Method and apparatus for generating a three-dimensional geometric model

Abstract

An embodiment of the present application provides a three-dimensional geometric model generation method and device, comprising: obtaining initial mesh data of a basic geometric body; performing noise disturbance processing on the basic geometric body based on the initial mesh data to obtain an initial deformation mesh; performing volume subtraction operation in the initial deformation mesh based on a hollowing control parameter to obtain a hollowing processing mesh with a target hollowing region; performing surface subdivision processing on the hollowing processing mesh to obtain a subdivided mesh with a target mesh density; and performing vertex displacement processing on the subdivided mesh based on surface displacement control data corresponding to the subdivided mesh to obtain a three-dimensional geometric model with a target geometric detail. In this way, the number of redundant patches and rendering overheads can be reduced, the gradual construction of a complex model from a macroscopic to a microscopic geometric detail can be realized, the generation efficiency and geometric precision are improved, and the geometric data storage and operation processing pressure can be relieved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to methods and apparatus for generating three-dimensional geometric models. Background Technology

[0002] In the field of computer graphics, procedural generation techniques are widely used for 3D modeling of natural scenes. These techniques typically use basic geometry as an initial template, offset mesh vertices through a noise function to simulate macroscopic shape changes, and employ normal mapping or uniform surface subdivision to enhance the model's visual details, thereby constructing 3D scene objects with natural characteristics, such as complex structures like rocks and landforms.

[0003] However, when generating complex geometric models with hollow structures, related technologies typically process each step independently, making it difficult to coordinate and optimize the macroscopic morphology, hollow topology, and surface details. Furthermore, conventional methods rely on normal mapping to simulate surface bumps and lack realistic volume displacement information. If global mesh refinement is used to improve geometric accuracy, redundant patches are easily generated, leading to data package expansion and increased rendering load. This increases the pressure on geometric data processing, making it difficult to balance generation efficiency with model accuracy. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and computer-readable storage medium for generating three-dimensional geometric models, in order to at least partially solve the aforementioned problems existing in the related art.

[0005] According to one aspect of this application, a method for generating a three-dimensional geometric model is provided, comprising: acquiring initial mesh data of a basic geometry; performing noise perturbation processing on the basic geometry based on the initial mesh data to obtain an initial deformed mesh; performing volume subtraction operation on the initial deformed mesh based on hollowing control parameters to obtain a hollowed-out mesh with a target hollowed-out region; performing surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with a target mesh density; and performing vertex displacement processing on the subdivided mesh based on the surface displacement control data corresponding to the subdivided mesh to obtain a three-dimensional geometric model with target geometric details.

[0006] According to one aspect of this application, a three-dimensional geometric model generation apparatus is provided, comprising: a data acquisition module for acquiring initial mesh data of a basic geometry; a noise perturbation module for performing noise perturbation processing on the basic geometry based on the initial mesh data to obtain an initial deformed mesh; a volume subtraction module for performing volume subtraction operations on the initial deformed mesh based on hollowing control parameters to obtain a hollowed-out mesh with a target hollowed-out region; a subdivision processing module for performing surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with a target mesh density; and a vertex displacement module for performing vertex displacement processing on the subdivided mesh based on surface displacement control data corresponding to the subdivided mesh to obtain a three-dimensional geometric model with target geometric details.

[0007] According to one aspect of this application, an electronic device is provided, including a processor and a memory; the memory stores a computer program; when the processor executes the computer program, it implements any of the above-described three-dimensional geometric model generation methods.

[0008] According to one aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements any of the above-described three-dimensional geometric model generation methods.

[0009] One embodiment of this application provides a method for generating a three-dimensional geometric model, comprising: acquiring initial mesh data of a basic geometric body; performing noise perturbation processing on the basic geometric body based on the initial mesh data to obtain an initial deformed mesh; performing volume subtraction operation on the initial deformed mesh based on hollowing control parameters to obtain a hollowed-out mesh with a target hollowed-out region; performing surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with a target mesh density; and performing vertex displacement processing on the subdivided mesh based on the surface displacement control data corresponding to the subdivided mesh to obtain a three-dimensional geometric model with target geometric details. In this way, integrating noise perturbation, volume subtraction, and vertex displacement into the same subdivision process helps reduce the number of redundant faces and rendering overhead, enables the progressive construction of complex models from macroscopic to microscopic geometric details, improves generation efficiency and geometric accuracy, and helps alleviate the pressure of geometric data storage and computation processing. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of a system architecture is shown in one exemplary embodiment of this application; Figure 2 This application shows a flowchart of a method in one exemplary embodiment; Figure 3 A schematic diagram of an ellipsoidal three-dimensional mesh model is shown in one exemplary embodiment of this application; Figure 4 A schematic diagram of an initial deformed mesh is shown in one exemplary embodiment of this application; Figure 5 A schematic diagram of a perforated mesh in one exemplary embodiment of this application is shown; Figure 6 This diagram illustrates a high-precision mesh after displacement mapping subdivision in one exemplary embodiment of this application; Figure 7 This diagram illustrates a rock mass slice thermographic analysis in one exemplary embodiment of this application; Figure 8 A schematic diagram illustrating the details of a bump map surface texture in one exemplary embodiment of this application; Figure 9 This diagram illustrates the fusion effect of macroscopic rock mass geometry and microscopic texture in one exemplary embodiment of this application; Figure 10 This illustration shows a schematic diagram of the structure of an electronic device according to one exemplary embodiment of the present application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0014] 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. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] The method in one embodiment of this application can run on a terminal device or a server. The terminal device can be a local terminal device. When the method in the embodiment runs on a server, the method can be implemented and executed based on a cloud interaction system, wherein the cloud interaction system includes a server and client devices. Figure 1 The figure shows a cloud interaction system architecture diagram provided in this application. As shown, the cloud interaction system may include: a client device 10 and a server 20, wherein the client device 10 can be connected to the server 20 via a network 30.

[0016] It should be noted that the information (including but not limited to user input information, such as information entered by the user into an input box), data (including but not limited to data used for analysis, stored data, and displayed data, such as context code, all code of the current project, service pressure corresponding to operations performed on all code of the current project, and code development status of the current project), and signals involved in the embodiments of this application are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards. For example, the context code, operations performed on all code of the current project, and the service pressure and code development status involved in the operations in the embodiments of this application were all obtained under full authorization.

[0017] A three-dimensional geometric model generation method according to one embodiment of this application, such as Figure 2 As shown, the method may include: Step S210: Obtain the initial mesh data of the basic geometry; Step S230: Based on the initial mesh data, perform noise perturbation processing on the basic geometry to obtain the initial deformed mesh; Step S250: Based on the cutout control parameters, perform volume subtraction operation in the initial deformation mesh to obtain a cutout processing mesh with the target cutout area; Step S270: Perform surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with the target mesh density; Step S290: Based on the surface displacement control data corresponding to the subdivided mesh, perform vertex displacement processing on the subdivided mesh to obtain a three-dimensional geometric model with target geometric details.

[0018] According to one embodiment of this application, integrating noise perturbation, volume subtraction, and vertex displacement into the same subdivision process helps reduce the number of redundant faces and rendering overhead, enables the progressive construction of complex models from macroscopic to microscopic geometric details, improves generation efficiency and geometric accuracy, and helps alleviate the pressure of geometric data storage and computation processing.

[0019] The embodiments of this application will be further described below.

[0020] In step S210, the initial mesh data of the basic geometry is obtained. In this way, obtaining regular and continuous initial mesh data from the standardized basic geometry can provide a topologically stable computational basis for subsequent noise perturbation, volume subtraction and surface subdivision, thereby improving the consistency of geometric deformation and structural integrity during the procedural rock mass generation process.

[0021] For example, when the procedural rock mass generation process starts, the geometric modeling engine first creates a closed ellipsoidal mesh composed of uniformly pieced triangular facets, and encodes the spatial coordinates, topological connectivity, and surface normal information of each vertex of the ellipsoid into initial mesh data; furthermore, the geometric parameters of the ellipsoid can be adaptively adjusted according to the span and height requirements of the rock mass to be generated, so as to obtain a computational basis that matches the target macroscopic contour. Figure 3 The diagram shows a three-dimensional mesh model of an ellipsoid, the surface of which is composed of dense triangular patches, presenting a basic geometric shape with uniform subdivision.

[0022] Optionally, the base geometry provides the initial closed volume for procedural rock mass generation, and its geometry can be adaptively set according to the macroscopic contour of the target rock mass. Optionally, the base geometry can be any regular geometry with a closed volume, such as a sphere, ellipsoid, or capsule, and its specific shape can be flexibly selected according to the macroscopic scale and spatial requirements of the rock mass to be generated. In one embodiment, the base geometry is configured as an ellipsoidal template stretched vertically to simulate the natural features of the longitudinal extension of a mountain. It should be noted that the surface mesh of the base geometry can be composed of triangular, quadrilateral, or mixed polygonal patches, as long as a continuous closed topological structure can be formed; this application does not intend to limit the mesh type. Furthermore, the initial subdivision density of the base geometry can be preset according to the accuracy requirements of the subsequent deformation algorithm. For example, when high-frequency noise perturbation is required, a higher initial subdivision density can be used to avoid surface tearing caused by insufficient subsequent subdivision, thereby enabling subsequent procedural deformation to unfold based on a stable computational basis with sufficient geometric resolution.

[0023] Optionally, the initial mesh data is used to characterize the spatial geometry and topological information of the basic geometry, providing an operable numerical computation basis for subsequent noise perturbation and surface subdivision. Optionally, the initial mesh data may include, but is not limited to, vertex coordinate sequences, face index arrays, vertex normal vectors, and texture coordinate information, and its data organization can be adaptively adjusted according to the adopted 3D engine or geometry processing library. In one specific implementation, the initial mesh data is stored in a half-edge data structure to facilitate quick querying of adjacent faces and neighboring vertices of any vertex, thereby improving the efficiency of neighborhood traversal in the subsequent noise perturbation stage. It should be noted that the initial mesh data can be obtained by directly instantiating the basic geometry and serializing it into mesh data through program code, or by loading pre-made regular geometric assets from a local resource library or network resources. Furthermore, after obtaining the initial mesh data, a topological consistency check can be performed on the data, such as detecting the presence of non-manifold edges, isolated vertices, or inconsistent face normal orientations. If anomalies are found, repair processing is performed to ensure that subsequent Boolean operations and surface subdivision can be reliably executed based on the correct and continuous mesh topology.

[0024] In step S230, based on the initial mesh data, noise perturbation processing is performed on the basic geometry to obtain an initial deformed mesh. In this way, by applying noise perturbation processing to the basic geometry, regular geometric surfaces can be transformed into irregular shapes with natural undulations, effectively enhancing the realism and organic variation of the macroscopic structure of the 3D model.

[0025] In one example, after acquiring the initial mesh data of the basic geometry, the system performs position offset processing on each vertex of the basic geometry according to the target noise field, causing the originally smooth, regular curved surface to produce uneven, natural undulations, thus forming the initial deformed mesh. The noise field can be generated using either a pseudo-noise function based on a random seed or a deterministic noise function based on a mathematical formula; this application is not limited to either. For example, when the target object is a mountain rock mass, the basic geometry can be a smooth ellipsoid. After noise perturbation processing, its surface can be transformed into an organic morphology with rock characteristics, achieving a highly realistic simulation of macroscopic geometric structures. Figure 4 The diagram shows an initial deformable mesh model. After applying noise perturbation to a base ellipsoid, the mesh surface exhibits an irregular, undulating organic morphology, transforming the overall outline from a smooth ellipsoid into a rugged mountain structure reminiscent of natural rock. The mesh topology remains intact, showcasing a programmed deformation effect.

[0026] The noise perturbation process is based on the offset of the geometric vertices by the noise field to generate an initial deformed mesh with irregular undulations.

