An Unreal Engine-based interactive three-dimensional convex hull Decal generation and ray interaction detection method

Abstract

This invention relates to the field of 3D model processing technology and discloses an interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine. The method includes: obtaining the local mesh vertex coordinates and normal vectors of the receptor model; calculating the directional offset of adjacent normal vectors to determine the local surface curvature and calculating weights; constructing an initial convex proxy using coordinates, adjusting the patch density according to the weights, and generating a virtual convex proxy that matches the topology of the receptor surface by driving vertex offset contraction; performing an intersection test with the interaction ray using the virtual convex proxy, and obtaining interaction feedback parameters by interpolating the receptor vertex normals based on the centroid coordinates of the hit point. This invention evolves the Decal from a projection mapping to a geometric entity, solving the projection stretching and depth conflict defects of high-curvature receptor surfaces, and achieving physical response consistency while reducing the computational load of interaction detection.

Description

Technical Field

[0001] This invention relates to an interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine, belonging to the field of 3D model processing technology. Background Technology

[0002] Currently, in the process of presenting local attribute mapping on the surface of a 3D model, the industry generally adopts the projection matrix method, defining Decal as a projection box. Its combination with the surface of the receptor model only exists in pixel mixing at the rendering level. With the evolution of high-fidelity graphics interaction requirements, the topological complexity of the receptor model has increased. For receptor models with high curvature surfaces, porous structures, or dynamic deformation characteristics, this method based on flat projection faces physical constraints.

[0003] Because the projection box uses a single sampling reference in a local space, texture stretching and pixel distortion occur when there is a large angle between the receptor surface normal and the projection direction. Furthermore, the lack of geometric topological association between the Decal and the receptor leads to a conventional approach of relying on triangular-face collision detection for the receptor mesh to achieve interactive judgment. This results in an exponential increase in computational load with the topological complexity of the receptor, creating an irreconcilable contradiction between interactive accuracy and real-time performance. In addition, in industrial digital twins or high-precision game scenarios, the receptor model often undergoes dynamic deformation due to skeletal animation or vertex offsets. The static projection area struggles to adapt to this topological evolution, causing a disconnect between the Decal and the receptor surface. These shortcomings reflect the limitations of software... The control logic of the device has shortcomings in terms of refined geometric perception and real-time response mechanism. For example, Chinese invention patent with publication number CN115409956A discloses a three-dimensional reconstruction and self-correction system based on multiple cameras and its application. It focuses on the overall environmental geometry restoration and long-cycle pose maintenance. When processing the Decal entity that needs to be deeply coupled with the local manifold of the receptor, the algorithm's computing power allocation is unbalanced. It lacks an adaptive energy contraction mechanism for the local curvature field of the receptor. There is a bridging gap between the geometric proxy and the receptor in the non-convex surface topology. It relies on image layer optimization and global correction logic calculation with high latency and cannot support sub-second physical response feedback. The Decal entity is difficult to perform high-precision pose evolution with the deformation of the receptor.

[0004] Therefore, how to construct a geometric proxy mechanism that has topological awareness and takes into account detection efficiency, so that Decal can evolve from projection information into a geometric entity isomorphic to the receptor model manifold, is the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of this invention is as follows: An interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine, comprising the following steps: Step S1: Obtain the local mesh data of the receptor model in the Unreal Engine scene space, and extract the spatial coordinates of the vertices and the corresponding normal vectors from the local mesh data; Step S2: Calculate the normal direction deviation between adjacent normal vectors in the local mesh data, and use the normal direction deviation to determine the surface curvature of the local surface of the receptor model, and calculate the patch subdivision weight based on the surface curvature. Step S3: Construct an initial convex proxy body using spatial coordinates. Perform non-uniform distribution adjustment on the surface density of the initial convex proxy body according to the surface subdivision weight. Specifically, increase the number of subdivision surfaces in areas where the surface curvature is greater than a preset curvature threshold, and drive the surface vertices of the initial convex proxy body to move a certain offset shrinkage distance towards the geometric center of the local mesh data. Perform surface merging in areas where the surface curvature is not greater than a preset curvature threshold to generate a virtual convex proxy body. Step S4: Obtain the emission parameters of the interactive ray, and use the virtual convex proxy to perform an intersection test on the interactive ray to obtain the hit point of the interactive ray on the surface of the virtual convex proxy. Step S5: Calculate the centroid coordinates of the hit point on the virtual convex proxy surface, perform weighted interpolation on the vertex normals of the receiver model based on the centroid coordinates, and obtain the interactive feedback parameters containing the hit point coordinates and the interpolated normal vectors.

