An edge-boundary conflict-based real scene three-dimensional model quality evaluation method, system, terminal and medium

Abstract

The application discloses a kind of based on edge-boundary conflict real scene three-dimensional model quality evaluation method, system, terminal and medium, method includes: obtaining real scene three-dimensional model, the gradient weighted normalization edge conflict index of each edge in real scene three-dimensional model is determined by texture gradient analysis, edge conflict index is used to reflect the degree that the geometric edge of real scene three-dimensional model is across domain semantic boundary in texture image space;Based on the edge conflict index of gradient weighted normalization, the face level edge conflict index of each triangular facet in real scene three-dimensional model is determined, face level edge conflict index is used to reflect the overall quality of triangular facet of real scene three-dimensional model;Establish grid level statistical evaluation framework, based on face level edge conflict index and grid level statistical evaluation framework, output quality evaluation result.The application is realized by constructing multilevel quality evaluation system, and overall quality characterization and low quality facet automatic identification from edge level, facet level to overall grid level of real scene three-dimensional model.

Description

Technical Field

[0001] This invention relates to the field of evaluation technology for 3D models, and in particular to a method, system, terminal, and medium for evaluating the quality of real-world 3D models based on edge-boundary conflict. Background Technology

[0002] With the rapid development of technologies such as oblique photogrammetry, UAV remote sensing, and multi-view stereo vision, image-based real-scene 3D reconstruction technology has become a key supporting means in fields such as urban modeling, digital twins, cultural heritage protection, and intelligent monitoring. Compared with traditional laser point cloud modeling, photogrammetric 3D models have advantages such as low cost, realistic texture, and continuous data, and have been widely used in practical scenarios such as urban planning, cadastral management, and semantic segmentation.

[0003] In the construction of photogrammetric 3D models, multi-view images are typically used for aerial triangulation and dense matching to generate 3D mesh models with realistic texture maps (Textured Photogrammetric 3D Mesh). These models not only contain geometric structural information but also incorporate high-resolution texture images, possessing strong visual realism and spatial representation capabilities. They are particularly suitable for automated scene understanding and intelligent analysis tasks, such as 3D semantic segmentation, target recognition, and change detection.

[0004] However, in the process of 3D reconstruction and mesh simplification, existing algorithms generally prioritize geometric accuracy or data compression rate as optimization goals, neglecting the consistency between geometric edges and texture semantic boundaries. For example... Figure 1 As shown, triangular faces often span multiple semantic regions at the intersection of building facades, road boundaries, green spaces, and impermeable surfaces, forming so-called "hybrid semantic faces" or "low-quality faces (LQFs)." These faces may not appear visually distorted, but they introduce significant noise and misjudgments in subsequent semantic analysis, severely impacting the usability of the model.

[0005] In summary, existing quality assessment techniques for real-world 3D models mainly focus on geometric errors or texture visual quality, lacking a dedicated modeling and quantitative evaluation mechanism for the inconsistency between geometric edges and texture semantic boundaries.

[0006] Therefore, existing technologies still have shortcomings. Summary of the Invention

[0007] To address the aforementioned shortcomings of existing technologies, this invention provides a method, system, terminal, and medium for evaluating the quality of real-world 3D models based on edge-boundary conflict. The technical solution adopted by this invention is as follows:

[0008] In a first aspect, the present invention provides a method for evaluating the quality of a real-world 3D model based on edge-boundary conflict, the method comprising:

[0009] A real-world 3D model is acquired, and the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined through texture gradient analysis. The edge conflict index is used to reflect the degree to which the geometric edges of the real-world 3D model cross the semantic boundary in the texture image space.

[0010] Based on the gradient-weighted normalized edge conflict index, the face-level edge conflict index of each triangular facet in the real-world 3D model is determined. The face-level edge conflict index is used to reflect the overall quality of the triangular facets in the real-world 3D model.

[0011] A grid-level statistical evaluation framework is established. Based on the surface-level edge conflict index and the grid-level statistical evaluation framework, the quality evaluation results of the real-scene 3D model are output. The quality evaluation results include the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model.

[0012] In one implementation, a realistic 3D model is obtained, and the gradient-weighted normalized edge conflict index of each edge in the realistic 3D model is determined through texture gradient analysis, including:

[0013] Texture gradients are extracted from the original texture image of the real-world 3D model;

[0014] Each edge of the real-world 3D model is mapped to the texture image space, and the Bresenham line algorithm is used for pixel-level sampling to obtain a gradient value sequence.

