An intelligent drawing output method and system based on three-dimensional model integration

Abstract

The application discloses an intelligent drawing method and system based on three-dimensional model integration. The method comprises the following steps: constructing a cross-platform three-dimensional model parser, analyzing three-dimensional model files of different sources, and generating a three-dimensional model assembly scene graph; performing bounding box analysis on the assembly scene graph to determine drawing parameters; generating an initial project plan according to the parameters, and performing contour integrity detection to obtain a detection result; completing defects according to the detection result to generate a complete project plan; and finally, performing automatic graphic annotation on the plan to output a two-dimensional drawing of the target project. The application can efficiently integrate multi-source three-dimensional model data, realize rapid and accurate output of two-dimensional drawings, improve design efficiency and intelligent level, and has high practical value and application prospect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering drawing, and in particular relates to an intelligent drawing method and system based on three-dimensional model integration. BACKGROUND

[0002] With the continuous improvement of informatization level in the fields of construction, machinery, electric power, municipal administration and the like, three-dimensional models are increasingly widely applied in engineering design, construction simulation, operation and maintenance management and the like. The three-dimensional model data formats, semantic structures and modeling methods generated by different platforms are different, which causes great difficulty in cross-platform data sharing and integration. In actual engineering projects, three-dimensional models from multiple sources often need to be integrated for unified analysis, management and drawing output to support engineering design deepening, construction briefing and review filing and the like application requirements.

[0003] Traditional two-dimensional drawing methods usually rely on manual selection of viewing angles, cross sections and details from three-dimensional models, and adopt semi-automatic or manual methods for graphic projection, line arrangement and label addition. This method is not only inefficient, but also depends on the experience level of the operator for drawing quality and consistency, which is difficult to meet the rapid drawing and batch drawing management requirements in large-scale projects.

[0004] Although there are some methods for converting three-dimensional models to two-dimensional models in the prior art, most of them are based on a single platform model and lack the ability to analyze and integrate cross-platform models. At the same time, there are problems such as low precision and insufficient automation in key links such as projection angle selection, contour recognition, defect completion and graphic labeling, which makes it difficult to guarantee the quality of the drawing results and limits its deep application in the whole process of engineering.

[0005] Therefore, there is an urgent need for an intelligent drawing method and system with the ability of cross-platform three-dimensional model analysis, spatial registration, drawing parameter optimization, contour detection and repair, and automatic graphic labeling, to realize efficient integration of multi-source three-dimensional models and automatic output of two-dimensional drawings in engineering projects, improve drawing efficiency and quality, and meet the actual needs of modern engineering digital design and management. SUMMARY

[0006] To solve at least one of the above technical problems, the present application provides an intelligent drawing method and system based on three-dimensional model integration.

[0007] The first aspect of the present application provides an intelligent drawing method based on three-dimensional model integration, comprising:

[0008] A cross-platform three-dimensional model parser is constructed, and three-dimensional model files of different sources of a target project are parsed according to the parser to construct a three-dimensional model assembly scene graph;

[0009] perform bounding box analysis on the three-dimensional model assembly scene graph to determine drawing parameters of the three-dimensional model assembly scene graph;

[0010] construct an initial project plan view of the target project according to the drawing parameters, perform contour integrity detection on the initial project plan view to obtain a detection result;

[0011] complete defects in the initial project plan view according to the detection result to construct a complete project plan view;

[0012] perform graphic annotation on the project plan view to output a two-dimensional drawing of the target project.

[0013] In this scheme, the cross-platform three-dimensional model parser is constructed, and different source three-dimensional model files of the target project are parsed according to the parser to construct a three-dimensional model assembly scene graph, specifically:

[0014] Construct a three-dimensional model parsing interface of different three-dimensional model platforms, integrate the three-dimensional model parsing interface, and construct a cross-platform three-dimensional model parser;

[0015] Obtain multi-source three-dimensional model files of the target project, import the multi-source three-dimensional model files into the cross-platform three-dimensional model parser for model format analysis, import each three-dimensional model file into the corresponding parsing interface according to the model format, identify feature points of each imported three-dimensional model based on Harris corner point detection, and determine feature points of each three-dimensional model;

[0016] Label the feature points based on two-dimensional circles, use the two-dimensional circles as model base points, and obtain position coordinates of the model base points according to the cross-platform three-dimensional model parser;

[0017] Perform spatial registration operation on each three-dimensional model of the target project according to the position coordinates, determine the base point offset of each three-dimensional model, and generate a coordinate compensation vector according to the base point offset;

[0018] Perform position calibration according to the coordinate compensation vector, integrate and visualize the multi-source three-dimensional models after position calibration based on the cross-platform three-dimensional model parser, and construct a three-dimensional model assembly scene graph of the target project.