[0006] Preferably, the logic for calculating the surface curvature of the local surface in step S2 follows the following calculation rules: using the formula Calculate the normal of the center vertex and adjacent vertex normals The sum of the dot products, where The normal direction deviation is used to characterize the gradient of surface curvature variation, and the gradient is mapped to patch subdivision weights; where... and All are unit vectors extracted from local grid data.

[0007] Preferably, in the process of generating the virtual convex proxy in step S3, the method further includes: extracting the boundary points of the local mesh data and using normal consistency verification to identify the orientation of the front normal of the receptor model; for receptor models with thickness values ​​lower than a preset thickness threshold, performing topological culling on the back face of the initial convex proxy using the orientation of the front normal, and shrinking the initial convex proxy into a one-sided open topological mesh that matches the front manifold of the receptor model.

[0008] Preferably, step S4, which involves performing the intersection test, further includes the following judgment rules: calculating the penetration path step length of the interactive ray in the local space of the receptor model, with the unit of the penetration path step length being mm; establishing an occlusion judgment model by combining the surface curvature features of the virtual convex proxy; comparing the penetration path step length with the preset occlusion threshold in the occlusion judgment model; and identifying and filtering false hit signals generated in the occlusion structure of the receptor model when the penetration path step length exceeds the occlusion threshold.

[0009] Preferably, before obtaining the local mesh data of the receptor model in step S1, the method includes: defining a three-dimensional projected bounding box with a preset geometric center in the coordinate space of the receptor model; detecting the intersection state of the mesh cells of the receptor model with the three-dimensional projected bounding box; and defining the set of mesh cells with the intersection state as local mesh data.

[0010] Preferably, after obtaining the interactive feedback parameters in step S5, the method further includes: extracting the vertex displacement vector of the receptor model under the skeleton deformation drive in real time; reusing the centroid coordinates of the hit point, performing displacement mapping on the face vertices of the virtual convex proxy body according to the vertex displacement vector, so that the virtual convex proxy body performs synchronous position evolution with the deformation of the receptor model.

[0011] Preferably, the process of the virtual convex proxy body performing synchronous position evolution includes: obtaining the mesh deformation amount before and after the deformation of the receptor model; and performing Laplacian smoothing on the facets of the virtual convex proxy body according to the mesh deformation amount to maintain the surface continuity of the virtual convex proxy body after deformation.

[0012] Preferably, in step S3, the offset shrinkage distance is proportional to the surface curvature, and the unit of the offset shrinkage distance is mm.

[0013] Preferably, step S5, which involves obtaining interactive feedback parameters, further includes: reconstructing the material rendering parameters of the virtual convex proxy using the centroid coordinates; and using the reconstructed material rendering parameters to fuse the Decal texture with the surface ambient occlusion data of the receptor model, thereby outputting a Decal geometric entity with consistent rendering performance.

[0014] Preferably, after obtaining the interaction feedback parameters in step S5, the method further includes: sending the interaction feedback parameters to the interaction event distribution interface of Unreal Engine; and using the interaction event distribution interface to trigger response instructions based on the interaction feedback parameters. The response instructions include modifying the material properties of the recipient model, transmitting physical impulse to the recipient model, or calling the audio module to play special effects audio.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. In the generation of 3D convex hull Decals, the physical fit between the Decal geometry and the receiver surface is improved. By constructing a virtual convex hull entity with normal constraints in the local space of the receiver, the traditional projection matrix method is replaced, allowing the Decal to evolve from pixel blending at the rendering level to a topological manifold with geometric features. By extracting the feature point cloud of the receiver model in the neighborhood of the hit point and generating a 3D convex proxy, a consistent alignment relationship between the proxy and the normal field of the receiver surface is established, eliminating the geometric stretching and sampling distortion phenomena that traditional projection boxes inevitably produce when facing high curvature or porous topologies. Combined with vector translation performed along the local average normal direction of the receiver, it is ensured that the convex proxy maintains a stable geometric spacing while covering the receiver surface, avoiding the uncertainty in depth cache determination and eliminating depth conflicts and flickering defects in the rendering process.