[0015] Based on the gradient value sequence obtained from sampling, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined.

[0016] In one implementation, extracting texture gradients based on the original texture image of the real-world 3D model includes:

[0017] The original texture image of the real-world 3D model is converted into a grayscale image using a weighted combination method.

[0018] The grayscale image is convolved using a σ-adaptive two-dimensional Gaussian filter to suppress illumination noise and preserve edge structure.

[0019] The Sobel operator is used to calculate the horizontal and vertical gradients to obtain the gradient magnitude, which is used to reflect the intensity of the change in image grayscale at the pixel.

[0020] In one implementation, based on the sampled gradient value sequence, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined, including:

[0021] The average gradient and maximum gradient are determined based on the gradient value sequence obtained from sampling, and the standardized edge conflict index is determined based on the average gradient and maximum gradient.

[0022] Based on the standardized edge conflict index, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined.

[0023] In one implementation, the face-level edge conflict index of each triangular facet in the real-world 3D model is determined based on a gradient-weighted normalized edge conflict index, including:

[0024] Obtain the gradient-weighted normalized edge conflict index of the three edges of each triangular facet in the real-world 3D model;

[0025] The face-level edge conflict index of each triangle is determined based on the gradient-weighted normalized edge conflict index of the three edges of each triangle.

[0026] In one implementation, a grid-level statistical evaluation framework is established. Based on the surface-level edge conflict index and the grid-level statistical evaluation framework, the quality evaluation results of the real-world 3D model are output, including:

[0027] A grid-level statistical evaluation framework is established, and basic statistical indicators are calculated based on the surface-level edge conflict index. The basic statistical indicators include: global average quality, standard deviation, and quantiles.

[0028] Based on basic statistical indicators, the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model are calculated to obtain the quality evaluation results.

[0029] In one implementation, based on basic statistical indicators, the proportion of low-quality facets and the proportion of low-quality region area of ​​the real-world 3D model are calculated, and the quality evaluation results are output, including:

[0030] Based on the quantiles in the aforementioned basic statistical indicators, and combined with the anomaly detection mechanism of the interquartile range, the anomaly detection threshold is calculated.

[0031] Triangular faces whose surface-level edge conflict index exceeds the anomaly detection threshold are identified as low-quality faces of the real-world 3D model. The proportion of low-quality faces and the area proportion of low-quality regions in the real-world 3D model are calculated to obtain the quality evaluation result.

[0032] Secondly, embodiments of the present invention also provide a quality evaluation system for real-scene 3D models based on edge-boundary conflict, wherein the system is used to implement the steps of the quality evaluation method for real-scene 3D models based on edge-boundary conflict as described in any of the above solutions, and the system includes:

[0033] The edge-level evaluation module is used to acquire a real-world 3D model and determine the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model through texture gradient analysis. The edge conflict index is used to reflect the degree to which the geometric edges of the real-world 3D model cross the semantic boundary in the texture image space.

[0034] The facet-level evaluation module is used to determine the facet-level edge conflict index of each triangular facet in the real-world 3D model based on the gradient-weighted normalized edge conflict index. The facet-level edge conflict index is used to reflect the overall quality of the triangular facets in the real-world 3D model.

[0035] The grid-level evaluation module is used to establish a grid-level statistical evaluation framework. Based on the face-level edge conflict index and the grid-level statistical evaluation framework, it outputs the quality evaluation results of the real-scene 3D model. The quality evaluation results include the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model.

[0036] Thirdly, embodiments of the present invention also provide a terminal, wherein the terminal includes a memory, a processor, and a real-scene 3D model quality evaluation program based on edge-boundary conflict stored in the memory and executable on the processor. When the processor executes the real-scene 3D model quality evaluation program based on edge-boundary conflict, it implements the steps of the real-scene 3D model quality evaluation method based on edge-boundary conflict of any of the above-mentioned schemes.

[0037] Fourthly, embodiments of the present invention also provide a computer-readable storage medium, wherein the computer-readable storage medium stores a real-scene 3D model quality evaluation program based on edge-boundary conflict, the real-scene 3D model quality evaluation program based on edge-boundary conflict implementing the steps of the real-scene 3D model quality evaluation method based on edge-boundary conflict as described in any of the above schemes on the computer-readable storage medium.

