A curtain wall reverse modeling method and system based on point cloud data

Abstract

The present invention provides a curtain wall reverse modeling method and system based on point cloud data, which are applied to the field of modeling data processing; the present invention can timely discover model defects and improve the overall quality by accurately identifying irregular areas and performing error analysis, and at the same time apply surface fitting algorithm and hierarchical structure reconstruction technology to enhance the smoothness and detail expression of the model to make it closer to the real object, and through local refinement guided by structural features, the geometric complexity is targetedly improved, and evenly distributed new vertices are inserted to reduce jagged and non-smooth areas, thereby improving the overall fluency and visual effects.

Description

Technical Field

[0001] The present invention relates to the field of modeling data processing, and in particular to a curtain wall reverse modeling method and system based on point cloud data. Background Art

[0002] Point cloud data is the basic data source in curtain wall reverse modeling and is widely used in the measurement and reconstruction of building structures. Point cloud data is usually obtained by laser scanning, photogrammetry or other 3D sensors. Its advantage is that it can accurately capture the geometric shape of the curtain wall.

[0003] However, curtain walls usually contain a large number of complex geometric shapes, such as curved surfaces, large angle changes, or multiple materials. As a result, the generated surface model may have irregular faces or sharp edges, making it difficult to ensure the smoothness of the model and unable to meet the appearance requirements of the actual curtain wall. Summary of the invention

[0004] The present invention aims to solve the problem that it is difficult to maintain the smoothness and realism of the model when converting from a point cloud to a mesh model or a surface model, and provides a curtain wall reverse modeling method and system based on point cloud data.

[0005] The present invention adopts the following technical means to solve the technical problem:

[0006] The present invention provides a curtain wall reverse modeling method based on point cloud data, comprising:

[0007] Based on the pre-converted mesh surface model of the building curtain wall, the model surface of the mesh surface model is scanned, and the corresponding irregular area is identified through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area, and an uneven area;

[0008] Determining whether the error value of the irregular area exceeds a preset threshold;

[0009] If so, the irregular area is marked on the mesh surface model, a preset surface fitting algorithm is applied to identify the model curvature change of the mesh surface model, the interpolation points of the irregular area are adjusted according to the model curvature change, the mesh surface model and the original point cloud data are subjected to error analysis to obtain model fitting parameters, and the hierarchical structure of the mesh surface model is reconstructed according to the interpolation points and the model fitting parameters;

[0010] Determining whether the model smoothness of the mesh surface model is consistent with a preset sense of reality;

[0011] If there is any inconsistency, based on the original point cloud data, the preset structural features of the building curtain wall are collected, and the local area to be refined of the mesh surface model is obtained according to the structural features. A preset refinement algorithm is applied to insert a number of new vertices in the local area to be refined, and the new vertices are evenly distributed to the local area to be refined. Each adjacent new vertices are connected, and a triangular mesh is constructed on the local area to be refined to complete the smooth refinement of the mesh surface model, wherein the structural features specifically include straight lines, planes, surfaces, edges and angles.

[0012] Furthermore, the step of performing error analysis on the mesh surface model and the original point cloud data further includes:

[0013] The mesh surface model is divided into different regions, corresponding region identifiers are assigned based on the different regions, points corresponding to the region identifiers are extracted from the original point cloud data, and error indicators of the points are calculated respectively, wherein the different regions specifically include important feature regions, boundary regions, and complex geometric regions, and the error indicators specifically include errors of feature edges, matching errors of adjacent structures, and curvature change errors;

[0014] Determining whether the error index exceeds a preset index;

[0015] If so, an error heat map corresponding to different regions is generated according to the error index, an error trend is collected according to the error heat map, and the error difference of the different regions is displayed through the error trend.

[0016] Furthermore, before the step of marking the irregular area on the mesh surface model, the method further includes:

[0017] Based on the density of neighboring points of the original point cloud data, identifying the density distance between points;

[0018] Determining whether the density distance exceeds a preset distance;

[0019] If so, the two points beyond the preset distance are defined as noise points, and the preset median filter is applied to remove the outliers of the noise points. According to the density change of the original point cloud data, the hole area corresponding to the density change is identified, and the point cloud boundary is extracted at the edge of the hole area. The area missing information is identified based on the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

[0020] Furthermore, the step of completing the smoothing and refinement of the mesh surface model further includes:

[0021] Based on the pre-extracted point coordinates in the local area to be refined, a preliminary fitting model is generated after the point coordinates are connected and combined, an error distance value between the mesh surface model and the preliminary fitting model is calculated, and a preset least square method is used to iteratively update the error distance value;

[0022] Determine whether the change in the error distance value after iterative update is less than a preset threshold;

[0023] If so, extract the vertex coordinates and the patch information from the preliminary fitting model, apply the preset Laplace smoothing according to the vertex coordinates and the patch information, and perform smoothing on the specific local area corresponding to the vertex coordinates and the patch information until the vertex coordinates and the patch information meet the preset smoothing convergence conditions.

[0024] Furthermore, the step of determining whether the error value of the irregular area exceeds a preset threshold value further includes:

[0025] Based on the error points pre-detected in the irregular area, obtaining the quantity value of the error points;

[0026] Determining whether the quantity value exceeds a preset value;

[0027] If so, the distribution information of the error point on the irregular area is identified, and the detailed information of the irregular area is generated on the pre-built visualization model according to the distribution information, wherein the detailed information specifically includes the normal vector, color, error type, timestamp and neighborhood point information.

[0028] Furthermore, the step of determining whether the model smoothness of the mesh surface model is consistent with the preset sense of reality further includes:

[0029] Based on a preset main viewing angle of the mesh surface model, the main viewing angle is switched to another preset secondary viewing angle, and a dynamic view test is performed on the mesh surface model under the other secondary viewing angle to collect dynamic viewing angle information;

[0030] Determining whether the dynamic viewing angle information meets a preset real condition;

[0031] If not, the viewing angle parameters of the other secondary viewing angles are recorded, and the visual defect information of the mesh surface model is obtained according to the viewing angle parameters, wherein the viewing angle parameters specifically include angle, distance and lighting.

[0032] Furthermore, the step of scanning the model surface of the mesh surface model based on the pre-converted building curtain wall, and identifying the corresponding irregular area through the model surface further includes:

[0033] Based on the point cloud features preset in the original point cloud data, the point cloud features are classified to identify geometric features, and a corresponding preliminary mesh model is generated by combining the point cloud features and the geometric features, wherein the point cloud features specifically include color, normal vector and density, and the geometric features specifically include edge, corner point and surface features;

[0034] Determining whether the preliminary grid model can cover the preset feature points;

[0035] If not, then obtain the uncovered feature surfaces of the preliminary mesh model, collect the hierarchical structure of the preliminary mesh model based on the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model according to the hierarchical structure, wherein the hierarchical levels specifically include a basic layer, an intermediate layer and an advanced layer.

[0036] The present invention also provides a curtain wall reverse modeling system based on point cloud data, comprising:

[0037] A scanning module is used to scan the model surface of the mesh surface model based on the pre-converted mesh surface model of the building curtain wall, and identify the corresponding irregular area through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area and an uneven area;

[0038] A judging module, used to judge whether the error value of the irregular area exceeds a preset threshold;

[0039] an execution module, for, if yes, marking the irregular area on the mesh surface model, applying a preset surface fitting algorithm to identify a model curvature change of the mesh surface model, adjusting the interpolation points of the irregular area according to the model curvature change, performing error analysis between the mesh surface model and the original point cloud data, analyzing to obtain model fitting parameters, and reconstructing the hierarchical structure of the mesh surface model according to the interpolation points and the model fitting parameters;

[0040] A second judgment module is used to judge whether the model smoothness of the mesh surface model is consistent with the preset realism;

[0041] The second execution module is used to collect preset structural features of the building curtain wall based on the original point cloud data if there is any inconsistency, obtain the local area to be refined of the mesh surface model according to the structural features, apply a preset refinement algorithm to insert a number of new vertices in the local area to be refined, evenly distribute the new vertices to the local area to be refined, connect each adjacent new vertices, construct a triangular mesh on the local area to be refined, and complete the smooth refinement of the mesh surface model, wherein the structural features specifically include straight lines, planes, surfaces, edges and angles.

[0042] Furthermore, the execution module also includes:

[0043] A division unit, used to divide the mesh surface model into different regions, assign corresponding region identifiers based on the different regions, extract points corresponding to the region identifiers from the original point cloud data, and calculate error indicators of the points respectively, wherein the different regions specifically include important feature regions, boundary regions and complex geometric regions, and the error indicators specifically include feature edge errors, matching errors of adjacent structures and curvature change errors;

[0044] A judging unit, used to judge whether the error index exceeds a preset index;

[0045] An execution unit is used to generate error heat maps corresponding to different areas according to the error indicators, collect error trends according to the error heat maps, and display the error differences between the different areas through the error trends.