[0016] 2. Achieve order reduction and decoupling between interaction detection overhead and model topology complexity. The triangular-face collision detection for complex receptor meshes is transformed into geometric determination for minimal convex surrogate bodies. By utilizing the separating axis theory followed by convex polyhedra in ray intersection calculation, the spatial retrieval depth and computational load in large-scale scenes are significantly reduced. During the interaction response, the coordinates of the ray hit point are directly mapped to the Decal texture coordinate space through the centroid coordinate transformation, avoiding logical delays under high-frequency interactions. This interaction method based on computational geometry features makes the system's operating efficiency no longer limited by the vertex size of the receptor mesh, achieving a balance between computational performance and physical feedback accuracy.

[0017] 3. Enhance geometric synchronization and logical stability in dynamic deformation environments. By establishing a topological anchor point weight binding mechanism, the geometric state of the convex proxy is transformed from static coordinate-driven to receptor topology-driven. By using the centroid coordinate weight to linearly associate the proxy vertices with the receptor triangles, when the receptor model undergoes dynamic deformation due to skeletal animation or vertex offset, the proxy vertices synchronously update their spatial coordinates according to the receptor displacement, eliminating the need for a time-consuming geometric reconstruction process and solving the clipping and suspension bottlenecks of Decal in dynamic scenes. This displacement synchronization method based on the topological conservation law ensures that the interactive entity maintains physical consistency throughout the receptor evolution process. Attached Figure Description

[0018] Figure 1 This is a flowchart of the interactive 3D convex hull Decal generation and ray detection execution process of the Unreal Engine in this invention; Figure 2 This is a schematic diagram of the data flow and closed-loop logic of the receptor topology-induced geometric proxy architecture of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to specific embodiments. However, the following embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0020] An interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine includes a local feature sampling unit, a normal divergence-guided convex hull construction unit, and a physical parameter backtracking unit based on centroid coordinates. The processing flow begins with the original mesh vertices and normal tensor field of the receptor model. Local receptor mesh data is extracted by defining a 3D projected bounding box in the scene space. The surface curvature is quantified by the directional deviation of adjacent normal vectors, thereby driving the non-uniform patch subdivision and vector contraction of the virtual convex proxy. After ray intersection testing, the texture space mapping and physical interaction normal reconstruction are realized by using the centroid coordinate algorithm, achieving physical conformity between the Decal and the high curvature receptor surface.

[0021] When processing local attribute mapping on the surface of a 3D model, to address sampling distortion caused by the non-uniformity of the normal distribution in the receptor model, a precise sampling reference is established. A 3D projection bounding box with a preset geometric center is defined within the coordinate space of the receptor model. Its size is determined based on the preset coverage area of ​​Decal, with the side length as... Taking a cube as an example, by detecting the intersection state between the mesh cells of the receptor model and the 3D projected bounding box, the set of intersecting mesh cells is determined as local mesh data, and the spatial coordinates of the vertices in the mesh cells are extracted. and the corresponding normal vector To characterize the complexity of the receptor surface, the system calculates the deviation of the normal direction between adjacent normal vectors. This deviation follows the calculation rules: using the formula... Calculate the normal of the center vertex and adjacent vertex normals The sum of the dot products, where, and All are unit vectors extracted from local grid data. The number of neighboring vertices. The normal deviation is used to characterize the gradient of surface curvature changes. The system maps this gradient to patch subdivision weights, enabling the system to acquire dense geometric sampling points in regions of dramatic topological changes. This ensures that the generated proxy captures the detailed features of the receptor surface, thus constructing the normal deviation. With vertex offset vector The nonlinear mapping is used to calculate the contraction displacement. benchmark value Based on Unreal Engine's camera near clipping plane depth tolerance experiments, the tolerance is determined to be 0.2mm. Logical judgment: If... Radius triggers a non-uniform subdivision operator, inserts sampling midpoints in the neighborhood and reconstructs the local convex hull patch index. For the redundant patches after subdivision, edge collapse processing based on the quadratic error metric QEM algorithm is performed. The collapse cost is determined by the sum of the squared distances from the vertex to the associated patch. By constraining the total number of patches on one side to not exceed 10% of the total number of local meshes, the energy minimization fitting of the normal field of the surrogate body and the recipient surface is achieved while maintaining the convex hull envelope characteristics, thus solving the texture stretching defect under high curvature surfaces.