[0038] Beneficial Effects: Compared with existing technologies, this invention provides a method for evaluating the quality of a realistic 3D model based on edge-boundary conflict. First, a realistic 3D model is acquired. Then, a gradient-weighted normalized edge conflict index is determined for each edge in the realistic 3D model through texture gradient analysis. This edge conflict index reflects the degree to which the geometric edges of the realistic 3D model cross semantic boundaries in the texture image space. Next, based on the gradient-weighted normalized edge conflict index, a facet-level edge conflict index is determined for each triangular facet in the realistic 3D model. This facet-level edge conflict index reflects the overall quality of the triangular facets in the realistic 3D model. Finally, a mesh-level statistical evaluation framework is established. Based on the facet-level edge conflict index and the mesh-level statistical evaluation framework, the quality evaluation result of the realistic 3D model is output. The quality evaluation result includes the proportion of low-quality facets and the proportion of low-quality regions in the realistic 3D model.

[0039] This invention maps the geometric edges of a 3D mesh from a real-world 3D model to a texture image space. By analyzing the texture gradient distribution characteristics along the edges, a multi-level quality evaluation system is constructed, enabling comprehensive quality characterization and automatic identification of low-quality patches from the edge level, patch level, to the overall mesh level. Furthermore, this invention requires no high-precision reference model or manual annotation, accurately identifies and locates low-quality patches and their spatial distribution, and sensitively detects quality changes brought about by mesh simplification and optimization. This provides a basis for automated quality control and subsequent optimization processing of photogrammetric 3D meshes. Attached Figure Description

[0040] Figure 1 This is a schematic diagram illustrating how triangular patches span multiple semantic regions during 3D reconstruction and mesh simplification in existing technologies.

[0041] Figure 2 This is a flowchart of a preferred embodiment of the method for evaluating the quality of a real-world 3D model based on edge-boundary conflict, provided by an embodiment of the present invention.

[0042] Figure 3 The flowchart illustrates the technical route of the method for evaluating the quality of real-world 3D models based on edge-boundary conflict, as provided in this embodiment of the invention.

[0043] Figure 4 The original texture map and grayscale gradient map are provided in the quality evaluation method for real-world 3D models based on edge-boundary conflict provided in the embodiments of the present invention.

[0044] Figure 5 This is a schematic diagram of the inward sampling strategy in the real-scene 3D model quality evaluation method based on edge-boundary conflict provided in the embodiments of the present invention.

[0045] Figure 6This is a schematic diagram illustrating the determination of the average gradient and the maximum gradient in the method for evaluating the quality of a real-world 3D model based on edge-boundary conflict, as provided in an embodiment of the present invention.

[0046] Figure 7 This is a schematic diagram of the basic statistical indicators in the real-scene 3D model quality evaluation method based on edge-boundary conflict provided in the embodiments of the present invention.

[0047] Figure 8 This is a visualization diagram of the low-quality patch recognition results of a real-scene 3D model in the edge-boundary conflict-based real-scene 3D model quality evaluation method provided in this embodiment of the invention.

[0048] Figure 9 This is a schematic diagram of the principle framework of a real-scene 3D model quality evaluation system based on edge-boundary conflict provided in an embodiment of the present invention.

[0049] Figure 10 A schematic diagram of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0051] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content, operations, or steps, nor does it require execution in the described order. For example, some operations or steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0052] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0053] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. For example, "first control information" and "second control information" are only used to distinguish different control information and do not limit their order.

[0054] Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or the order of execution, and that the words "first" and "second" do not necessarily imply that they are different.

[0055] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0056] Existing quality assessment techniques for real-world 3D models primarily focus on geometric errors or texture visual quality, lacking specialized modeling and quantitative evaluation mechanisms for inconsistencies between geometric edges and texture semantic boundaries. Especially in real-world production environments without reference models or manual annotation, existing methods struggle to serve as automated quality control tools for effectively screening large-scale 3D meshes, and cannot reveal the spatial distribution characteristics of low-quality patches. Consequently, they fail to provide reliable data for subsequent semantic analysis, mesh optimization, and model refinement.