[0046] Furthermore, it also includes:

[0047] An identification module, used for identifying the density distance between points based on the density of neighboring points of the original point cloud data;

[0048] A third judging module is used to judge whether the density distance exceeds a preset distance;

[0049] The third execution module is used to define the two points beyond the preset distance as noise points, apply the preset median filter to remove the outliers of the noise points, identify the hole area corresponding to the density change according to the density change of the original point cloud data, extract the point cloud boundary at the edge of the hole area, and identify the area missing information according to the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

[0050] The present invention provides a curtain wall reverse modeling method and system based on point cloud data, which has the following beneficial effects:

[0051] The present invention can timely discover model defects and improve the overall quality by accurately identifying irregular areas and performing error analysis. At the same time, it applies surface fitting algorithms and hierarchical structure reconstruction technology to enhance the smoothness and detail expression of the model, making it closer to the real object. It also specifically improves the geometric complexity through local refinement guided by structural features, inserts evenly distributed new vertices to reduce jagged and uneven areas, thereby improving the overall fluency and visual effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic diagram of a process of an embodiment of a curtain wall reverse modeling method based on point cloud data of the present invention;

[0053] Figure 2 The present invention is a structural block diagram of an embodiment of a curtain wall reverse modeling system based on point cloud data. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. The implementation of the objectives, functional features and advantages of the present invention will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Reference Figure 1 , is a curtain wall reverse modeling method based on point cloud data in one embodiment of the present invention, comprising:

[0057] S1: Based on the pre-converted mesh surface model of the building curtain wall, scan the model surface of the mesh surface model, and identify the corresponding irregular area through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area, and an uneven area;

[0058] S2: Determine whether the error value of the irregular area exceeds a preset threshold;

[0059] S3: If yes, mark the irregular area on the mesh surface model, apply a preset surface fitting algorithm to identify the model curvature change of the mesh surface model, adjust the interpolation points of the irregular area according to the model curvature change, perform error analysis on the mesh surface model and the original point cloud data, analyze and obtain model fitting parameters, and reconstruct the hierarchical structure of the mesh surface model according to the interpolation points and the model fitting parameters;

[0060] S4: Determine whether the model smoothness of the mesh surface model is consistent with the preset sense of reality;

[0061] S5: If there is inconsistency, based on the original point cloud data, the preset structural features of the building curtain wall are collected, and the local area to be refined of the mesh surface model is obtained according to the structural features. A preset refinement algorithm is applied to insert a number of new vertices in the local area to be refined, and the new vertices are evenly distributed to the local area to be refined, and each adjacent new vertices are connected to construct a triangular mesh on the local area to be refined, thereby completing the smooth refinement of the mesh surface model, wherein the structural features specifically include straight lines, planes, surfaces, edges and angles.

[0062] In this embodiment, the system scans the model surface of the mesh surface model based on the pre-converted mesh surface model of the building curtain wall, and identifies the corresponding irregular areas through the model surface. The irregular areas specifically include jagged areas, distorted areas, and non-smooth areas. Then the system determines whether the error values ​​of these irregular areas exceed the preset error thresholds to execute the corresponding steps; for example, when the system determines that the error value of a certain irregular area does not exceed the preset error threshold, the system will consider that the geometric features of the area are consistent with the expected model and basically meet the design requirements. The system will keep monitoring the area and re-evaluate its status regularly to ensure that the system is updated or changed in the subsequent model. In the model, the stability of its geometric features is still maintained, while other irregular areas are continuously evaluated to ensure the overall quality of the model and avoid ignoring other possible problem areas. After confirming that the error value of the area meets the requirements, it can be decided not to modify it and maintain the existing model state, thereby saving time and resources. For example, when the system determines that the error value of an irregular area exceeds the preset error threshold, the system will consider that the geometric features of the area are inconsistent with the expected model and cannot meet the design requirements. The system will mark these irregular areas on the mesh surface model and apply the pre-set surface fitting algorithm to identify the model curvature changes of the mesh surface model. The interpolation points in the irregular area are adjusted according to the different changes in the model curvature, and the mesh surface model and the original point cloud data are subjected to error analysis to obtain the model fitting parameters. Finally, the interpolation points and model fitting parameters are combined to reconstruct the hierarchical structure of the mesh surface model. The system can deeply analyze the geometric features of the model by applying the preset surface fitting algorithm and identify the deviation from the expected surface. This refined processing helps to optimize the geometric shape of the model to make it more in line with the design requirements, thereby improving the overall quality of the model. At the same time, the interpolation points in the irregular area are adjusted according to the different changes in curvature, making the geometric features of the area more delicate and precise. This adjustment helps to achieve in key areas Smoother transitions avoid jagged or uneven phenomena caused by improper interpolation, improve visual effects, and by performing error analysis on the mesh surface model and the original point cloud data, more detailed model information can be obtained to identify which areas have obvious deviations. This process enables the model fitting parameters to be optimized based on actual data, enhances the accuracy of the model, and ensures that it can better reflect the characteristics of real objects in practical applications. The hierarchical structure of the reconstructed mesh surface model enables the model to show richer details visually. The system then detects whether the model smoothness of the mesh surface model is consistent with the pre-set sense of reality, and then executes the corresponding steps.For example, when the system determines that the model smoothness of the mesh surface model is consistent with the pre-set sense of reality, the system will consider that the model has met the design requirements in terms of visual effects and geometric features. The system will record the current state of the model to confirm that it meets the design standards and quality requirements, laying the foundation for subsequent operations. At the same time, after the model is confirmed to be qualified, it can enter the next project stage, such as engineering drawing generation, rendering, animation production or other subsequent design work. Even if the current state of the model is good, it should be continuously monitored and re-evaluated regularly to ensure that it maintains its quality in subsequent modifications or environmental changes. For example, when the system determines that the model smoothness of the mesh surface model is inconsistent with the pre-set sense of reality, the system will consider that the model still cannot meet the design requirements in terms of visual effects. The system will collect the pre-set structural features of the building curtain wall based on the original point cloud data. The structural features specifically include straight lines, planes, surfaces, edges and angles. According to these structural features, the local area of ​​the mesh surface model to be refined is obtained, and the pre-set The refinement algorithm inserts several new vertices in the local area to be refined, evenly distributes these new vertices to the local area to be refined, connects each adjacent new vertices, and constructs a triangular mesh on the local area to be refined to complete the smooth refinement of the mesh surface model; the system significantly improves the surface smoothness of the model by inserting new vertices in the local area to be refined and rebuilding the corresponding triangular mesh. This smoothing process reduces visible geometric irregularities and makes the model visually softer and more natural. At the same time, adding new vertices according to the structural features of the building curtain wall (such as straight lines, planes, curved surfaces, etc.) during the refinement process helps to more accurately express the complex shape of the building. This detail enhancement enables the model to more realistically reflect the actual appearance of the building, improve the accuracy and visual appeal of the design, and refinement based on the preset structural features of the building curtain wall can ensure that the model remains geometrically consistent and coherent after refinement, which means that while adding details, the shape and function of the model still conform to the original design intent, avoiding new geometric errors caused by refinement. ;

[0063] It should be noted that a preset surface fitting algorithm is applied to identify the model curvature change of the mesh surface model, and the interpolation points of the irregular area are adjusted according to the model curvature change. The specific examples are as follows:

[0064] Suppose that the exterior curtain wall modeling of a modern building is being carried out. The system obtains the point cloud data of the building's exterior wall. This data is obtained by laser scanning and contains a large number of details and geometric features.

[0065] Point cloud data: Contains point cloud data of the building's exterior wall surface, which may have noise and irregularities. By creating a smooth and realistic mesh surface model that meets the design requirements, the system then selects the B-spline surface fitting algorithm because the B-spline algorithm is suitable for processing complex geometric shapes and can provide high flexibility and control;

[0066] The B-spline algorithm is used to perform surface fitting on the original point cloud data to generate a preliminary mesh model. This process involves selecting appropriate control points and parameters to obtain a smoother surface. On the generated mesh model, the curvature of the model is calculated using the Gaussian curvature and mean curvature calculation methods. Through analysis, it is found that the Gaussian curvature values ​​of some areas are as high as 0.12, which exceeds the preset threshold (such as 0.1), indicating that these areas have large bends or protrusions.

[0067] Irregular areas are then identified. Based on the curvature analysis results, the system automatically marks these high curvature areas as irregular areas, for example:

[0068] Area A: A protruding edge of the exterior wall with a clear change in curvature;

[0069] Area B: The curved part of the corner, which may be slightly twisted;

[0070] The interpolation points can be adjusted. Assume that there are originally 5 interpolation points in area A. After the interpolation points are adjusted, the number is increased to 15 to better capture details. These new interpolation points are evenly distributed on the edges to ensure that each part can accurately reflect the actual shape. Area B may have fewer adjustments and only needs to be increased to 7 interpolation points to ensure smoothness even in the distorted parts.

[0071] Finally, the mesh is regenerated and the mesh models of area A and area B are rebuilt using the newly adjusted interpolation points. This process uses the interpolation points to build new triangular meshes, thereby improving the geometric accuracy of these areas. The Laplace smoothing algorithm can be applied to the newly generated mesh to eliminate possible jagged edges and small depressions to ensure the smoothness of the overall model. The processed model is rechecked to see if its visual effect meets the design standards. The processing of curvature changes can ensure the visual smoothness and realism of the exterior wall.

[0072] In summary, the system enables the final mesh surface model to more realistically reproduce the appearance characteristics of the building curtain wall, with higher visual smoothness and richer details, which meets the design requirements. Such a model will show better results in subsequent rendering and analysis.

[0073] It should be added that a preset thinning algorithm is applied to insert a number of new vertices in the local area to be thinned, and the new vertices are evenly distributed to the local area to be thinned. The specific example is as follows:

[0074] Suppose you are working on a modern building's exterior wall model, specifically a curved corner. After preliminary analysis, you find that the details in this area are insufficient, resulting in an unsmooth and unrealistic effect when visualizing.

[0075] First, determine the refinement area. Through curvature analysis or error analysis, determine that the corner part is the local refinement area. This area is a triangle formed by four vertices (A, B, C, D), forming a corner surface. Select the Loop refinement algorithm because it can effectively increase the number of triangles and improve the surface details.

[0076] Then insert new vertices. The original triangle is set to △ABCD, with sides AB, BC, CD, and DA. Insert a new vertex in each side: new vertex E is at the midpoint of side AB; new vertex F is at the midpoint of side BC; new vertex G is at the midpoint of side CD; new vertex H is at the midpoint of side DA.