[0019] In this scheme, the bounding box analysis is performed on the three-dimensional model assembly scene graph to determine the drawing parameters of the three-dimensional model assembly scene graph, specifically:

[0020] A plurality of axial sets of the three-dimensional model assembly scene graph are preset, projection calculation is performed on all vertices in the three-dimensional model assembly scene graph according to the axial sets, and extreme value data is constructed by extracting maximum and minimum projection values of the vertices on each axis.

[0021] Difference calculation is performed on the maximum projection value and the minimum projection value in each axis according to the extreme value data, a bounding box boundary range in each axis is generated, and a discrete oriented polyhedral bounding box of the three-dimensional model assembly scene graph is constructed based on the bounding box boundary range;

[0022] Obtaining the drawing view angle requirement information of the target project, extracting the initial projection direction of the three-dimensional model assembly scene graph according to the drawing view angle requirement information, performing visibility analysis based on the projection profile of the discrete oriented polyhedral bounding box in the initial projection direction, calculating the area of the projection profile and the complexity of the profile boundary;

[0023] If the area of the projection profile reaches a preset complete display threshold and the complexity of the profile boundary is lower than a preset complexity threshold, the initial projection direction is output as the optimal projection view angle as the drawing parameter;

[0024] Otherwise, the initial projection direction is angularly offset according to the principal axis direction of the discrete oriented polyhedral bounding box to generate a candidate projection direction set, each direction in the candidate projection direction set is iterated and the projection profile area and the profile boundary complexity are recalculated, and the candidate projection direction with the maximum projection profile area and the lowest profile boundary complexity is selected as the optimal projection view angle and output as the drawing parameter.

[0025] In the scheme, the initial project plan view of the target project is constructed according to the drawing parameter, and contour integrity detection is performed on the initial project plan view to obtain a detection result, specifically:

[0026] A projection transformation matrix of the three-dimensional model assembly scene graph is constructed according to the drawing parameter, an initial two-dimensional point cloud distribution graph is generated by performing orthogonal projection mapping on the three-dimensional vertex set based on the projection transformation matrix, and the initial two-dimensional point cloud distribution graph is grid processed to construct the initial project plan view of the target project by using a Delaunay triangulation algorithm.

[0027] The contour feature pixel set is extracted by performing Gaussian-Laplacian edge detection on the initial project plan view, the vector contour chain set is generated by performing straight line segment fitting on the contour feature pixel set by using Hough transformation, the Euclidean distance between the starting point and the ending point of each contour chain is calculated by iterating the vector contour chain set, and if the Euclidean distance exceeds a preset breaking threshold, it is determined that the contour chain has a breaking defect, and the local interruption label of contour integrity detection is generated according to the breaking defect.

[0028] According to the topological relationship of each sub-model in the three-dimensional model assembly scene graph, an occlusion analysis tree is constructed, the vector contour chains in the initial project plan are sorted in a depth-first manner based on the occlusion analysis tree, contour chains with a depth value greater than a preset visible threshold value after sorting are extracted as potential occlusion regions, a region growing algorithm is used to analyze the neighborhood pixel connectivity of the potential occlusion regions, and if there is an isolated connected domain not covered by the contour chain, it is determined as a contour missing region and a global occlusion label is generated;

[0029] According to the local interruption label and the global occlusion label, a defect distribution heat map is constructed, the contour integrity score is calculated based on the area ratio of the labeled region in the defect distribution heat map, and the contour integrity detection result of the initial project plan is obtained.

[0030] In the scheme, the initial project plan is defect-completed according to the detection result, and a complete project plan is constructed, specifically:

[0031] According to the end point coordinates of the broken contour chain extracted from the local interruption label in the detection result, a path search space is constructed based on the vertex topological relationship of the corresponding sub-model in the three-dimensional model assembly scene graph;

[0032] An A* algorithm is used to plan the shortest connection path between the end points in the path search space, and a B-spline curve fitting is used to smooth the shortest connection path to generate a completed contour segment;

[0033] According to the centroid coordinates of the isolated connected domain positioned by the global occlusion label in the detection result, the projection vertex set of the occluded sub-model is extracted based on the occlusion analysis tree of the three-dimensional model assembly scene graph, a minimum enclosing contour is generated by performing convex hull calculation on the projection vertex set, and the minimum enclosing contour is fused by performing Boolean operation with the existing contour chain in the initial project plan;

[0034] According to the completed contour segment and the contour chain fused by the Boolean operation, the initial project plan is updated, and a complete project plan is constructed.

[0035] In the scheme, the project plan is graphically annotated, and a two-dimensional drawing of the target project is output, specifically:

[0036] Legend data of a target engineering project plan is obtained, a project plan annotation model is constructed based on a convolutional neural network, and the convolutional neural network structure is constructed, including an input layer, a convolutional layer and a pooling layer;

[0037] The legend data is imported into the project plan annotation model for training, and the model parameters are optimized through a loss function and an optimizer, including a learning rate, a training round, and a training batch size;

[0038] The project plan is imported into the trained project plan annotation model for pattern recognition and name annotation, and the annotated project plan is output to obtain a two-dimensional drawing of the target project.