[0027] Optionally, the aforementioned noise perturbation processing can be based on applying a random offset in a uniform direction to all vertices of the basic geometry using a single-channel noise function, or it can be based on applying perturbations of different intensities along the vertex normal and tangential directions using a multi-channel noise function. In one possible implementation, this processing synthesizes low-frequency and high-frequency noise components, enabling the surface of the basic geometry to exhibit natural microscopic variations in texture while maintaining overall structural stability. The low-frequency noise components can be used to control the undulation direction of the overall contour of the mountain, while the high-frequency noise components are used to depict fine cracks and erosion marks on the rock surface, thus giving the initial deformed mesh both a realistic macroscopic form and rich microscopic details. It should be noted that the intensity and frequency parameters of the noise perturbation processing can be dynamically configured according to the degree of rock weathering in the target environment; this application does not impose any limitations on this.

[0028] The initial deformation mesh is formed by noise perturbation of the geometry and is used to characterize irregular surface morphology with natural rock characteristics.

[0029] Optionally, the initial deformable mesh can be a three-dimensional mesh structure with an irregular surface obtained from a regular base geometry after noise perturbation, where the coordinates of its surface vertices have been systematically shifted relative to their original positions. In one embodiment, when the initial deformable mesh is used to characterize a mountain rock mass, its overall outline can be transformed from a smooth ellipsoid to an uneven shape with natural rock characteristics; when used to characterize stalactite structures, its regular prisms can be transformed into realistic sedimentary morphologies with conical bodies and a sense of biological growth. It should be understood that the topology of the initial deformable mesh can remain consistent with the base geometry, or a more optimized patch distribution can be obtained after retopology processing. Furthermore, the initial deformable mesh can not only be directly used for subsequent volume subtraction operations, but also serve as input for volume property field calculations, thereby providing a geometric basis for determining the location of material interfaces.

[0030] In an optional implementation, noise perturbation processing is performed on the basic geometry, including: generating a target noise field based on a multi-scale noise function, which includes low-frequency, mid-frequency, and high-frequency noise components; and offsetting the vertex positions of the basic geometry according to the target noise field to obtain an initial deformed mesh. Thus, through layered modulation of low-frequency, mid-frequency, and high-frequency components, not only can the overall contour, local undulations, and surface roughness of the geometric shape be controlled separately, but the rock surface can also exhibit organic weathering characteristics that conform to natural laws, improving the realism and controllability of the three-dimensional geometric model.

[0031] For example, in one embodiment of generating natural rock mass, an ellipsoid can be first obtained as the basic geometry, and then a target noise field is generated based on a multi-scale noise function to perform noise perturbation processing on the ellipsoid. Specifically, low-frequency noise components can be configured to control the macroscopic uplift and depression morphology of the rock wall, mid-frequency noise components can be configured to superimpose secondary folds and ridges on the surface, and high-frequency noise components can be configured to characterize the fine erosion texture and fracture edges of the rock surface. After generating the target noise field, based on the sampled values ​​of the noise field at each vertex of the basic geometry, each vertex is offset along its normal direction, thereby transforming the originally smooth ellipsoid surface into an initial deformable mesh with natural rock texture.

[0032] Optionally, the aforementioned multi-scale noise function is used to fuse spatial disturbance information from different frequency bands to generate a composite noise field that combines macroscopic morphology and microscopic roughness. Optionally, the aforementioned multi-scale noise function can be implemented by layering and superimposing continuous noise from different frequency bands. In a specific implementation, multiple basic noise functions can be superimposed according to an overtone relationship, where the low-frequency noise component has a longer wavelength and a relatively larger amplitude coefficient, used to control the large fluctuations in the overall contour of the rock mass, such as simulating the main peak orientation of a mountain range or the expansion shape of a large karst cave; the mid-frequency noise component has a wavelength in the middle range, used to superimpose secondary structures on the main contour, such as gullies, terraces, or the main conical gradient of stalactites on the rock surface; the high-frequency noise component has a shorter wavelength and a smaller amplitude coefficient, used to depict fine erosion marks, layered spalling textures, and the small unevenness of the rock's fractured edges. It should be noted that the above description of the three-band division is only an example. In actual implementation, more subdivided frequency bands can be introduced, or the noise function can be replaced with a discrete element field based on the simulation of the physical erosion process. This application does not intend to limit the mathematical form of the noise function. By weighted mixing of noise components from different frequency bands, not only can the efficiency of procedural generation be preserved, but the final output geometry can also present multi-layered details that conform to the natural geological laws.

[0033] Optionally, considering the different surface texture requirements of different rock types, the frequency components of the aforementioned multi-scale noise function can also be configured using arbitrary combinations of non-harmonic frequencies. For example, when generating sedimentary rock bedding structures, the frequency range of the mid-frequency noise components can be set closer to that of the low-frequency components to produce continuous and gentle layered undulations, while suppressing the amplitude of high-frequency components to avoid introducing too many disordered impurities onto the bedding surface. When generating volcanic rocks or weathered granite, the weight of the high-frequency noise components can be increased and a turbulent noise mode can be introduced to simulate the sharp and chaotic crystal erosion textures on the surface of these rocks. Furthermore, the weight coefficients of the aforementioned frequency components can be adjusted in real time according to the target environmental parameters. For example, when simulating a humid water erosion environment, the anisotropic stretching weight along the water flow direction in the mid-frequency noise can be dynamically increased, causing the rock surface to exhibit a grooved structure shaped by directional erosion. Thus, by flexibly configuring the weights and frequencies of each frequency band in the noise function, diverse rock morphologies suitable for different geological ages and climatic environments can be quickly generated without changing the basic geometric template.

[0034] Optionally, the vertex position offset processing is used to perform spatial coordinate transformation on each vertex according to the target noise field to generate an initial deformed mesh with irregular surface undulations.

[0035] Optionally, the vertex offset processing described above can be performed by shifting the vertices of the basic geometry surface along their original normal direction, tangential plane direction, or a combination of both, based on the values ​​of each sampling point in the target noise field. For example, when generating the macroscopic undulations of a mountain rock wall, each vertex can be shifted primarily outward or inward along the normal, with the shift amount proportional to the scalar value of the target noise field at that vertex multiplied by a preset deformation intensity coefficient. Simultaneously, to avoid excessive local stretching or compression of the mesh due to displacement of all vertices in only a single direction, a tangential perturbation component can be introduced, causing relative slippage between adjacent vertices, thereby simulating the shear deformation of rock caused by the release of internal stress during natural weathering. It should be understood that the above offset processing can be applied not only to the original vertices of the basic geometry but also to the mesh after preliminary subdivision; this application does not limit this. By flexibly adjusting the offset direction and deformation intensity coefficient, the procedurally generated rock structure can possess both the diversity of macroscopic contours and the organic chaos of local surfaces while maintaining the integrity of the original mesh topology.

[0036] In an optional implementation, after performing noise perturbation on the basic geometry, the method further includes: obtaining the volumetric property field of the initial deformation mesh and calculating the gradient distribution of the volumetric property field; determining the material interface location based on the gradient distribution, whereby the material interface location is used to characterize the transition region between different material properties. Thus, determining the material interface location through gradient calculation of the volumetric property field allows the rock mass model to possess both external morphology and internal bedding structure, providing a spatial boundary basis for subsequent fracture structures or differentiated material processing, thereby enhancing the overall geological realism. In one implementation, after the basic geometry is perturbed by noise to form an irregular mountain, the system can further extract the volumetric property field of the initial deformation mesh. Specifically, the process may include: (1) discretizing the internal space of the grid into a voxel array along the cross-sectional direction of the mountain, with each voxel recording the corresponding density or material property value, and constructing a volume property field; (2) calculating the gradient distribution of the volume property field in three-dimensional space, generating a segmented heat map, with the color from blue to red representing the gradient change from low to high; (3) determining the area where the gradient value exceeds the target gradient threshold as the material interface location, and outlining the transition boundary of different rock layers where the contour lines are dense, thereby clarifying the spatial direction of the material properties inside the rock mass.

[0037] like Figure 7 This diagram illustrates a rock mass slice thermographic analysis, visualizing a color thermographic map generated along the cross-section of the mountain. A thermographic spectrum from blue to yellow to red is used to visually represent the density / material property gradients of different layers within the rock mass. Contour lines clearly delineate the interfaces between different rock layers, demonstrating the stratified distribution characteristics of volumetric slope properties.

[0038] Optionally, the volumetric attribute field is used to characterize the material properties at various locations within the mesh and can be constructed based on a voxel array to represent the differences within the rock mass. Optionally, the aforementioned volumetric attribute field can be a three-dimensional scalar field, and its spatial sampling method includes, but is not limited to, regular voxel meshes, adaptive octree structures, or tetrahedral meshes. Each sampling point in this field stores a single material density value, or a multi-dimensional vector containing porosity, hardness, and water content, thereby more finely distinguishing the physical states of different geological structures within the rock mass. In one embodiment, the system can slice the noise-perturbed ellipsoidal mesh along the cross-sectional direction of the mountain, filling each layer with voxels at a fixed resolution and assigning initial density values, so that the internal regions far from the surface exhibit denser properties different from the surface weathered rock. It should be noted that the generation of the volumetric attribute field is not limited to voxelization; it can also directly utilize three-dimensional textures or procedural noise for continuous interpolation within the mesh. This application does not intend to limit the construction form of the field. One objective of this approach is to establish an internal property description that is decoupled from surface geometry, so that subsequent generation of layered textures or insertion of fracture structures can obtain a reliable spatial reference, thereby enhancing the geological realism of the rock mass while maintaining high computational flexibility.

[0039] Optionally, gradient distribution is used to describe the changes in the volumetric property field in various spatial directions, which can help identify regions of abrupt property changes within the rock mass.

[0040] Optionally, the gradient distribution described above can be obtained by calculating the three-dimensional spatial derivative of the volumetric attribute field. The calculation methods include, but are not limited to, the central difference method, three-dimensional difference filtering, or differentiation after Gaussian smoothing. The magnitude of the gradient value reflects the drastic change in the attribute field, while the gradient direction points to the normal direction where the attribute value increases the fastest. Together, they determine whether a clear geological boundary exists within the rock mass. In one embodiment, a heat map generated along a mountain cross-section can visually represent this gradient distribution: the color transitions from a smooth blue area to a high-gradient red area, with dense contour lines indicating abrupt attribute changes. It should be noted that the gradient threshold can be dynamically adjusted according to the rock mass type. For example, the interlayer transition of sedimentary rocks exhibits a gentle trend, so the threshold sensitivity can be appropriately reduced; while the boundary between volcanic rocks and surrounding rocks exhibits sharp abrupt changes, requiring a higher threshold to accurately capture the boundary. One purpose of this processing step is to transform continuously changing volumetric attributes into quantifiable boundary criteria, avoiding the tedious operation of manually calibrating layer interfaces and providing a reliable mechanical basis for the subsequent automatic insertion of fracture structures in geologically weak zones, thereby improving the overall realism and automation of programmed rock mass generation.

[0041] Optionally, the material interface location is used to indicate the transition boundary between different materials within the rock mass, and it is located based on gradient peak regions or contour lines.

[0042] Optionally, once the material interface location is determined, it can serve as a reference for various subsequent geometric or texturing operations. This location is not an infinitely thin mathematical plane, but rather a transition zone with a certain thickness, the width of which can be adaptively adjusted according to the gradient value: in areas with extremely steep gradients, the interface can be treated as a sharp rock fracture edge; in areas with gentle gradients, it is treated as a gradually changing mineral contamination zone. In one embodiment, after determining the interface between sedimentary rock and mudstone, the system can insert a fracture structure parallel to the gradient direction along this interface. The fracture depth varies with the undulation of the interface, thereby simulating the characteristic of joints easily developing at the junction of different lithologies under real geological processes. It should be noted that the use of the material interface location is not limited to inserting fractures; it can also include assigning different bump map intensities to both sides of the interface, adjusting the rock layer color blending weights of the surface shader, or assigning differentiated fracture parameters to different rock layers in the physics engine. This application does not limit this application. One objective of this approach is to elevate a simple surface undulation model into a layered rock mass with internal structural logic through joint analysis of internal properties and geometric morphology, so that the final rendering result can present a believable geological structure in both profile display and physical interaction.