[0077] Then, the new vertices E, F, G, and H are evenly distributed in the middle of their respective edges, so that each edge introduces a new detail point; that is, the original 4 vertices are now expanded to 8 vertices (the original 4 vertices + 4 new vertices), forming a more detailed mesh structure; the newly inserted vertices are used to connect the original vertices to form new triangular faces; for example:

[0078] New triangle 1: △ABE (including new vertex E);

[0079] New triangle 2: △BCF (including new vertex F);

[0080] New triangle 3: △CDG (including new vertex G);

[0081] New triangle 4: △DAH (including new vertex H);

[0082] New triangle 5: △EFG (connecting new vertices E, F and G);

[0083] New triangle 6: △EHA (connecting new vertices E, H and original vertex A);

[0084] Finally, with these new triangles, the model's corner details are significantly improved; that is, these new details make the surface look smoother and the visual effect is closer to the appearance of the actual building;

[0085] In summary, by inserting new vertices and evenly distributing them, the system can make the model of the corner part more realistically reflect the architectural design; in the subsequent rendering, the refined model can not only capture more changes in light and shadow, but also make the building's performance at different angles more coordinated, thereby improving the overall visual effect and realism.

[0086] In this embodiment, the step S3 of performing error analysis on the mesh surface model and the original point cloud data further includes:

[0087] S31: Divide the mesh surface model into different regions, assign corresponding region identifiers based on the different regions, extract points corresponding to the region identifiers from the original point cloud data, and calculate error indicators of the points respectively, wherein the different regions specifically include important feature regions, boundary regions, and complex geometric regions, and the error indicators specifically include feature edge errors, matching errors of adjacent structures, and curvature change errors;

[0088] S32: Determine whether the error index exceeds a preset index;

[0089] S33: If yes, then generate error heat maps corresponding to different regions according to the error index, collect error trends according to the error heat maps, and display the error differences between the different regions through the error trends.

[0090] In this embodiment, the system divides the mesh surface model into different areas, which specifically include important feature areas, boundary areas, and complex geometric areas. Based on these different areas, corresponding area identifiers are assigned, and points corresponding to the area identifiers are extracted from the original point cloud data. The error indicators of these points are calculated respectively. The error indicators specifically include the error of the feature edge, the matching error of the adjacent structure, and the curvature change error. Then the system determines whether these error indicators exceed the preset indicator upper limit to execute the corresponding steps; for example, when the system determines that the error indicator of a certain point does not exceed the preset indicator upper limit, the system will consider that the points in the area meet the expected accuracy requirements, indicating that the reconstruction quality of the mesh surface model is good. The system will mark the area as a valid area, keep it unchanged in subsequent processing, continue to optimize the details of the area or merge it into a larger model to ensure the accuracy and integrity of the entire model, and store the error indicator and other related data of the point in the database for subsequent performance analysis and model evaluation, and continue to evaluate the error indicators of other areas to ensure the quality of the entire mesh model and promptly discover and deal with possible problems; for example, when the system determines that When the error index of a certain point exceeds the preset upper limit of the index, the system will consider that the points in the area do not meet the expected accuracy requirements, resulting in poor reconstruction quality of the mesh surface model. The system will generate error heat maps corresponding to different areas based on these error indicators, collect the corresponding error trends based on the error heat maps, and display the error differences in different areas through the error trends; by monitoring the error indicators, the system can identify potential accuracy problems in real time to avoid greater errors in subsequent processing, which helps to improve the reliability of the overall model. At the same time, based on the error indicators, the system can generate error heat maps for different areas, which intuitively show the error distribution of each area. This visualization effect helps engineers quickly identify problem areas and make targeted adjustments. By collecting and analyzing the trend data of the error heat map, the system can identify the error changes in different areas. This trend analysis can help developers understand the causes of errors under specific conditions, and then optimize the processing flow and reconstruction algorithm of point cloud data. By analyzing the error differences, engineers can make targeted adjustments to the data collection and processing flow, optimize the design of model reconstruction, and improve the efficiency and quality of subsequent processing.

[0091] It should be noted that the mesh surface model is divided into different regions, corresponding region identifiers are assigned based on the different regions, points corresponding to the region identifiers are extracted from the original point cloud data, and error indicators of the points are calculated respectively. The specific examples are as follows:

[0092] Suppose you are rebuilding a 3D model of a garage door, using laser scanning or photogrammetry to obtain point cloud data, and the system wants to process this point cloud data to ensure that the final model is accurate;

[0093] 1. Regional division of the mesh surface model: Divide the mesh surface model into the following areas:

[0094] Important feature area (F1): including the edge and lock part of the garage door;

[0095] Boundary area (B1): upper and side boundaries of the garage door;

[0096] Complex geometry area (C1): the hinge part of the garage door and the surface recessed area;

[0097] 2. Assignment of regional identification, assign identification to each region:

[0098] Important feature area: marked as "F1"

[0099] Boundary area: marked as "B1"

[0100] Complex geometry area: marked as "C1"

[0101] 3. Extract points from the original point cloud data. Assume that the following points are extracted from the point cloud data:

[0102] F1 (important feature area):

[0103] Point 1: (1.0, 2.0, 3.0)

[0104] Point 2: (1.1, 2.1, 3.1)

[0105] B1 (Border Zone):

[0106] Point 3: (1.5, 2.0, 3.0)

[0107] Point 4: (1.5, 2.1, 3.0)

[0108] C1 (complex geometry area):

[0109] Point 5: (1.2, 2.5, 2.9)

[0110] Point 6: (1.3, 2.4, 2.8)

[0111] 4. Calculate the error index. For each point extracted, calculate the error index;

[0112] F1 (important feature area), error of feature edge:

[0113] Reference points for ideal edge models (1.0, 2.0, 3.0):

[0114] The distance between point 1 and the reference point,

[0115] =0.0 (perfect match);

[0116] The distance between point 2 and the reference point,

[0117] =√((1.1-1.0)²+(2.1-2.0)²+(3.1-3.0)²)=√(0.01+0.01+0.01)=0.173;

[0118] The matching error of adjacent structures,

[0119] The distance between point 1 and point 2 = √((1.1-1.0)²+(2.1-2.0)²+(3.1-3.0)²)=0.173, which is in line with expectations;

[0120] Curvature variation error: In this area, if the calculated average curvature is 0.1, it exceeds the preset error range compared with the design required curvature (0.05);

[0121] B1 (boundary area), error at the edge of the feature:

[0122] Reference points for ideal boundaries (1.5, 2.0, 3.0):

[0123] The distance between point 3 and the reference point,

[0124] =0.0 (perfect match);

[0125] The distance between point 4 and the reference point,

[0126] =√((1.5-1.5)²+(2.1-2.0)²+(3.0-3.0)²)=0.1;

[0127] The matching error of adjacent structures,

[0128] The distance between point 3 and point 4 = 0.1, which is in line with expectations;

[0129] Curvature change error: The curvature calculation of this area is normal and meets the preset requirements;

[0130] C1 (complex geometry area), error at feature edges:

[0131] Reference points of ideal geometric model (1.2, 2.5, 3.0):

[0132] The distance between point 5 and the reference point,

[0133] =√((1.2-1.2)²+(2.5-2.5)²+(2.9-3.0)²)=0.1;

[0134] The distance between point 6 and the reference point,

[0135] =√((1.3-1.2)²+(2.4-2.5)²+(2.8-3.0)²)=√(0.01+0.01+0.04)=0.173;

[0136] The matching error of adjacent structures,

[0137] The distance between point 5 and point 6 = √((1.3-1.2)²+(2.4-2.5)²+(2.8-2.9)²)=0.173;

[0138] Curvature change error: The curvature change of the area exceeds the preset range;

[0139] Finally, error analysis and processing are performed. The system will generate an error heat map based on the above error indicators to show the error differences in different areas, as follows:

[0140] The error in the F1 region is large and exceeds the design requirements, and the system can be marked as needing to be rebuilt or optimized;

[0141] The B1 area behaves normally and can continue to be used;

[0142] The complex geometric area of ​​the C1 region requires special attention and may require further analysis;

[0143] In summary, by performing error analysis on different areas, the system can identify accuracy problems and take corresponding measures, which not only helps to improve the reconstruction quality of the mesh surface model, but also provides important reference data for future design and use.

[0144] In this embodiment, before step S3 of marking the irregular area on the mesh surface model, the method further includes:

[0145] S301: Identifying density distances between points based on the density of neighboring points of the original point cloud data;

[0146] S302: Determine whether the density distance exceeds a preset distance;

[0147] S303: If so, the two points beyond the preset distance are defined as noise points, and the preset median filter is applied to remove the outliers of the noise points. According to the density change of the original point cloud data, the hole area corresponding to the density change is identified, and the point cloud boundary is extracted at the edge of the hole area. The area missing information is identified based on the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

[0148] In this embodiment, the system identifies the density distance between points based on the density of neighboring points of the original point cloud data, and then the system determines whether these density distances exceed a preset distance to execute the corresponding steps; for example, when the system determines that the density distance between points does not exceed the preset distance, the system will consider that the point distribution of the original point cloud data is relatively uniform and meets the accuracy requirements of model reconstruction. The system will continue with subsequent calculation steps, such as error index calculation, area division or model reconstruction, because the density distribution of the point cloud meets expectations. At the same time, the results of the density distance are recorded in the database for subsequent analysis and data quality monitoring, which helps to establish a more complete data quality assessment system, and regularly monitors the density changes of point cloud data to ensure that it remains within the preset density range in subsequent operations, so as to promptly discover and solve potential problems; for example, when the system determines that the density distance between points exceeds the preset distance, the system will consider that the point distribution of the original point cloud data is scattered and does not meet the accuracy requirements of model reconstruction. The system will define two points that exceed the preset distance as noise points, apply the preset median filter to remove the outliers of the noise points, identify the hole area corresponding to the density change according to the real-time density change of the original point cloud data, extract the point cloud boundary at the edge of the hole area, and identify the regional missing information based on these point cloud boundaries. The regional missing information specifically includes discontinuous areas and vacant areas. The system can smooth the point cloud data and reduce the impact of random noise on the data by using the median filtering algorithm. This filtering method can retain edge features and avoid damage to the data structure, which helps to improve the accuracy of the model. At the same time, real-time monitoring of the density changes of the point cloud data can help the system dynamically identify the hole areas, which may cause vacancies or discontinuities in subsequent modeling, thereby affecting the overall integrity and reliability of the model. By extracting the point cloud boundary at the edge of the hole area, the system can clearly identify the discontinuous areas and vacant areas. This process can not only provide a basis for subsequent point cloud reconstruction, but also provide an important reference for model optimization.

