Construction project digital display method and system based on artificial intelligence

Abstract

The invention aims to provide a construction project digital display method and system based on artificial intelligence, and belongs to the technical field of construction information.The method comprises the steps that building information model data and a construction parameter set are obtained, a feature coding matrix is extracted by using a three-dimensional convolutional neural network in combination with an attention mechanism, and a construction parameter set is obtained; and constructing a multi-target constraint condition set to perform parameter optimization, generating an optimized parameter matrix, inputting the optimized parameter matrix into a real-time rendering engine to reconstruct a dynamic three-dimensional model, finally constructing an interactive display interface through a WebGL technology, and integrating a collision detection module and a parameter feedback module. The system comprises a data acquisition module, a feature extraction module, an optimization processing module, a visual rendering module and an interactive display module. According to the invention, the digital display of the whole life cycle of the construction project is realized, the synchronism, the feature expression ability and the multi-objective optimization effect of the model and the construction site are improved, the visualization quality and the interaction experience are enhanced, and an efficient solution is provided for the management of the construction project.

Description

Technical Field

[0001] The present invention relates to the field of building information technology, and in particular, to a digital display method and system for building projects based on artificial intelligence. Background Art

[0002] In the construction industry, traditional building project display methods mainly rely on two-dimensional drawings, static three-dimensional models, and on-site physical displays. These methods have many limitations. Two-dimensional drawings are difficult for non-professionals to intuitively understand the overall view and details of a building; although static three-dimensional models improve the visual effect to a certain extent, they cannot reflect the dynamic changes during the building construction process in real time, such as construction progress, material usage, etc.; on-site physical displays are restricted by time and space, and the effects of unfinished parts cannot be shown in advance. In addition, the traditional display methods have poor interactivity during the construction stage and are difficult to adjust and optimize in real time according to the actual situation during the construction process, resulting in problems such as construction progress delays and cost overruns. With the increasing complexity and refinement of building projects, higher requirements are put forward for the real-time, interactive, and accurate display of building projects.

[0003] In recent years, with the rapid development of artificial intelligence technology, its application in the construction field has gradually attracted attention. However, there are still deficiencies in the current solutions for applying artificial intelligence technology to building project displays. Although some existing methods attempt to use artificial intelligence for the generation or optimization of building models, they often lack comprehensive consideration of construction parameters and cannot organically combine building information model data with multi-dimensional parameters such as construction progress, material properties, and budgets, making it difficult to meet the requirements of the full life cycle display of building projects. In terms of dynamic display, the three-dimensional models generated by existing technologies still need to be improved in terms of real-time performance, detail level adjustment, and physical effect simulation, and cannot provide users with an immersive interactive experience. In addition, existing technologies also have defects in collision detection and feedback optimization, and cannot timely detect potential problems during the construction process and make effective adjustments, affecting the construction quality and efficiency of building projects.

[0004] With the deepening of the digital transformation of the construction industry, Building Information Modeling (BIM) technology has become an industry standard. However, there are currently many technical bottlenecks restricting its further development. Existing building information models are difficult to keep in sync with the actual construction status, resulting in poor "digital twin" effects. Traditional feature extraction methods are mainly based on geometric segmentation and rule recognition, lacking effective expression of the semantic information of building structures, and thus performing poorly in the analysis of complex buildings. In addition, building project optimization often uses single-objective or simple weighted methods and cannot handle the collaborative optimization problems of multi-dimensional constraints such as structural safety, construction progress, and cost control.

[0005] Existing rendering techniques mostly adopt traditional rasterization methods, lacking physical realism and experiencing a significant drop in frame rate when dealing with large-scale building data, which affects the user experience. At the same time, collision detection algorithms are mostly based on simplified geometries and rough detections, making it difficult to balance accuracy and efficiency, resulting in frequent component conflicts and design rework in practical applications, increasing project costs and cycles.

[0006] Artificial intelligence technologies, especially deep learning methods, have made significant breakthroughs in the fields of computer vision and graphics, but their applications in the construction field have not been fully developed. Existing research mostly focuses on single-point technological breakthroughs, lacking a systematic solution that integrates data collection, feature extraction, multi-objective optimization, visual rendering, and interactive display, and cannot meet the needs of digital management throughout the life cycle of construction projects. Summary of the Invention

[0007] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a digital display method for construction projects based on artificial intelligence. The present invention uses a three-dimensional convolutional neural network and an attention mechanism to extract high-dimensional features of the building model, constructs a multi-objective constraint optimization framework for joint parameter optimization, and combines a WebGL interactive interface to provide an efficient, accurate, and highly interactive digital display solution for construction projects.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] In the first aspect, the present invention provides a digital display method for construction projects based on artificial intelligence, including the following steps:

[0010] Step S1, obtaining building information model data and construction parameter sets of the target construction project, where the construction parameter sets include material attribute parameters, construction progress parameters, and budget parameters;

[0011] Step S2, inputting the building information model data into a pre-trained three-dimensional convolutional neural network to generate a feature encoding matrix including geometric features and semantic features, where the three-dimensional convolutional neural network includes a residual connection structure and an attention mechanism module;

[0012] Step S3, constructing a multi-objective constraint condition set based on the construction parameter sets, inputting the feature encoding matrix into a parameter optimization model for joint optimization, and generating an optimized parameter matrix that meets construction feasibility;

[0013] Step S4, inputting the optimized parameter matrix into a real-time rendering engine for dynamic model reconstruction to generate a visualized three-dimensional model with temporal attributes;

[0014] Step S5: Based on the visualized 3D model, construct an interactive 3D display interface through WebGL technology, integrating a collision detection module and a parameter feedback module. The parameter feedback module is used to feedback the results of interactive operations into the multi-objective constraint condition set for iterative optimization.