[0039] The second aspect of the application also provides an intelligent drawing system based on three-dimensional model integration, which comprises a memory and a processor, the memory comprising a program of an intelligent drawing method based on three-dimensional model integration, and the program being executed by the processor to implement the following steps:

[0040] A cross-platform three-dimensional model parser is constructed, and three-dimensional model files of different sources of a target project are parsed according to the parser to construct a three-dimensional model assembly scene graph;

[0041] A bounding box analysis is performed on the three-dimensional model assembly scene graph to determine drawing parameters of the three-dimensional model assembly scene graph;

[0042] An initial project plan of the target project is constructed according to the drawing parameters, and contour integrity detection is performed on the initial project plan to obtain a detection result;

[0043] Defects are completed on the initial project plan according to the detection result to construct a complete project plan;

[0044] The project plan is annotated to output a two-dimensional drawing of the target project.

[0045] The application discloses an intelligent drawing method and system based on three-dimensional model integration, which comprises the following steps: constructing a cross-platform three-dimensional model parser, parsing three-dimensional model files of different sources, and generating a three-dimensional model assembly scene graph; performing a bounding box analysis on the assembly scene graph to determine drawing parameters; generating an initial project plan according to the parameters and performing contour integrity detection to obtain a detection result; completing defects according to the detection result to generate a complete project plan; and finally performing automatic graphic annotation on the project plan to output a two-dimensional drawing of the target project. The application can efficiently integrate three-dimensional model data of multiple sources, realize rapid and accurate output of two-dimensional drawings, improve design efficiency and intelligent level, and has high practical value and application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 A flowchart of the intelligent drawing method based on three-dimensional model integration is shown;

[0047] Figure 2 A flowchart of the contour integrity detection is shown;

[0048] Figure 3 A flowchart of the construction of the complete project plan is shown;

[0049] Figure 4 A block diagram of an intelligent drawing system based on three-dimensional model integration is shown. DETAILED DESCRIPTION

[0050] In order to enable a more clear understanding of the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0051] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other manners different from those described herein, and therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0052] Figure 1 A flowchart of an intelligent drawing method based on three-dimensional model integration is shown.

[0053] As Figure 1 shown, the first aspect of the present application provides an intelligent drawing method based on three-dimensional model integration, comprising:

[0054] S102, a cross-platform three-dimensional model parser is constructed, and three-dimensional model files of different sources of a target project are parsed according to the parser to construct a three-dimensional model assembly scene graph;

[0055] S104, bounding box analysis is performed on the three-dimensional model assembly scene graph to determine drawing parameters of the three-dimensional model assembly scene graph;

[0056] S106, an initial project plan view of the target project is constructed according to the drawing parameters, and contour integrity detection is performed on the initial project plan view to obtain a detection result;

[0057] S108, defect completion is performed on the initial project plan view according to the detection result to construct a complete project plan view;

[0058] S110, graphic annotation is performed on the project plan view to output a two-dimensional drawing of the target project.

[0059] It should be noted that the construction of the cross-platform three-dimensional model parser can realize unified analysis and format conversion of multi-source three-dimensional model files from different design platforms or software (such as Revit, CATIA, SolidWorks, etc.), and ensure the intercommunication and compatibility of model data; the three-dimensional model assembly scene graph constructed by the parser can realize feature point extraction, spatial position alignment and overall visualization of multi-source models; the bounding box analysis of the three-dimensional model assembly scene graph can automatically obtain the optimal projection view angle, ensuring that the view angle is comprehensive and the redundant information is reduced when drawing two-dimensional drawings; the initial project plan can be constructed according to the drawing parameters, and through contour integrity detection, the automatic identification of the two-dimensional projection contour can be realized, and the accuracy and integrity of the model projection result can be improved; based on the detection result, the defect completion can repair the contour interruption or shielding area through automatic path search and fitting algorithm, avoiding the error and low efficiency caused by manual repair; finally, the project plan is graphically annotated, and the component name and attribute are automatically identified by intelligent algorithms such as convolutional neural network, realizing efficient, standardized and standardized two-dimensional drawing output, and greatly improving the drawing production efficiency and quality in the process of engineering design, review and construction.

