A three-dimensional structure reconstruction system based on artificial intelligence topology optimization

Abstract

The application relates to the technical field of three-dimensional reconstruction, and discloses a three-dimensional structure reconstruction system based on artificial intelligence topology optimization, a data extraction module is used for carrying out regional screening according to a coordinate point distribution density value, combining surface mesh boundary data, and generating a component partition judgment value; a structure identification module is used for comparison and judgment to determine whether a condition is met, and a boundary regularity coefficient is obtained; a topology adjustment module is used for screening a coordinate set with an offset value in a relative boundary range and a height variation less than a critical height variation coefficient, and a boundary topology adjustment coefficient is obtained; a feature reconstruction module is used for calculating a component form deviation degree according to the product of a node distribution density and an angle variation value, and generating a reconstructed feature prediction value; and a reconstruction prediction module is used for calculating an equilibrium factor between a component structure state value and a reliable level coefficient, and obtaining a structure reconstruction trend degree; the application reduces the need for manual intervention and saves cost.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of three-dimensional reconstruction, and in particular to a three-dimensional structure reconstruction system based on artificial intelligence topology optimization. BACKGROUND

[0002] Three-dimensional reconstruction technology is committed to collecting information from various sensors and data sources to construct a three-dimensional digital model of the physical world. Currently, the commonly used core technologies include computer vision, laser scanning, and stereophotogrammetry. Although these technologies can capture spatial information and appearance details of complex environments and objects, and generate three-dimensional models for virtual reality, building information modeling (BIM), and industrial design, there are still many problems in practical applications. Traditional three-dimensional reconstruction technology requires a large amount of time and computing resources when dealing with large-scale or particularly complex data sets. For example, in the process of building information modeling (BIM), traditional methods are slow in processing large amounts of building data, leading to delays in project progress. In the field of industrial design, the high consumption of computing resources increases the design cost when dealing with complex component data. Moreover, traditional methods are difficult to adapt to rapidly changing task requirements and lack precision, which can cause inaccurate design. In addition, traditional three-dimensional reconstruction methods have low automation and intelligence. Each time a task is executed, professional personnel need to manually adjust and correct it, which is complex and prone to errors. This seriously limits the widespread application of three-dimensional reconstruction technology in modern application scenarios that require rapid and continuous data processing. SUMMARY

[0003] The purpose of the present application is to solve the above problems, and a three-dimensional structure reconstruction system based on artificial intelligence topology optimization is designed.

[0004] The present application provides a three-dimensional structure reconstruction system based on artificial intelligence topology optimization, which comprises:

[0005] A data extraction module is used to obtain building component point cloud image data, face mesh boundary data, and spatial scale parameters. According to the coordinate point distribution density value in the component point cloud image data, combined with the face mesh boundary data, the region is screened, the component region with closed boundary and data continuity is taken as the target region, and the component partition judgment value is generated.

[0006] A structure recognition module is used to call the sub-region boundary line data in the face mesh boundary data according to the component partition judgment value, compare the position parameters of the gap region, the corresponding edge length, and the number of corners in the boundary line, judge whether the boundary continuity and regularity judgment conditions are met, and obtain the boundary regularity coefficient.

[0007] The topology adjustment module is configured to call spatial coordinate points in the point cloud image data according to the boundary regularity coefficient, to jointly calculate each coordinate point according to a height difference value and a relative offset distance of the boundary contour, to screen a coordinate set in which an offset value is within a relative boundary range and a height variation amount is less than a critical height variation coefficient, and to obtain a boundary topology adjustment coefficient.

[0008] The feature reconstruction module is configured to call each component boundary contour line based on the boundary topology adjustment coefficient, to obtain a component contour line node number, a node distribution density and a corresponding angle variation value of each node, to calculate a component form deviation degree according to a product of the node distribution density and the angle variation value, and to generate a reconstructed feature prediction value.

[0009] The reconstruction prediction module is configured to call the reconstructed feature prediction value, to perform difference calculation on the prediction value and a standard feature value of each component category form, to calculate an equilibrium factor between a component structure state value and a confidence level coefficient according to a reconstruction confidence level coefficient corresponding to an interval in which the difference value is located, to average the equilibrium factor values of each component to establish a component overall structure reconstruction trend degree, and to obtain the structure reconstruction trend degree.

[0010] Optionally, in the first implementation manner of the present application, the data extraction module comprises:

[0011] The traversal submodule is configured to set a screening standard, to traverse each coordinate point in the building component point cloud image data, and to obtain a region preliminarily screened out.

[0012] The determination submodule is configured to determine, in the region preliminarily screened out, a boundary line related to each dense point in the face mesh boundary data.

[0013] The first calculation submodule is configured to record each closed boundary region possibly meeting the requirement, to randomly select two adjacent points in the region, to calculate a distance between the two points, and to continue to check distances of other adjacent points in the region if the distance is less than or equal to a continuity distance standard.

[0014] The determination submodule is configured to determine that the data of the region is continuous when distances between all adjacent points in the region all satisfy the condition of being less than or equal to the continuity distance standard.

[0015] The generation submodule is configured to generate a component partition determination value by taking a component region with a closed boundary and continuous data as a target region.