[0022] To address the manifold mismatch problem caused by phase misalignment between the proxy patch and the receptor surface normal, the system utilizes spatial coordinates. An initial convex proxy is constructed, and its patch density is adjusted non-uniformly according to the patch subdivision weights. The number of subdivided patches is increased in regions where the surface curvature is greater than a preset curvature threshold, which is set to [value missing]. The radius is determined by the radius, and the initial convex surface vertex is moved by an offset shrinkage distance towards the geometric center of the local mesh data, which is proportional to the surface curvature. The reference value is set to... At the peak of curvature, the shrinkage amount is adaptively increased to offset the bridging gaps generated by the convex hull on the non-convex surface. In regions where the surface curvature is no greater than a preset curvature threshold, patch merging is performed to simplify the topology, generating a virtual convex proxy that matches the receptor surface topology. This adaptive collapse mechanism ensures a high fit between the virtual convex proxy and the receptor normal field, solving the texture stretching problem on high-curvature surfaces. To address computational redundancy in low-curvature regions, the system performs topology merging in regions where the surface curvature is no greater than a preset curvature threshold. This is achieved by calculating the quadratic error matrix of the patch vertices and identifying the topological edge with the minimum folding cost for collapse, until the patch density of the virtual convex proxy meets a preset simplification benchmark, i.e., the total number of patches on a single side is no greater than the total number of patches in the local mesh. This ensures that the topology size of interaction decisions is reduced while maintaining the convex hull envelope characteristics.

[0023] To achieve efficient physical interaction, the system acquires the emission parameters of the interaction ray, including the ray's origin coordinates, direction vector, and maximum detection distance. It then uses a virtual convex proxy to perform intersection tests on the interaction ray, employs a separating axis algorithm to determine the geometric intersection state, and obtains the coordinates of the hit point on the virtual convex proxy surface. The system then calculates the weight of the centroid coordinates of this hit point on its corresponding facet. , and We perform weighted interpolation on the original vertex normals of the receptor model using the barycentric coordinate weights. If the normal vectors of the three vertices are respectively... , , The reconstructed physical interaction normals satisfy ,in, , and Let be the centroid coordinate components of the hit point relative to the vertices of the face, and let their algebraic sum be . The system uses barycentric coordinates to map the hit point to the Decal texture coordinate space, specifically using the formula... Determine the texture space mapping result of the interaction hit point, where, These are the mapped texture coordinates. , and The system uses preset texture coordinates for the vertices of the facet, combined with the reconstructed physical interaction normals. Output interactive feedback parameters and send them to the interactive event distribution interface to trigger response commands such as modifying material properties, transmitting physical impulse, or playing special effects audio, and establish a step size based on the penetration path. The occlusion determination model uses the formula Real-time calculation of dynamic occlusion threshold Physical thickness The value is 2mm, and the local surface inclination angle is... By normal deviation The mapping is generated, and the correlation coefficient is derived from the X-ray penetration test of standard aerospace skin parts. The execution logic determines whether the X-ray hit point satisfies the following conditions: The signal was identified as a false trigger signal caused by back-side self-occlusion. The masking operator automatically blocked the issuance of instructions from the interactive event distribution interface, performed unilateral open topology reconstruction on the thin-walled receptor, and calculated the surrogate patch normals. With the receptor's frontal normal If the direction cosine is less than 0, the redundant patch is removed, causing the initial convex surrogate to shrink into a one-sided open mesh, thus eliminating transmission defects on the thin-walled acceptor.

[0024] When dealing with thin-walled structures or complex occlusion conditions, the system extracts the boundary points of the local mesh data and uses normal consistency verification to identify the orientation of the frontal normal of the receptor model. For receptor models with a thickness value lower than a preset thickness threshold, this preset thickness threshold is set to... The system utilizes the orientation of the front normal to perform topological culling on the back facets of the initial convex proxy, shrinking it into a single-sided open topological mesh to address the transmission defects of Decal on thin-walled receptors. The system calculates the penetration path step size of the interactive rays in the local space of the receptor and establishes an occlusion judgment model based on surface curvature features. It compares the penetration path step size with the preset occlusion threshold in the occlusion judgment model. When the penetration path step size exceeds the occlusion threshold, the system identifies and filters false hit signals generated by the receptor model from the occlusion structure. In addition, for receptor models with dynamic deformation characteristics, the system extracts the vertex displacement vectors of the receptor model in real time under the drive of skeleton deformation, repeatedly uses the centroid coordinates of the hit points to map these displacements to the vertices of the virtual convex proxy facets, and uses Laplacian smoothing to maintain the surface continuity after deformation, so that the virtual convex proxy evolves synchronously with the receptor model.