[0057] To address the shortcomings of existing technologies, this embodiment provides a method for evaluating the quality of realistic 3D models based on edge-boundary conflict. The core idea is to map the geometric edges of the 3D mesh of the realistic 3D model to a texture image space, and by analyzing the texture gradient distribution characteristics along the edges, construct a multi-level quality evaluation system to achieve comprehensive quality characterization and automatic identification of low-quality faces from the edge level, facet level, to the overall mesh level. In specific applications, this embodiment first acquires the realistic 3D model and determines the gradient-weighted normalized edge conflict index of each edge in the realistic 3D model through texture gradient analysis. The edge conflict index reflects the degree to which the geometric edges of the realistic 3D model cross semantic boundaries in the texture image space. Then, based on the gradient-weighted normalized edge conflict index, the facet-level edge conflict index of each triangular facet in the realistic 3D model is determined. The facet-level edge conflict index reflects the overall quality of the triangular facets of the realistic 3D model. Finally, a grid-level statistical evaluation framework is established. Based on the surface-level edge conflict index and the grid-level statistical evaluation framework, the quality evaluation results of the real-scene 3D model are output. The quality evaluation results include the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model.

[0058] Specifically, the edge-boundary conflict-based real-scene 3D model quality evaluation method of this embodiment can be applied to terminals, including intelligent product terminals such as computers. Specifically, as... Figure 2 As shown in the figure, the method for evaluating the quality of a real-world 3D model based on edge-boundary conflict in this embodiment includes the following steps:

[0059] Step S100: Obtain a real-world 3D model and determine the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model through texture gradient analysis. The edge conflict index is used to reflect the degree to which the geometric edges of the real-world 3D model cross the semantic boundary in the texture image space.

[0060] Combination Figure 3 As shown, this embodiment first performs step 1: edge-level conflict index calculation. Specifically, this embodiment extracts texture gradients based on the original texture image of the real-world 3D model. In practical applications, this embodiment uses a weighted combination method to convert the original texture image of the real-world 3D model into a grayscale image. The weighted combination form is as follows:

[0061] ,

[0062] The original texture image is an RGB texture image. Represents pixel coordinates The gray intensity value at that location, express The red channel value at that location, express The green channel value at that location express The blue channel value at that location. The weighting coefficient for the red channel. This represents the weighting coefficient for the green channel. This represents the weighting coefficient for the blue channel. This weighting coefficient takes into account the human eye's sensitivity to different color bands, with the human eye being most sensitive to green, followed by red, and least sensitive to blue. The weighting coefficient satisfies... .

[0063] Next, in this embodiment, a σ-adaptive two-dimensional Gaussian filter is used to convolve the grayscale image to suppress illumination noise and preserve edge structure, as detailed below:

[0064] ,

[0065] in, Represents pixel coordinates The grayscale value after Gaussian filtering. is the x-coordinate of the pixel. The vertical coordinate is the pixel coordinate. This represents the lateral offset of the convolution kernel relative to the center. This represents the vertical offset of the convolution kernel relative to the center. The radius of the convolution kernel determines the size of the filter window, and the window size is (2... +1)×(2 +1). The two-dimensional Gaussian kernel function is calculated using the following formula: σ is the standard deviation of the Gaussian kernel, which controls the smoothness of the filter. The original grayscale image in coordinates The pixel value at that location. By adjusting the window size (2... +1) and standard deviation σ can achieve a balance between smoothing and noise reduction and detail preservation.

[0066] Furthermore, this embodiment uses the Sobel operator to calculate the horizontal and vertical gradients to obtain the gradient magnitude. The gradient magnitude reflects the intensity of the grayscale change at a pixel. Figure 4 As shown in the diagram. The Sobel operator is one of the most important operators in pixel image edge detection, playing a crucial role in information technology fields such as machine learning, digital media, and computer vision. The gradient magnitude is expressed as:

[0067] ,

[0068] in, Represents pixel coordinates The gradient magnitude (i.e., gradient strength) at a given point is the pixel coordinate. The intensity of the grayscale change at a given location in the image; a larger value indicates more pronounced edge features at that location. is the x-coordinate of the pixel. The vertical coordinate is the pixel coordinate. Represents pixel coordinates along The gradient components in the direction are calculated by Sobel horizontal operator convolution, reflecting the rate of gray-level change of the image in the horizontal direction. Represents pixel coordinates along The gradient component in the direction is calculated by convolution using the Sobel vertical operator, reflecting the rate of change of grayscale in the image along the vertical direction. In this embodiment, the gradient magnitude ranges from [0, +∞), and the unit is related to the image grayscale value.