[0043] In an optional implementation, determining the material interface location based on the gradient distribution includes: identifying regions where the gradient value is greater than a target gradient threshold as material interface locations; and inserting fracture structures at the material interface locations, with the fracture structure orientation determined based on the gradient value direction. This method, by locating the material interface through gradient threshold comparison and inserting fractures along the gradient direction, not only achieves the programmed generation of rock mass joints but also ensures the consistency between the fracture orientation and the material transition zone, significantly enhancing the geological realism and structural rationality of the rock mass model. For example, after the system completes the gradient calculation of the volumetric attribute field, it iterates through the gradient modulus of each voxel or vertex of the rock mass; if the gradient value in a certain region is greater than a preset target gradient threshold (for example, this threshold can be dynamically set to 0.35 based on the statistical mean of rock layer density differences), then that region is marked as a material interface location. Furthermore, the system inserts a fracture structure at the interface location along the direction of the gradient value (i.e., the normal direction where the property changes most drastically). The opening width and extension depth of the fracture structure can be adaptively scaled according to the current gradient modulus, thereby simulating the interlayer joints and fracture zones formed in sedimentary rocks due to geological compression or water erosion.

[0044] Optionally, a target gradient threshold is used to quantize and filter the gradient distribution to distinguish between uniform material regions and transitional material regions.

[0045] Optionally, the target gradient threshold can be a normalized value determined based on the statistical standard deviation of rock mass density, porosity, or mineral composition, or it can be a custom threshold manually input by the user for a specific geological stratum. Considering the often significant differences in physical properties between different rock strata, as a possible implementation, the target gradient threshold can be configured as an adaptive coefficient that dynamically changes with depth or rock stratum type. For example, for the boundary between dense granite and loose sandstone / conglomerate, the target gradient threshold can be set to a higher value to capture drastic density changes; while for the weak bedding transitions within sedimentary shale, the threshold can be correspondingly lowered to ensure that delicate material interfaces are not missed. One objective of this embodiment is that, by introducing an adjustable target gradient threshold, it avoids over-segmentation or missed detection caused by a fixed threshold, and also adapts to the automatic generation needs of various complex geological structures, making the determination of material interfaces more consistent with geophysical laws.

[0046] Optionally, the gradient value is used to characterize the rate of change of the volumetric attribute field in local space, providing a quantitative basis for the identification of material interfaces. Optionally, the gradient value can be obtained by calculating the central difference, edge detection operator, or higher-order partial derivatives of the volumetric attribute field, and its magnitude reflects the contrast intensity of material attributes between adjacent regions. It should be noted that the gradient value is not limited to the form of scalar modulus; in some embodiments, it can also be decomposed into horizontal and vertical components to evaluate the rate of change of rock strata in the lateral distribution and longitudinal stacking directions, respectively. To avoid false interfaces caused by abnormally high gradient values ​​due to local noise points, the above method can also introduce neighborhood smoothing filtering or median filtering as preprocessing when calculating the gradient value; that is, the volumetric attribute field is first subjected to Gaussian blurring with a preset radius, and then the gradient calculation is performed, thereby improving the stability of material interface identification. In actual implementation, the calculation accuracy of the gradient value can be dynamically adjusted according to the grid resolution and rock mass scale, and this application is not limited to this.

[0047] Optionally, the fracture structure is designed to insert fissures or fracture zones at material interfaces to simulate rock joints and enhance the micro-geometric complexity of the surface. Optionally, the fracture structure can be one or more narrow, elongated depressions or through-cracks extending along the gradient value direction, with a V-shaped, U-shaped, or irregular polygonal cross-section. In one possible implementation, the insertion depth of the fracture structure is positively correlated with the gradient modulus at the current location: if the gradient modulus is large, it indicates a significant difference in material between the two rock layers, and the fracture structure can be given a larger width and deeper incision depth to simulate obvious joints formed by differential weathering or tectonic tension; if the gradient modulus is small, the fracture structure tends to be fine and shallow, appearing as a network of micro-fractures on the bedding plane. Furthermore, the edges of the fracture structure can be further modified with displacement and normal maps to visually present a weathered ridge effect of rock peeling and debris accumulation. One objective of this embodiment is to enable the rock mass to not only possess a macroscopic cave and tunnel structure, but also a natural damage morphology that conforms to the logic of geomechanics at the mesoscopic level, by programmatically inserting fracture structures at the material interface.

[0048] Optionally, besides inserting fracture structures along a single gradient direction, secondary fracture groups can also be generated based on the principal stress direction or bedding dip angle; this application is not limited to this. In some cases, if two adjacent material interfaces are close to each other and have opposite gradient directions, rock bridges or clusters of broken rock blocks can be generated between them to simulate the joint system in a real rock mass. It should be noted that the number and density of fracture structures are not only related to the target gradient threshold but can also be modulated secondaryly based on preset fragmentation parameters; for example, in exposed areas near cave entrances, the fragmentation parameter is higher, resulting in denser fracture structures and greater surface roughness, while in intact sections deep within the rock mass, the fractures are relatively sparse. Considering the user's sensitivity to close-up details of the rock face in a virtual environment, as a possible implementation, the geometric subdivision level of the fracture structure can be dynamically increased at interfaces closer to the viewing point, thereby presenting a realistic rock stratification effect and sharp weathered ridges in close-up shots.

[0049] Optionally, the direction of the gradient value is used to determine the three-dimensional extension direction of the fracture structure, so that it conforms to the normal characteristics of the local property changes of the rock mass.

[0050] Optionally, the direction of the gradient value is determined by the ratio of the partial derivatives of the volumetric property field in each spatial dimension, pointing towards the normal direction where the rock mass property increases most dramatically. In one implementation, the principal axis of the fracture structure is set to be parallel to this direction, meaning the fracture will extend perpendicularly to the isosurface, precisely corresponding to the normal fracturing pattern of bedding planes or lithological boundaries in geology. For example, when the rock mass transitions from mudstone to sandstone, if the gradient direction is approximately horizontal and points to the right, the system will generate a set of fractures extending mainly to the right in the horizontal direction at this interface, simulating interlayer slip fractures formed by differential deposition or lateral compression. It should be understood that at the interface of a three-dimensional curved surface, the direction of the gradient value may change continuously with spatial position. In this case, the orientation of the fracture structure can also be configured to bend accordingly or extend in a segmented, broken line manner, rather than being limited to a strictly straight line. In addition, the opening orientation of the fracture can also be fine-tuned based on the gradient direction to ensure that the lighting calculations of the bump map and displacement map can correctly reflect the orientation differences of the fracture sidewalls, enhancing the volumetric and fragmented feel of the microscopic surface.

[0051] like Figure 8 This diagram illustrates the surface texture details of a bump map, showing a close-up of a rock surface after applying a bump map (Bump / Normal Map). The surface exhibits obvious layered peeling, cracks, and weathering erosion marks. The normal details simulate the microscopic unevenness of the rock, achieving photorealistic surface texture while maintaining a low polygon count.

[0052] In an optional implementation, the method further includes: obtaining the reference geometry of the auxiliary structure; performing noise perturbation processing on the reference geometry of the auxiliary structure based on a multi-scale noise function to obtain a deformed mesh of the auxiliary structure; and performing surface subdivision processing and vertex displacement processing on the deformed mesh of the auxiliary structure to obtain the geometric model of the target auxiliary structure. In this way, by reusing noise perturbation, surface subdivision, and vertex displacement on the reference geometry of the auxiliary structure, components such as stalactites can simultaneously obtain the macroscopic form and microscopic details of natural features along with the main mountain, thereby unifying the procedural generation standard and improving the efficiency of scene modeling.

[0053] In one implementation, the system selects a regularly extending quadrangular prism as the reference geometry for the stalactite's auxiliary structure. A multi-scale noise function is then applied to the vertices of this prism, applying low-frequency structural offsets and high-frequency surface roughness perturbations. This transforms the originally regular prism into a deformable mesh with natural taper and irregular undulations. Subsequently, based on the stalactite's large curvature at its tip, the system performs adaptive surface subdivision of the deformable mesh to increase mesh density at the tip. A displacement map is then applied to the subdivided vertices, pushing the vertices along the normal direction to generate layered depositional textures. This results in a target auxiliary structure geometric model with a sharp tip, a thick base, and surface traces of water erosion.

[0054] Optionally, this baseline geometry provides the initial closed volume profile and topological basis for the attached structure, and a regular geometric template can be flexibly selected according to the type of attached structure. Optionally, the baseline geometry of the attached structure can be a simple geometric template such as a regular quadrangular prism, cuboid, or capsule. Taking stalactites or stalagmites as an example, considering that such structures grow vertically along the direction of gravity in real geological environments, using a uniformly subdivided quadrangular prism as the initial template can better simulate its downward extension trend. It should be noted that the size, aspect ratio, and subdivision density of this baseline geometry are not limited to fixed values, but can be dynamically set according to the scale of the target cave space and the intensity of subsequent noise disturbances. For example, when generating large dome-shaped stalactites, a taller initial prism can be selected and the number of lateral segments can be appropriately increased to retain sufficient geometric details in the subsequent deformation stage. This baseline geometry maintains a closed manifold topologically, thereby ensuring that non-manifold edges are not generated during subsequent volume calculations or mesh trimming, maintaining the stability of mesh calculations.

[0055] Alternatively, besides using regular polyhedra as the reference geometry for the auxiliary structures, local surface patches from the main mountain mesh can be directly extracted as the growth basis. For example, when generating draperies or curtains attached to cave walls, a certain thickness can be stretched outward from a triangular facet of a specific area on the rock surface to form an initial shell with a naturally fitting contour. Furthermore, the reference geometry can also be a composite structure formed by combining multiple simple geometries through union operations, used to create more complex branched stalagmites or clustered stalactites. In practical implementation, it should be ensured that the reference geometry has sufficient initial subdivision so that subsequent multi-scale noise perturbations can exhibit rich surface undulations; if the subdivision is too low, high-frequency noise components will not be effectively geometrically represented on the mesh, resulting in an overly smooth and unrealistic surface for the auxiliary structures.

[0056] Optionally, the deformable mesh is the intermediate geometric state of the reference geometry of the auxiliary structure after noise perturbation, used to carry macroscopic shape offset and surface undulation information.

[0057] Optionally, the generation process of the auxiliary structure deformation mesh shares the same multi-scale noise function as the deformation logic of the main mountain, but the range of its noise parameters is independent of that of the main mountain. When simulating real geological structures, the deformation law of geological structures such as stalactites can be set as a conical transition feature with a thick root and a thin tip. As a possible implementation, when the system performs vertex offset on the reference geometry, a deformation intensity coefficient decreasing along the height direction can be introduced, so that the top region is subjected to strong noise stretching to form a slender end, while the bottom region maintains a relatively stable support section. It should be noted that the topological connection relationship of the deformation mesh must be maintained during the offset process. Local patch overlap caused by large vertex offset can be eliminated by small correction steps or local retopology. In addition, the mesh density after deformation should meet the minimum curvature sampling requirements of subsequent subdivision steps. If the patches in a local area are too narrow due to excessive stretching, an edge splitting operation can be inserted in advance at that point to improve the mesh quality.

[0058] Optionally, this processing is used to refine the mesh cells in key geometric regions of the deformed mesh of the attached structure to provide the base subdivision density required for the displacement map. Optionally, the surface subdivision processing performed on the deformed mesh of the attached structure can be adaptively adjusted based on the morphological characteristics of the attached structure itself. Taking stalactites as an example, their tip regions often have large surface curvature variations, while the root regions are relatively flat. To avoid redundant patches caused by uniform subdivision across the entire mesh, the system can detect the curvature variation rate of each region of the deformed mesh, performing local mesh refinement only at the tips or depressions, while maintaining a coarser base topology at the root. This reduces the consumption of video memory and computational resources, while ensuring that high-frequency displacement details are mainly concentrated in visually sensitive areas. It should be noted that this adaptive subdivision strategy is not limited to a single metric based on curvature; it can also be combined with the distance from the vertex to the root, the local surface area stretching rate, or the local amplitude of multi-scale noise for comprehensive judgment. In one specific approach, if the attached structure is a vertically growing stalagmite, the subdivision thresholds for different axial heights can be set differentially according to linear or nonlinear gradients to match the differences in the degree of physical erosion between upper and lower sections during geological deposition.