[0149] It should be noted that two points beyond the preset distance are defined as noise points, and a preset median filter is applied to remove the outliers of the noise points. According to the density change of the original point cloud data, the hole area corresponding to the density change is identified, and the point cloud boundary is extracted at the edge of the hole area. The area missing information is marked according to the point cloud boundary. The specific examples are as follows:

[0150] Suppose there is a set of original point cloud data, containing the following points:

[0151] A(1,2,3);

[0152] B(10,10,10);

[0153] C(2,2,3);

[0154] D(1,2,4);

[0155] The system preset distance threshold is 5 units; calculate the distance between point A and point B:

[0156] Distance = √((10-1)²+(10-2)²+(10-3)²)≈12.73; Since 12.73 is greater than 5, these two points are defined as noise points;

[0157] Median filter application, near point A, there are the following neighboring points:

[0158] C(2,2,3);

[0159] D(1,2,4);

[0160] The median filter calculates the median of these two neighboring points, for each coordinate axis:

[0161] X-axis median = 1.5;

[0162] Y-axis median = 2;

[0163] Z-axis median = 3.5;

[0164] Therefore, the noise point A will be updated to (1.5, 2, 3.5) to reduce the impact of noise;

[0165] After density change identification, the processed point cloud data is now:

[0166] A'(1.5,2,3.5);

[0167] C(2,2,3);

[0168] D(1,2,4);

[0169] Suppose that in a certain area, the system calculates the density of each point and finds that there are only points A' and C in a certain area, which are far away, while other surrounding areas have dense point cloud data; the system will determine that the density of this area changes significantly and mark it as a void area;

[0170] Point cloud boundary extraction, after identifying the hole area, the system will extract the point cloud boundary at the edge of the hole area; for example, if the other points in the area are:

[0171] E(5,5,5);

[0172] F(6,6,6);

[0173] Then the boundary extraction may obtain points E and F as boundary points, forming the outline of the hole area;

[0174] Identification of missing information in the region. The system identifies missing information in the region through the extracted boundary points. Assuming that there is no other point cloud data around boundary points E and F, this means that there are discontinuities and gaps in the region. Discontinuous regions may be marked as “missing boundary” and gaps as “missing data”.

[0175] In summary, through the above steps, the system cleans up the noise data, identifies the void areas, and extracts the boundaries. This process ensures the accuracy and completeness of the point cloud data and provides a solid foundation for subsequent model reconstruction. For example, this point cloud data represents the building curtain wall. The processed data will make the three-dimensional model of the building curtain wall more accurate and avoid design errors or structural instability caused by noise and missing data.

[0176] In this embodiment, the step S5 of completing the smoothing and refinement of the mesh surface model further includes:

[0177] S51: Based on the pre-extracted point coordinates in the local area to be refined, a preliminary fitting model is generated after the point coordinates are connected and combined, an error distance value between the mesh surface model and the preliminary fitting model is calculated, and a preset least square method is used to iteratively update the error distance value;

[0178] S52: Determine whether the change in the error distance value after iterative update is less than a preset threshold;

[0179] S53: If so, extract the vertex coordinates and the patch information from the preliminary fitting model, apply the preset Laplace smoothing according to the vertex coordinates and the patch information, and perform smoothing on the specific local area corresponding to the vertex coordinates and the patch information until the vertex coordinates and the patch information meet the preset smoothing convergence conditions.

[0180] In this embodiment, the system generates a preliminary fitting model obtained by connecting and combining the point coordinates pre-extracted in the local area to be refined, calculates the error distance value between the mesh surface model and the preliminary fitting model, applies the preset least squares method to iteratively update the error distance value, and then the system determines whether the change in the error distance value after the iterative update is less than the preset threshold value to execute the corresponding steps; for example, when the system determines that the change in the error distance value between the mesh surface model and the preliminary fitting model after the iterative update is not less than the preset threshold value, the system will consider that the geometric shape of the local area to be refined still has a significant deviation and cannot meet the design requirements. The system will identify which specific areas cause large errors through comparative analysis, and achieve this through a visual error distribution map to help the system locate the problem. At the same time, in the area with large errors, more points are added to improve the accuracy of the fitting model. These points can be extracted based on the surrounding geometric features, and the updated preliminary fitting model is compared with the mesh surface model to recalculate the error distance value; for example, when the system determines that the error distance value between the mesh surface model and the preliminary fitting model is completely If the change after iterative update is less than the preset threshold, the system will consider that the geometry of the local area to be refined does not have significant deviation and can meet the design requirements. The system will extract vertex coordinates and patch information from the preliminary fitting model, and apply the preset Laplace smoothing to the specific local area corresponding to these vertex coordinates and patch information according to these vertex coordinates and patch information, until these vertex coordinates and patch information can meet the preset smooth convergence conditions; Laplace smoothing smoothes the surface by calculating the neighborhood average of each vertex to eliminate noise and irregularities. This treatment makes the surface of the model smoother, avoids sharp edges or abrupt shapes, and improves the overall aesthetics. At the same time, smoothing makes the transition between the local area and the surrounding area more natural. This effect can effectively reduce the visual incoordination of the model from different perspectives, provide a more unified visual experience, and enhance the beauty and professionalism of the overall model. In addition, despite the smoothing process, the Laplace algorithm can effectively retain important geometric features, such as edges and corners, which is particularly important for architectural models because these features usually represent key elements in the design, such as window frames, facade changes, etc.

[0181] It should be noted that the error distance value between the mesh surface model and the preliminary fitting model is calculated, and the error distance value is iteratively updated using a preset least squares method. The specific example is as follows:

[0182] Assuming there is a mesh surface model and a preliminary fitting model, each of which contains 3 points, the system calculates the error distance value and uses the least squares method to iteratively optimize to make the preliminary fitting model closer to the mesh surface model;

[0183] Point P of the mesh surface model = [(1,2,3), (2,3,4), (3,4,5)];

[0184] The points of the initial fitting model are F=[(1.1,2.1,3.1),(1.9,3.1,4.1),(3.2,4.1,5.2)];

[0185] First, the error distance is calculated to calculate the error between the mesh surface model and the preliminary fitting model;

[0186] Error calculation:

[0187] For points P1=(1,2,3) and F1=(1.1,2.1,3.1);

[0188] but

[0189] For points P2=(2,3,4) and F2=(1.9,3.1,4.1);

[0190] but

[0191] For points P3=(3,4,5) and F3=(3.2,4.1,5.2);

[0192] but

[0193] The overall error (RMSE) is calculated as:

[0194]

[0195] 2. Apply the least squares method to iteratively update the error distance value by setting the objective function. The objective function E is defined as the sum of squares of the errors between each pair of points:

[0196]

[0197] The system selects the initial vertex coordinates of the preliminary fitting model.

[0198] F=[(1.1,2.1,3.1),(1.9,3.1,4.1),(3.2,4.1,5.2)];

[0199] Since it is necessary to calculate the impact of each point on the objective function, the gradient of the objective function relative to each coordinate is obtained:

[0200]

[0201] The update formula using the least squares method is:

[0202]

[0203] Where a is the learning rate;

[0204] Assuming that the learning rate a=0.01 is set, in the first iteration, the current error is calculated and points F1, F2, and F3 are updated;

[0205]

[0206] Perform corresponding calculations to obtain new coordinates;

[0207] Finally, the convergence is judged, and the updated error value is calculated after each iteration. If the change is less than the preset threshold (for example, 0.001), the iteration is stopped;

[0208] In summary, through the above steps, the system can calculate the error between the mesh surface model and the preliminary fitting model, and iteratively update the model parameters using the least squares method until the error converges.

[0209] It should be added that the preset Laplace smoothing is applied to smooth the specific local area corresponding to the vertex coordinates and the patch information until the vertex coordinates and the patch information meet the preset smoothing convergence condition. The specific example is as follows:

[0210] Assume that the system is processing a mesh surface model of a building curtain wall, where the vertices in a certain local area have obvious jagged irregularities, so it is hoped that the Laplace smoothing technology can be used to improve the smoothness and visual effect of this area;

[0211] First, select the local area to be smoothed. Assume that the selected local area contains the following four vertices:

[0212] V=[(1,1,1),(1,1.5,1),(1,1.2,1.2),(1.1,1.3,1.3)];

[0213] The coordinates of these vertices show jagged features, especially uneven changes in height (z coordinate);

[0214] By extracting the coordinates of the above vertices,

[0215] Vertex coordinates: V = [(1,1,1), (1,1.5,1), (1,1.2,1.2), (1.1,1.3,1.3)];

[0216] Assume that these vertices form two triangular patches:

[0217] Patch 1: (1,1,1), (1,1.5,1), (1,1.2,1.2);

[0218] Patch 2: (1,1.5,1), (1.1,1.3,1.3), (1,1.2,1.2);

[0219] Then, the Laplace smoothing algorithm is applied to calculate the average position of the surrounding vertices for each vertex;

[0220] For the vertex (1,1,1)(1, 1, 1)(1,1,1), assume its neighborhood is:

[0221] N((1,1,1))=[(1,1.5,1),(1,1.2,1.2)];

[0222] Then calculate the average position:

[0223]

[0224] Then update the vertex position (assuming λ=0.5)

[0225]

[0226] Then check the convergence conditions and calculate the updated change:

[0227]

[0228] Then set the convergence threshold to 0.01. If the change is greater than this value, continue to iterate;

[0229] Finally, update the other vertices and repeat the same process for the other vertices. The final updated position of each vertex is as follows:

[0230] (1,1.175,1.05);

[0231] (1,1.5,1) (assuming no change);

[0232] (1,1.2,1.2);

[0233] (1.1,1.3,1.3);

[0234] Update the patch information and recalculate the patch according to the updated vertex coordinates:

[0235] Then the new coordinates of patch 1 are (1,1.175,1.05), (1,1.5,1), (1,1.2,1.2);

[0236] The new coordinates of patch 2 are (1,1.5,1), (1.1,1.3,1.3), (1,1.2,1.2);

[0237] In summary, after Laplace smoothing, the vertex positions in the local area are more uniform, the visual effect of the patch is improved, the jagged irregularities are significantly reduced, and the smoothness and realism of the building curtain wall mesh surface model can be effectively improved.