[0069] In other implementations, this embodiment can also use the Canny operator, Scharr operator, or Laplacian operator to replace the Sobel operator. The Canny operator is an image edge detection operator; the Scharr operator is an improved edge detection operator proposed in image analysis, mainly used to calculate the gradient of an image; and the Laplacian operator is a second-order differential operator in n-dimensional Euclidean space. Furthermore, this embodiment can also use a deep convolutional network (such as U-Net) to output the gradient magnitude end-to-end with mean squared error loss supervision, suitable for scenarios with significant texture noise.

[0070] Since the gradient magnitude of each edge in the real-world 3D model can be obtained based on the above steps, this embodiment can further map each edge of the real-world 3D model to a texture image space, and use the Bresenham line algorithm to perform pixel-level sampling on each edge, thereby obtaining a gradient value sequence. The Bresenham line algorithm is an algorithm used to draw a line segment between two points on the screen. Its characteristic is that it can be completed using only basic addition, subtraction, and comparison operations, making it an efficient line drawing algorithm.

[0071] Specifically, in combination Figure 5 As shown, to avoid the influence of vertex noise, an inward pixel shrinkage strategy is adopted, which shrinks the vertex coordinates inward by a fixed pixel distance before sampling. Figure 5 The diagram shows a comparison between using the original edge gradient sampling and the indented edge gradient sampling, where L1, L2, and L3 are three edges. For N sampling points on the edges, the correspondence between UV coordinates and pixel coordinates is established:

[0072]

[0073] Indicates the first Texture coordinates (UV coordinates) for each sampling point. UV coordinates are a coordinate system used in computer graphics to map two-dimensional textures to the surface of a three-dimensional model. In this system, U represents the horizontal direction and V represents the vertical direction. For the first The x-coordinate of each sampling point in the texture image space. For the first The vertical coordinate of each sampling point in the texture image space. The index number is the sampling point number, with a value range of 1, 2, ..., N. N is the total number of sampling points along a certain edge, which is determined by the length of the edge. In this embodiment, the number of sampling points and the indentation strategy can be adaptively adjusted according to the edge length to ensure that the long edge is not undersampled and the short edge is not oversampled. This represents the UV coordinates of the edge origin in the texture image space. This represents the UV coordinates of the edge endpoint in the texture image space. This represents the horizontal pixel distance from the starting point to the end point. This represents the vertical pixel distance from the start point to the end point. The pixel coordinates of the edge starting point. Here are the pixel coordinates of the edge endpoint. This formula calculates the texture coordinates of uniformly distributed sampling points on the edge using linear interpolation, realizing the spatial mapping relationship between the three-dimensional geometric edge and the two-dimensional texture image. In this embodiment, gradient value sequence is obtained by sampling the edge gradient after a certain edge is shrunk. ,in, This is the first gradient value sampled. This is the second gradient value sampled. This is the Nth gradient value sampled.

[0074] Furthermore, this embodiment determines the gradient-weighted normalized edge conflict index for each edge in the real-world 3D model by sampling the gradient value sequence for each edge. Specifically, this embodiment determines the average gradient and maximum gradient based on the sampled gradient value sequence, and determines the standardized edge conflict index based on the average gradient and maximum gradient. Figure 6 As shown in Figure 6, the triangular facet is composed of sides AB, AC, and BC. In this embodiment, the average gradient and maximum gradient of sides AB, AC, and BC of the triangle are calculated. The average gradient is expressed as: The maximum gradient is expressed as: Next, the standardized edge conflict index is defined as follows:

[0075]

[0076] in, To prevent the stable term from being divided by zero.

[0077] Furthermore, in other implementations, the standardized edge conflict index can also be replaced by a median-maximum normalization, expressed as:

[0078] ,in, This is the median of the gradient.

[0079] Furthermore, this embodiment determines the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model based on a standardized edge conflict index, expressed as:

[0080]

[0081] The standardized edge conflict index (NECI) serves as a normalized indicator reflecting the relative contrast of edges, while It provides absolute gradient intensity information. The product of the two can simultaneously capture the relative salience and absolute magnitude of texture changes, ensuring that only edges with significant local contrast and sufficient intensity receive higher weight values.