[0060] According to the embodiment of the application, the cross-platform three-dimensional model parser is constructed according to the analysis of the three-dimensional model files of different sources of the target project by the parser, and a three-dimensional model assembly scene graph is constructed, specifically:

[0061] The three-dimensional model analysis interface of different three-dimensional model platforms is constructed, and the cross-platform three-dimensional model parser is constructed by integrating the three-dimensional model analysis interface;

[0062] The multi-source three-dimensional model files of the target project are obtained, the multi-source three-dimensional model files are imported into the cross-platform three-dimensional model parser for model format analysis, each three-dimensional model file is imported into the corresponding analysis interface according to the model format, the feature points of each imported three-dimensional model are identified based on Harris corner detection, and the feature points of each three-dimensional model are determined;

[0063] The feature points are labeled based on two-dimensional circles, the two-dimensional circles are used as model base points, and the position coordinates of the model base points are obtained according to the cross-platform three-dimensional model parser;

[0064] According to the position coordinates, a spatial registration operation is performed on each three-dimensional model of the target project, the base point offset of each three-dimensional model is determined, and a coordinate compensation vector is generated according to the base point offset;

[0065] According to the coordinate compensation vector, the position is calibrated, the multi-source three-dimensional models after position calibration are integrated and visualized based on the cross-platform three-dimensional model parser, and a three-dimensional model assembly scene graph of the target project is constructed.

[0066] It should be noted that by constructing a three-dimensional model analysis interface of different three-dimensional model platforms and uniformly integrating, efficient analysis of multi-source three-dimensional model files can be realized, and the problem of incompatible model formats and difficult analysis in the prior art is solved; the imported three-dimensional model is identified by Harris corner point detection, the key geometric features of the model can be accurately extracted, and the key geometric features are used for subsequent spatial positioning and alignment, wherein the "feature point" refers to a landmark point on the model surface with significant curvature change and geometric stability, such as edges and corners, hole edges and the like; the feature points are defined as "model base points" by a two-dimensional circle marking method, so that they have a unique coordinate identifier in three-dimensional space, facilitating subsequent spatial registration operation; further, by comparing the spatial positions of the model base points, the offset of the model base points is calculated, and a corresponding coordinate compensation vector is generated to perform high-precision position calibration, avoiding the common misalignment and overlapping problems in multi-model splicing; finally, the multi-source three-dimensional models after position calibration are integrated and visualized; the three-dimensional model analysis interface is connected to different three-dimensional model platforms, so that three-dimensional models of different formats can be compatible in the present application.

[0067] According to the embodiment of the present application, the bounding box analysis is performed on the three-dimensional model assembly scene graph to determine the drawing parameters of the three-dimensional model assembly scene graph, specifically:

[0068] A plurality of axial sets of the three-dimensional model assembly scene graph are preset, all vertices in the three-dimensional model assembly scene graph are projected and calculated according to the axial sets, and the maximum projection value and the minimum projection value of the vertices on each axis are extracted to construct extreme value data;

[0069] The maximum projection value and the minimum projection value on each axis are difference calculated according to the extreme value data, the bounding box boundary range on each axis is generated, and the discrete oriented polyhedral bounding box of the three-dimensional model assembly scene graph is constructed based on the bounding box boundary range;

[0070] The drawing perspective requirement information of the target project is obtained, the initial projection direction of the three-dimensional model assembly scene graph is extracted according to the drawing perspective requirement information, the visibility analysis is performed on the projection profile of the discrete oriented polyhedral bounding box in the initial projection direction, and the area of the projection profile and the complexity of the profile boundary are calculated;

[0071] If the area of the projection profile reaches a preset complete display threshold and the complexity of the profile boundary is lower than a preset complexity threshold, the initial projection direction is output as an optimal projection perspective as the drawing parameter;

[0072] Otherwise, the initial projection direction is angularly offset according to the principal axis direction of the discrete directional polyhedron bounding box to generate a candidate projection direction set, each direction in the candidate projection direction set is iterated and the projection profile area and profile boundary complexity are recalculated, and the candidate projection direction with the maximum projection profile area and the lowest profile boundary complexity is selected as the optimal projection view angle and output as the drawing parameter.

[0073] It should be noted that when a complex three-dimensional model is converted into a two-dimensional drawing, problems such as unreasonable selection of view angle, occlusion of key details, incomplete display, or line redundancy are prone to occur. By presetting multiple axial directions (for example, X, Y, Z axes and their combined directions), projection calculation is performed on the model vertices, the maximum and minimum projection values are extracted to generate 'extreme value data' (referring to the vertex projection range in each axial direction), and then the axial difference is calculated and a 'discrete directional polyhedron bounding box' (referring to an approximate polyhedron shell based on the boundary range of each axial direction, which is used to tightly wrap the overall model) is constructed. The bounding box describes the overall spatial range of the model. Combined with the drawing view angle requirement of the project, the area of the profile (referring to the total coverage area projected on the two-dimensional plane) and the complexity of the profile boundary (referring to the number of profile line segments and curvature changes) are automatically calculated through visibility analysis of the profile of the initial projection direction, thereby realizing the optimization of the projection view angle. When the initial projection direction cannot meet the preset complete display and complexity requirements, the candidate directions are generated based on the angular offset of the principal axis direction of the bounding box, and the view angle with the maximum projection area and the simplest profile boundary is selected through iteration screening, so that the output drawing parameter has both display completeness and drawing simplicity, thereby significantly improving the automation, intelligence and professionalization level of two-dimensional drawing.