[0238] In this embodiment, the step S2 of determining whether the error value of the irregular area exceeds a preset threshold value further includes:

[0239] S21: based on the error points pre-detected in the irregular area, obtaining the quantity value of the error points;

[0240] S22: Determine whether the quantity value exceeds a preset value;

[0241] S23: If yes, then identify the distribution information of the error point on the irregular area, and generate detailed information of the irregular area on a pre-built visualization model according to the distribution information, wherein the detailed information specifically includes a normal vector, color, error type, timestamp and neighborhood point information.

[0242] In this embodiment, the system obtains the quantity values ​​of the error points based on the error points pre-detected in the irregular area, and then the system determines whether the quantity values ​​exceed the preset values ​​to execute the corresponding steps; for example, when the system determines that the quantity values ​​of the error points do not exceed the preset values, the system will consider that the number of error points is within an acceptable range, indicating that the geometry of the model is consistent with the expected design and the overall quality is good. The system can perform a more in-depth error analysis, such as checking the specific location and error value of the error points to ensure that their distribution does not affect the application of the overall model. At the same time, if the model can be further optimized in some features, local refinement or smoothing can be performed to enhance the overall effect of the model, and the error analysis results are presented in a visual manner to generate a report to facilitate team members to understand the current state of the model and ensure the transparency of design decisions; for example, when the system determines that the quantity values ​​of the error points exceed the preset values, the system will consider that the number of error points is within an acceptable range, indicating that the geometry of the model is consistent with the expected design and the overall quality is good. The system can perform a more in-depth error analysis, such as checking the specific location and error value of the error points to ensure that their distribution does not affect the application of the overall model. At the same time, if the model can be further optimized in some features, local refinement or smoothing can be performed to enhance the overall effect of the model, and the error analysis results are presented in a visual manner to generate a report to facilitate team members to understand the current state of the model and ensure the transparency of design decisions. value, the system will think that the geometry of the model is inconsistent with the expected design, and the system will identify the distribution information of these error points on the irregular area, and generate detailed information of the irregular area on the pre-built visualization model according to different distribution information. The detailed information includes normal vector, color, error type, timestamp and neighborhood point information; the system displays the normal vector, color, error type and other detailed information of the irregular area through the visualization model, and the user can intuitively understand which parts have problems and the nature of the problems. At the same time, different types of errors (such as geometric errors, measurement errors, etc.) provide an in-depth understanding of the root causes of the problems, and recording timestamps enables the team to track the occurrence of problems and analyze the trend of model quality changes. In addition, the neighborhood point information provides detailed data of local geometric features. By analyzing and adjusting irregular areas, engineers can ensure that the final model meets the preset design standards, which can effectively reduce problems that may arise in future use.

[0243] In this embodiment, the step S4 of judging whether the model smoothness of the mesh surface model is consistent with the preset realism further includes:

[0244] S41: based on a preset main viewing angle of the mesh surface model, switching the main viewing angle to another preset secondary viewing angle, performing a dynamic view test on the mesh surface model under the other secondary viewing angle, and collecting dynamic viewing angle information;

[0245] S42: Determine whether the dynamic viewing angle information satisfies a preset real condition;

[0246] S43: If not, record the viewing angle parameters of the other secondary viewing angles, and obtain the visual defect information of the mesh surface model according to the viewing angle parameters, wherein the viewing angle parameters specifically include angle, distance and lighting.

[0247] In this embodiment, the system switches the main observation perspective to other preset secondary observation perspectives based on the pre-set main observation perspective of the mesh surface model, performs dynamic view tests on the mesh surface model under other secondary observation perspectives, collects the dynamic perspective information, and then the system determines whether the dynamic perspective information can meet the pre-set real conditions to execute the corresponding steps; for example, when the system determines that the dynamic perspective information can meet the pre-set real conditions, the system will consider that the performance of the mesh surface model at different observation angles is consistent with expectations and can effectively show the realism and details of the design. The system will archive the dynamic perspective information that meets the real conditions for future reference or use in presentations, and conduct a comprehensive evaluation, including a detailed review of the overall quality, details and visual effects of the model to ensure that all aspects meet the expected standards, and regularly monitor the model performance to ensure its applicability under various conditions ... dynamic perspective information can meet the pre-set real conditions, and the system will regularly monitor the model performance to ensure its applicability under various conditions. If the dynamic viewing angle information cannot meet the pre-set real conditions, the system will consider that the performance of the mesh surface model at different viewing angles deviates from expectations, and the system will record the viewing angle parameters of other secondary viewing angles, which specifically include angle, distance and lighting, and obtain the visual defect information of the mesh surface model according to different viewing angle parameters; by recording the viewing angle parameters (such as angle, distance and lighting) of other secondary viewing angles, the system can timely locate the visual defects of the model under different viewing conditions, and help engineers quickly identify areas that need improvement. At the same time, by analyzing the visual defects under different viewing angles, the system can provide a more comprehensive model evaluation to ensure that the model maintains quality under various circumstances, and can formulate targeted repair strategies for different viewing angle parameters (such as defects that occur under specific lighting conditions) to improve the effectiveness of the repair. By recording the viewing angle parameters and visual defect information, a feedback loop can be formed to help the team continuously optimize the model in subsequent iterations.

[0248] In this embodiment, based on the pre-converted mesh surface model of the building curtain wall, the model surface of the mesh surface model is scanned, and the step S1 of identifying the corresponding irregular area through the model surface further includes:

[0249] S11: Based on the point cloud features preset in the original point cloud data, classify the point cloud features to identify geometric features, and generate a corresponding preliminary mesh model according to the point cloud features and the geometric features, wherein the point cloud features specifically include color, normal vector and density, and the geometric features specifically include edge, corner point and surface features;

[0250] S12: Determine whether the preliminary grid model can cover the preset feature points;

[0251] S13: If not, then obtain the uncovered feature surfaces of the preliminary mesh model, collect the hierarchical structure of the preliminary mesh model according to the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model according to the hierarchical structure, wherein the hierarchical levels specifically include a basic layer, an intermediate layer and an advanced layer.

[0252] In this embodiment, the system classifies different point cloud features to identify geometric features based on the point cloud features pre-set in the original point cloud data, which specifically include color, normal vector and density. The geometric features specifically include edge, corner point and surface features. According to these point cloud features and geometric features, a corresponding preliminary mesh model is generated by combination, and then the system determines whether the preliminary mesh model can cover the pre-set feature points to execute the corresponding steps; for example, when the system determines that the preliminary mesh model can cover the pre-set feature points, the system will consider that the mesh model is consistent with the geometric features of the original point cloud data and can better reflect the shape and details of the point cloud. The system will perform error analysis and calculate the error between the model and the original point cloud data to ensure that there is no obvious geometric deviation. At the same time, the classification method of the point cloud features and geometric features used, the model generation process and any optimization steps are recorded to provide a reference for future projects, and the visual effect of the model is verified through different observation angles and conditions to ensure that it still maintains good performance under different circumstances; for example, when the system determines that the preliminary mesh model cannot cover the pre-set feature points, the system will consider that the mesh model is consistent with the original point cloud data. If the geometric features of the data do not match, the system will obtain the uncovered feature surfaces of the preliminary mesh model, collect the hierarchical structure of the preliminary mesh model based on the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model through the hierarchical structure, which specifically include basic layer, intermediate layer and advanced layer; by obtaining the uncovered feature surfaces, the system can accurately identify the deficiencies of the model in terms of geometric features, which is helpful for targeted optimization and correction, and visualize the uncovered feature surfaces, so that engineers and designers can more intuitively understand the defects of the model. At the same time, dividing the hierarchical structure (basic layer, intermediate layer, advanced layer) makes the problem easier to manage and analyze, that is, according to the hierarchical level, the basic layer problems with greater impact are dealt with first, and then the intermediate layer and advanced layer are gradually optimized, so as to improve the overall model quality, and according to the information and hierarchical structure of the uncovered feature surfaces, the system can formulate more targeted refinement or reconstruction strategies to ensure that the model can better conform to the original point cloud data, and through the division of the hierarchical structure, the system can more effectively identify and deal with problem areas, reduce redundant work, and improve reconstruction efficiency, because a clear hierarchy can reduce the risk of errors in the model reconstruction process and ensure that problems at each level are properly handled.