[0059] Optionally, this processing aims to apply controlled displacement along the vertex normal direction, creating a volumetric micro-layering and sedimentary structure on the surface of the attached structure. Optionally, this vertex displacement processing can be driven by surface displacement control data (displacement map data), pushing each vertex of the deformed mesh of the attached structure a target distance along its normal direction. Unlike the main mountain, the layered textures of attached structures such as stalactites typically exhibit thinner, denser, and more periodic accumulation characteristics; therefore, their displacement maps can correspondingly employ higher spatial frequencies and narrower amplitude ranges. For example, when simulating the surface of stalactites formed by long-term dripping water deposition, the displacement data can generate a series of concentric ring-shaped ridges arranged along the axial direction through procedural noise, with each ring slightly lower than the previous one, thus representing the visual effect of water carrying minerals and depositing them layer by layer. It should be noted that, to avoid excessive geometric stretching or face flipping at the layer boundaries of the replaced geometry, the system can introduce stability constraints on the vertex tangent space during the displacement calculation stage, performing local smoothing filtering on the displacement vector along the normal direction. In addition, after vertex displacement processing is completed, a normal map can be superimposed to enhance the light and shadow undulations of the microscopic surface, so that rich detail texture can still be obtained while maintaining a relatively controllable number of geometric faces.

[0060] Optionally, the model represents the final 3D geometry of the appendage structure after procedural generation, used for combined rendering with the main rock mass scene. Optionally, the target appendage structure geometry model can include, but is not limited to, geological components with decorative or connecting functions such as stalactites, stalagmites, flowstones, curtains, or isolated rock columns. The model achieves visual consistency and scale harmony with the main rock mass by multi-level fusion of large structures determined by multi-scale noise, mesoscopic topology provided by adaptive subdivision, and microscopic details imparted by displacement mapping. For example, a group of stalactite models hanging from the cave ceiling can serve as a significant visual focal point in the hollowed-out area of ​​the main cave, their densely arranged tips helping to guide the viewer's gaze deeper into the cave. It should be noted that the relationship between the target appendage structure geometry model and the main rock mass is not a simple attachment; in one possible implementation, the system can generate layered fusion transition surfaces at the junction of the appendage structure's root and the rock mass, making the appendage structure appear to grow naturally from the main rock mass rather than being an externally inserted independent object. This fusion transition surface can be achieved by expanding the local mesh of the main mountain and performing a soft Boolean union operation with the root of the auxiliary structure, thereby eliminating the hard edges and floating feeling at the joint.

[0061] In an optional implementation, generating the target noise field based on a multi-scale noise function further includes: dynamically adjusting the weighting coefficients of each frequency component in the multi-scale noise function based on target environmental parameters, whereby the target environmental parameters characterize the degree of influence of the target environment on the geometric morphology; regenerating the target noise field based on the updated weighting coefficients, and offsetting the vertex positions of the basic geometry based on the regenerated target noise field to obtain an initial deformation mesh with target erosion characteristics. In this way, by dynamically adjusting the weighting coefficients of each frequency in the multi-scale noise function using target environmental parameters, the programmed rock mass geometry can adapt to the external force conditions of different virtual environments, effectively improving the realism and adaptability of the initial deformation mesh to environmental erosion characteristics such as water erosion and weathering.

[0062] In one implementation, when the virtual scene needs to construct a cave wall subjected to long-term erosion by strong water flow, the target environmental parameters can be configured as a composite vector representing high humidity and strong runoff velocity. Based on this, the system assigns a first weighting coefficient to the low-frequency noise component in the multi-scale noise function to expand the overall concave region, while assigning a second weighting coefficient to the high-frequency noise component to characterize the microscopic water flow patterns and fine grooves on the rock wall surface. The mid-frequency component maintains its baseline weight to preserve the natural transition of the mid-section undulations. After the weighting coefficients are updated, the resynthesized target noise field performs vertex offset processing on the basic ellipsoid, causing the initial deformed mesh to exhibit an organic undulating shape with obvious water erosion traces, providing a basic geometric structure with environmental narrative for subsequent Boolean subtraction and surface subdivision processing.

[0063] Optionally, the target environmental parameters are used to quantify the influence of the target virtual environment on the geometric morphology, and can be single-dimensional parameters or a set of composite parameters. Optionally, the target environmental parameters can include, but are not limited to, quantitative indicators such as environmental humidity, fluid erosion intensity, wind erosion level, or temperature cycle amplitude. Considering the differentiated requirements of different virtual scenes for rock mass morphology, in one specific implementation, when simulating a humid karst cave environment, the target environmental parameters can be set as a composite vector characterized by high humidity and high water flow velocity. The system accordingly lowers the weight of mid-to-high frequency noise to suppress sharp protrusions, thereby generating a rounded rock surface with water erosion marks. In another dry rift valley environment, the target environmental parameters are adjusted accordingly to be characterized by low humidity and strong diurnal temperature variation, and the system increases the weight of high-frequency noise to generate more surface fracturing and weathering erosion details. It should be noted that the above specific category description of the target environmental parameters is only one example. Any quantifiable data that can objectively reflect the shaping effect of the virtual scene environment on the rock mass morphology can be included in this parameter set; this application does not intend to exhaustively limit this.

[0064] Optionally, to avoid low production efficiency caused by repeated manual adjustments of noise parameters, the target environmental parameters can be obtained in multiple ways. Besides having scene designers manually input preset environment type identifiers in the 3D editor, the system can also automatically extract and update this parameter based on dynamic weather systems or intermediate results of fluid simulations within the virtual scene. As one possible implementation, when the rainfall intensity value in the virtual scene continuously increases, the fluid scour component in the target environmental parameters can dynamically increase accordingly, thereby adjusting the noise weighting coefficient in real time to generate an erosion morphology that matches the current environmental state. By introducing this dynamic feedback mechanism, the procedural rock mass generation system can continuously respond to environmental changes during scene operation, ensuring that the geometric morphology always maintains logical consistency with the surrounding virtual ecosystem.

[0065] Optionally, the weighting coefficient is used to adjust the contribution ratio of different frequency band noise components in the multi-scale noise synthesis stage, and its numerical update directly determines the spectral composition and deformation intensity of the target noise field. Optionally, the weighting coefficient can be assigned an independent scalar value to each frequency component, or it can be updated in batches based on the target environmental parameters through a preset mapping curve. Considering the layered control effect of different frequency band noise on the macroscopic morphology and microscopic texture of the rock mass, in one embodiment, the weighting coefficient of the low-frequency noise component can be set to a first value to control the undulation amplitude of the macroscopic contour of the mountain; the weighting coefficient of the mid-frequency noise component is set to a second value to shape the medium to large-scale concave-convex structure of the rock surface; and the weighting coefficient of the high-frequency noise component is set to a third value to characterize the microscopic cracks and weathering particles on the rock surface. Furthermore, when the target environmental parameters indicate a highly erosive environment, the first and third values ​​can be adjusted upwards simultaneously, while the second value can be appropriately adjusted downwards to form a larger-span main morphological depression and richer surface fine texture. It should be noted that the specific value range of each weight coefficient mentioned above is only one example. In actual implementation, it can be flexibly adjusted according to the rock mass type, art style or performance budget. This application is not intended to limit it.

[0066] Optionally, erosion features are used to describe the morphology of the geometry after disturbance by environmental forces, and are simulated directionally by adjusting the noise spectrum weights. Optionally, erosion features may include, but are not limited to, longitudinal grooves formed by water erosion, arc-shaped depressions created by wind abrasion, layered erosion caused by thermal fission, and fractured cliffs caused by gravity collapse. In one embodiment, when the system identifies that the water erosion component is dominant in the target environmental parameters, the erosion features can be directionally configured as a groove system with a clear confluence direction. In this case, the weighting coefficients of high-frequency and mid-frequency noise in a specific direction are differentially enhanced, causing the vertex offset to form an extended depression along the direction of the water flow vector in a local area. In another embodiment, to simulate arid weathered rock masses, erosion features can be configured as a surface-wide crack texture and sharp spalling edges. In this case, the global weight of high-frequency noise is significantly increased, and a random micro-offset is applied along the local surface orientation to enhance the sense of volumetric dislocation of the fragmented blocks. It should be noted that the specific forms of erosion characteristics are not limited to the two types mentioned above. By combining different weight adjustment strategies, diverse traces of geological external forces can be derived, thereby endowing the programmed rock mass with a high degree of environmental narrative and visual realism.

[0067] In step S250, based on the hollowing control parameters, a volume subtraction operation is performed on the initial deformed mesh to obtain a hollowed-out mesh with the target hollowed-out region. Thus, by performing a volume subtraction operation on the mountain mesh after noise deformation and using the hollowing control parameters for region orientation, a three-dimensional rock cavity with a natural tunnel structure can be quickly generated, significantly improving the modeling efficiency and morphological controllability of complex geometric hollowing processing. In one embodiment, hollowing control parameters can be applied to the deformed main mountain mesh first. These parameters instruct the generation of a vertically stretched ellipsoidal shell along a horizontal path in the middle of the mountain as a subtraction tool, with the major axis of the shell's cross-section set to 40% of the mountain's width and the minor axis to 60% of its height. Subsequently, a Boolean subtraction operation is performed on this shell from the mountain mesh, thereby forming an elliptical tunnel cavity in the middle of the main mountain body, resulting in a hollowed-out mesh containing the target hollowed-out region and maintaining a manifold structure at both its inner and outer boundaries.

[0068] Optionally, the cutout control parameters are used to define the shape, size, and spatial location of the subtracted area to adjust the geometric contour of the target cutout area. Optionally, the cutout control parameters can be directly input numerical combinations, configuration items in a script file, or structural variables automatically derived from the path curve. For example, when constructing an observable tunnel in the middle of a mountain body, the cutout control parameters can indicate that the center coordinates of the subtracted area are distributed along the transverse central axis of the mountain, and specify that the major axis of the shell section is 40% of the mountain's width and the minor axis is 60% of its height, thereby creating an elliptical passage that fits the overall proportions within the mountain. In addition to the aforementioned size and location information, these parameters can also include curvature control information or the number of stretch segments for the subtracted path, in order to generate a polygonal or arc-shaped walkway cutout effect in complex terrain. This application does not intend to limit the data format and source of the cutout control parameters.

[0069] Optionally, the volume subtraction operation aims to remove the shell from the host mesh by constructing solid geometry to form a continuous internal cavity. Optionally, the volume subtraction operation can be implemented using the Construct Solid Geometry (CSG) algorithm, where the shell is the subtractive body and the initial deformed mesh is the subtracted body; any spatially overlapping portions of the two are removed. During the operation, the intersection line between the shell surface and the mountain mesh surface is first calculated, and the triangular facets on both sides are re-divided based on this intersection line. Then, the mesh cells located inside the shell are removed, and finally, the newly generated edge loops at the boundary are stitched together to maintain the topological closure of the resulting mesh. To accommodate the complex boundary of the mountain surface after noise perturbation, the volume subtraction operation can also introduce a tolerance detection mechanism. For vertices within a preset threshold range from the shell surface, local re-division is performed, thereby avoiding broken surfaces or non-manifold edges caused by slight vertex penetration or floating-point errors, ensuring that the final hollowed-out mesh has complete geometric closure.

[0070] Optionally, the target hollow area represents the continuous cavity formed inside the rock mass after subtraction, and its shape corresponds to the spatial occupancy range of the subtracted shell.

