, constructing a Coulomb-Hooke particle system based on the virtual charge particles corresponding to each vertex, and obtaining the discrete point cloud mesh model based on the Coulomb-Hooke particle system.

[0034] Specifically, this invention reconstructs a point cloud mesh model based on a preset discretization algorithm, such as Delaunay triangulation, spherical pivot algorithm, or Poisson mesh, and assigns the mesh model the physical properties of a Coulomb-Hooke particle system, wherein the particles carry virtual charges and are interconnected through Hooke springs, Hooke support springs, and Coulomb repulsion forces. As one possible approach, a 3D point cloud in a unified coordinate system is used as input. A pre-defined discretization algorithm (such as Delaunay triangulation, spherical pivot algorithm, or Poisson mesh reconstruction) is invoked. Based on the spatial distribution density and topological relationship of the point cloud, a triangular mesh model is generated. In this model, each vertex of the triangular mesh corresponds one-to-one with a discrete point in the unified coordinate 3D point cloud. The mesh edges and triangular faces conform to the local morphology of the original 3D surface, ensuring that the model can completely reproduce the spatial structural features of the surface. Each vertex of the triangular mesh model is mapped to a particle with a virtual charge, and the spatial coordinates of the vertex are directly used as the initial coordinates of the corresponding virtual charge particle. Simultaneously, based on the material properties of the surface to be flattened, an adaptive virtual charge is assigned to each virtual charge particle. Based on the topological structure of the triangular mesh model, a Coulomb-Hooke particle system is constructed for the virtual charge particles. The geometric structure of the triangular mesh and the physical properties of the Coulomb-Hooke particle system are integrated to form the final discrete point cloud mesh model.

[0035] In some embodiments, each particle interacts with the others through Coulomb repulsion.

[0036] Furthermore, in some embodiments, generating a two-dimensional flattened image of the target object based on the particle displacement field and the target three-dimensional dense point cloud includes: interpolating the particle displacement field and the target three-dimensional dense point cloud according to a preset interpolation algorithm to obtain a flattened two-dimensional point cloud; projecting the flattened two-dimensional point cloud onto a flattened reference plane to obtain an initial two-dimensional image; and repairing the initial two-dimensional image based on a pre-constructed image inpainting neural network to generate a two-dimensional flattened image of the target object.

[0037] Specifically, this embodiment of the invention simulates the process of two virtual parallel plates closing towards each other from the top and bottom sides of a pre-flattened curved mesh, gradually compressing the curved mesh to a flattened reference plane. During this process, the movement speed of the virtual plates can be flexibly adjusted, effectively preventing structural deformation of the curved mesh due to over-compression. Simultaneously, the compression process is further iteratively optimized using an energy minimization algorithm to obtain an initial two-dimensional image. Then, a pre-constructed image inpainting neural network (such as one based on CNN, GAN, LaMa networks, or U-Net structures) is used to repair missing regions in the initial image. Finally, illumination correction algorithms, such as the Retinex algorithm and homomorphic filtering, are used to eliminate shadows and color distortion, ultimately generating a high-fidelity two-dimensional flattened image of the target object.

[0038] Furthermore, in some embodiments, before compressing the point cloud mesh model to the flattened reference plane based on a preset curved mesh compression method, the method further includes: eliminating folded and overlapping point clouds in the discrete point cloud mesh model based on Coulomb repulsion and Hooke support forces until the discrete point cloud mesh model meets the preset flattening conditions of the point cloud mesh model, thereby obtaining the point cloud mesh model.

[0039] Specifically, in this embodiment of the invention, the Coulomb repulsion force generated by particles carrying the same virtual charge is used to actively push away overlapping or crowded particles in space to eliminate point cloud overlap. With the support force provided by the Hooke support spring between the particles and the flattening reference plane (which increases as the distance of the particles from the reference plane increases), the particles at the wrinkled protrusions are lifted to flatten them and eliminate wrinkles, while avoiding the collapse of the grid structure, until the preset flattening strip is met, and finally a point cloud grid model with no obvious wrinkles and overlaps and a complete topological structure is obtained. To enable those skilled in the art to better understand the three-dimensional surface flattening method of the present invention, the following explanation will be provided in conjunction with specific embodiments.