[0016] Optionally, in the second implementation manner of the present application, the set screening standard comprises:

[0017] A standard value for judging the density of the coordinate point distribution is set, denoted as the density judgment standard; a minimum value for measuring the size of the closed boundary is set as the minimum boundary standard; and a maximum distance for judging whether the data is continuous is set as the continuity distance standard.

[0018] Optionally, in the third implementation manner of the present application, the structure recognition module comprises:

[0019] The extraction sub-module is configured to extract boundary line data of each sub-region from the surface mesh boundary data according to the component partition determination value;

[0020] The first recognition sub-module is configured to use an edge detection algorithm to process the extracted boundary line data, recognize a discontinuous part in the boundary line, determine the discontinuous part as a gap region, record a specific position of each gap region on the boundary line, and obtain a position parameter of the gap region;

[0021] The measurement sub-module is configured to start from a starting point of the boundary line, measure each line segment of the boundary line in sequence, and add up lengths of all the line segments to obtain a total length of the edge corresponding to the boundary line after traversing the entire boundary line;

[0022] The movement sub-module is configured to move along the boundary line, form a position of a corner at each position where a change in each direction is obvious, and determine a number of corners of the boundary line;

[0023] The arrangement sub-module is configured to arrange the obtained position parameter of the gap region, the length of the edge, and the number of corners into a data set containing boundary feature information;

[0024] The first comparison sub-module is configured to compare the arranged boundary feature data set with a pre-set boundary continuity and regularity judgment condition;

[0025] The input sub-module is configured to input the boundary feature data after the pre-set condition comparison into a support vector machine algorithm, classify and judge the boundary feature, output an evaluation result, and obtain a boundary regularity coefficient.

[0026] Optionally, in the fourth implementation manner of the present application, the first recognition sub-module comprises:

[0027] A convolution operation is performed on the extracted boundary line data by using different scale gradient operators to generate a multi-scale gradient map;

[0028] A gradient direction of each pixel point is calculated, each pixel point is divided into several direction intervals, a gradient amplitude consistency check is performed on adjacent pixels in the same direction interval, and if a difference in gradient direction amplitude of the adjacent pixels exceeds a threshold value, the adjacent pixels are marked as suspicious gap points;

[0029] Introducing the time dimension, constructing the space-time correlation matrix, checking whether there is a similar gradient feature in the corresponding position in the adjacent frame for each suspicious gap point in the current frame, and if there is, enhancing the confidence of the real gap;

[0030] Based on the gradient statistical characteristics of the local region, a dynamic threshold is calculated, the gradient amplitude image is adaptively threshold segmented, and a binary edge image is generated;

[0031] The broken edge in the binary edge image is subjected to morphological analysis, the edge center line is extracted through skeletonization operation, the curvature change rate of the center line is calculated, the suspicious gap region is searched near the curvature mutation point, and the geometric feature of each suspicious gap region is verified;

[0032] For the confirmed gap region, the topological prediction is performed using the edge line segment direction information at both ends, the potential edge path of the gap region is fitted through the Bezier curve, and the complete boundary topological structure is reconstructed to determine the gap region.

[0033] Optionally, in the fifth implementation manner of the present application, the topological adjustment module comprises:

[0034] The hierarchical index submodule is configured to perform hierarchical indexing on the three-dimensional point cloud data according to the boundary regularity coefficient, add a multi-scale geometric feature descriptor to each spatial coordinate point, and form an enhanced point cloud feature space;

[0035] The adjustment submodule is configured to generate an adaptive reference plane based on the geometric center and the main direction of the boundary contour line, and dynamically adjust the tolerance of the reference plane according to the boundary regularity coefficient;

[0036] The second identification submodule is configured to construct a space-time cube, perform context analysis on each coordinate point, extract three-dimensional space features and time sequence features, and automatically identify the feature dimension of the boundary topological adjustment through an attention mechanism;

[0037] The fusion submodule is configured to perform double-path CNN feature fusion, generate a joint feature representation, derive a critical height change coefficient and a relative boundary range parameter according to the statistical characteristics of the input three-dimensional point cloud data, construct a decision network to classify each point, and output a probability distribution;

[0038] The screening submodule is configured to establish a point cloud graph structure, with vertices being coordinate points and edges being topological relationships between points, screen a coordinate set in the point cloud graph structure, which has an offset value within the relative boundary range and a height change amount less than the critical height change coefficient, and obtain a boundary topological adjustment coefficient.

[0039] Optionally, in the sixth implementation manner of the present application, the double-path CNN feature fusion comprises:

[0040] extracting structural information in the vertical direction of the three-dimensional point cloud data to obtain a height difference path, wherein the structural information at least includes a change feature, capturing steps and cracks;

[0041] analyzing a horizontal offset relationship between the three-dimensional point cloud data and the boundary contour to identify an edge blur area to obtain an offset path;

[0042] performing nonlinear combination on the height difference path and the offset path through a feature fusion module of a CNN network to generate a joint feature representation.