[0071] Optionally, the target hollowed-out area can take various forms, such as a tunnel penetrating the rock mass, a semi-open lateral niche, or a ventilation shaft located at the top of the rock mass. The specific shape of this area is determined by the geometric definition of the subtracted shell and the spatial coupling relationship between the original mountain grid; for example, when the subtracted shell is an ellipsoid stretched along a horizontal path, the target hollowed-out area presents as an elliptical cross-section channel with an arc-shaped top wall in the middle. In practical applications, the target hollowed-out area is not limited to a single connected cavity; it can also be a branching cave system or layered cavities formed by the sequential subtraction of multiple shells. It should be noted that the boundary surface of the target hollowed-out area inherits the subtle undulations of the initial deformed grid, so its inner wall is not an ideal smooth curved surface, but rather a rough interface with the texture of natural rock.

[0072] Optionally, the hollowed-out mesh is a three-dimensional mesh entity obtained after volume subtraction, which combines the external mountain outline with the internal hollowed-out boundary.

[0073] Optionally, the perforated mesh can be a manifold mesh data structure composed of triangular, quadrilateral, or hybrid polygons. After volume subtraction, the mesh retains both the noise deformation characteristics of the external rock surface and the internal boundary loops exposed by the subtraction operation, allowing the mountain to maintain a closed volume macroscopically while possessing an internal surface available for observation and lighting calculation. At the data organization level, the inner and outer surfaces of the perforated mesh can be assigned different material identifiers or surface normal orientations to distinguish the lighting response of the outer rock wall from the inner tunnel during subsequent rendering. Furthermore, if surface subdivision or displacement mapping needs to be added to the perforated mesh later, the boundary loop density and topological orientation of the mesh will directly affect the geometric accuracy after refinement; therefore, the generation of the perforated mesh should balance topological quality and subdivision adaptability.

[0074] In an optional implementation, based on the hollowing control parameters, a volume subtraction operation is performed in the initial deformation mesh, including: generating a hollowed-out shell according to the hollowing control parameters, the hollowed-out shell being a reference geometry stretched along the target path; and performing a Boolean subtraction operation on the hollowed-out shell from the initial deformation mesh to obtain a hollowed-out processed mesh. In this way, by using a reference geometry stretched along the target path as the hollowed-out shell for Boolean subtraction, not only can the shape and orientation of the target hollowed-out area be precisely controlled, but it can also ensure that the rock mass body forms a spatially connected region with clear boundaries while maintaining topological integrity.

[0075] like Figure 5 The diagram shows a hollowed-out mesh, the result of performing a Boolean subtraction operation on a deformed mountain model. A distinct elliptical tunnel cavity is formed in the middle of the main mountain body. The subtraction area is highlighted in dark blue, clearly demonstrating the spatial relationship between the tunnel entrance and the main rock mass, verifying the correctness of the CSG (Constructive Solid Geometry) algorithm.

[0076] In one implementation, the system first defines a target path extending from the outside to the inside of the mountain according to the hollowing control parameters, and then stretches an ellipsoidal reference geometry along this path to generate a hollow, capsule-shaped hollow shell. Next, the system calls a solid geometry construction algorithm to perform a Boolean subtraction operation between the hollow shell and the initial deformed mesh, so that the main rock mass is precisely removed from the area occupied by the shell, ultimately forming a well-defined, contour-controlled tunnel structure and obtaining a hollowed-out mesh.

[0077] Optionally, the hollow shell is used to define the outline and location of the target hollow area before Boolean subtraction, and it is stretched along the target path. Optionally, as a subtraction tool in Boolean subtraction, the geometry of the hollow shell directly determines the entrance shape and internal channel outline of the hollow area in the final 3D rock mass. Depending on the application scenario, the reference geometry of the hollow shell can be selected from various basic shapes such as ellipsoids, cylinders, or capsules, and the cross-sectional dimensions can be kept consistent during stretching along the target path. It can also be non-uniformly scaled according to the needs of erosion simulation or artificial design. For example, when generating the entrance of a mountain tunnel, the hollow shell can be set as a capsule-like structure obtained by stretching an ellipsoid along a straight path, thereby forming a rounded and smooth-edged arch-shaped cavity on the front wall of the mountain; while when generating the internal connecting channels of a karst cave, the hollow shell can be an ellipsoid with a smaller diameter stretched along a curved path to simulate the narrow channels formed by natural water erosion. It should be noted that the generation of hollow shells is not limited to a single stretch along a single path. More complex spatial shapes can also be constructed by sweeping and merging multiple paths or by variable cross-section lofting, as long as a clear and closed subtraction boundary can be formed with the main mesh of the rock mass in the subsequent Boolean subtraction operation.

[0078] Optionally, to avoid non-manifold boundary surfaces or micro-fragments affecting the subsequent surface subdivision quality after Boolean subtraction, the hollow shell can be pre-configured with the boundary fusion band width or minimum spacing threshold between it and the initial deformed mesh during the generation stage. When the hollow shell is stretched along a complex curved path, if its local radius of curvature is too small, it may result in extremely narrow and elongated triangular facets after mesh subtraction. Therefore, the reference geometry can be further adaptively subdivided between the path insertion points, making the mesh density of the hollow shell denser in curved sections and sparser in straight sections. Furthermore, before performing Boolean subtraction between the hollow shell and the rock mass, the boundary frame of the hollow shell can be expanded or a transition bevel can be added, thereby forming a natural chamfer or weathering fracture effect at the edge of the subtracted area, rather than a harsh mechanical cut. In other words, the hollow shell is not limited to a simple subtraction tool; its own minor adjustments to its shape can be directly transformed into an artistic detail representation of the final rock mass in terms of macroscopic structure.

[0079] Optionally, the target path defines the spatial extension trajectory of the hollow shell, and its shape can be determined by parametric curves or discrete point interpolation. Optionally, the target path serves as the skeleton or guide line for generating the hollow shell, and its geometric expression can be flexibly selected according to actual application requirements. In one possible implementation, the target path can be defined as a parametric curve driven by multiple control vertices, allowing artists to interactively drag and adjust control points to preview the curvature of the tunnel in real time. In another implementation, the target path can also be generated by spline interpolation from a set of discrete anchor points in three-dimensional space, suitable for scenarios where procedural terrain automatically plans the passage direction based on height maps. It should be noted that the target path is not limited to a two-dimensional planar projection curve; it can be a spatial curve undulating and spiraling in a full three-dimensional environment to meet the connectivity requirements of multi-layered karst cave systems within mountains. Besides being statically set, the target path can also be dynamically evolved in conjunction with erosion simulation or physical destruction algorithms. For example, in a water erosion scenario, the local subsidence or lateral erosion offset of the target path can be adjusted in real time based on fluid simulation results, resulting in a tunnel that exhibits natural, non-artificial meandering characteristics.

[0080] Optionally, considering that the hardness and density distribution of rock masses at different heights or depths is usually non-uniform, the setting of the target path often needs to take into account both geological rationality and visual composition requirements. As a possible implementation method, the above-mentioned target path generation process may include: determining the orientation of weak layers based on the location of material interfaces within the rock mass, and adjusting at least a local segment of the target path to the interior or near the interface of that weak layer, thereby simulating the geological law in nature where water flow preferentially infiltrates and dissolves along weak rock layers to form channels. Furthermore, if during the path planning stage, it is detected that the target path intersects with or is too close to other existing hollow areas or fracture structures, a path offset strategy can be automatically triggered to shift local nodes of the target path away from the conflict area by a preset safe distance, in order to avoid large areas of rock mass being suspended or topologically degraded after Boolean subtraction. It should be understood that the degree of curvature change of the target path is directly related to the mesh complexity of the final hollow shell. Therefore, after the path is generated, its maximum radius of curvature can be calculated, and the reference geometry of the hollow shell can be pre-subdivided accordingly to ensure that there is sufficient vertex density in areas with severe curvature to support subsequent high-quality mesh subtraction operations.

[0081] Optionally, Boolean subtraction aims to remove the space occupied by the hollow shell to form a clearly defined target hollow area. Optionally, Boolean subtraction is a basic mesh operation based on constructing a solid geometric framework, and its execution process can be implemented based on various mesh Boolean libraries or a custom mesh cutting algorithm. In one implementation, the system can first spatially voxelize or octree segment the initial deformed mesh and the hollow shell to quickly locate the local area where they interfere; then, within this local area, the intersection calculation is performed on the triangular facets of the initial deformed mesh and the boundary of the hollow shell. Based on the determination result of the center point of the facet relative to the boundary of the hollow shell, the facets completely located inside the hollow shell are deleted, and the partially intersecting boundary triangles are re-subdivided to generate a cutting boundary consistent with the surface of the hollow shell. It should be noted that Boolean subtraction is not limited to explicit trimming methods based on surface meshes; it can also achieve equivalent volume subtraction by using implicit surface or voxelized distance field interpolation. When using implicit field fusion, the calculation results are not affected by topological defects in the original mesh, and can more stably handle complex uneven surfaces caused by noise disturbances. In other words, regardless of whether explicit mesh cutting or implicit field fusion is used, as long as the spatial region defined by the hollow shell can be completely removed from the initial deformed mesh while maintaining the closedness of the remaining rock mass surface, it should be included in the scope of Boolean subtraction.

[0082] In step S270, surface subdivision processing is performed on the hollowed-out mesh to obtain a subdivided mesh with the target mesh density. Performing surface subdivision after hollowing-out processing ensures sufficient vertex density in critical areas, providing data structure support for the accurate loading of subsequent geometric details, while also ensuring the rationality of the overall topology.

[0083] In practical applications, once the cutout mesh is generated, the system can immediately initiate surface subdivision processing. The system refines the mesh in all or part of the cutout mesh according to preset subdivision rules, for example, subdividing the basic triangular face into multiple smaller-scale face units, or generating a uniform quadrilateral structure through retopology. After this processing, the system outputs a subdivided mesh with the target mesh density. The vertex distribution density of this mesh is sufficient to support the computational accuracy required for subsequent vertex displacement processing, achieving a balance between storage capacity and rendering overhead.

[0084] Optionally, the target mesh density is used to characterize the density of vertex distribution in the subdivided mesh and is dynamically determined based on the subsequent geometric detail loading requirements. Optionally, considering that displacement mapping requires sufficient geometric subdivision to present a realistic sense of volume and rock undulation, the target mesh density can be set differently according to the geometric complexity of different regions. For example, in areas where high-frequency micro-details need to be rendered later, such as tunnel entrance edges, rock strata peeling surfaces, or stalactite tips, the target mesh density can be configured to a relatively high value to ensure that the region has a sufficient number of vertices to support the displacement generated by the vertices along the normal direction; while in the gentle back of the mountain contour or in the inner regions far from the viewpoint, the target mesh density can be appropriately reduced to alleviate the load pressure on the graphics processor during the vertex displacement calculation stage. It should be noted that the above-mentioned regional division method for the target mesh density is only one example. In actual implementation, it can also be comprehensively determined based on the distance field, viewpoint direction, or material importance. This application does not intend to limit the method of determining the target mesh density.

[0085] Optionally, the subdivided mesh refers to the intermediate mesh data output after surface subdivision processing, used to provide geometric accuracy support for vertex displacement processing. Optionally, the subdivided mesh is a key intermediate product connecting macroscopic deformation and microscopic details; its data organization can include, but is not limited to, discrete meshes based on triangular facets, uniform topologies dominated by quadrilaterals, or hybrid structures of both. In one implementation, the subdivided mesh maintains the same object space coordinate system as the cutout mesh, but its vertex and face unit numbers have been significantly increased according to the surface subdivision processing rules; furthermore, the normal directions of each vertex in the subdivided mesh have been updated synchronously to accurately reflect the orientation of the refined local surface. It should be noted that the above description of the subdivided mesh topology is only one example; in actual implementation, the corresponding mesh type can be selected according to the compatibility requirements of the rendering pipeline. If the density of a local area cannot meet the requirements of subsequent geometric detail processing after evaluation, the surface subdivision processing step can be returned for local iterative refinement.

[0086] In an optional implementation, surface subdivision processing is performed on the hollowed-out mesh, including: determining the surface curvature of each region of the hollowed-out mesh; determining the curvature change of each region based on the surface curvature; adaptively refining the mesh in regions where the curvature change is greater than a target threshold; and maintaining the basic topological density in regions where the curvature change is less than the target threshold, thus obtaining the subdivided mesh. In this way, through an adaptive refinement strategy based on curvature changes, mesh resources can be concentrated in regions with drastic features, while maintaining low density in smooth regions. This approach ensures the accuracy of surface details while suppressing the total number of facets, balancing model accuracy and computational overhead.