[0074] Figure 2 A flowchart of the profile integrity detection performed by the present application is shown.

[0075] According to the embodiment of the present application, the initial project plan view of the target project is constructed according to the drawing parameter, the profile integrity detection is performed on the initial project plan view, and the detection result is obtained, specifically:

[0076] S202, a projection transformation matrix of a three-dimensional model total assembly scene graph is constructed according to the drawing parameter, an initial two-dimensional point cloud distribution graph is generated by orthogonally projecting and mapping a three-dimensional vertex set based on the projection transformation matrix, and a Delaunay triangulation algorithm is used to grid process the initial two-dimensional point cloud distribution graph to construct an initial project plan view of the target project.

[0077] S204, Gaussian-Laplace edge detection is performed on the initial project plan to extract a contour feature pixel set, Hough transformation is used to perform line segment fitting on the contour feature pixel set to generate a vector contour chain set, the vector contour chain set is traversed and the Euclidean distance between the start point and the end point of each contour chain is calculated, if the Euclidean distance exceeds a preset breaking threshold, it is determined that the contour chain has a breaking defect, and a local interruption label for contour integrity detection is generated according to the breaking defect;

[0078] S206, an occlusion analysis tree is constructed according to the topological relationship of each sub-model in the three-dimensional model assembly scene graph, the vector contour chains in the initial project plan are depth-first ordered based on the occlusion analysis tree, the contour chains with a depth value greater than a preset visible threshold after ordering are extracted as potential occlusion areas, a region growing algorithm is used to analyze the neighborhood pixel connectivity of the potential occlusion areas, if there is an isolated connected domain not covered by the contour chains, it is determined as a contour missing area and a global occlusion label is generated;

[0079] S208, a defect distribution heat map is constructed according to the local interruption label and the global occlusion label, a contour integrity score is calculated based on the area proportion of the labeled area in the defect distribution heat map, and the contour integrity detection result of the initial project plan is obtained.

[0080] It should be noted that when the multi-source three-dimensional model is integrated to generate a two-dimensional drawing, due to problems such as coordinate system deviation, insufficient projection view angle optimization and model occlusion, defects such as contour breakage, detail loss or key structure occlusion may occur in the plan view. Therefore, by constructing a projection transformation matrix of a three-dimensional model assembly scene graph according to drawing parameters, an initial two-dimensional point cloud is generated by orthogonally projecting the three-dimensional vertex set, and an initial project plan view is generated by using a Delaunay triangulation algorithm to complete meshing. The contour pixels are extracted by using the Gauss-Laplace edge detection, and the straight line segments are fitted to form a vector contour chain by using the Hough transform, and the Euclidean distance between the start point and the end point is calculated to judge the breakage and automatically mark the local interruption. By combining the topological relationship of the sub-model to construct an occlusion analysis tree, the contour chain is depth sorted, the occlusion area with a high depth value is extracted, and the isolated connected domain is analyzed by using a region growing algorithm to judge the global occlusion defect, generate a defect heat map, and finally calculate the contour integrity score. The projection transformation matrix is a mathematical tool for three-dimensional to two-dimensional conversion, the Delaunay triangulation is used for constructing the optimal grid of the point set, the Gauss-Laplace edge detection is an operator for identifying the boundary features of the image, the Hough transform is an algorithm for fitting the pixel points to a geometric straight line, the Euclidean distance is the straight line distance between two points, and the occlusion analysis tree is used to describe the visibility and occlusion relationship between the sub-models. The region growing algorithm is used to analyze the connectivity of the pixels in the image. The occlusion analysis tree is based on the spatial position, size and hierarchical relationship of each sub-model in the three-dimensional model assembly scene graph, and nodes are established layer by layer according to the visibility order from front to back. Each sub-model is taken as a tree node and the relative depth value is recorded, and the branch structure is formed through the visibility and occlusion relationship between the parent and child nodes, so that the occlusion relationship of each part in the whole scene is completely represented.

[0081] Figure 3 A flow chart for constructing a complete project plan view is shown.

[0082] According to the embodiments of the present application, the initial project plan view is completed according to the detection result, and a complete project plan view is constructed, specifically:

[0083] S302, according to the end point coordinates of the broken contour chain extracted from the local interruption mark in the detection result, a path search space is constructed based on the vertex topological relationship of the corresponding sub-model in the three-dimensional model assembly scene graph;

[0084] S304, the shortest connection path between the end points in the path search space is planned by using the A* algorithm, and the shortest connection path is smoothed by B-spline curve fitting to generate a completed contour segment;

[0085] S306, positioning the centroid coordinates of the isolated connected domain according to the global occlusion label in the detection result, extracting a projection vertex set of the occluded sub-model based on an occlusion analysis tree of the three-dimensional model general assembly scene graph, performing convex hull calculation on the projection vertex set to generate a minimum enclosing contour, and performing Boolean operation fusion on the minimum enclosing contour and an existing contour chain in the initial project plan view;

[0086] S308, updating the initial project plan view according to the completed contour segment and the contour chain after the Boolean operation fusion, and constructing a complete project plan view.