[0082] As can be seen, this embodiment maps the geometric edges of the 3D mesh of the real-world 3D model to the texture image space, performs pixel-level gradient sampling along the edges, and constructs an edge conflict index using the normalized difference between the maximum gradient value and the average gradient value, thereby quantitatively representing the degree to which edges cross semantic boundaries. This mechanism can effectively distinguish between high-quality edges located in textured uniform regions and low-quality edges that cross semantic boundaries. Furthermore, based on the standardized edge conflict index (NECI), the maximum gradient value is introduced as a weighting factor, while also considering relative gradient contrast and absolute gradient strength, ensuring that only edges with significant local contrast and sufficient gradient strength can obtain high conflict scores.

[0083] Step S200: Based on the gradient-weighted normalized edge conflict index, determine the face-level edge conflict index of each triangular facet in the real-world 3D model. The face-level edge conflict index is used to reflect the overall quality of the triangular facets in the real-world 3D model.

[0084] Combination Figure 3 As shown, step 2 is performed: calculation of the facet-level edge conflict index. In specific applications, this embodiment obtains the gradient-weighted normalized edge conflict index of the three edges of each triangular facet in the real-world 3D model. Since the quality of a triangular facet is determined by its lowest quality edge, the facet-level edge conflict index of each triangular facet is determined based on the gradient-weighted normalized edge conflict index of the three edges of each triangular facet, expressed as:

[0085] in, , , These are the gradient-weighted edge conflict indices of the three edges of the triangular facet. This embodiment is based on the "weakest link," taking the edge index with the highest degree of conflict among the three edges of the triangular facet as the overall quality characterization of the facet. This design ensures that quality defects on any edge can be reflected in the facet-level evaluation.

[0086] In other implementations, the face-level edge conflict index in this embodiment can also be replaced by the mean of the three-sided gradient-weighted edge conflict index, expressed as:

[0087]

[0088] Alternatively, the edge conflict index can be changed to a three-sided gradient-weighted index, weighted by the side length, and expressed as:

[0089]

[0090] in, For the first The side lengths of the sides, For the first The gradient-weighted edge conflict index for each edge.

[0091] In this embodiment, the maximum conflict index among the three edges of a triangular facet is used as the overall quality characterization of the facet, ensuring that the quality defects of any edge can be accurately captured, thus achieving reliable identification of mixed semantic facets.

[0092] Step S300: Establish a grid-level statistical evaluation framework. Based on the face-level edge conflict index and the grid-level statistical evaluation framework, output the quality evaluation results of the real-scene 3D model. The quality evaluation results include: the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model.

[0093] Combination Figure 3 As shown, step 3 is performed: mesh-level statistical evaluation. In practical applications, this embodiment establishes a mesh-level statistical evaluation framework, which can be used to achieve mesh-level quality evaluation of real-world 3D models. Its basic statistical indicators include: global average quality, standard deviation, and quantiles. In practical applications, the basic statistical indicators can be calculated based on the aforementioned surface-level edge conflict index. The basic statistical indicators are as follows: Figure 7 As shown, the global average quality is represented as follows: ;

[0094] Standard deviation is expressed as: ;

[0095] Quantiles are represented as: .

[0096] in, For the first The edge conflict index of a triangular facet, where inf is an abbreviation for infimum, represents the infimum, and the quantile formula indicates that it satisfies the cumulative distribution function. The smallest value among all real numbers x. R is the set of real numbers. The quantile (i.e., the proportion corresponding to the quantile) is represented. In this embodiment, the quantile values ​​are 25%, 50%, and 75%.

[0097] Next, based on basic statistical indicators, the proportion of low-quality faces and the proportion of low-quality areas in the real-world 3D model are calculated to obtain the quality evaluation results. Specifically, based on the quantiles in the basic statistical indicators, combined with the interquartile range anomaly detection mechanism, the anomaly detection threshold is calculated, expressed as:

[0098] This represents the anomaly detection threshold calculated based on the interquartile range method. This represents the 75th percentile (upper quartile) of the distribution of the surface-level edge conflict index. This represents the 25th percentile (lower quartile) of the distribution of the surface-level edge conflict index. This is called the Interquartile Range (IQR). In this embodiment, the value of the face-level edge conflict index exceeding the anomaly detection threshold can be used. Triangular facets are identified as low-quality facets (LQF). Mapping the facet-level edge conflict index values ​​to a color space visually displays the mesh quality distribution, such as... Figure 8 As shown.