[0043] Optionally, in the seventh implementation manner of the present application, the feature reconstruction module comprises:

[0044] a sorting sub-module configured to sort the component boundary contour lines according to the boundary topology adjustment coefficients, and establish a contour line space index structure;

[0045] a forming sub-module configured to uniformly sample an initial node set on the contour lines, record node coordinates and normal directions, and apply an adaptive sampling strategy to increase the sampling density in areas with large curvature changes to form a multi-scale node representation;

[0046] a statistical sub-module configured to define a local neighborhood centered on each node, count the number of nodes in the neighborhood, calculate a normalized node distribution density value according to the neighborhood radius and the number of nodes, and apply Gaussian smoothing processing to eliminate local density fluctuation noise;

[0047] a filtering sub-module configured to calculate a tangent direction change angle between adjacent nodes as a node angle change value, and adopt a bidirectional angle calculation strategy to perform gradient filtering on the angle change value;

[0048] a weighting sub-module configured to multiply the node distribution density and the angle change value point by point, apply a spatial attention mechanism to weight the product results of key areas, aggregate the product values of all nodes through area integration to obtain an overall morphological deviation degree;

[0049] a mapping sub-module configured to map the morphological deviation degree to a preset discrete level, construct a multi-dimensional feature vector in combination with the geometric features of the contour lines, apply principal component analysis for dimension reduction to generate a reconstructed feature prediction value.

[0050] Optionally, in the eighth implementation manner of the present application, if there are multiple frames of time series data, a corresponding relationship between the nodes of the current frame and the historical frames is established, a time derivative of the node angle change is calculated, a spatiotemporal feature vector is constructed, and Kalman filtering is applied to predict the future position of the node.

[0051] Optionally, in the ninth implementation manner of the present application, the reconstruction prediction module comprises:

[0052] The second comparison sub-module is used for classifying and arranging the reconstructed feature prediction value according to the component category, sequentially comparing the reconstructed feature prediction value of each component with the corresponding component category standard feature value, and collecting the differences of all feature dimensions to obtain the comprehensive difference value of the component;

[0053] The second calculation sub-module is used for setting an initial component structure state value for each component according to the reconstruction confidence level coefficient corresponding to the interval where the difference value is located, calculating the component structure state value and the reconstruction confidence level coefficient, and obtaining the balancing factor between the component structure state value and the reconstruction confidence level coefficient.

[0054] The adding sub-module is used for adding the balancing factor values of all components, dividing the sum by the total number of components, obtaining the average value, and determining the overall structure reconstruction trend of the components.

[0055] In the technical scheme provided by the application, the building component point cloud image data, the surface mesh boundary data and the spatial scale parameters are obtained, the component partition determination value is generated by screening the regions according to the coordinate point distribution density value existing in the component point cloud image data and in combination with the surface mesh boundary data, taking the component region with a closed boundary and data continuity as a target region, the boundary regularity coefficient is obtained by calling the sub-region boundary line data in the surface mesh boundary data, comparing the position parameters of the notch region existing in the boundary line, the corresponding edge length and the number of corners, judging whether the boundary continuity and regularity determination conditions are met, calling the spatial coordinate points in the point cloud image data according to the boundary regularity coefficient, jointly calculating each coordinate point according to the height difference value and the relative offset distance of the boundary contour, screening the coordinate set in which the offset value is within the relative boundary range and the height change amount is less than the critical height change coefficient, obtaining the boundary topology adjustment coefficient, calling the component boundary contour line based on the boundary topology adjustment coefficient, obtaining the component contour line node number, node distribution density and angle change value corresponding to each node, calculating the component shape deviation degree according to the product of the node distribution density and the angle change value, generating the reconstructed feature prediction value, calling the reconstructed feature prediction value, calculating the difference between the prediction value and the component category standard feature value, calculating the balancing factor between the component structure state value and the reconstruction confidence level coefficient according to the reconstruction confidence level coefficient corresponding to the interval where the difference value is located, averaging the balancing factor values of all components to establish the overall structure reconstruction trend of the components, and obtaining the structure reconstruction trend; the application automatically adjusts and optimizes the data processing steps, reduces the need for manual intervention, continuously optimizes the reconstruction result when the same type of component is reconstructed multiple times, makes the reconstruction result closer to the actual physical structure, greatly enhances the practicability and reliability of the model, greatly reduces the number of computing devices and the computing time required when processing three-dimensional reconstruction data of large urban buildings, and saves costs. BRIEF DESCRIPTION OF DRAWINGS

[0056] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not intended to be limiting.

[0057] Figure 1 A structural schematic diagram of a three-dimensional structure reconstruction system based on artificial intelligence topology optimization is provided for an embodiment of the present application;

[0058] Figure 2 A structural schematic diagram of a data extraction module is provided for an embodiment of the present application;

[0059] Figure 3 A structural schematic diagram of a structure identification module is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0060] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application, if any, of the above-described drawings, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so construed can interchange, under appropriate circumstances, without departing from the scope of the present application as described and claimed. Furthermore, the term "comprising" or "containing" and variations thereof as used herein is used generically and inclusively, in connection with explaining processes, methods, articles or apparatuses, and contains, but is not limited to, any process, method, article, or apparatus that includes a list of steps or elements as explained without necessarily being limited to only those steps or elements.