[0087] In one implementation, for the rock face contour region with drastic curvature changes at the mountain entrance, the system calculates the surface curvature values ​​of each vertex in this region and filters out boundary zones where the curvature changes exceed a preset target threshold. Adaptive mesh refinement is then performed on these boundary zones to generate a fine topology that mixes quadrilateral and triangular faces. Simultaneously, for the gently curving backslope surface at the mountain top, the original sparse base topology density is maintained without further subdivision. This achieves a subdivision density that adequately expresses the micro-geometric undulations at the entrance edge, while maintaining only the necessary number of base faces on the gentle rock surface far from the entrance. Optionally, the aforementioned surface curvature is used to quantify the degree of curvature of the surface in each region of the hollowed-out mesh, serving as a basis for determining whether to perform mesh refinement.

[0088] Optionally, the surface curvature mentioned above reflects the degree of geometric curvature of various local regions on the hollowed-out mesh. This value can be obtained by calculating the change in the angle between the normal vectors in the vertex neighborhood or by fitting a local quadratic surface. In a specific implementation, the surface curvature is generally larger in areas where geometric features change drastically, such as the edge of a mountain entrance, while it approaches zero in flatter areas such as the top back slope. Considering the significant differences in geometric complexity between different regions, the system evaluates the surface curvature vertex-by-vertex or face-by-face patch, which can accurately identify the locations where mesh density needs to be increased. This avoids performing indiscriminate uniform subdivision on the overall mesh, ensuring that the allocation of mesh resources matches the spatial distribution of geometric features.

[0089] Optionally, the calculation objects and quantification methods for the aforementioned surface curvature are not limited to a single mode. As another implementation, the deviation angle between the vertex normal vector and the face normal vector of a triangular facet can be statistically analyzed directly, and a weighted average can be used to characterize the surface curvature of the corresponding region. Furthermore, a comprehensive analysis can be performed by combining the maximum and minimum principal curvature within the region, using a dual-indicator approach of Gaussian curvature or average curvature to improve the distinguishability between saddle-shaped and bowl-shaped surfaces. In this way, even when facing local surfaces in complex rock masses that simultaneously possess both convex and concave features, the degree of geometrical variation can be accurately obtained, thus providing a stable and reliable judgment benchmark for subsequent adaptive mesh refinement.

[0090] Optionally, the target threshold is used to define the critical criteria for curvature changes, distinguishing between regions with drastic features requiring encryption and regions with smooth surfaces that should maintain their basic density. Optionally, the target threshold can be a pre-configured static value based on the model category, or a relative threshold dynamically calculated based on the overall curvature distribution of the hollowed-out mesh. In one possible implementation, the system can traverse the curvature change values ​​of all regions on the mesh and perform normalized statistics, automatically identifying the top 30% of high curvature change regions as target regions requiring adaptive mesh encryption, while the remaining 70% of smooth regions maintain their basic topology density. Furthermore, this target threshold can also be linked to the hardware performance budget of the current computing task. When available video memory or processing threads become scarce, the system automatically increases the target threshold to reduce the proportion of encrypted regions, thereby constraining the overall mesh size within a manageable computing power range while preserving the details of key visual regions.

[0091] Optionally, adaptive mesh refinement aims to dynamically increase the patch density in areas where curvature changes exceed a target threshold, thereby improving the accuracy of local geometric representation. Optionally, the aforementioned adaptive mesh refinement can dynamically increase the number of vertices and patches according to a preset subdivision level in local areas of the hollowed-out mesh where curvature changes drastically. In one example, for the edge region of a mountain entrance, the system can perform one or more rounds of cyclic subdivision on the existing quadrilateral or triangular facet topology, increasing the patch density in this region to several times the base density, thus generating a high-density subdivided surface sufficient to represent microscopic geometric undulations. It should be noted that the refinement level and the final patch density are not fixed; they can be dynamically adjusted according to the distance between the region and the camera's viewpoint or the expected close-up observation requirements, ensuring that mesh resources are always concentrated on the geometric contour transitions where visual attention is most focused, maintaining image detail sharpness while reasonably controlling the overall data volume.

[0092] Optionally, besides mesh refinement based on cyclic subdivision, adaptive mesh refinement can also directly rearrange the logical structure of the mesh through local retopology. For example, at rock joints where the curvature change exceeds a target threshold, the system can insert streamlined quadrilaterals aligned with the principal curvature direction and compress the proportion of narrow, unhelpful triangular faces, allowing the mesh edges to naturally extend along the rock surface. To avoid topological defects such as non-manifold vertices or mesh flipping during the refinement process, connectivity detection and local mesh cleanup can be further performed after the above refinement operations. This enhances the continuity of surface details during subsequent vertex displacement processing and reduces the risk of normal tearing and texture stretching caused by poor topology during animation and rendering stages.

[0093] Optionally, the base topology density is used to define the baseline level of patch distribution maintained in flat regions to suppress excessive expansion of the overall mesh size.

[0094] Optionally, the aforementioned base topology density can be the original patch distribution density retained after noise perturbation and volume subtraction operations, or it can be the baseline mesh density formed after performing primary uniform subdivision. In gentle rock slopes or large cave roof areas where the curvature change does not exceed the target threshold, the system directly inherits this base topology density without performing additional densification operations, ensuring that these visually less noticeable areas maintain a smaller number of vertices and patches. It should be noted that the base topology density does not require global consistency; it can be set in zones based on the initial geometric complexity of the target area. For example, a lower patch baseline can be configured for large ellipsoidal cave roofs, while a slightly higher density can be retained for slightly undulating ground, thus establishing a balance between global limits and local feature preservation that meets actual artistic needs.

[0095] In step S290, based on the surface displacement control data corresponding to the subdivided mesh, vertex displacement processing is performed on the subdivided mesh to obtain a three-dimensional geometric model with target geometric details. In this way, by driving vertex displacement through surface displacement control data, microscopic surface details with realistic volume can be generated on the basis of subdivided mesh. Compared with the method of simply relying on texture lighting simulation, it can better reflect geometric volume and make the rock mass model have high-precision realism.

[0096] In one implementation, the system first acquires surface displacement control data corresponding to the subdivided mesh, which includes the target displacement of each vertex. Then, based on this data, vertex displacement processing is performed on the subdivided mesh, moving each vertex along a preset direction to alter local surface undulations, thereby generating a 3D geometric model with target geometric details such as rock textures and erosion grooves. For example, when the surface displacement control data for a certain area of ​​a mountain indicates a large positive displacement, the vertices in that area extend outward to form prominent rock edges; when the surface displacement control data indicates a negative displacement, the corresponding vertices contract inward to form weathering pits. Thus, based on the sufficient vertex density provided by mesh subdivision, vertex displacement further shapes microscopic surface undulations with a realistic sense of volume.

[0097] Optionally, surface displacement control data is used to characterize the target displacement of each vertex in the subdivided mesh, guiding the vertex displacement processing. Optionally, the surface displacement control data can be displacement map data, which records the target displacement of each vertex through grayscale or vector channels. White channel values ​​correspond to the maximum positive displacement, black channel values ​​correspond to the maximum negative displacement, and gray transition values ​​represent intermediate displacement. It should be understood that the aforementioned surface displacement control data is not limited to two-dimensional textures; it can also be a vertex displacement distribution table obtained through physical erosion simulation calculations, which includes the vertex erosion calculated from fluid flow on the mesh surface and the subsequent vertex deposition. It should be noted that the specific source and storage format of this data can be determined based on actual computing power and accuracy requirements. In practical applications, either program-generated noise textures can be used to characterize random micro-undulations, or hand-drawn height textures can be used to depict the depositional direction of specific rock layers, thus providing a flexible and highly controllable deformation basis for subsequent vertex displacement processing, ensuring that the generated three-dimensional geometric model presents realistic rock layer deposition or water erosion effects at the microscale.

[0098] Optionally, vertex displacement processing aims to adjust the vertex spatial coordinates of the subdivided mesh based on surface displacement control data to generate the target geometric details.

[0099] Optionally, vertex displacement processing can be a process of geometrically displacing each vertex of the subdivided mesh along its normal direction based on surface displacement control data. In one implementation, the system reads the displacement amount corresponding to each vertex, and moves it a corresponding distance along the positive or negative direction of the normal, using the vertex's original coordinates as a reference, thereby forming surface undulations. Considering that the subdivided mesh already has sufficient vertex density in areas of severe curvature, in the above processing, more displacement details can be retained in high-curvature areas, while low-curvature areas maintain a relatively smooth transition. Furthermore, in addition to performing single-axis displacement along the normal direction, the vertex displacement processing can also be combined with multi-axis offset vectors in the local surface framework for composite displacement to simulate the complex fracturing and spalling effects of the rock surface. It should be understood that the magnitude of the displacement can not only be directly determined by the surface displacement control data, but can also be weighted and corrected according to the hardness of the base material or the thickness of the sedimentary layer in the area where the vertex is located, so that the same surface displacement control data produces different deformation depths at different rock layer locations, enhancing the geological realism and detail richness of the three-dimensional geometric model.

[0100] In an optional implementation, based on the surface displacement control data corresponding to the subdivided mesh, vertex displacement processing is performed on the subdivided mesh, including: acquiring displacement map data, which is used to characterize the target displacement amount of each vertex of the subdivided mesh; and displacing the vertices of the subdivided mesh along the normal direction according to the displacement map data to obtain a three-dimensional geometric model. In this way, by driving the vertices to offset along the normal direction through displacement map data, the micro-texture is transformed into realistic geometric undulations while maintaining the basic topology. This not only significantly improves the volumetric feel of the rock surface but also effectively avoids the visual flattening defects caused by relying solely on normal maps.

[0101] For example, in one specific implementation, the displacement map data is generated in the form of a high-density grayscale texture, where different brightness values ​​correspond to the target displacement of each vertex on the rock surface; in the vertex displacement processing stage, the system traverses each vertex of the subdivided mesh, queries the grayscale value of the vertex's mapped coordinate position in the displacement map, and converts it into an actual offset vector along the normal direction of that vertex, thereby generating layered sedimentary textures and weathering details with realistic thickness and depth illusions on the surface of the cave rock mass. Figure 6 This diagram illustrates a high-precision mesh after displacement mapping subdivision, showing a high-density subdivided mesh resulting from displacement mapping. The model surface exhibits micron-level rock undulations and erosion textures, significantly improving geometric accuracy. This surpasses the limitations of traditional normal mapping, which only alters lighting, achieving truly volumetric microscopic surface details.

[0102] Optionally, the displacement map data may include height textures or simulated displacement fields to provide the target displacement amount for each vertex.

[0103] Optionally, displacement map data is essentially scalar or vector displacement information stored in texture form. This data can be generated offline during the preprocessing stage of the 3D engine or dynamically loaded at runtime according to a call command. In one specific implementation, a two-dimensional texture coordinate parameterized space is first generated for the subdivided mountain mesh, and then height values ​​reflecting the microscopic surface undulations are filled into this parameterized space. It should be noted that, in addition to program noise, height data can also be obtained by scanning the real rock surface and downsampling. The above description is not intended to limit the data source. For example, for the layered sedimentary structure of the rock surface, a grayscale image with continuous gradient characteristics can be generated based on a multi-level noise function, where the brighter areas correspond to the outward protrusions of the rock layers, and the lower-brightness areas correspond to pits or water traces formed by erosion. When the displacement map data is loaded, the geometry processing unit performs sampling according to the texture coordinates corresponding to the current vertex, and performs a linear transformation by combining the height value with a preset scaling factor and an offset reference value, thereby obtaining the actual distance that the vertex needs to be offset in the normal direction. This processing method creates realistic and perceptible volumetric undulations on the originally smooth, subdivided mesh surface, which in turn creates self-occlusion, projected edges, and broken edges under the interaction of light and shadow, significantly different from the planar visual illusion presented by normal maps that only change the lighting.