[0087] It should be noted that the A* algorithm combines the cost function and the heuristic function, can quickly search the global optimal path from the starting point to the ending point under complex topological constraints, and avoid invalid traversal, thereby providing a connection path for the broken contour chain which is short and consistent with the model space structure; and through B-spline curve fitting, the polyline type shortest path obtained by the A* algorithm can be smoothed, so that the completed contour segment not only satisfies the continuity, but also keeps the curve natural and smooth, avoiding the appearance of harsh polyline connection or geometric mutation. For the global occlusion defect, the convex hull calculation of the projection vertex set is because the convex hull can accurately describe the circumscribed boundary of the occluded area projection contour, and ensure that the completed area covers the missing part and does not expand excessively; the minimum enclosing contour and the existing contour chain are fused through Boolean operation (such as set operation), which can integrate the occluded area and the known area in geometry, realize topological closure and boundary unification. Therefore, through a series of steps based on computational geometry and topological optimization such as path search, smoothing fitting, convex hull calculation and Boolean fusion, the local interruption and global loss can be effectively restored, the integrity of the initial project plan view in geometric structure and expression semantics is ensured, and finally the complete project plan view which is accurate in expression, closed in boundary and can be used for drawing is generated.

[0088] According to the embodiment of the present application, the project plan view is graphically annotated, and a two-dimensional drawing of the target project is output, specifically:

[0089] Obtaining legend data of the target engineering project plan view, constructing a project plan view annotation model based on a convolutional neural network, and constructing the convolutional neural network structure, including an input layer, a convolutional layer and a pooling layer;

[0090] Training the legend data in the project plan view annotation model, optimizing the model parameters including learning rate, training rounds and training batch size through a loss function and an optimizer;

[0091] Importing the project plan view into the trained project plan view annotation model for graph recognition and name annotation, outputting the annotated project plan view, and obtaining the two-dimensional drawing of the target project.

[0092] It should be noted that the automatic graphic recognition and name annotation of the project plan by the convolutional neural network model not only greatly improves the annotation efficiency and accuracy, but also avoids the problems of missing annotation and wrong annotation in manual annotation; through the training and model optimization of the legend data, the model has good graphic feature extraction and classification ability, can accurately distinguish different types of building, structure or equipment symbols, and automatically match the corresponding annotation name; the finally output two-dimensional drawing has standard, clear and complete annotation information, meets the use requirements of engineering design, construction, acceptance and other links, and improves the automation level and engineering application value of the drawing.

[0093] Figure 4 A block diagram of an intelligent drawing system based on three-dimensional model integration is shown.

[0094] The second aspect of the application also provides an intelligent drawing system 4 based on three-dimensional model integration, which comprises a memory 41 and a processor 42, the memory comprises an intelligent drawing method program based on three-dimensional model integration, and the intelligent drawing method program based on three-dimensional model integration is executed by the processor to realize the following steps:

[0095] A cross-platform three-dimensional model parser is constructed, and three-dimensional model files of different sources of a target project are parsed according to the parser to construct a three-dimensional model assembly scene graph;

[0096] A bounding box analysis is performed on the three-dimensional model assembly scene graph to determine drawing parameters of the three-dimensional model assembly scene graph;

[0097] An initial project plan of the target project is constructed according to the drawing parameters, and contour integrity detection is performed on the initial project plan to obtain a detection result;

[0098] Defects of the initial project plan are completed according to the detection result to construct a complete project plan;

[0099] Graphic annotation is performed on the project plan to output a two-dimensional drawing of the target project.

[0100] The application discloses an intelligent drawing method and system based on three-dimensional model integration. The method comprises the following steps: constructing a cross-platform three-dimensional model parser, analyzing three-dimensional model files of different sources, and generating a three-dimensional model assembly scene graph; performing bounding box analysis on the assembly scene graph to determine drawing parameters; generating an initial project plan according to the parameters, and performing contour integrity detection to obtain a detection result; performing defect completion according to the detection result to generate a complete project plan; and finally, performing automatic graphic annotation on the plan to output a two-dimensional drawing of the target project. The application can efficiently integrate multi-source three-dimensional model data, realize rapid and accurate output of two-dimensional drawings, improve design efficiency and intelligent level, and has high practical value and application prospect.

[0101] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other manners. The above described device embodiments are only schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, direct coupling or communication connection between the components can be indirect coupling or communication connection through some interfaces, or electrical, mechanical or other forms.