[0025] Example 1: In the digital twin inspection of high-pressure turbine blades for aero-engines, the turbine blades have a hyperbolic surface and an array of cooling holes. When the surface of the receiver is attached using a projection matrix method, the divergence of the local normal distribution and the geometric jumps at the hole edges cause stretching distortion and texture drift in the projection information under a single sampling reference. Furthermore, performing triangular-face ray detection on this mesh results in a system response delay exceeding [a certain value]. This cannot meet the real-time physical feedback requirements of the simulated environment. For the above conditions, the system defines a side length of [missing information] in the coordinate space of the receptor model. The 3D projected bounding box is used to extract the spatial coordinates of vertices from the local mesh data. and the corresponding normal vector Using the formula Calculate the deviation of the normal direction, where, This refers to the deviation in the normal direction. The normal to the center vertex, The normal vectors of adjacent vertices. The number of adjacent vertices. For vertex index.

[0026] The system determines the surface curvature based on the deviation of the normal direction. When the surface curvature is greater than... When dealing with radians, the number of subdivision patches on the initial convex surrogate is increased, and the vertices of the patches are driven to move towards the geometric center by an offset contraction distance, which increases with the surface curvature. The baseline value is set to... Perform ray intersection tests using the generated virtual convex proxy and calculate the centroid coordinate weights of the hit points on the face. , and Using the formula Reconstruct the physical interaction normals, where, For the reconstructed physical interaction normals, , and For the weight of the centroid coordinates, , and The system uses the original vertex normals of the receptor model. By converting the mesh-based decision into a virtual convex surrogate body decision based on the normal field fitting, the system reduces the ray detection load to that of the original mesh detection load of the receptor model while maintaining an interaction accuracy no lower than the original mesh detection standard. The following avoids the phase mismatch phenomenon caused by traditional techniques on porous surfaces. Even in dynamic conditions where the receptor performs animation deformation, the physical consistency of the Decal surface is maintained by using the centroid coordinates of the hit point to achieve synchronous movement between the virtual convex proxy and the receptor vertex, thus eliminating depth conflicts of the rendered surface and feedback deviations of the interaction logic.

[0027] Example 2: In a geometric simulation environment with multidimensional topological gradients, the system tests the physical conformal performance of the receptor topology-induced geometric surrogate architecture on receptor surfaces with different curvatures and the ray interaction response delay. The experimental data originates from a set of sampling points built on a three-dimensional geometric computing platform, simulating... The physical coordinate field with spatial resolution is actively superimposed on the experimental signal source, and the signal-to-noise ratio is... The coordinate position jitter noise is used to simulate sampling error. The offset shrinkage distance is set to balance the molding risk and geometric conformal accuracy of the Decal surface. The decision logic of this parameter follows the positive correlation mapping rule between surface curvature and shrinkage vector, and the deviation in the normal direction is adjusted accordingly. As the distance between the vertex of the patch and the geometric center increases, the shrinkage distance tends towards a set upper limit, where, The deviation is in the normal direction; in the benchmark test, the side length of the three-dimensional projection bounding box is selected as... The preset curvature threshold is set to Radius; The experiment setup included the present invention sample group, control group 1, and control group 2. The present invention sample group adopted a complete scheme based on adjusting the surface density according to normal divergence. Control group 1 removed the normal divergence adjustment step and only used a basic convex polyhedron as a proxy. Control group 2 set the preset curvature threshold to [value missing]. The curvature is adjusted to allow the curvature recognition range to exceed the defined optimal range. The system collects data under three difficulty gradients: low curvature, medium curvature, and high curvature, and observes the root mean square error of geometric deviation and interaction delay.

[0028] Table 1: Comparison of Geometric Characteristics and Interaction Performance under Different Receptor Curvatures

[0029] See Table 1, as the receptor surface curvature from The radius increases to To minimize the geometric deviation root mean square error, the present invention dynamically adjusts the number of subdivisions of the facets and simultaneously increases the offset shrinkage distance. The following demonstrates stable geometric conformal capability when the curvature exceeds a preset curvature threshold. After curvature, due to the lack of topological collapse compensation induced by normal divergence, the root mean square error of geometric deviation in control group 1 shows a linear increasing trend with increasing curvature. At a curvature of [missing information], [missing information]. The error in radians reached This results in a floating texture effect. Control group 2, due to an excessively high curvature threshold, exhibits this effect when the curvature is at a certain level. radians to The failure to trigger non-uniform patch subdivision within the radian range, and the similarity in fitting accuracy to control group 1, proves... The role of radian as a performance inflection point parameter in capturing surface details, while the interactive detection delay of the sample group of this invention remains constant under various curvature gradients. The following is a comparison of the results generated by performing ray detection on the original high-precision mesh of the receptor model. The above delay reduces the computational overhead of the system.