[0040] Figure 3 This is a schematic diagram of a three-dimensional surface flattening method according to a specific embodiment of the present invention, as shown below. Figure 3 As shown, firstly, multi-view images of the target's three-dimensional surface are acquired, and corresponding three-dimensional point clouds are obtained using image reconstruction technology, providing raw spatial data for subsequent processing. Next, the three-dimensional point clouds are preprocessed (removing redundant points and aligning directions), and fitted to a predefined flattening reference plane to establish a unified spatial operation reference. Then, the processed point cloud is discretized into particles with virtual charges, and a triangular mesh is constructed to form a continuous medium model, transforming the discrete point cloud into a carrier capable of transmitting mechanical forces. Subsequently, pre-flattening is performed based on the Coulomb-Hooke particle system: the Coulomb repulsion between particles eliminates point cloud overlap, and the Hooke support force smooths out wrinkles, placing the mesh in a "stable compressible" state. Then, by simulating the process of two parallel virtual plates closing from top to bottom, the pre-flattened mesh is compressed into a two-dimensional plane under the constraint of the Hooke spring force, achieving the transformation from three-dimensional to two-dimensional form. Finally, combining the particle displacement field during the compression process, the original point cloud is interpolated to obtain a two-dimensional point cloud, which is then projected to generate an initial image. After image restoration and illumination correction, a complete two-dimensional flattened image of the target object is obtained.

[0041] The three-dimensional surface flattening method provided by the present invention acquires multi-view surface images of a target object and reconstructs a three-dimensional dense point cloud. After preprocessing, planar fitting to a flattening reference plane, and aligning the point cloud to its coordinate system to obtain a unified coordinate point cloud, a discrete point cloud mesh model is constructed and compressed to the reference plane to obtain a particle displacement field. Finally, a two-dimensional flattened image of the target object is generated based on this displacement field and the target three-dimensional dense point cloud. This method solves the problems of incomplete display of two-dimensional flattened images and low flattening efficiency when dealing with surfaces with severe wrinkles, tears, or complex curvatures in related technologies, improves the accuracy of two-dimensional images of three-dimensional surface flattening, and increases the efficiency of three-dimensional flattening.

[0042] Next, the three-dimensional surface flattening device according to an embodiment of the present invention is described with reference to the accompanying drawings.

[0043] Figure 4 This is a block diagram of a three-dimensional curved surface flattening device provided according to an embodiment of the present invention.

[0044] like Figure 4 As shown, the three-dimensional curved surface flattening device 10 includes: an acquisition module 100, a processing module 200, and a generation module 300.

[0045] The module 100 is used to acquire surface images of the target object from multiple perspectives and reconstruct a three-dimensional dense point cloud based on the surface images from multiple perspectives. The processing module 200 is used to preprocess the three-dimensional dense point cloud to obtain a preprocessed point cloud, and to perform plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm to obtain a flattening reference plane. The three-dimensional dense point cloud is then aligned to the plane coordinate system of the flattening reference plane to obtain a three-dimensional point cloud with unified coordinates. The generation module 300 is used to construct a discrete point cloud mesh model based on the three-dimensional point cloud with unified coordinates according to a preset discretization algorithm, and to compress the point cloud mesh model to the flattening reference plane based on a preset curved surface mesh compression method to obtain the particle displacement field during the compression process. Based on the particle displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated.

[0046] Furthermore, in some embodiments, the generation module 300 is specifically used for: constructing a triangular mesh model based on a three-dimensional point cloud with unified coordinates according to a preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the three-dimensional point cloud with unified coordinates; obtaining virtual charge particles corresponding to each vertex based on each vertex in the triangular mesh model, constructing a Coulomb-Hooke particle system based on the virtual charge particles corresponding to each vertex, and obtaining a discrete point cloud mesh model based on the Coulomb-Hooke particle system.

[0047] Furthermore, in some embodiments, each particle interacts with the others through Coulomb repulsion. Furthermore, in some embodiments, the generation module 300 is also used to: interpolate based on the particle displacement field and the target three-dimensional dense point cloud according to a preset interpolation algorithm to obtain a flattened two-dimensional point cloud; project the flattened two-dimensional point cloud onto a flattened reference plane to obtain an initial two-dimensional image; and repair the initial two-dimensional image based on a pre-built image inpainting neural network to generate a two-dimensional flattened image of the target object.

[0048] Furthermore, in some embodiments, before compressing the point cloud mesh model to the flattened reference plane based on a preset curved mesh compression method, the generation module 300 is also used to: eliminate the folded point cloud and overlapping point cloud in the discrete point cloud mesh model based on Coulomb repulsion force and Hooke support force until the discrete point cloud mesh model meets the preset point cloud mesh model flattening conditions, thereby obtaining the point cloud mesh model.

[0049] It should be noted that the foregoing explanation of the three-dimensional surface flattening method embodiment also applies to the three-dimensional surface flattening device of this embodiment, and will not be repeated here.

[0050] The three-dimensional surface flattening device provided in this embodiment of the invention acquires multi-view surface images of a target object and reconstructs a three-dimensional dense point cloud. After preprocessing, plane fitting to a flattening reference plane, and aligning the point cloud to its coordinate system to obtain a unified coordinate point cloud, a discrete point cloud mesh model is constructed and compressed to the reference plane to obtain a particle displacement field. Finally, based on this displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated. This solves the problems of incomplete display of two-dimensional flattened images and low flattening efficiency when dealing with surfaces with severe wrinkles, tears, or complex curvatures in related technologies. It improves the accuracy of two-dimensional images of three-dimensional surface flattening and increases the efficiency of three-dimensional flattening.