[0061] For the purpose of facilitating understanding, the specific flow of the embodiments of the present application is described below, please refer to Figure 1 The structural schematic diagram of the three-dimensional structure reconstruction system based on artificial intelligence topology optimization provided by the embodiments of the present application, the system comprises:

[0062] The data extraction module is used for acquiring building component point cloud image data, surface mesh boundary data and spatial scale parameters, screening regions according to the coordinate point distribution density value existing in the component point cloud image data, combining the surface mesh boundary data, taking the component region with closed boundary and data continuity as a target region, and generating a component partition determination value;

[0063] The structure identification module is used for calling each sub-region boundary line data in the surface mesh boundary data according to the component partition determination value, comparing the position parameters of the notch region existing in the boundary line, the corresponding edge length and the number of corners, judging whether the boundary continuity and regularity determination conditions are satisfied, and acquiring a boundary regularity coefficient;

[0064] The topology adjustment module is configured to call spatial coordinate points in the point cloud image data according to the boundary regularity coefficient, to jointly calculate each coordinate point according to a height difference value and a relative offset distance of the boundary contour, to screen a coordinate set in which an offset value is within a relative boundary range and a height variation amount is less than a critical height variation coefficient, and to obtain a boundary topology adjustment coefficient.

[0065] The feature reconstruction module is configured to call each component boundary contour line based on the boundary topology adjustment coefficient, to obtain a component contour line node number, a node distribution density, and a corresponding angle variation value of each node, to calculate a component form deviation degree according to a product of the node distribution density and the angle variation value, and to generate a reconstructed feature prediction value.

[0066] The reconstruction prediction module is configured to call the reconstructed feature prediction value, to perform difference calculation on the prediction value and a standard feature value of each component category form, to calculate an equilibrium factor between a component structure state value and a confidence level coefficient according to a reconstruction confidence level coefficient corresponding to an interval in which the difference value is located, to average the equilibrium factor values of each component to establish a component overall structure reconstruction trend degree, and to obtain the structure reconstruction trend degree.

[0067] In this embodiment, please refer to Figure 2 The data extraction module includes:

[0068] The traversal submodule is configured to set a screening standard, to traverse each coordinate point in the building component point cloud image data, and to obtain a region preliminarily screened out.

[0069] The determination submodule is configured to determine, in the region preliminarily screened out, a boundary line related to each dense point in the face mesh boundary data.

[0070] The first calculation submodule is configured to record each closed boundary region that may meet the requirement, to randomly select two adjacent points in the region, to calculate a distance between the two points, and to continue to check distances of other adjacent points in the region if the distance is less than or equal to a continuity distance standard.

[0071] The determination submodule is configured to determine that the data of the region is continuous when distances between all adjacent points in the region all meet the requirement of being less than or equal to the continuity distance standard.

[0072] The generation submodule is configured to generate a component partition determination value by taking a component region that has a closed boundary and continuous data as a target region.

[0073] In this embodiment, the set screening standard includes:

[0074] A standard value for judging the density of the coordinate point distribution is set, denoted as the density judgment standard; a minimum value for measuring the size of the closed boundary is set as the minimum boundary standard; and a maximum value for judging whether the data is continuous is set as the continuity distance standard.

[0075] In this embodiment, please refer to Figure 3 The structure recognition module includes:

[0076] The extraction submodule is configured to extract boundary line data of each sub-region from the face mesh boundary data according to the component partition determination value.

[0077] The first recognition submodule is configured to use an edge detection algorithm to process the extracted boundary line data, identify a discontinuous part in the boundary line, determine the discontinuous part as a gap region, record the specific position of each gap region on the boundary line, and obtain a position parameter of the gap region.

[0078] The measurement submodule is configured to start from the starting point of the boundary line, measure each line segment of the boundary line in sequence, and add up the lengths of all the line segments to obtain a total edge length of the boundary line after traversing the entire boundary line.

[0079] The movement submodule is configured to move along the boundary line, form a corner at a position where a significant change occurs in each direction, and determine the number of corners of the boundary line.

[0080] The arrangement submodule is configured to arrange the obtained gap region position parameter, edge length, and corner number into a data set containing boundary feature information.

[0081] The first comparison submodule is configured to compare the arranged boundary feature data set with a pre-set boundary continuity and regularity judgment condition.

[0082] The input submodule is configured to input the boundary feature data after the pre-set condition comparison into a support vector machine algorithm, classify and judge the boundary features, output an evaluation result, and obtain a boundary regularity coefficient.

[0083] In this embodiment, the first recognition submodule includes:

[0084] Different scale gradient operators are applied to the extracted boundary line data for convolution operation to generate a multi-scale gradient map.

[0085] The gradient direction of each pixel point is calculated, and each pixel point is divided into several direction intervals. Adjacent pixels in the same direction interval are checked for gradient amplitude consistency. If the difference in gradient direction amplitude between adjacent pixels exceeds a threshold value, the adjacent pixels are marked as suspicious gap points.

[0086] Introducing the time dimension, constructing the space-time correlation matrix, for each suspicious gap point in the current frame, checking whether there is a similar gradient feature in the corresponding position in the adjacent frame, if there is, enhancing its confidence as a real gap;

[0087] Based on the gradient statistical characteristics of the local region, the dynamic threshold is calculated, the adaptive threshold segmentation is carried out on the gradient amplitude image, and the binary edge graph is generated;

[0088] The broken edge in the binary edge graph is subjected to morphological analysis, the edge center line is extracted through skeletonization operation, the curvature change rate of the center line is calculated, the suspicious gap region is searched near the curvature mutation point, and the geometric feature verification is carried out on each suspicious gap region.