[0104] Optionally, in practical implementation, in addition to using a single-channel grayscale height map, the displacement map data can also be in the form of a multi-channel vector displacement map. This vector displacement map not only records the offset along the normal direction but also includes lateral perturbation information in the tangential space, thereby generating detachment bedding and lateral erosion grooves with lateral extension characteristics on the rock surface. Furthermore, considering the differences in processing capabilities of different display hardware, the above displacement map data can dynamically adjust its resolution and bit depth according to the performance indicators of the target platform. For example, in mobile devices or virtual reality scenes, a compressed low-bit-depth map can be used to reduce memory usage and vertex computation load. On host platforms that support real-time ray tracing, a medium-precision 16-bit floating-point height map can be loaded, while in high-end workstation rendering environments, an uncompressed 32-bit floating-point height map can be loaded to obtain sub-millimeter-level geometric precision. Based on this, the vertex displacement processing mentioned above can specifically include: dynamically adjusting the sampling level of the displacement map according to the view distance during the shader stage; reducing the map resolution and correspondingly reducing the displacement amplitude when the observation point is far from the rock surface; and gradually restoring full-precision sampling and complete displacement as the observation point gets closer, thereby maintaining frame rate stability while ensuring visual realism.

[0105] Optionally, the target displacement is used to quantify the normal offset distance of each vertex, and can be determined based on texture sampling values ​​or erosion simulation results.

[0106] Optionally, the target displacement is a pre-calculated geometric offset quantization index for each vertex in the subdivided mesh. This index can be derived from the texture sampling values ​​of the displacement map or from the erosion depth field output by the computational fluid dynamics simulation. In one specific implementation, when using grayscale displacement map data, the system first establishes a mapping relationship between mesh vertices and texture pixels, and then converts the sampled normalized grayscale values ​​into a target displacement with actual length units using a linear mapping formula. The slope and intercept of the mapping formula can be dynamically adjusted according to the weathering degree of the rock mass material. Severely weathered rock surfaces typically correspond to a larger displacement range, while newly exposed hard rock layers correspond to a smaller displacement range. When generating the target displacement using physical erosion simulation output, the system calculates the scouring and deposition amounts at each vertex based on the flow trajectory of virtual water on the subdivided mesh surface, and uses the algebraic sum of these two amounts as the target displacement for that vertex. It should be noted that the above two generation methods can be applied independently to the entire rock mass surface or superimposed and merged for the same area. This application does not intend to limit the generation logic of the target displacement.

[0107] In an optional implementation, after displacing the vertices of the subdivided mesh along the normal direction based on the displacement map data, the method further includes: performing surface normal perturbation processing on the 3D geometric model based on the normal map, whereby the normal map is used to characterize the microscopic unevenness variations on the surface of the 3D geometric model. Thus, by further perturbing the surface normal using the normal map, based on the geometric volume details established by the displacement map, the microscopic unevenness and lighting can be enhanced without significantly increasing the face count, effectively improving the realism and operational efficiency of rock rendering. In one implementation, after the vertices of the subdivided mesh have been displaced along the normal direction to generate a 3D geometric model with realistic cuts and broken edges, the rendering pipeline loads a tangent space normal map with a size of 2048×2048. This map pre-simulates the layered texture and joint structure of the sedimentary rock surface through procedural noise. In the pixel-level rendering calculation stage, the rendering program samples the normal map according to the texture coordinates corresponding to the current pixel, obtaining a normal offset vector representing local microscopic bump changes. This vector is then mapped to the world coordinate system using a transformation matrix composed of the tangents, binormals, and original normals at each vertex. Subsequently, this vector is fused and perturbed with the geometric surface interpolation normal after displacement mapping, so that the normals used for lighting calculations are no longer smooth and continuous, but instead exhibit subtle undulations consistent with the rock layer orientation. Furthermore, this normal map can independently control the joint depth of different orientations using red, green, and blue channels. The red channel controls the concavity and inclination of transverse bedding, the green channel controls the weathering and erosion degree of longitudinal fractures, and the blue channel participates in the normalization correction of the tangent space as a height reference. In this way, while maintaining a low geometric facet number, the rock surface can exhibit micron-level weathering traces with a volumetric illusion under lighting and shadow interaction, effectively compensating for the limitations of displacement mapping in sub-pixel-level geometric detail.

[0108] Optionally, the normal map is used to store the local normal offset vector of the rock surface, which contains at least multi-channel direction information in tangent space, and is intended to render stage perturbation interpolated normals to simulate micro-bump texture.

[0109] Optionally, the normal map can be a directional texture in tangent space or a directional texture aligned with the local coordinates of the model surface. In one implementation, each color channel of the normal map corresponds to a normal offset component along a different axis. For example, the red channel stores offset information in the horizontal tangent direction, the green channel stores offset information in the vertical subnormal direction, and the blue channel stores height reference component information related to the original surface normal. One objective of this embodiment is to enable the rock mass model to exhibit depth illusions through the collaborative work of the normal map and the displacement map, while simultaneously creating cut and broken edge effects at the geometric edges due to real vertex displacement. In other words, the displacement map is responsible for changing the macroscopic volume contour of the rock mass at the geometric level, while the normal map is responsible for simulating microscopic surface details such as rock layer peeling, weathering erosion, and joint structures through pixel-by-pixel or vertex-by-vertex normal perturbation without adding extra polygons. It should be noted that the above description of the normal map channel allocation and spatial coordinates is only one example, and this application does not intend to limit the specific encoding format of the normal map. In actual implementation, the normal map can be a high-precision texture baked offline, or a dynamic texture generated in real time by procedural noise. It can also be adaptively enhanced by combining the density gradient obtained from the analysis of the rock mass slice thermal map. This application does not limit this.

[0110] Optionally, considering the significant differences in weathering degree and depositional age among different rock masses, the generation method and resolution of the normal map can be dynamically adjusted according to the target environmental parameters. As one possible implementation, for relatively smooth limestone walls subjected to long-term water erosion, high-frequency noise-dominated normal details can be used to represent dense, layered watermarks and mineral crystal reflections. As another implementation, for severely weathered and significantly eroded volcanic rock surfaces, mid-to-low frequency breakage noise and crack masks can be integrated to generate a normal map with large areas of irregular depressions. To avoid obvious periodic patterns on the rock surface due to repeated tiling of a single texture, a macroscopic variability function based on world spatial coordinates can be introduced during the normal map processing, allowing the same type of rock mass in different spatial locations to exhibit unique local microscopic unevenness distributions. Furthermore, during the pixel-level calculation stage of the rendering pipeline, the sampling precision of the normal map can be dynamically graded based on the viewing distance. High-resolution normal sampling can be enabled for close-up viewing to render sharp textures, while low-resolution or blended detail layer normal textures can be switched for long-distance viewing to reduce memory bandwidth usage. It should be understood that normal maps and displacement maps can be stored as independent texture resources or merged into a composite texture containing both height and normal information. If stored merged, the rendering pipeline needs to separately resolve the displacement and normal perturbation amounts after sampling, and care must be taken to avoid distortion of the tangent space vector normalization due to compression algorithms.

[0111] Optionally, surface normal perturbation processing aims to correct the original normal direction of each pixel or vertex on the surface of the 3D geometric model based on the sampling results of the normal map, so as to enhance the lighting and shadow levels of the rock surface while maintaining a low face number.

[0112] Optionally, surface normal perturbation can be performed immediately after vertex displacement processing, or it can be delayed until before the lighting calculation stage, depending on the batch optimization strategy of the rendering pipeline. In one specific approach, the rendering calculation program first samples the perturbation normal vector in tangent space from the normal map using texture coordinates. Then, it transforms this vector into model space or world space using a transformation matrix composed of vertex tangents, binormals, and original normals. After the transformation, the perturbation normal vector is superimposed or replaced with the interpolated normals of the geometry after subdivision and permutation, according to specified weights, to obtain the final surface normals used for lighting calculations. Considering that cracks and joints on the rock surface should exhibit significant differences in brightness and darkness under different lighting angles, the above perturbation processing can also dynamically adjust the intensity coefficient of the normal perturbation in conjunction with the direction of the main light source. For example, when the line of sight is nearly parallel to the direction of the main light source, the amplitude of the normal perturbation can be appropriately increased to highlight the unevenness of the rock layer peeling, while when the angle between the line of sight and the light source direction is large, the basic perturbation amount can be maintained to avoid overexposure. It should be noted that the above-mentioned normal perturbation can not only be applied to the basic lighting model, but also further affect the sampling direction of specular reflection, ambient light occlusion and even semi-transparent scattering, thereby improving the realism of the rock surface in multiple dimensions.

[0113] In an optional implementation, acquiring displacement map data includes: performing physical erosion simulation on the subdivided mesh to obtain the displacement distribution of each vertex of the subdivided mesh; the physical erosion simulation includes calculating the vertex erosion amount based on fluid flow on the mesh surface and determining the vertex deposition amount based on the vertex erosion amount; and generating displacement map data based on the displacement distribution. In this way, generating displacement map data through physical erosion simulation gives microscopic details the physical plausibility of geological processes, utilizes the dynamic relationship between erosion and deposition to shape natural layers, and significantly enhances the realism of the rock surface and the credibility of the landform.

[0114] In one implementation, for the aforementioned subdivided mesh with the target mesh density, the system initiates a physical erosion simulation. Specifically, the simulation simulates water flow along the surface of the rock mass mesh: in steep, convex regions, the scouring intensity of the water flow is calculated, and the amount of material carried away by each vertex in that region is taken as the vertex erosion amount; in concave regions where the water flow velocity slows down, the material accumulation thickness is calculated according to the rules of material deposition, and this is taken as the vertex deposition amount. The net displacement of each vertex is obtained by combining the data from each iteration to form a displacement distribution, and finally, displacement map data is generated based on this. Optionally, the physical erosion simulation is used to calculate the scouring and material transfer of fluids on the rock mass surface to generate a geometric displacement basis that conforms to natural laws.

[0115] Optionally, physical erosion simulation uses numerical calculations to simulate the long-term erosion and material transfer process of fluids on the rock surface, including key steps such as fluid path tracking, erosion intensity assessment, and material redistribution. In this simulation, the system can establish virtual fluid particle emission sources on the surface of a subdivided mesh, allowing the fluid particles to slide along the tangential plane of the mesh normal and naturally converge under gravity, following a topographically driven rule from high to low. As the fluid particles move, their kinetic energy exerts varying degrees of stripping effect on the mesh vertices they pass through; the intensity of this effect can be determined comprehensively based on fluid velocity, local curvature, and a preset rock hardness coefficient. It should be noted that the fluid particles can be simulated using either the mesh tracking method under the continuous medium assumption or the discrete particle method for calculating individual trajectories; this application does not limit the specific fluid solution framework. Furthermore, during a complete simulation cycle, the material stripped from the rock surface does not disappear into thin air, but continues to migrate with the fluid particles until it settles when the fluid kinetic energy is insufficient to support the material transport. This forms a closed-loop calculation logic of erosion and deposition, providing displacement distribution data for the displacement map that is both physically reasonable and visually layered.

[0116] Optionally, considering the discontinuous media characteristics such as joints and fissures in the local geological structure of the cave rock mass, to avoid erosion simulations that do not conform to the laws of gravity at steep cliffs or suspended rock walls, the above simulation process can also incorporate height field constraints and collapse detection mechanisms. For example, the system can pre-generate a height field buffer for the subdivided mesh, and before each erosion iteration, determine whether the normal direction of the target vertex's region meets the deposition angle condition: when the local dip angle exceeds a preset repose angle threshold, the eroded material, even if not yet transported by fluids, will slide directly under gravity and form a natural deposit at the toe of the slope. In this way, by coupling the fluid erosion logic with the solid collapse logic, the micro-surface morphology carried by the displacement map not only includes brush marks like flowing water but also reproduces the rough fracture edges caused by rock fragmentation. It should be noted that the above collapse detection mechanism can be executed as a post-processing step in each erosion iteration or can be performed in parallel with the fluid erosion calculation; this application does not limit this.