[0253] Reference Figure 2 , is a curtain wall reverse modeling system based on point cloud data in one embodiment of the present invention, comprising:

[0254] The scanning module 10 is used to scan the model surface of the mesh surface model based on the pre-converted mesh surface model of the building curtain wall, and identify the corresponding irregular area through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area and an uneven area;

[0255] A judging module 20, used to judge whether the error value of the irregular area exceeds a preset threshold;

[0256] An execution module 30 is used for, if yes, marking the irregular area on the mesh surface model, applying a preset surface fitting algorithm to identify a model curvature change of the mesh surface model, adjusting the interpolation points of the irregular area according to the model curvature change, performing error analysis between the mesh surface model and the original point cloud data, analyzing to obtain model fitting parameters, and reconstructing the hierarchical structure of the mesh surface model according to the interpolation points and the model fitting parameters;

[0257] A second judgment module 40 is used to judge whether the model smoothness of the mesh surface model is consistent with the preset realism;

[0258] The second execution module 50 is used to collect preset structural features of the building curtain wall based on the original point cloud data if there is any inconsistency, obtain the local area to be refined of the mesh surface model according to the structural features, apply a preset refinement algorithm to insert a number of new vertices in the local area to be refined, evenly distribute the new vertices to the local area to be refined, connect each adjacent new vertices, construct a triangular mesh on the local area to be refined, and complete the smooth refinement of the mesh surface model, wherein the structural features specifically include straight lines, planes, surfaces, edges and angles.

[0259] In this embodiment, the scanning module 10 scans the model surface of the mesh surface model based on the pre-converted mesh surface model of the building curtain wall, and identifies the corresponding irregular areas through the model surface. The irregular areas specifically include jagged areas, distorted areas and non-smooth areas. Then the judgment module 20 judges whether the error values ​​of these irregular areas exceed the preset error threshold value to execute the corresponding steps; for example, when the system determines that the error value of a certain irregular area does not exceed the preset error threshold value, the system will consider that the geometric features of the area are consistent with the expected model and basically meet the design requirements. The system will keep monitoring the area and regularly re-evaluate its status to ensure that the model is updated in the subsequent model. or changes, the stability of its geometric features is still maintained, while other irregular areas are continuously evaluated to ensure the overall quality of the model and avoid ignoring other possible problem areas. After confirming that the error value of the area meets the requirements, it can be decided not to modify it and maintain the existing model state, thereby saving time and resources. For example, when the system determines that the error value of a certain irregular area exceeds the preset error threshold, the execution module 30 will consider that the geometric features of the area are inconsistent with the expected model and cannot meet the design requirements. The system will mark these irregular areas on the mesh surface model and apply a pre-set surface fitting algorithm to identify the model curvature of the mesh surface model. Changes, adjust the interpolation points of irregular areas according to different model curvature changes, perform error analysis on the mesh surface model and the original point cloud data, analyze the model fitting parameters, and finally reconstruct the hierarchical structure of the mesh surface model by combining the interpolation points and model fitting parameters; the system can deeply analyze the geometric features of the model by applying the preset surface fitting algorithm and identify the deviation from the expected surface. This refined processing helps to optimize the geometric shape of the model to make it more in line with the design requirements, thereby improving the overall quality of the model. At the same time, the interpolation points of irregular areas are adjusted according to different curvature changes, making the geometric features of the area more delicate and precise. This adjustment helps to achieve smoother surfaces in key areas. Smooth transition can be achieved to avoid jagged or uneven phenomena caused by improper interpolation, improve visual effects, and by performing error analysis on the mesh surface model and the original point cloud data, more detailed model information can be obtained to identify which areas have obvious deviations. This process enables the model fitting parameters to be optimized based on actual data, enhances the accuracy of the model, and ensures that it can better reflect the characteristics of real objects in practical applications. The hierarchical structure of the reconstructed mesh surface model enables the model to visually show richer details; then the second judgment module 40 determines whether the model smoothness of the mesh surface model is consistent with the preset sense of reality through detection, and then executes the corresponding steps;For example, when the system determines that the model smoothness of the mesh surface model is consistent with the pre-set sense of reality, the system will consider that the model has met the design requirements in terms of visual effects and geometric features. The system will record the current state of the model to confirm that it meets the design standards and quality requirements, laying the foundation for subsequent operations. At the same time, after the model is confirmed to be qualified, it can enter the next project stage, such as engineering drawing generation, rendering, animation production or other subsequent design work. Even if the current state of the model is good, it should be continuously monitored and the model should be re-evaluated regularly to ensure that it maintains its quality in subsequent modifications or environmental changes. For example, when the system determines that the model smoothness of the mesh surface model is inconsistent with the pre-set sense of reality, the second execution module 50 will consider that the model still cannot meet the design requirements in terms of visual effects. The system will collect the pre-set structural features of the building curtain wall based on the original point cloud data. The structural features specifically include straight lines, planes, surfaces, edges and angles. According to these structural features, the local area to be refined of the mesh surface model is obtained, and the pre-set The refinement algorithm is designed to insert several new vertices in the local area to be refined, evenly distribute these new vertices to the local area to be refined, connect each adjacent new vertex, construct a triangular mesh on the local area to be refined, and complete the smooth refinement of the mesh surface model; the system significantly improves the surface smoothness of the model by inserting new vertices in the local area to be refined and rebuilding the corresponding triangular mesh. This smoothing process reduces visible geometric irregularities and makes the model visually softer and more natural. At the same time, new vertices are added according to the structural features of the building curtain wall (such as straight lines, planes, curved surfaces, etc.) during the refinement process, which helps to more accurately express the complex shape of the building. This detail enhancement enables the model to more realistically reflect the actual appearance of the building, improve the accuracy and visual appeal of the design, and refinement based on the preset structural features of the building curtain wall can ensure that the model remains geometrically consistent and coherent after refinement, which means that while adding details, the shape and function of the model still conform to the original design intent, avoiding new geometric errors caused by refinement. ;

[0260] In this embodiment, the execution module further includes:

[0261] A division unit, used to divide the mesh surface model into different regions, assign corresponding region identifiers based on the different regions, extract points corresponding to the region identifiers from the original point cloud data, and calculate error indicators of the points respectively, wherein the different regions specifically include important feature regions, boundary regions and complex geometric regions, and the error indicators specifically include feature edge errors, matching errors of adjacent structures and curvature change errors;

[0262] A judging unit, used to judge whether the error index exceeds a preset index;

[0263] An execution unit is used to generate error heat maps corresponding to different areas according to the error indicators, collect error trends according to the error heat maps, and display the error differences between the different areas through the error trends.

[0264] In this embodiment, the system divides the mesh surface model into different areas, which specifically include important feature areas, boundary areas, and complex geometric areas. Based on these different areas, corresponding area identifiers are assigned, and points corresponding to the area identifiers are extracted from the original point cloud data. The error indicators of these points are calculated respectively. The error indicators specifically include the error of the feature edge, the matching error of the adjacent structure, and the curvature change error. Then the system determines whether these error indicators exceed the preset indicator upper limit to execute the corresponding steps; for example, when the system determines that the error indicator of a certain point does not exceed the preset indicator upper limit, the system will consider that the points in the area meet the expected accuracy requirements, indicating that the reconstruction quality of the mesh surface model is good. The system will mark the area as a valid area, keep it unchanged in subsequent processing, continue to optimize the details of the area or merge it into a larger model to ensure the accuracy and integrity of the entire model, and store the error indicator and other related data of the point in the database for subsequent performance analysis and model evaluation, and continue to evaluate the error indicators of other areas to ensure the quality of the entire mesh model and promptly discover and deal with possible problems; for example, when the system determines that When the error index of a certain point exceeds the preset upper limit of the index, the system will consider that the points in the area do not meet the expected accuracy requirements, resulting in poor reconstruction quality of the mesh surface model. The system will generate error heat maps corresponding to different areas based on these error indicators, collect the corresponding error trends based on the error heat maps, and display the error differences in different areas through the error trends; by monitoring the error indicators, the system can identify potential accuracy problems in real time to avoid greater errors in subsequent processing, which helps to improve the reliability of the overall model. At the same time, based on the error indicators, the system can generate error heat maps for different areas, which intuitively show the error distribution of each area. This visualization effect helps engineers quickly identify problem areas and make targeted adjustments. By collecting and analyzing the trend data of the error heat map, the system can identify the error changes in different areas. This trend analysis can help developers understand the causes of errors under specific conditions, and then optimize the processing flow and reconstruction algorithm of point cloud data. By analyzing the error differences, engineers can make targeted adjustments to the data collection and processing flow, optimize the design of model reconstruction, and improve the efficiency and quality of subsequent processing.

[0265] In this embodiment, it also includes:

[0266] An identification module, used for identifying the density distance between points based on the density of neighboring points of the original point cloud data;

[0267] A third judging module is used to judge whether the density distance exceeds a preset distance;

[0268] The third execution module is used to define the two points beyond the preset distance as noise points, apply the preset median filter to remove the outliers of the noise points, identify the hole area corresponding to the density change according to the density change of the original point cloud data, extract the point cloud boundary at the edge of the hole area, and identify the area missing information according to the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

[0269] In this embodiment, the system identifies the density distance between points based on the density of neighboring points of the original point cloud data, and then the system determines whether these density distances exceed a preset distance to execute the corresponding steps; for example, when the system determines that the density distance between points does not exceed the preset distance, the system will consider that the point distribution of the original point cloud data is relatively uniform and meets the accuracy requirements of model reconstruction. The system will continue with subsequent calculation steps, such as error index calculation, area division or model reconstruction, because the density distribution of the point cloud meets expectations. At the same time, the results of the density distance are recorded in the database for subsequent analysis and data quality monitoring, which helps to establish a more complete data quality assessment system, and regularly monitors the density changes of point cloud data to ensure that it remains within the preset density range in subsequent operations, so as to promptly discover and solve potential problems; for example, when the system determines that the density distance between points exceeds the preset distance, the system will consider that the point distribution of the original point cloud data is scattered and does not meet the accuracy requirements of model reconstruction. The system will define two points that exceed the preset distance as noise points, apply the preset median filter to remove the outliers of the noise points, identify the hole area corresponding to the density change according to the real-time density change of the original point cloud data, extract the point cloud boundary at the edge of the hole area, and identify the regional missing information based on these point cloud boundaries. The regional missing information specifically includes discontinuous areas and vacant areas. The system can smooth the point cloud data and reduce the impact of random noise on the data by using the median filtering algorithm. This filtering method can retain edge features and avoid damage to the data structure, which helps to improve the accuracy of the model. At the same time, real-time monitoring of the density changes of the point cloud data can help the system dynamically identify the hole areas, which may cause vacancies or discontinuities in subsequent modeling, thereby affecting the overall integrity and reliability of the model. By extracting the point cloud boundary at the edge of the hole area, the system can clearly identify the discontinuous areas and vacant areas. This process can not only provide a basis for subsequent point cloud reconstruction, but also provide an important reference for model optimization.