[0089] For the confirmed gap region, the topological prediction is carried out by using the edge line segment direction information at both ends, the potential edge path of the gap region is fitted through the Bezier curve, and the complete boundary topological structure is reconstructed to determine the gap region.

[0090] In this embodiment, the topological adjustment module comprises:

[0091] The hierarchical index submodule is used for hierarchical indexing of three-dimensional point cloud data according to the boundary regularity coefficient, adding a multi-scale geometric feature descriptor to each spatial coordinate point, and forming an enhanced point cloud feature space;

[0092] The adjustment submodule is used for generating an adaptive reference plane based on the geometric center and the main direction of the boundary contour line, and dynamically adjusting the tolerance of the reference plane according to the boundary regularity coefficient;

[0093] The second identification submodule is used for constructing a space-time cube to perform context analysis on each coordinate point, extracting three-dimensional space features and time sequence features, and automatically identifying the feature dimension of the boundary topological adjustment through an attention mechanism;

[0094] The fusion submodule is used for double-path CNN feature fusion to generate a joint feature representation, deriving a critical height change coefficient and a relative boundary range parameter according to the statistical characteristics of the input three-dimensional point cloud data, constructing a decision network to classify each point, and outputting a probability distribution;

[0095] The screening submodule is used for establishing a point cloud graph structure, with vertices being coordinate points and edges being topological relationships between points, screening a coordinate set in the point cloud graph structure whose offset value is within the relative boundary range and whose height change amount is less than the critical height change coefficient, and obtaining a boundary topological adjustment coefficient.

[0096] In this embodiment, the double-path CNN feature fusion comprises:

[0097] Extracting structural information of the three-dimensional point cloud data in the vertical direction to obtain a height difference path, wherein the structural information at least includes a change feature, capturing steps and cracks;

[0098] Analyzing a horizontal offset relationship between the three-dimensional point cloud data and the boundary contour to identify an edge blur area to obtain an offset path;

[0099] Nonlinearly combining the height difference path and the offset path through a feature fusion module of a CNN network to generate a joint feature representation.

[0100] In this embodiment, the feature reconstruction module includes:

[0101] A sorting submodule is configured to sort the component boundary contour lines according to the importance of the boundary topology adjustment coefficients, and establish a contour line space index structure;

[0102] A forming submodule is configured to uniformly sample an initial node set on the contour line, record the node coordinates and normal directions, apply an adaptive sampling strategy to increase the sampling density in areas with large curvature changes, and form a multi-scale node representation;

[0103] A statistical submodule is configured to define a local neighborhood centered on each node, count the number of nodes in the neighborhood, calculate a normalized node distribution density value according to the neighborhood radius and the number of nodes, and apply Gaussian smoothing processing to eliminate local density fluctuation noise;

[0104] A filtering submodule is configured to calculate the tangent direction change angle between adjacent nodes as a node angle change value, and adopt a bidirectional angle calculation strategy to gradient filter the angle change value;

[0105] A weighting submodule is configured to multiply the node distribution density and the angle change value point by point, apply a spatial attention mechanism to weight the product results of key areas, aggregate the product values of all nodes through area integration to obtain an overall morphological deviation degree;

[0106] A mapping submodule is configured to map the morphological deviation degree to a preset discrete level, construct a multi-dimensional feature vector in combination with the geometric features of the contour line, apply principal component analysis for dimension reduction, and generate a reconstructed feature prediction value.

[0107] In this embodiment, if there are multiple frames of time series data, a corresponding relationship between the current frame nodes and the historical frames is established, a time derivative of the node angle change is calculated, a spatiotemporal feature vector is constructed, and Kalman filtering is applied to predict the future position of the node.

[0108] In this embodiment, the reconstruction prediction module includes:

[0109] The second comparison submodule is configured to classify and arrange the reconstructed feature prediction values according to the component categories, compare the reconstructed feature prediction values of each component with the corresponding component category standard feature value in sequence, and obtain the comprehensive difference value of the component by summing up the differences of all feature dimensions.

[0110] The second calculation submodule is configured to set an initial component structure state value for each component according to the reconstruction confidence level coefficient corresponding to the interval in which the difference value is located, calculate the component structure state value of each component and the reconstruction confidence level coefficient, and obtain the balance factor between the component structure state value and the confidence level coefficient.

[0111] The adding submodule is configured to add the balance factor values of all components, divide the sum by the total number of components, and obtain the average value, so as to determine the overall structure reconstruction trend degree of the components.