[0051] Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0052] When processor 502 executes the program, it implements the three-dimensional surface flattening method provided in the above embodiments.

[0053] Furthermore, electronic devices / vehicles also include: Communication interface 503 is used for communication between memory 501 and processor 502.

[0054] The memory 501 is used to store computer programs that can run on the processor 502.

[0055] The memory 501 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0056] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0057] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0058] Processor 502 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention.

[0059] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described three-dimensional surface flattening method.

[0060] In addition, embodiments of the present invention also provide a computer program product, including a computer program, which is executed to implement the above-described three-dimensional surface flattening method.

[0061] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0062] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0063] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for flattening a three-dimensional curved surface, characterized in that, Includes the following steps: Acquire surface images of the target object from multiple perspectives, and reconstruct a three-dimensional dense point cloud based on the surface images from the multiple perspectives; The three-dimensional dense point cloud is preprocessed to obtain a preprocessed point cloud, and a flattened reference plane is obtained by performing plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm. The three-dimensional dense point cloud is then aligned to the plane coordinate system of the flattened reference plane to obtain a three-dimensional point cloud with unified coordinates. Based on the unified coordinates of the three-dimensional point cloud, a discrete point cloud mesh model is constructed according to a preset discretization algorithm. The point cloud mesh model is then compressed to the flattening reference plane based on a preset curved surface mesh compression method to obtain the particle displacement field during the compression process. Based on the particle displacement field and the target three-dimensional dense point cloud, a two-dimensional flattened image of the target object is generated.

2. The method according to claim 1, characterized in that, The three-dimensional point cloud based on the unified coordinates is used to construct a discrete point cloud mesh model according to a preset discretization algorithm, including: Based on the unified coordinate 3D point cloud, a triangular mesh model is constructed according to the preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the unified coordinate 3D point cloud. Based on each vertex in the triangular mesh model, a virtual charge particle corresponding to each vertex is obtained, and a Coulomb-Hooke particle system is constructed based on the virtual charge particles corresponding to each vertex. The discrete point cloud mesh model is then obtained based on the Coulomb-Hooke particle system.

3. The method according to claim 2, characterized in that, Each particle interacts with the others through Coulomb repulsion.

4. The method according to claim 1, characterized in that, The step of generating a two-dimensional flattened image of the target object based on the particle displacement field and the target three-dimensional dense point cloud includes: The two-dimensional point cloud after flattening is obtained by interpolating based on the particle displacement field and the target three-dimensional dense point cloud according to a preset interpolation algorithm. The flattened two-dimensional point cloud is projected onto the flattened reference plane to obtain an initial two-dimensional image; The initial two-dimensional image is repaired based on a pre-built image inpainting neural network to generate a two-dimensional flattened image of the target object.

5. The method according to claim 1, characterized in that, Before compressing the point cloud mesh model to the flattened reference plane using the preset curved surface mesh compression method, the method further includes: The point cloud mesh model is obtained by eliminating folded and overlapping point clouds in the discrete point cloud mesh model based on Coulomb repulsion and Hooke support forces until the discrete point cloud mesh model meets the preset point cloud mesh model flattening conditions.

6. A three-dimensional curved surface flattening device, characterized in that, include: The acquisition module is used to acquire surface images of the target object from multiple perspectives and reconstruct a three-dimensional dense point cloud based on the surface images from multiple perspectives. The processing module is used to preprocess the three-dimensional dense point cloud to obtain a preprocessed point cloud, and to perform plane fitting on the preprocessed point cloud according to a preset plane fitting algorithm to obtain a flattened reference plane, and to align the three-dimensional dense point cloud to the plane coordinate system of the flattened reference plane to obtain a three-dimensional point cloud with unified coordinates. The generation module is used to construct a discrete point cloud mesh model based on the three-dimensional point cloud with the unified coordinates according to a preset discretization algorithm, and compress the point cloud mesh model to the flattening reference plane based on a preset curved surface mesh compression method to obtain the particle displacement field during the compression process, and generate a two-dimensional flattened image of the target object based on the particle displacement field and the target three-dimensional dense point cloud.

7. The apparatus according to claim 6, characterized in that, The generation module is specifically used for: Based on the unified coordinate 3D point cloud, a triangular mesh model is constructed according to the preset discretization algorithm, wherein the vertices of the triangular mesh model correspond to discrete points in the unified coordinate 3D point cloud. Based on each vertex in the triangular mesh model, a virtual charge particle corresponding to each vertex is obtained, and a Coulomb-Hooke particle system is constructed based on the virtual charge particles corresponding to each vertex. The discrete point cloud mesh model is then obtained based on the Coulomb-Hooke particle system.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the three-dimensional surface flattening method as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the three-dimensional surface flattening method as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the three-dimensional surface flattening method as described in any one of claims 1-5.

Images