[0117] Optionally, fluid is used to form directional flow paths on the mesh surface to drive the calculation of vertex erosion and vertex deposition.

[0118] Optionally, fluid flow on the grid surface refers to the diffusion and confluence of virtual liquid or gas media along the subdivided grid surface following the terrain gradient. The flow path can be determined by calculating the gradient direction from the scalar field of the grid surface. In one implementation, the system can convert the local elevation information of the subdivided grid into a two-dimensional flow field driving source, using stream functions or potential functions to calculate the tangential velocity vector at each vertex, causing the fluid to converge into streams in concave areas and disperse into curtains in convex areas. It should be noted that the fluid is not limited to a single aqueous medium; it can also be set to groundwater rich in organic acids, high-viscosity mud flow, or high-speed airflow containing abrasive mineral sands, depending on the cave environment. Different fluid media correspond to different erosion coefficients and deposition thresholds. Furthermore, to reduce the computational cost of the global simulation, the fluid activity range can be constrained to a spatial area centered on the target cave entrance, with high-fidelity flow calculations only initiated for the grid segments visible to the player or those about to be reached, thus ensuring immersion while meeting the performance requirements of real-time generation.

[0119] Optionally, the vertex erosion is used to quantify the height of material lost at the vertex due to fluid erosion, and the value is determined based on the fluid kinetic energy and rock hardness.

[0120] Optionally, vertex erosion is used to quantify the material height or volume lost by a single vertex in the subdivided mesh due to fluid erosion during the simulation cycle. Its value is determined by the ratio of the fluid kinetic energy flowing through the vertex to the erosion resistance of the rock mass at that vertex. In one implementation, the vertex erosion can be expressed as the product of the projected component of the fluid velocity vector along the vertex normal and the rock mass hardness coefficient. The higher the flow velocity and the softer the rock mass, the greater the erosion value of that vertex. It should be noted that to prevent continuous high-intensity erosion from causing holes or topological fractures in the mesh, the system can set a maximum erosion depth limit for a single vertex. When the cumulative erosion reaches this limit, the erodible amount of that vertex is reduced to zero or switched to a fracture and collapse mode in subsequent iterations. Furthermore, the calculation of vertex erosion is not limited to being affected only by local flow velocity; the upstream catchment area can also be used as a gain factor to reflect the natural hydrological law that the greater the convergence flow, the stronger the erosion, allowing for a more profound erosion characterization of key geomorphic features such as drip lines at the top of caves or spillways on cliff walls.

[0121] Optionally, the vertex deposition is used to quantify the incremental height of material carried by the fluid that settles and accumulates at the vertex due to the decrease in flow velocity.

[0122] Optionally, the vertex deposition amount characterizes the incremental height of material carried by fluid erosion from the upstream region, which settles and accumulates at the vertex due to reduced flow velocity. Its value is positively correlated with the current sediment concentration carried by the fluid and the local topographic resistance coefficient. In one implementation, when fluid flows into a low-lying area along the grid surface or encounters a reverse slope, its kinetic energy decays, causing suspended material to gradually unload. At this point, the system calculates the deposition height at the vertex based on the remaining kinetic energy and the angle of repose, and adds this height to the vertex deposition amount. It should be noted that the deposited material can cover the original rock surface to form new texture layers, or partially fill existing fissures, thereby changing the normal orientation and micro-surface curvature of the local grid. Furthermore, to prevent the infinite accumulation of deposition from causing unreasonable topographic protrusions, the system can introduce a compaction coefficient to gradually attenuate the historical deposition amount, simulating the cyclical state of sediments consolidating into rock and being eroded again over time in real geological processes, so that the displacement field output by the displacement map tends to reach dynamic equilibrium in long-term iterations.

[0123] Optionally, the displacement distribution describes the set of net displacement heights of each vertex after erosion and deposition calculations, serving as the basis for generating the displacement map. Optionally, the displacement distribution refers to the collective description of the net displacement heights of each vertex in the subdivided mesh relative to its original position after completing the physical erosion simulation. This distribution data is typically encoded and stored in the form of a two-dimensional texture or vertex attribute array. In one implementation, the system can assign a scalar displacement value to each mesh vertex. This value is the net result of the algebraic sum of the vertex erosion and deposition amounts: a positive value indicates that material accumulation at the vertex causes surface bulging, a negative value indicates that erosion causes surface concavity, and a zero value maintains the original surface elevation. It should be noted that, in addition to being directly used to generate displacement map data, the above displacement distribution can also be simultaneously converted into color mapping information for visualization and adjustment. For example, warm colors can represent depositional areas, and cool colors can represent erosion areas, helping creators intuitively assess the rationality of micro-topography. Furthermore, the displacement distribution can be sampled and interpolated between different resolution levels, using a low-resolution coarse-grained displacement field to guide high-resolution detail sculpting, thereby ensuring that the displacement map maintains spatial continuity and visual consistency during the switching of multiple detail levels.

[0124] like Figure 9 This diagram illustrates the fusion of macroscopic rock mass geometry and microscopic texture, providing a global view of the complete rock mass model and showcasing the final output of the procedural generation system. The model seamlessly integrates macroscopic cave and tunnel structures, mesoscopic surface undulations and deformations, and microscopic textural details to create a highly realistic natural rock mass.

[0125] According to one embodiment of the present application, the three-dimensional geometric model generation apparatus may include: The data acquisition module is used to acquire the initial mesh data of the basic geometry; The noise perturbation module is used to perform noise perturbation processing on the basic geometry based on the initial mesh data to obtain the initial deformed mesh; The volume subtraction module is used to perform volume subtraction operations on the initial deformed mesh based on the hollowing control parameters to obtain a hollowed-out mesh with the target hollowed-out area. The subdivision processing module is used to perform surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with the target mesh density; The vertex displacement module is used to perform vertex displacement processing on the subdivided mesh based on the surface displacement control data corresponding to the subdivided mesh, so as to obtain a three-dimensional geometric model with target geometric details.

[0126] In this way, integrating noise perturbation, volume reduction, and vertex displacement into the same subdivision process helps reduce the number of redundant faces and rendering overhead, enabling the progressive construction of complex models from macroscopic to microscopic geometric details, improving generation efficiency and geometric accuracy, and helping to alleviate the pressure on geometric data storage and computation processing.

[0127] The specific details of each part of the above-mentioned device have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0128] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the exemplary embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0129] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0130] The following is a detailed reference. Figure 10The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 1201, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1202 or a program loaded from memory 1208 into random access memory (RAM) 1203. The RAM 1203 also stores various programs and data required for the operation of the electronic device. The processor 1201, ROM 1202, and RAM 1203 are interconnected via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.

[0131] Typically, the following devices can be connected to I / O interface 1205: input devices 1206 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 1207 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 1208 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1209. Communication device 1209 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 10 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0132] In particular, according to one embodiment of this application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, one embodiment of this application includes a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via communication device 1209, or installed from memory 1208, or installed from ROM 1202. When the computer program is executed by processor 1201, it performs the functions defined in the methods described above in various embodiments of this application.

[0133] Figure 10 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0134] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium may also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0135] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0136] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for generating a three-dimensional geometric model, characterized in that, include: Obtain the initial mesh data for the basic geometry; Based on the initial mesh data, noise perturbation processing is performed on the basic geometry to obtain an initial deformed mesh; Based on the cutout control parameters, a volume subtraction operation is performed on the initial deformed mesh to obtain a cutout processed mesh with the target cutout region. Perform surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with the target mesh density; Based on the surface displacement control data corresponding to the subdivided mesh, vertex displacement processing is performed on the subdivided mesh to obtain a three-dimensional geometric model with target geometric details.

2. The method according to claim 1, characterized in that, The noise perturbation processing of the basic geometry includes: The target noise field is generated based on a multi-scale noise function, which includes low-frequency noise components, mid-frequency noise components, and high-frequency noise components. The vertex positions of the basic geometry are offset according to the target noise field to obtain the initial deformed mesh.

3. The method according to claim 1, characterized in that, The volume subtraction operation performed in the initial deformed mesh based on the hollowing control parameters includes: A hollow shell is generated according to the hollow control parameters, and the hollow shell is a reference geometry stretched along the target path; The hollowed-out shell is subjected to a Boolean subtraction operation from the initial deformed mesh to obtain the hollowed-out processed mesh.

4. The method according to claim 1, characterized in that, The surface subdivision processing of the hollowed-out mesh includes: Determine the surface curvature of each region of the hollowed-out mesh; Based on the surface curvature, the curvature change of each region is determined. Adaptive mesh refinement is performed in regions where the curvature change is greater than the target threshold, while the basic topology density is maintained in regions where the curvature change is less than the target threshold, thus obtaining the subdivided mesh.

5. The method according to claim 1, characterized in that, The step of performing vertex displacement processing on the subdivided mesh based on the surface displacement control data corresponding to the subdivided mesh includes: Obtain displacement map data, which is used to characterize the target displacement of each vertex of the subdivided mesh; Based on the displacement map data, the vertices of the subdivided mesh are displaced along the normal direction to obtain the three-dimensional geometric model.

6. The method according to claim 5, characterized in that, After displacing the vertices of the subdivided mesh along the normal direction based on the displacement map data, the method further includes: The surface normal perturbation of the three-dimensional geometric model is performed based on the normal map, which is used to characterize the microscopic unevenness of the surface of the three-dimensional geometric model.

7. The method according to claim 1, characterized in that, After performing noise perturbation processing on the basic geometry, the method further includes: Obtain the volumetric property field of the initial deformed mesh, and calculate the gradient distribution of the volumetric property field; The material interface position is determined based on the gradient distribution, and the material interface position is used to characterize the transition region of different material properties.

8. The method according to claim 7, characterized in that, Determining the material interface location based on the gradient distribution includes: The region in the gradient distribution whose gradient value is greater than the target gradient threshold is defined as the material interface location; A crack structure is inserted at the material interface, and the orientation of the crack structure is determined based on the direction of the gradient value.

9. The method according to claim 2, characterized in that, The method further includes: Obtain the reference geometry of the auxiliary structure; Based on the multi-scale noise function, noise perturbation processing is performed on the reference geometry of the auxiliary structure to obtain the deformed mesh of the auxiliary structure. The surface subdivision and vertex displacement processes are performed on the deformed mesh of the attached structure to obtain the geometric model of the target attached structure.

10. The method according to claim 2, characterized in that, The method of generating the target noise field based on a multi-scale noise function further includes: The weighting coefficients of each frequency component in the multi-scale noise function are dynamically adjusted based on the target environment parameters, which are used to characterize the degree of influence of the target environment on the geometric shape. The target noise field is regenerated based on the updated weighting coefficients, and the vertex positions of the basic geometry are offset based on the regenerated target noise field to obtain the initial deformed mesh with target erosion characteristics.

11. The method according to claim 5, characterized in that, The acquisition of displacement mapping data includes: Physical erosion simulation is performed on the subdivided mesh to obtain the displacement distribution of each vertex of the subdivided mesh. The physical erosion simulation includes calculating the vertex erosion amount based on the fluid flow on the mesh surface and determining the vertex deposition amount based on the vertex erosion amount. The displacement map data is generated based on the displacement distribution.

12. A three-dimensional geometric model generation device, characterized in that, include: The data acquisition module is used to acquire the initial mesh data of the basic geometry; The noise perturbation module is used to perform noise perturbation processing on the basic geometry based on the initial mesh data to obtain an initial deformed mesh; The volume subtraction module is used to perform a volume subtraction operation on the initial deformed mesh based on the hollowing control parameters to obtain a hollowed-out mesh with the target hollowing area. The subdivision processing module is used to perform surface subdivision processing on the hollowed-out mesh to obtain a subdivided mesh with a target mesh density; The vertex displacement module is used to perform vertex displacement processing on the subdivided mesh based on the surface displacement control data corresponding to the subdivided mesh, so as to obtain a three-dimensional geometric model with target geometric details.

Images