[0102] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units; they can be located in one place, or distributed on multiple network units; and some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0103] In addition, each functional unit in each embodiment of the application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be realized in the form of hardware or hardware plus software functional unit.

[0104] Those skilled in the art can understand that all or part of the steps of the above method embodiments can be completed by program instruction related hardware, and the foregoing program can be stored in a computer readable storage medium, and the program is executed to perform steps including the above method embodiments; and the foregoing storage medium includes mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic discs or optical discs, and various storage media that can store program codes.

[0105] Alternatively, the above-mentioned integrated unit of the present application, if realized in the form of a software function module and sold or used as an independent product, can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: mobile storage devices, ROM, RAM, magnetic disks or optical disks, and various media that can store program codes.

[0106] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent drawing output method based on three-dimensional model integration, characterized in that, The method comprises the following steps: A cross-platform three-dimensional model parser is constructed, different source three-dimensional model files of a target project are parsed according to the parser, a three-dimensional model assembly scene graph is constructed, and specifically: A three-dimensional model parsing interface of different three-dimensional model platforms is constructed, the three-dimensional model parsing interface is integrated, and a cross-platform three-dimensional model parser is constructed; Multi-source three-dimensional model files of a target project are obtained, the multi-source three-dimensional model files are imported into the cross-platform three-dimensional model parser for model format parsing, each three-dimensional model file is imported into a corresponding parsing interface according to the model format, feature points of each imported three-dimensional model are identified based on Harris corner point detection, and feature points of each three-dimensional model are determined; The feature points are marked based on two-dimensional circles, the two-dimensional circles are used as model base points, and position coordinates of the model base points are obtained according to the cross-platform three-dimensional model parser; Space registration operations are performed on each three-dimensional model of the target project according to the position coordinates, base point offsets of each three-dimensional model are determined, and a coordinate compensation vector is generated according to the base point offsets; Position calibration is performed according to the coordinate compensation vector, multi-source three-dimensional models after position calibration are integrated and visualized based on the cross-platform three-dimensional model parser, and a three-dimensional model assembly scene graph of the target project is constructed; A bounding box analysis is performed on the three-dimensional model assembly scene graph, and drawing parameters of the three-dimensional model assembly scene graph are determined; Initial project plan views of the target project are constructed according to the drawing parameters, contour integrity detection is performed on the initial project plan views, and detection results are obtained; Defects of the initial project plan views are completed according to the detection results, and complete project plan views are constructed; Graphical annotations are performed on the project plan views, and two-dimensional drawings of the target project are output. 2.The intelligent drawing method based on three-dimensional model integration of claim 1, wherein, The bounding box analysis performed on the three-dimensional model assembly scene graph to determine the drawing parameters of the three-dimensional model assembly scene graph specifically comprises: A plurality of axial sets of the three-dimensional model assembly scene graph are preset, projection calculation is performed on all vertices in the three-dimensional model assembly scene graph according to the axial sets, maximum projection values and minimum projection values of the vertices on each axis are extracted to construct extreme value data; Difference calculation is performed on the maximum projection values and the minimum projection values on each axis according to the extreme value data, bounding box boundary ranges on each axis are generated, and a discrete directional polyhedron bounding box of the three-dimensional model assembly scene graph is constructed based on the bounding box boundary ranges; Drawing visual angle requirement information of the target project is obtained, an initial projection direction of the three-dimensional model assembly scene graph is extracted according to the drawing visual angle requirement information, visibility analysis is performed on a projection contour of the discrete directional polyhedron bounding box in the initial projection direction, and the area of the projection contour and the complexity of the contour boundary are calculated; If the area of the projection contour reaches a preset complete display threshold and the complexity of the contour boundary is lower than a preset complexity threshold, the initial projection direction is output as an optimal projection visual angle as the drawing parameter. Otherwise, the initial projection direction is angularly offset according to a principal axis direction of the discrete oriented polytope bounding box to generate a candidate projection direction set, each direction in the candidate projection direction set is iterated and the projection profile area and profile boundary complexity are recalculated, and a candidate projection direction with the largest projection profile area and the lowest profile boundary complexity is selected as an optimal projection view angle output as the drawing parameter. 3.The intelligent drawing method based on three-dimensional model integration of claim 1, wherein, The initial project plan is constructed according to the drawing parameter, and contour integrity detection is performed on the initial project plan to obtain a detection result, specifically: A projection transformation matrix of the three-dimensional model assembly scene graph is constructed according to the drawing parameter, an initial two-dimensional point cloud distribution graph is generated by orthogonally projecting a three-dimensional vertex set based on the projection transformation matrix, and an initial project plan of the target project is constructed by grid processing the initial two-dimensional point cloud distribution graph using a Delaunay triangulation algorithm; A contour feature pixel set is extracted by performing Gaussian-Laplacian edge detection on the initial project plan, a vector contour chain set is generated by performing Hough transformation on the contour feature pixel set, and the Euclidean distance between the start point and the end point of each contour chain in the vector contour chain set is calculated, if the Euclidean distance exceeds a preset breaking threshold, it is determined that the contour chain has a breaking defect, and a local interruption label for contour integrity detection is generated according to the breaking defect; An occlusion analysis tree is constructed according to the topological relationship of each sub-model in the three-dimensional model assembly scene graph, the vector contour chains in the initial project plan are depth-first ordered based on the occlusion analysis tree, and the contour chains with a depth value greater than a preset visible threshold are extracted as potential occlusion areas, and a region growing algorithm is used to analyze the neighborhood pixel connectivity of the potential occlusion areas, if there is an isolated connected domain not covered by the contour chain, it is determined as a contour missing area and a global occlusion label is generated; A defect distribution heat map is constructed according to the local interruption label and the global occlusion label, a contour integrity score is calculated based on the area proportion of the labeled area in the defect distribution heat map, and a contour integrity detection result of the initial project plan is obtained.