[0030] Example 3: This example combines Figures 1 to 2 This paper describes an interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine. Figure 1 As shown, step S1 is executed to obtain local mesh data of the receptor model in the Unreal Engine scene, extract the spatial coordinates and normal vectors of the vertices, and proceed to step S2 to calculate the directional deviation of adjacent normal vectors to determine the surface curvature. Based on the surface curvature, the subdivision weight of the facets is calculated. Then, in step S3, an initial convex proxy is constructed, and the facet density is adjusted according to the weights to drive vertex offset contraction to generate a virtual convex proxy that matches the topology of the receptor surface. Then, step S4 is executed to obtain the interaction ray emission parameters, and the intersection test is performed using the virtual convex proxy to obtain the hit point of the ray on the proxy surface. Finally, in step S5, the centroid coordinates of the hit point are calculated and the normals of the receptor vertices are weighted and interpolated to obtain the interaction feedback parameters containing the coordinates and normal vectors.

[0031] like Figure 2As shown, the logical architecture runs within the Unreal Engine computing environment domain and consists of multiple functional modules that work together to form a closed-loop system. The geometry proxy builder is responsible for performing normal divergence curvature analysis, non-uniform subdivision of convex hull patches, and vertex offset shrinkage, and provides virtual convex proxy bodies to the ray interaction detector. The ray interaction detector receives user interaction input, performs interaction ray intersection calculation and occlusion misjudgment filtering, and passes the generated hit points to the feedback parameter reconstructor. The feedback parameter reconstructor performs hit point centroid coordinate calculation and physical normal and texture interpolation, and inputs the backtracking physical interaction parameters to the dynamic receptor data pool, which includes mesh vertex flow and bone deformation field, to drive topology data to the geometry proxy builder to achieve real-time synchronization of interaction and model deformation.

[0032] Example 4: In the application of non-destructive testing of fuselage skin in the aerospace manufacturing field, the system needs to inspect fuselage skin with a thickness of... The aluminum alloy thin-walled structure is annotated with Decal for surface defects. Because the physical thickness of the receptor model is lower than the depth deviation tolerance of the general rendering engine, Decal projection produces a shadow on the inner and outer surfaces of the receptor skin. Furthermore, during interactive detection at the skin edge, rays penetrate the thin wall, causing a misjudgment of the self-occlusion logic, resulting in false triggering of the interactive feedback signal on the opposite side. To address the physical constraints of the thin-walled structure, the system extracts the physical thickness of the receptor skin when activating the local feature sampling unit. for And obtain the direction vector of the interactive ray. To eliminate backface projection, a one-sided open topology reconstruction is performed, extracting the normal vectors of the vertices from the local mesh data. When constructing a virtual convex proxy, a normal consistency filtering rule is introduced, and the system calculates the normals of the proxy facets. With the receptor's frontal normal The dot product, if the dot product is less than The facet is determined to be a redundant topology on the back side of the receptor, and the facet is then deleted, transforming the initial convex surrogate into a one-sided open mesh conformal to the front side of the receptor.

[0033] In the X-ray interactive detection process, an occlusion determination model is used to filter false hit signals. A mapping logic based on the penetration path step size and surface curvature weight is employed to calculate the path step size of the X-ray penetrating the receptor locally. and using the formula Determine the dynamic occlusion threshold, where, For dynamic occlusion threshold, The physical thickness of the receptor model is given by a value of [value missing]. , The deviation from the normal direction The local surface tilt angle obtained by mapping, in radians, is compared with the system's path step size. With dynamic occlusion threshold The numerical value of the path step size satisfy When the condition is met, the hit signal is determined to be generated by the receptor self-occlusion structure and is removed to ensure that the interaction response acts on the front-visible topology, achieving consistency between the interaction logic and physical orientation; by introducing the above-mentioned thickness-aware dynamic threshold calculation procedure, the system in Decal ghosting rate on skin receptors decreased to Below, the false trigger rate of ray detection is suppressed in the region of abrupt change in edge curvature. The reconstructed virtual convex proxy maintains the topological features on one side while decoupling the physical parameters on both sides of the thin wall through backtracking of the centroid coordinates, thus completing the physical-level interaction of the complex receptor under the constraint of extreme space.