[0099] Specifically, the low-quality surface area ratio of the real-world 3D model Represented as:

[0100]

[0101] Low-quality area proportion Represented as:

[0102]

[0103] in, The threshold for determining low-quality patches (in this embodiment, the anomaly detection threshold is used). ): Indicates the first The value of the facet-level edge conflict index exceeds the anomaly detection threshold. The area of ​​the triangular facet. For the first The area of ​​the triangular facets with the edge conflict index at the face level. Based on the proportion of low-quality facets and the proportion of low-quality regions in the real-world 3D model, the quality evaluation result of the real-world 3D model is obtained.

[0104] In summary, the real-scene 3D model quality evaluation method proposed in this embodiment does not require a high-precision reference model or manual annotation. It can accurately identify and locate low-quality patches and their spatial distribution, and sensitively detect quality changes brought about by mesh simplification and optimization. This provides a basis for automated quality control and subsequent optimization processing of photogrammetric 3D meshes.

[0105] Based on the above embodiments, the present invention also provides a quality evaluation system for real-scene 3D models based on edge-boundary conflict. This system is used to implement the steps of the above method embodiments, such as... Figure 9As shown, the system includes: an edge-level evaluation module 10, a patch-level evaluation module 20, and a mesh-level evaluation module 30. Specifically, the edge-level evaluation module 10 acquires a real-world 3D model and determines the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model through texture gradient analysis. The edge conflict index reflects the degree to which the geometric edges of the real-world 3D model cross semantic boundaries in the texture image space. The patch-level evaluation module 20 determines the face-level edge conflict index of each triangular facet in the real-world 3D model based on the gradient-weighted normalized edge conflict index. The face-level edge conflict index reflects the overall quality of the facets in the real-world 3D model. The mesh-level evaluation module 30 establishes a mesh-level statistical evaluation framework and outputs the quality evaluation result of the real-world 3D model based on the face-level edge conflict index and the mesh-level statistical evaluation framework. The quality evaluation result includes the proportion of low-quality facets and the proportion of low-quality area in the real-world 3D model.

[0106] The principles of each module in the real-scene 3D model quality evaluation system based on edge-boundary conflict in this embodiment are the same as the principles of each step in the above method embodiment, and will not be elaborated further here.

[0107] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 10 As shown. The terminal may include one or more processors 100 ( Figure 10 (Only one is shown in the image), memory 101, and computer program 102 stored in memory 101 and executable on one or more processors 100. For example, a quality assessment program for a real-world 3D model based on edge-boundary conflict. When one or more processors 100 execute computer program 102, they can implement the various steps in the embodiment of the method for quality assessment of a real-world 3D model based on edge-boundary conflict. Alternatively, when one or more processors 100 execute computer program 102, they can implement the functions of various modules / units in the embodiment of the system for quality assessment of a real-world 3D model based on edge-boundary conflict, which is not limited here.

[0108] In one embodiment, the processor 100 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0109] In one embodiment, memory 101 may be an internal storage unit of an electronic device, such as a hard drive or RAM. Memory 101 may also be an external storage device of the electronic device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital Card (SD), or Flash Card. Furthermore, memory 101 may include both internal and external storage units. Memory 101 is used to store computer programs and other programs and data required by the terminal. Memory 101 can also be used to temporarily store data that has been output or will be output.

[0110] Those skilled in the art will understand that Figure 10 The schematic diagram shown is only a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. The specific terminal may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0111] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), direct memory bus RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the quality of a realistic 3D model based on edge-boundary conflict, characterized in that, The method includes: A real-world 3D model is acquired, and the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined through texture gradient analysis. The edge conflict index is used to reflect the degree to which the geometric edges of the real-world 3D model cross semantic boundaries in the texture image space. Based on the gradient-weighted normalized edge conflict index, the face-level edge conflict index of each triangular facet in the real-world 3D model is determined. The face-level edge conflict index is used to reflect the overall quality of the triangular facets in the real-world 3D model. A grid-level statistical evaluation framework is established. Based on the face-level edge conflict index and the grid-level statistical evaluation framework, the quality evaluation results of the real-scene 3D model are output. The quality evaluation results include the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model. Obtain a realistic 3D model, and determine the gradient-weighted normalized edge conflict index of each edge in the realistic 3D model through texture gradient analysis, including: Texture gradients are extracted from the original texture image of the real-world 3D model; Each edge of the real-world 3D model is mapped to the texture image space, and the Bresenham line algorithm is used for pixel-level sampling to obtain a gradient value sequence. Based on the gradient value sequence obtained from sampling, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined; The edge conflict index at the face level is determined for each triangular facet in the real-world 3D model based on gradient-weighted normalization, including: Obtain the gradient-weighted normalized edge conflict index of the three edges of each triangular facet in the real-world 3D model; The face-level edge conflict index of each triangle is determined based on the gradient-weighted normalized edge conflict index of the three edges of each triangle.