4. The intelligent drawing method based on three-dimensional model integration according to claim 1, characterized in that, The initial project plan is defect-completed according to the detection result to construct a complete project plan, specifically: The end point coordinates of the broken contour chain are extracted according to the local interruption label in the detection result, and a path search space is constructed based on the vertex topological relationship of the corresponding sub-model in the three-dimensional model assembly scene graph; An A* algorithm is used to plan the shortest connection path between the end points in the path search space, and a B-spline curve fitting is used to smooth the shortest connection path to generate a completed contour segment; The centroid coordinates of the isolated connected domain are located according to the global occlusion label in the detection result, the projection vertex set of the occluded sub-model is extracted based on the occlusion analysis tree of the three-dimensional model assembly scene graph, a minimum enclosing contour is generated by performing convex hull calculation on the projection vertex set, and Boolean operation is performed between the minimum enclosing contour and the existing contour chain in the initial project plan for fusion. According to the contour segment after the completion of the Boolean operation fusion, the initial project plan is updated, and a complete project plan is constructed.

5. The intelligent plot generation method based on three-dimensional model integration according to claim 1, characterized in that, The project plan is graphically annotated, and a two-dimensional drawing of the target project is output, specifically: Obtain the legend data of the target engineering project plan, and construct a project plan annotation model based on a convolutional neural network, the structure of the convolutional neural network includes an input layer, a convolutional layer and a pooling layer; The legend data is imported into the project plan annotation model for training, and the model parameters are optimized through the loss function and the optimizer, the model parameters include learning rate, training rounds, and training batch size; The project plan is imported into the trained project plan annotation model for graphic recognition and name annotation, and the annotated project plan is output to obtain the two-dimensional drawing of the target project.

6. An intelligent drawing system based on three-dimensional model integration, characterized in that, The intelligent drawing system based on three-dimensional model integration includes a memory and a processor, the memory includes an intelligent drawing method based on three-dimensional model integration program, when the intelligent drawing method based on three-dimensional model integration program is executed by the processor, the following steps are realized: A cross-platform three-dimensional model parser is constructed, and different source three-dimensional model files of the target project are parsed according to the parser to construct a three-dimensional model assembly scene graph, specifically: A three-dimensional model parsing interface of different three-dimensional model platforms is constructed, and the three-dimensional model parsing interface is integrated to construct a cross-platform three-dimensional model parser; Obtain the multi-source three-dimensional model file of the target project, import the multi-source three-dimensional model file into the cross-platform three-dimensional model parser for model format analysis, import each three-dimensional model file into the corresponding parsing interface according to the model format, identify the feature points of each imported three-dimensional model based on Harris corner detection, and determine the feature points of each three-dimensional model. Based on the two-dimensional circle, the feature points are labeled, the two-dimensional circle is used as the model base point, and the position coordinates of the model base point are obtained according to the cross-platform three-dimensional model parser. According to the position coordinates, a spatial registration operation is performed on each three-dimensional model of the target project to determine the base point offset of each three-dimensional model, and a coordinate compensation vector is generated according to the base point offset. According to the coordinate compensation vector, the position is calibrated, the multi-source three-dimensional model after position calibration is integrated and visualized based on the cross-platform three-dimensional model parser, and a three-dimensional model assembly scene graph of the target project is constructed. The three-dimensional model assembly scene graph is analyzed by the bounding box, and the drawing parameters of the three-dimensional model assembly scene graph are determined. According to the drawing parameters, an initial project plan of the target project is constructed, the contour integrity of the initial project plan is detected, and a detection result is obtained. According to the detection result, the initial project plan is defect-completed, and a complete project plan is constructed. The project plan is graphically annotated, and a two-dimensional drawing of the target project is output.

Images