[0112] In this embodiment, the boundary feature data after the preset condition comparison is normalized to ensure that different feature dimensions have the same scale range, avoiding unreasonable influence on the model due to some feature factors being too large or too small; based on the boundary feature data, the data features are expanded through feature engineering; for example, the ratio of edge length to average edge length, the difference between the number of corners and the specific standard number of corners, and other derived features are calculated to enrich the expression ability of the data and provide more information for subsequent classification and judgment; the distribution characteristics of the boundary feature data are analyzed, and various kernel functions are designed, such as polynomial kernel function, radial basis kernel function, Gaussian kernel function, etc.; according to the local density and dimension characteristics of the data, the appropriate kernel function or combined kernel function is dynamically selected; for the region where the data is more concentrated, the kernel function that can better fit the local characteristics is selected; for the data with more dispersed distribution, the kernel function that can map in a wider space is selected; through this adaptive selection, the adaptability of the model to different data distribution is improved; a dynamic weight is assigned to each boundary feature data sample; considering the distance of the sample to the classification hyperplane, the feature difference degree of the sample, and other factors, the sample with a larger feature difference and a farther distance to the hyperplane is given a higher weight, and vice versa; in this way, during the model training and judgment process, more attention can be paid to the samples that have a key influence on the determination of the classification boundary, and the accuracy of the model is improved; multiple support vector machine classifiers are constructed, and each classifier is trained based on a different feature subset; for example, one classifier focuses on edge length related features, and another classifier focuses on corner number related features, etc.; in the classification and judgment stage, multiple classifiers work together; the judgment results of multiple classifiers are integrated, different confidence weights are assigned to each classifier according to its performance in the training stage, and finally the comprehensive judgment result is obtained through weighted fusion, which enhances the robustness and accuracy of the model; after obtaining the preliminary boundary regularity coefficient, if new boundary feature data is added, the online learning mechanism is started; the improved support vector machine model is incrementally trained using the new data to update the parameters of the model; according to the characteristics of the new data and the feedback of the model, the previously set weights, kernel function selection strategies, etc. are dynamically adjusted, so that the model can be continuously optimized as the data changes, better adapt to different boundary feature situations, and continuously output more accurate boundary regularity coefficients.

[0113] In this embodiment, the height difference path and the offset distance path capture different features of the point cloud related to the boundary in the vertical and horizontal directions, respectively; the joint feature representation fuses the features of these two aspects, which can more comprehensively and comprehensively reflect the relationship between the coordinate points and the boundary; for example, when processing point cloud data of complex building structures, it can not only reflect the structural changes at different heights of the building facade (such as the height difference features caused by window sills, waist lines, etc.), but also reflect the offset features of the wall edge from the design boundary due to construction errors or erosion, providing rich information for subsequent accurate judgment of whether the coordinate points meet the boundary topology adjustment requirements; the joint feature representation after fusion can make the subsequent decision network more accurately distinguish different types of points; through such comprehensive feature representation, the decision network can better identify coordinate points whose offset value is within the relative boundary range and the height change is less than the critical height change coefficient, reducing the situation of misjudgment and omission; for example, when processing point cloud data with noise interference, relying solely on height difference or offset distance features may misjudge some noise points as valid points, while joint feature representation can use multi-aspect information to more accurately exclude noise and filter out the truly qualified coordinate point set, thereby improving the accuracy of boundary topology adjustment coefficient calculation; joint feature representation can improve the adaptability of the model to different scenes and data characteristics; different three-dimensional reconstruction scenes, such as industrial part reconstruction and terrain reconstruction, have large differences in point cloud data features and boundary conditions; joint feature representation can flexibly fuse features in different aspects, so that the model can effectively work in the face of various complex situations; for example, in industrial scenes, small imperfections on the surface of equipment may exhibit comprehensive features of height difference and offset, and joint feature representation can help the model better adapt to such fine feature changes, accurately perform boundary topology adjustment, and enhance the versatility and robustness of the model.

[0114] In this embodiment, the contour lines are weighted and sorted according to the boundary topology adjustment coefficient, and the higher the coefficient, the more stable the regional topology structure and the greater the contribution to the reconstruction result. A multi-index sorting strategy is adopted, and the priority queue is constructed by considering the boundary length, closeness and relevance to the neighborhood. High-priority contour lines are efficiently processed. Hybrid indexing is adopted for contour lines of different complexities. Quadtree indexing is used for regular regions to improve region query efficiency, and KD-tree indexing is used for complex regions to optimize the performance of near-neighbor search. A multi-level index structure is established, the top-level index quickly locates the approximate region, and the bottom-level index accurately matches the specific contour. Based on the contour line length and the preset sampling interval, an initial node set is generated. The equi-arc-length sampling method is adopted to ensure the uniform distribution of nodes on the curve, and the three-dimensional coordinates and local surface normal direction of each node are recorded. The curvature values of the points of the contour line are calculated, and the regions with sharp curvature changes (such as corners and breakpoints) are identified. The sampling density is increased in the high-curvature region, the two-point insertion method is used to refine the node distribution, and a multi-resolution node hierarchy is constructed. The coarse-scale nodes capture the overall shape, and the fine-scale nodes retain the detail features. A variable radius neighborhood is defined with each node as the center, the radius size is inversely proportional to the local curvature, the number of nodes in the neighborhood is counted, the original density value is calculated in combination with the neighborhood volume, the kernel density estimation method is applied, and the density distribution is smoothed. The density values are normalized by the z-score method to eliminate the scale effect, the local density fluctuations are eliminated by applying Gaussian filtering, the macroscopic density change trend is retained, the density threshold is introduced, and the outliers and low-density noise regions are filtered. The tangent direction change angle between adjacent nodes is calculated, considering both clockwise and counterclockwise directions. The sliding window method is used to calculate the angle change rate, and the window size is adaptive to the local curvature. The angle change value is periodically processed and mapped to the [0, π] interval. The Sobel operator is used to calculate the gradient of the angle change, highlight the shape mutation region, design a direction-sensitive filter to retain the angle change perpendicular to the main direction of the contour, construct an angle change pyramid, and capture the shape features at different scales.The normalized node distribution density is multiplied point by point with the angle change value to generate an initial deviation map. A spatial attention mechanism is applied to give higher weights to key areas such as corners and endpoints. Topological constraints are introduced to punish deviations that destroy the connectivity of the original contour. The contour line is divided into multiple semantic regions such as walls and doors and windows. The deviation of each region is calculated separately. A weighted integral method is used to aggregate the deviations of each region. The weights are based on the importance and area of the region. A deviation cumulative distribution map is constructed to visually display the spatial distribution of shape abnormalities. Continuous shape deviation values are mapped to a preset discrete level (such as 1-5 levels). The K-means clustering method is used to automatically determine the threshold for level division. Each level is assigned a semantic label (such as "normal", "slight deformation", "severe damage"). The contour geometric features (perimeter, area, circularity) and topological features (hole number, branch number) are combined to extract shape descriptors (such as Fourier descriptors, Zernike moments) to enhance feature expression capability. A spatiotemporal feature vector is constructed to fuse the change trend of multiple frames of data. A feature covariance matrix is calculated to determine the contribution rate of each feature dimension. The principal components with cumulative contribution rates exceeding a threshold (such as 95%) are selected to generate a low-dimensional feature vector to remove redundant information.