[0270] In this embodiment, the second execution module further includes:

[0271] A generating unit, configured to generate, based on the pre-extracted point coordinates in the local area to be refined, a preliminary fitting model obtained by connecting and combining the point coordinates, calculate an error distance value between the mesh surface model and the preliminary fitting model, and iteratively update the error distance value using a preset least squares method;

[0272] A second judgment unit is used to judge whether the change amount of the error distance value after iterative update is less than a preset threshold;

[0273] The second execution unit is used to extract vertex coordinates and patch information from the preliminary fitting model, and apply preset Laplace smoothing according to the vertex coordinates and the patch information to smooth the specific local area corresponding to the vertex coordinates and the patch information until the vertex coordinates and the patch information meet preset smoothing convergence conditions.

[0274] In this embodiment, the system generates a preliminary fitting model obtained by connecting and combining the point coordinates pre-extracted in the local area to be refined, calculates the error distance value between the mesh surface model and the preliminary fitting model, applies the preset least squares method to iteratively update the error distance value, and then the system determines whether the change in the error distance value after the iterative update is less than the preset threshold value to execute the corresponding steps; for example, when the system determines that the change in the error distance value between the mesh surface model and the preliminary fitting model after the iterative update is not less than the preset threshold value, the system will consider that the geometric shape of the local area to be refined still has a significant deviation and cannot meet the design requirements. The system will identify which specific areas cause large errors through comparative analysis, and achieve this through a visual error distribution map to help the system locate the problem. At the same time, in the area with large errors, more points are added to improve the accuracy of the fitting model. These points can be extracted based on the surrounding geometric features, and the updated preliminary fitting model is compared with the mesh surface model to recalculate the error distance value; for example, when the system determines that the error distance value between the mesh surface model and the preliminary fitting model is completely If the change after iterative update is less than the preset threshold, the system will consider that the geometry of the local area to be refined does not have significant deviation and can meet the design requirements. The system will extract vertex coordinates and patch information from the preliminary fitting model, and apply the preset Laplace smoothing to the specific local area corresponding to these vertex coordinates and patch information according to these vertex coordinates and patch information, until these vertex coordinates and patch information can meet the preset smooth convergence conditions; Laplace smoothing smoothes the surface by calculating the neighborhood average of each vertex to eliminate noise and irregularities. This treatment makes the surface of the model smoother, avoids sharp edges or abrupt shapes, and improves the overall aesthetics. At the same time, smoothing makes the transition between the local area and the surrounding area more natural. This effect can effectively reduce the visual incoordination of the model from different perspectives, provide a more unified visual experience, and enhance the beauty and professionalism of the overall model. In addition, despite the smoothing process, the Laplace algorithm can effectively retain important geometric features, such as edges and corners, which is particularly important for architectural models because these features usually represent key elements in the design, such as window frames, facade changes, etc.

[0275] In this embodiment, the judgment module further includes:

[0276] An acquisition unit, configured to acquire a quantity value of the error points based on the error points pre-detected in the irregular area;

[0277] A third judging unit, used to judge whether the quantity value exceeds a preset value;

[0278] The third execution unit is used to identify the distribution information of the error point on the irregular area, and generate detailed information of the irregular area on a pre-built visualization model according to the distribution information, wherein the detailed information specifically includes a normal vector, color, error type, timestamp and neighborhood point information.

[0279] In this embodiment, the system obtains the quantity values ​​of the error points based on the error points pre-detected in the irregular area, and then the system determines whether the quantity values ​​exceed the preset values ​​to execute the corresponding steps; for example, when the system determines that the quantity values ​​of the error points do not exceed the preset values, the system will consider that the number of error points is within an acceptable range, indicating that the geometry of the model is consistent with the expected design and the overall quality is good. The system can perform a more in-depth error analysis, such as checking the specific location and error value of the error points to ensure that their distribution does not affect the application of the overall model. At the same time, if the model can be further optimized in some features, local refinement or smoothing can be performed to enhance the overall effect of the model, and the error analysis results are presented in a visual manner to generate a report to facilitate team members to understand the current state of the model and ensure the transparency of design decisions; for example, when the system determines that the quantity values ​​of the error points exceed the preset values, the system will consider that the number of error points is within an acceptable range, indicating that the geometry of the model is consistent with the expected design and the overall quality is good. The system can perform a more in-depth error analysis, such as checking the specific location and error value of the error points to ensure that their distribution does not affect the application of the overall model. At the same time, if the model can be further optimized in some features, local refinement or smoothing can be performed to enhance the overall effect of the model, and the error analysis results are presented in a visual manner to generate a report to facilitate team members to understand the current state of the model and ensure the transparency of design decisions. value, the system will think that the geometry of the model is inconsistent with the expected design, and the system will identify the distribution information of these error points on the irregular area, and generate detailed information of the irregular area on the pre-built visualization model according to different distribution information. The detailed information includes normal vector, color, error type, timestamp and neighborhood point information; the system displays the normal vector, color, error type and other detailed information of the irregular area through the visualization model, and the user can intuitively understand which parts have problems and the nature of the problems. At the same time, different types of errors (such as geometric errors, measurement errors, etc.) provide an in-depth understanding of the root causes of the problems, and recording timestamps enables the team to track the occurrence of problems and analyze the trend of model quality changes. In addition, the neighborhood point information provides detailed data of local geometric features. By analyzing and adjusting irregular areas, engineers can ensure that the final model meets the preset design standards, which can effectively reduce problems that may arise in future use.

[0280] In this embodiment, the second determination module further includes:

[0281] A collection unit, configured to switch the main viewing angle to another preset secondary viewing angle based on a main viewing angle preset by the mesh surface model, perform a dynamic view test on the mesh surface model under the other secondary viewing angle, and collect the dynamic viewing angle information;

[0282] A fourth judgment unit, used to judge whether the dynamic viewing angle information meets a preset real condition;

[0283] The fourth execution unit is used to record the viewing angle parameters of the other secondary viewing angles, and obtain the visual defect information of the mesh surface model according to the viewing angle parameters, wherein the viewing angle parameters specifically include angle, distance and lighting.

[0284] In this embodiment, the system switches the main observation perspective to other preset secondary observation perspectives based on the pre-set main observation perspective of the mesh surface model, performs dynamic view tests on the mesh surface model under other secondary observation perspectives, collects the dynamic perspective information, and then the system determines whether the dynamic perspective information can meet the pre-set real conditions to execute the corresponding steps; for example, when the system determines that the dynamic perspective information can meet the pre-set real conditions, the system will consider that the performance of the mesh surface model at different observation angles is consistent with expectations and can effectively show the realism and details of the design. The system will archive the dynamic perspective information that meets the real conditions for future reference or use in presentations, and conduct a comprehensive evaluation, including a detailed review of the overall quality, details and visual effects of the model to ensure that all aspects meet the expected standards, and regularly monitor the model performance to ensure its applicability under various conditions ... dynamic perspective information can meet the pre-set real conditions, and the system will regularly monitor the model performance to ensure its applicability under various conditions. If the dynamic viewing angle information cannot meet the pre-set real conditions, the system will consider that the performance of the mesh surface model at different viewing angles deviates from expectations, and the system will record the viewing angle parameters of other secondary viewing angles, which specifically include angle, distance and lighting, and obtain the visual defect information of the mesh surface model according to different viewing angle parameters; by recording the viewing angle parameters (such as angle, distance and lighting) of other secondary viewing angles, the system can timely locate the visual defects of the model under different viewing conditions, and help engineers quickly identify areas that need improvement. At the same time, by analyzing the visual defects under different viewing angles, the system can provide a more comprehensive model evaluation to ensure that the model maintains quality under various circumstances, and can formulate targeted repair strategies for different viewing angle parameters (such as defects that occur under specific lighting conditions) to improve the effectiveness of the repair. By recording the viewing angle parameters and visual defect information, a feedback loop can be formed to help the team continuously optimize the model in subsequent iterations.

[0285] In this embodiment, the scanning module further includes:

[0286] A classification unit, configured to classify the point cloud features based on the preset point cloud features of the original point cloud data to identify geometric features, and generate a corresponding preliminary mesh model according to the point cloud features and the geometric features, wherein the point cloud features specifically include color, normal vector and density, and the geometric features specifically include edge, corner point and surface features;

[0287] A fifth judging unit, used to judge whether the preliminary grid model can cover the preset feature points;

[0288] The fifth execution unit is used to obtain the uncovered feature surfaces of the preliminary mesh model if not, collect the hierarchical structure of the preliminary mesh model according to the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model according to the hierarchical structure, wherein the hierarchical levels specifically include a basic layer, an intermediate layer and an advanced layer.