2. The method for evaluating the quality of real-scene 3D models based on edge-boundary conflict according to claim 1, characterized in that, Extracting texture gradients from the original texture image of the real-world 3D model includes: The original texture image of the real-world 3D model is converted into a grayscale image using a weighted combination method. The grayscale image is convolved using a σ-adaptive two-dimensional Gaussian filter to suppress illumination noise and preserve edge structure. The Sobel operator is used to calculate the horizontal and vertical gradients to obtain the gradient magnitude, which is used to reflect the intensity of the change in image grayscale at the pixel.

3. The method for evaluating the quality of a real-world 3D model based on edge-boundary conflict according to claim 2, characterized in that, Based on the sampled gradient value sequence, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined, including: The average gradient and maximum gradient are determined based on the gradient value sequence obtained from sampling, and the standardized edge conflict index is determined based on the average gradient and maximum gradient. Based on the standardized edge conflict index, the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model is determined.

4. The method for evaluating the quality of a real-world 3D model based on edge-boundary conflict according to claim 1, characterized in that, A grid-level statistical evaluation framework is established. Based on the surface-level edge conflict index and the grid-level statistical evaluation framework, the quality evaluation results of the real-scene 3D model are output, including: A grid-level statistical evaluation framework is established, and basic statistical indicators are calculated based on the surface-level edge conflict index. The basic statistical indicators include: global average quality, standard deviation, and quantiles. Based on basic statistical indicators, the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model are calculated to obtain the quality evaluation results.

5. The method for evaluating the quality of a real-world 3D model based on edge-boundary conflict according to claim 4, characterized in that, Based on basic statistical indicators, the proportion of low-quality patches and the proportion of low-quality areas in the real-world 3D model are calculated, and the quality evaluation results are output, including: Based on the quantiles in the aforementioned basic statistical indicators, and combined with the anomaly detection mechanism of the interquartile range, the anomaly detection threshold is calculated. Triangular faces whose surface-level edge conflict index exceeds the anomaly detection threshold are identified as low-quality faces of the real-world 3D model. The proportion of low-quality faces and the area proportion of low-quality regions in the real-world 3D model are calculated to obtain the quality evaluation result.

6. A quality evaluation system for real-world 3D models based on edge-boundary conflict, characterized in that, The system is used to implement the steps of the real-scene 3D model quality evaluation method based on edge-boundary conflict as described in any one of claims 1-5, and the system includes: The edge-level evaluation module is used to acquire a real-world 3D model and determine the gradient-weighted normalized edge conflict index of each edge in the real-world 3D model through texture gradient analysis. The edge conflict index is used to reflect the degree to which the geometric edges of the real-world 3D model cross semantic boundaries in the texture image space. The facet-level evaluation module is used to determine the facet-level edge conflict index of each triangular facet in the real-world 3D model based on the gradient-weighted normalized edge conflict index. The facet-level edge conflict index is used to reflect the overall quality of the triangular facets in the real-world 3D model. The grid-level evaluation module is used to establish a grid-level statistical evaluation framework. Based on the face-level edge conflict index and the grid-level statistical evaluation framework, it outputs the quality evaluation results of the real-scene 3D model. The quality evaluation results include the proportion of low-quality facets and the proportion of low-quality area of ​​the real-scene 3D model.

7. A terminal, characterized in that, The terminal includes a memory, a processor, and a real-world 3D model quality evaluation program based on edge-boundary conflict stored in the memory and executable on the processor. When the processor executes the real-world 3D model quality evaluation program based on edge-boundary conflict, it implements the steps of the real-world 3D model quality evaluation method based on edge-boundary conflict as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a quality evaluation program for a real-world 3D model based on edge-boundary conflict, and the quality evaluation program for a real-world 3D model based on edge-boundary conflict implements the steps of the quality evaluation method for a real-world 3D model based on edge-boundary conflict as described in any one of claims 1-5 on the computer-readable storage medium.

Images