[0034] Example 5: When the system faces the pre-deployment calibration scenario for a novel receptor model with variable-scale topological features, the system executes a geometric parameter calibration process to determine the baseline value of the offset contraction distance. Within the local coordinate system of the receptor model, a curvature calibration block with known geometric properties is selected as a reference object, and the normal vector of the surface of the curvature calibration block is extracted using local feature sampling units. The deviation of the normal direction is calculated through the measurement program. With physical tilt angle The ratio coefficient between Determine if the mapping function satisfies ,in, This is the baseline value for the offset contraction distance. It is the normal vector. This refers to the deviation in the normal direction. This is the physical tilt angle, expressed in radians. As a scaling factor, the Decal projection state was tested at different viewing distances. The alignment accuracy between the Decal patch and the normal of the receptor surface was observed. The shrinkage offset at which the minimum geometric ghosting rate within a preset range was determined as the baseline value of the offset shrinkage distance. The parameters are then written into the system's preset parameter configuration table to determine the initial baseline for different grid densities in the interactive response.

[0035] In the pre-integration process for digital twin maintenance of industrial equipment, the system constructs a correlation database between normal divergence and patch subdivision weights through geometric reduction experiments. On a standard spherical topology, it observes the gradient of triangle area changes by discretizing and adjusting the subdivision step size, calculating the subdivision density distribution required to maintain surface smoothness under different curvature gradients, and then calculates the normal direction deviation. Divided into Each step is The quantization interval is determined and the corresponding subdivision weight values ​​for each interval are recorded. Simultaneously, a step size calibration factor based on physical thickness is introduced into the occlusion determination model. X-ray penetration tests are performed using standard parts with different wall thicknesses to determine the dynamic occlusion threshold. As the angle of inclination tends to The numerical convergence boundary at time , where, This refers to the deviation in the normal direction. For dynamic occlusion threshold, Using pi as the circumference of a circle, the system solidifies the resulting mapping curve into geometric agent execution logic. The virtual convex agent generation process of the heterogeneous receptor follows predetermined physical constraint standards in the absence of prior topological information.

[0036] Example 6: In the scene deployment calibration procedure containing a real-time animation-driven receptor model, the system executes a topological adjacency pre-calculation process to construct a vertex index mapping table. It extracts the adjacent face information of each vertex using the original mesh data structure. By establishing a topological association matrix between receptor model vertices and virtual convex proxy face vertices, it determines the influence of surrounding faces on each proxy vertex. The normalized weighting coefficients for the influence of the displacement of each receptor vertex are calculated according to the formula. ,in, For the first The receptor vertex pairs with the first The displacement of each proxy vertex affects the weight. For the first The proxy vertex and the first Euclidean distance between the receptor vertices The number of neighboring vertices. For the neighborhood vertex index, this process eliminates the computational redundancy of searching the neighborhood frame by frame in a dynamic environment and determines the spatial mapping benchmark for the agent body to perform synchronous evolution as the skeleton deforms.

[0037] When performing surface continuity maintenance calibration for proxy patch self-intersection defects caused by animation stretching, the system executes an offline optimization process for Laplacian smoothing parameters to determine the smoothing coefficient. With the number of iterations Displacement simulation tests were performed on the parts of the receptor model with deformation tension. The mean deviation of the normal angle between the smoothed surrogate surface and the receptor surface was calculated. When the mean deviation of the normal angle is within... radians to Within the radian range and with a single calculation time of less than At that time, the system determines the corresponding smoothing coefficient. for And set the number of iterations. for The Laplacian operator is used to iteratively correct the vertex coordinates of the virtual convex proxy. The corrected coordinates are... satisfy The mapping relationship, where, These are the corrected vertex coordinates. These are the vertex coordinates before correction. For smoothing coefficients, For the first The coordinates of the adjacent vertices, The number of adjacent vertices. By indexing adjacent vertices, this method eliminates the jagged effect on the surface of the geometric proxy body under dynamic deformation conditions.