[0289] In this embodiment, the system classifies different point cloud features to identify geometric features based on the point cloud features pre-set in the original point cloud data, which specifically include color, normal vector and density. The geometric features specifically include edge, corner point and surface features. According to these point cloud features and geometric features, a corresponding preliminary mesh model is generated by combination, and then the system determines whether the preliminary mesh model can cover the pre-set feature points to execute the corresponding steps; for example, when the system determines that the preliminary mesh model can cover the pre-set feature points, the system will consider that the mesh model is consistent with the geometric features of the original point cloud data and can better reflect the shape and details of the point cloud. The system will perform error analysis and calculate the error between the model and the original point cloud data to ensure that there is no obvious geometric deviation. At the same time, the classification method of the point cloud features and geometric features used, the model generation process and any optimization steps are recorded to provide a reference for future projects, and the visual effect of the model is verified through different observation angles and conditions to ensure that it still maintains good performance under different circumstances; for example, when the system determines that the preliminary mesh model cannot cover the pre-set feature points, the system will consider that the mesh model is consistent with the original point cloud data. If the geometric features of the data do not match, the system will obtain the uncovered feature surfaces of the preliminary mesh model, collect the hierarchical structure of the preliminary mesh model based on the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model through the hierarchical structure, which specifically include basic layer, intermediate layer and advanced layer; by obtaining the uncovered feature surfaces, the system can accurately identify the deficiencies of the model in terms of geometric features, which is helpful for targeted optimization and correction, and visualize the uncovered feature surfaces, so that engineers and designers can more intuitively understand the defects of the model. At the same time, dividing the hierarchical structure (basic layer, intermediate layer, advanced layer) makes the problem easier to manage and analyze, that is, according to the hierarchical level, the basic layer problems with greater impact are dealt with first, and then the intermediate layer and advanced layer are gradually optimized, so as to improve the overall model quality, and according to the information and hierarchical structure of the uncovered feature surfaces, the system can formulate more targeted refinement or reconstruction strategies to ensure that the model can better conform to the original point cloud data, and through the division of the hierarchical structure, the system can more effectively identify and deal with problem areas, reduce redundant work, and improve reconstruction efficiency, because a clear hierarchy can reduce the risk of errors in the model reconstruction process and ensure that problems at each level are properly handled.

[0290] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A curtain wall reverse modeling method based on point cloud data, characterized in that: The following steps are involved: Based on the pre-converted mesh surface model of the building curtain wall, the model surface of the mesh surface model is scanned, and the corresponding irregular area is identified through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area, and an uneven area; Determining whether the error value of the irregular area exceeds a preset threshold; If so, the irregular area is marked on the mesh surface model, a preset surface fitting algorithm is applied to identify the model curvature change of the mesh surface model, the interpolation points of the irregular area are adjusted according to the model curvature change, the mesh surface model and the original point cloud data are subjected to error analysis to obtain model fitting parameters, and the hierarchical structure of the mesh surface model is reconstructed according to the interpolation points and the model fitting parameters; Determining whether the model smoothness of the mesh surface model is consistent with a preset sense of reality; If they are inconsistent, based on the original point cloud data, the preset structural features of the building curtain wall are collected, and the local area to be refined of the mesh surface model is obtained according to the structural features. A preset refinement algorithm is applied to insert a number of new vertices in the local area to be refined, and the new vertices are evenly distributed to the local area to be refined, and each adjacent new vertex is connected. A triangular mesh is constructed on the local area to be refined, and the smooth refinement of the mesh surface model is completed, wherein the structural features specifically include straight lines, planes, surfaces, edges and angles; Wherein, the step of performing error analysis on the mesh surface model and the original point cloud data further includes: The mesh surface model is divided into different regions, corresponding region identifiers are assigned based on the different regions, points corresponding to the region identifiers are extracted from the original point cloud data, and error indicators of the points are calculated respectively, wherein the different regions specifically include important feature regions, boundary regions, and complex geometric regions, and the error indicators specifically include errors of feature edges, matching errors of adjacent structures, and curvature change errors; Determining whether the error index exceeds a preset index; If so, an error heat map corresponding to different regions is generated according to the error index, an error trend is collected according to the error heat map, and the error difference of the different regions is displayed through the error trend.

2. The curtain wall reverse modeling method based on point cloud data according to claim 1, characterized in that: Before the step of marking the irregular area on the mesh surface model, the method further includes: Based on the density of neighboring points of the original point cloud data, identifying the density distance between points; Determining whether the density distance exceeds a preset distance; If so, the two points beyond the preset distance are defined as noise points, and the preset median filter is applied to remove the outliers of the noise points. According to the density change of the original point cloud data, the hole area corresponding to the density change is identified, and the point cloud boundary is extracted at the edge of the hole area. The area missing information is identified based on the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

3. The curtain wall reverse modeling method based on point cloud data according to claim 1, characterized in that: The step of completing the smoothing and refinement of the mesh surface model further includes: Based on the pre-extracted point coordinates in the local area to be refined, a preliminary fitting model is generated after the point coordinates are connected and combined, an error distance value between the mesh surface model and the preliminary fitting model is calculated, and a preset least square method is used to iteratively update the error distance value; Determine whether the change in the error distance value after iterative update is less than a preset threshold; If so, extract the vertex coordinates and the patch information from the preliminary fitting model, apply the preset Laplace smoothing according to the vertex coordinates and the patch information, and perform smoothing on the specific local area corresponding to the vertex coordinates and the patch information until the vertex coordinates and the patch information meet the preset smoothing convergence conditions.

4. The curtain wall reverse modeling method based on point cloud data according to claim 1, characterized in that: The step of determining whether the error value of the irregular area exceeds a preset threshold value further includes: Based on the error points pre-detected in the irregular area, obtaining the quantity value of the error points; Determining whether the quantity value exceeds a preset value; If so, the distribution information of the error point on the irregular area is identified, and the detailed information of the irregular area is generated on the pre-built visualization model according to the distribution information, wherein the detailed information specifically includes the normal vector, color, error type, timestamp and neighborhood point information.

5. The curtain wall reverse modeling method based on point cloud data according to claim 1, characterized in that: The step of judging whether the model smoothness of the mesh surface model is consistent with the preset sense of reality also includes: Based on a preset main viewing angle of the mesh surface model, the main viewing angle is switched to another preset secondary viewing angle, and a dynamic view test is performed on the mesh surface model under the other secondary viewing angle to collect dynamic viewing angle information; Determining whether the dynamic viewing angle information meets a preset real condition; If not, the viewing angle parameters of the other secondary viewing angles are recorded, and the visual defect information of the mesh surface model is obtained according to the viewing angle parameters, wherein the viewing angle parameters specifically include angle, distance and lighting.

6. The curtain wall reverse modeling method based on point cloud data according to claim 1, characterized in that: The step of scanning the model surface of the mesh surface model based on the pre-converted building curtain wall, and identifying the corresponding irregular area through the model surface also includes: Based on the point cloud features preset in the original point cloud data, the point cloud features are classified to identify geometric features, and a corresponding preliminary mesh model is generated by combining the point cloud features and the geometric features, wherein the point cloud features specifically include color, normal vector and density, and the geometric features specifically include edge, corner point and surface features; Determining whether the preliminary grid model can cover the preset feature points; If not, then obtain the uncovered feature surfaces of the preliminary mesh model, collect the hierarchical structure of the preliminary mesh model based on the uncovered feature surfaces, and divide the hierarchical levels of the preliminary mesh model according to the hierarchical structure, wherein the hierarchical levels specifically include a basic layer, an intermediate layer and an advanced layer.

7. A curtain wall reverse modeling system based on point cloud data, characterized in that: include: A scanning module is used to scan the model surface of the mesh surface model based on the pre-converted mesh surface model of the building curtain wall, and identify the corresponding irregular area through the model surface, wherein the irregular area specifically includes a jagged area, a distorted area and an uneven area; A judging module, used to judge whether the error value of the irregular area exceeds a preset threshold; an execution module, for, if yes, marking the irregular area on the mesh surface model, applying a preset surface fitting algorithm to identify a model curvature change of the mesh surface model, adjusting the interpolation points of the irregular area according to the model curvature change, performing error analysis between the mesh surface model and the original point cloud data, analyzing to obtain model fitting parameters, and reconstructing the hierarchical structure of the mesh surface model according to the interpolation points and the model fitting parameters; A second judgment module is used to judge whether the model smoothness of the mesh surface model is consistent with the preset realism; The second execution module is used for collecting the preset structural features of the building curtain wall based on the original point cloud data if there is any inconsistency, obtaining the local area to be refined of the mesh surface model according to the structural features, inserting a number of new vertices in the local area to be refined by applying a preset refinement algorithm, evenly distributing the new vertices to the local area to be refined, connecting each adjacent new vertices, constructing a triangular mesh on the local area to be refined, and completing the smooth refinement of the mesh surface model, wherein the structural features specifically include straight lines, planes, curved surfaces, edges and angles; Wherein, the execution module also includes: A division unit, used to divide the mesh surface model into different regions, assign corresponding region identifiers based on the different regions, extract points corresponding to the region identifiers from the original point cloud data, and calculate error indicators of the points respectively, wherein the different regions specifically include important feature regions, boundary regions, and complex geometric regions, and the error indicators specifically include errors of feature edges, matching errors of adjacent structures, and changes in curvature; A judging unit, used to judge whether the error index exceeds a preset index; An execution unit is used to generate error heat maps corresponding to different areas according to the error indicators, collect error trends according to the error heat maps, and display the error differences between the different areas through the error trends.

8. The curtain wall reverse modeling system based on point cloud data according to claim 7, characterized in that: Also includes: An identification module, used for identifying the density distance between points based on the density of neighboring points of the original point cloud data; A third judging module is used to judge whether the density distance exceeds a preset distance; The third execution module is used to define the two points beyond the preset distance as noise points, apply the preset median filter to remove the outliers of the noise points, identify the hole area corresponding to the density change according to the density change of the original point cloud data, extract the point cloud boundary at the edge of the hole area, and identify the area missing information according to the point cloud boundary, wherein the area missing information specifically includes discontinuous areas and vacant areas.

Images