[0115] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional structure reconstruction system based on artificial intelligence topology optimization, characterized in that, The system includes: The data extraction module is used to acquire point cloud image data, surface mesh boundary data and spatial scale parameters of building components. Based on the coordinate point distribution density values ​​in the component point cloud image data, combined with the surface mesh boundary data, the module performs region filtering, selects component regions with closed boundaries and data continuity as target regions, and generates component partitioning judgment values. The structure recognition module is used to call the boundary line data of each sub-region in the surface mesh boundary data according to the component partition judgment value, compare the position parameters, corresponding edge length and number of corners of the gap area in the boundary line, determine whether the boundary continuity and regularity judgment conditions are met, and obtain the boundary regularity coefficient. The topology adjustment module is used to call spatial coordinate points in point cloud image data according to the boundary regularity coefficient, calculate the relative offset distance of each coordinate point by the height difference and the boundary contour, and filter the set of coordinates whose offset value is within the relative boundary range and whose height change is less than the critical height change coefficient to obtain the boundary topology adjustment coefficient. The feature reconstruction module is used to call the boundary contour lines of each component based on the boundary topology adjustment coefficient, obtain the number of nodes, node distribution density and the corresponding angle change value of each node, calculate the degree of deviation of the component shape by the product of node distribution density and angle change value, and generate the reconstructed feature prediction value. The reconstruction prediction module is used to call the reconstruction feature prediction value, calculate the difference between the prediction value and the standard feature value of each component category, calculate the balance factor between the component structural state value and the confidence level coefficient based on the reconstruction confidence level coefficient corresponding to the interval of the difference value, and establish the overall structural reconstruction trend degree of the component by averaging the balance factor values ​​of each component.

2. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 1, characterized in that, The data extraction module includes: The traversal submodule is used to set the filtering criteria and traverse every coordinate point in the point cloud image data of building components to obtain the initially filtered areas. The determination submodule is used to determine the associated boundary line for each dense point in the surface mesh boundary data within the initially selected area; The first calculation submodule is used to record each closed boundary region that may meet the requirements. It randomly selects two adjacent points within the region and calculates the distance between the two points. If the distance is less than or equal to the continuity distance standard, it continues to check the distances of other adjacent points within the region. The determination submodule is used to determine whether the data in a region is continuous when the distance between all adjacent points in the region is less than or equal to the continuity distance criterion. The generation submodule is used to take the component region with closed boundaries and data continuity as the target region and generate component partitioning judgment values.

3. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 2, characterized in that, The set screening criteria include: Set a standard value for judging the density of coordinate points, and denote it as the density judgment standard; set a minimum value for measuring the size of the closed boundary, and denote it as the minimum boundary standard; set a maximum distance for judging whether the data is continuous, i.e., the continuity distance standard.

4. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 1, characterized in that, The structure recognition module includes: The extraction submodule is used to extract the boundary line data of each sub-region from the surface mesh boundary data based on the component partitioning judgment value. The first identification submodule is used to process the extracted boundary line data using an edge detection algorithm, identify discontinuous parts in the boundary line, determine them as gap areas, record the specific position of each gap area on the boundary line, and obtain the position parameters of the gap area. The measurement submodule is used to measure each segment of the boundary line sequentially, starting from the starting point of the boundary line. After traversing the entire boundary line, the lengths of all segments are accumulated to obtain the total edge length of the corresponding boundary line. The move submodule is used to move along the boundary line, form the position of the corner where there is a significant change in direction, and determine the number of corners along the boundary line; The sorting submodule is used to sort the obtained gap region location parameters, edge lengths and number of corners into a dataset containing boundary feature information; The first comparison submodule is used to compare the organized boundary feature dataset with the pre-defined boundary continuity and regularity determination conditions. The input submodule is used to input the boundary feature data after comparison under preset conditions into the support vector machine algorithm, classify and judge the boundary features, and output the evaluation results to obtain the boundary regularity coefficient.

5. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 4, characterized in that, The first identification submodule includes: The extracted boundary line data are convolved using gradient operators of different scales to generate multi-scale gradient maps. Calculate the gradient direction of each pixel and divide each pixel into several directional intervals. Perform gradient magnitude consistency checks on adjacent pixels within the same directional interval. If the difference in gradient direction magnitude between adjacent pixels exceeds a threshold, it is marked as a suspicious gap point. By introducing a time dimension and constructing a spatiotemporal correlation matrix, for each suspicious gap point in the current frame, we check whether there are similar gradient features at its corresponding position in adjacent frames. If so, we enhance its confidence as a real gap. Dynamic thresholds are calculated based on the gradient statistical characteristics of local regions, and adaptive threshold segmentation is performed on the gradient magnitude image to generate a binary edge map. Morphological analysis is performed on the fracture edges in the binary edge map. The edge centerline is extracted through skeletonization operation, the curvature change rate of the centerline is calculated, and suspicious gap regions are searched near the curvature abrupt change points. Geometric features are verified for each suspicious gap region. For a confirmed gap region, topological prediction is performed using the direction information of the edge segments at both ends. By fitting the potential edge paths of the gap region with Bézier curves, the complete boundary topology is reconstructed to identify the gap region.

6. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 1, characterized in that, The topology adjustment module includes: The hierarchical indexing submodule is used to perform hierarchical indexing of 3D point cloud data based on boundary regularity coefficients, adding multi-scale geometric feature descriptors to each spatial coordinate point to form an enhanced point cloud feature space. The adjustment submodule is used to generate an adaptive reference plane based on the geometric center and main direction of the boundary contour line, and dynamically adjust the tolerance of the reference plane according to the boundary regularity coefficient. The second identification submodule is used to construct a spatiotemporal cube to perform contextual analysis on each coordinate point, extract three-dimensional spatial features and time series features, and automatically identify the feature dimensions of boundary topology adjustments through an attention mechanism. The fusion submodule is used to perform dual-path CNN feature fusion, generate joint feature representation, derive critical height change coefficient and relative boundary range parameter based on the statistical characteristics of the input 3D point cloud data, construct a decision network to classify each point, and output probability distribution; The filtering submodule is used to establish a point cloud structure, where vertices are coordinate points and edges are the topological relationships between points. It filters the set of coordinates in the point cloud structure whose offset values ​​are within the relative boundary range and whose height change is less than the critical height change coefficient, thus obtaining the boundary topology adjustment coefficient.

7. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 6, characterized in that, The dual-path CNN feature fusion includes: Extract the structural information of the 3D point cloud data in the vertical direction to obtain the height difference path. The structural information includes at least the change features, capturing steps and cracks. Analyze the horizontal offset relationship between 3D point cloud data and boundary contours, identify blurred edge areas, and obtain the offset path; The feature fusion module of the CNN network nonlinearly combines the height difference path and the offset path to generate a joint feature representation.

8. The three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 1, characterized in that, The feature reconstruction module includes: The sorting submodule is used to sort the importance of component boundary contours according to the boundary topology adjustment coefficient and establish a spatial index structure for the contours. A submodule is formed to uniformly sample the initial node set on the contour line, record the node coordinates and normal directions, and apply an adaptive sampling strategy to increase the sampling density in areas with large curvature changes to form a multi-scale node representation. The statistics submodule is used to define a local neighborhood centered on each node, count the number of nodes in the neighborhood, calculate the normalized node distribution density value based on the neighborhood radius and the number of nodes, and apply Gaussian smoothing to eliminate local density fluctuation noise. The filtering submodule is used to calculate the angle of change of the tangent direction between adjacent nodes, which is used as the node angle change value. A two-way angle calculation strategy is used to perform gradient filtering on the angle change value. The weighted submodule is used to multiply the node distribution density with the angle change value point by point. It applies a spatial attention mechanism to weight the product results of key areas and aggregates the product values ​​of all nodes through regional integration to obtain the overall shape deviation. The mapping submodule is used to map the morphological deviation to a preset discrete level, construct a multi-dimensional feature vector by combining the geometric features of the contour, apply principal component analysis to reduce dimensionality, and generate reconstructed feature prediction values.

9. A three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 8, characterized in that, If multiple frames of time-series data exist, establish the correspondence between the current frame node and the historical frames, calculate the time derivative of the node angle change, construct the spatiotemporal feature vector, and apply Kalman filtering to predict the future position of the node.

10. A three-dimensional structure reconstruction system based on artificial intelligence topology optimization as described in claim 1, characterized in that, The reconstruction prediction module includes: The second comparison submodule is used to classify and organize the reconstructed feature prediction values ​​according to the component category. For each component, its reconstructed feature prediction value is compared with the corresponding component category morphological standard feature value in turn, and the differences of all feature dimensions are summarized to obtain the comprehensive difference value of the component. The second calculation submodule is used to set an initial structural state value for each component based on the reconstruction confidence level coefficient corresponding to the interval where the difference value is located, and to calculate the structural state value of each component and the reconstruction confidence level coefficient to obtain the balance factor between the structural state value and the confidence level coefficient. The summation submodule is used to sum the equilibrium factor values ​​of all components, divide by the total number of components, and obtain the average value to determine the overall structural reconstruction trend of the components.

Images