[0038] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for interactive 3D convex hull decal generation and ray interaction detection based on Unreal Engine, characterized in that, Includes the following steps: Step S1: Obtain the local mesh data of the receptor model in the Unreal Engine scene space, and extract the spatial coordinates of the vertices and the corresponding normal vectors from the local mesh data; Step S2: Calculate the normal direction deviation between adjacent normal vectors in the local mesh data, and use the normal direction deviation to determine the surface curvature of the local surface of the receptor model, and calculate the patch subdivision weight based on the surface curvature. Step S3: Construct an initial convex proxy body using spatial coordinates. Perform non-uniform distribution adjustment on the surface density of the initial convex proxy body according to the surface subdivision weight. Specifically, increase the number of subdivision surfaces in areas where the surface curvature is greater than a preset curvature threshold, and drive the surface vertices of the initial convex proxy body to move a certain offset shrinkage distance towards the geometric center of the local mesh data. Perform surface merging in areas where the surface curvature is not greater than a preset curvature threshold to generate a virtual convex proxy body. Step S4: Obtain the emission parameters of the interactive ray, and use the virtual convex proxy to perform an intersection test on the interactive ray to obtain the hit point of the interactive ray on the surface of the virtual convex proxy. Step S5: Calculate the centroid coordinates of the hit point on the virtual convex proxy surface, perform weighted interpolation on the vertex normals of the receiver model based on the centroid coordinates, and obtain the interactive feedback parameters containing the hit point coordinates and the interpolated normal vectors.

2. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, The logic for calculating the surface curvature of the local surface in step S2 follows the following calculation rules: using the formula Calculate the normal of the center vertex and adjacent vertex normals The sum of the dot products, where The deviation in the normal direction is used to characterize the gradient of surface curvature variation, and this gradient is mapped to patch subdivision weights; where... and All are unit vectors extracted from local grid data.

3. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, Step S3, in the process of generating the virtual convex proxy, also includes: extracting the boundary points of the local mesh data and using normal consistency verification to identify the orientation of the front normal of the receptor model; for receptor models with thickness values ​​lower than a preset thickness threshold, performing topological culling on the back face of the initial convex proxy using the orientation of the front normal, shrinking the initial convex proxy into a one-sided open topological mesh that matches the front manifold of the receptor model.

4. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 3, characterized in that, Step S4, which involves performing the intersection test, also includes the following judgment rules: calculating the penetration path step length of the interactive ray in the local space of the receptor model, with the unit of the penetration path step length being mm; establishing an occlusion judgment model by combining the surface curvature features of the virtual convex proxy; comparing the penetration path step length with the preset occlusion threshold in the occlusion judgment model; and identifying and filtering false hit signals generated in the self-occlusion structure of the receptor model when the penetration path step length exceeds the occlusion threshold.

5. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, Before obtaining the local mesh data of the receptor model in step S1, the process includes: defining a three-dimensional projected bounding box with a preset geometric center in the coordinate space of the receptor model; detecting the intersection state between the mesh cells of the receptor model and the three-dimensional projected bounding box; and defining the set of mesh cells with the intersection state as local mesh data.

6. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, After completing step S5 to obtain the interactive feedback parameters, the process also includes: extracting the vertex displacement vector of the receptor model under the skeleton deformation drive in real time; reusing the centroid coordinates of the hit point, performing displacement mapping on the face vertices of the virtual convex proxy body according to the vertex displacement vector, so that the virtual convex proxy body performs synchronous position evolution with the deformation of the receptor model.

7. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 6, characterized in that, The process of virtual convex proxy body performing synchronous position evolution includes: obtaining the mesh deformation amount before and after the deformation of the receptor model; and performing Laplacian smoothing on the facets of the virtual convex proxy body according to the mesh deformation amount to maintain the surface continuity of the virtual convex proxy body after deformation.

8. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, In step S3, the offset shrinkage distance is proportional to the surface curvature, and the unit of the offset shrinkage distance is mm.

9. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, Step S5, which involves obtaining interactive feedback parameters, also includes: reconstructing the material rendering parameters of the virtual convex proxy using the centroid coordinates; and using the reconstructed material rendering parameters to fuse the Decal texture with the surface ambient occlusion data of the receptor model, outputting a Decal geometric entity with consistent rendering performance.

10. The interactive 3D convex hull Decal generation and ray interaction detection method based on Unreal Engine according to claim 1, characterized in that, After obtaining the interaction feedback parameters in step S5, the process also includes: sending the interaction feedback parameters to the Unreal Engine's interaction event distribution interface; and using the interaction event distribution interface to trigger response instructions based on the interaction feedback parameters. The response instructions include modifying the material properties of the recipient model, passing physical impulse to the recipient model, or calling the audio module to play special effects audio.

Images