[0015] Advantages: By integrating building information model data and construction parameter sets, using a 3D convolutional neural network to extract features, optimizing parameters based on multi-objective constraints, combining real-time rendering technology with a WebGL interactive interface, the present invention constructs a complete digital display process for construction projects, which can significantly improve the visualization effect of construction projects, optimize building design schemes, enhance construction management efficiency, and achieve interactive real-time parameter adjustment and feedback.

[0016] Further, step S1 specifically includes:

[0017] Step S1.1: Collect on-site point cloud data through a laser scanning device, and register the point cloud data with the building information model; the point cloud data includes spatial coordinate and reflection intensity information.

[0018] Step S1.2: Use a data cleaning algorithm to eliminate outliers and noise data in the point cloud data. The data cleaning algorithm calculates the point cloud density threshold based on the following formula:

[0019] ρ threshold =μ ρ -α·σ ρ

[0020] where ρ threshold represents the point cloud density threshold, μ ρ represents the regional average point cloud density, σ ρ represents the point cloud density standard deviation, and α represents the adjustment coefficient.

[0021] Step S1.3: Perform a difference analysis on the cleaned point cloud data and the building information model to generate model correction parameters.

[0022] Step S1.4: Update the building information model data based on the model correction parameters to make it consistent with the actual on-site construction status.

[0023] Advantages: By collecting point cloud data through a laser scanning device and registering it with the building information model, using a data cleaning algorithm to eliminate noise, performing difference analysis and model correction, the building information model is made consistent with the actual on-site construction status, which can effectively solve the problem that traditional building information models cannot reflect on-site construction changes in a timely manner, improve the accuracy and timeliness of the model, and lay a data foundation for subsequent feature extraction and parameter optimization.

[0024] Furthermore, the generation process of the feature encoding matrix in step S2 includes:

[0025] Step S2.1: Perform multi-scale voxelization on the building information model to generate hierarchical features including local geometric details and global structure. The multi-scale voxelization process uses an octree recursive segmentation algorithm.

[0026] Step S2.2: Calculate the feature weight values of each voxel region through an attention mechanism. The feature weight values are calculated according to the following formula:

[0027]

[0028] where w i represents the feature weight value of the i-th voxel region; q i represents the query vector; K represents the key matrix; V represents the value matrix; d represents the feature dimension.

[0029] Step S2.3: Dynamically adjust the convolution kernel parameters in the three-dimensional convolutional neural network according to the feature weight values. The convolution kernel parameters include the convolution kernel weight matrix, bias value, and stride parameter. The adjustment method is as follows:

[0030]

[0031] where represents the adjusted convolution kernel weight; represents the original convolution kernel weight; γ represents the adjustment factor; w i represents the feature weight value.

[0032] Step S2.4: Generate the feature encoding matrix by fusing shallow geometric features and deep semantic features through a residual connection structure.

[0033] Beneficial effects: By using multi-scale voxelization and an attention mechanism, dynamically adjusting the convolution kernel parameters, and fusing shallow geometric features and deep semantic features through a residual connection, a high-quality feature encoding matrix is generated, greatly improving the expression ability for complex building structures, enabling the system to capture the detailed features and global structure information of the building model, and providing rich feature representations for subsequent optimization processing.

[0034] Furthermore, the construction method of the multi-objective constraint condition set in step S3 includes:

[0035] Step S3.1: Establish a mechanical constraint equation between material strength and structural stress. The mechanical constraint equation satisfies:

[0036] σ max ≤φ·f y

[0037] Among them, σ max represents the maximum stress; f y represents the material yield strength; φ represents the safety factor;

[0038] Step S3.2: Construct a progress constraint relationship graph of construction processes and time nodes. The relationship graph is represented by a directed acyclic graph, where:

[0039] G = (V, E, T)

[0040] Among them, G represents the relationship graph; V represents the set of construction processes; E represents the set of process dependency relationships; T represents the set of time constraints; for any process v i , v j ∈V, if process v i must be completed before process v j , then there is a directed edge e ij ∈E, and it satisfies the time constraint t j -t i ≥d i , where t i and t j respectively represent the start times of processes v i and v j ; d i represents the duration of process v i ;

[0041] Step S3.3: Set the economic constraint threshold of project cost and material usage;

[0042] Step S3.4: Integrate the mechanical constraint equation, the progress constraint relationship graph, and the economic constraint threshold into a multi-objective constraint condition set.

[0043] Beneficial effects: A multi-objective constraint condition set including a mechanical constraint equation, a construction progress relationship graph, and an economic constraint threshold is constructed, realizing a comprehensive consideration of structural safety, construction feasibility, and economic rationality, breaking through the limitation that traditional optimization methods are difficult to simultaneously meet multiple constraints, making the optimization result more in line with the actual project requirements, and improving the feasibility of the construction plan.

[0044] Furthermore, the specific dynamic model reconstruction in step S4 includes:

[0045] Step S4.1: Perform feature decoding on the optimization parameter matrix to generate triangular patch data with material attributes;

[0046] Step S4.2: Calculate the lighting effect and shadow information based on the physically based rendering algorithm. The physically based rendering algorithm includes physically based rendering path tracing and global illumination models;

[0047] Step S4.3: Dynamically adjust the model detail level of LOD according to the viewing distance;

[0048] Step S4.4: Combine the triangular patch data with material attributes with the lighting effects and shadow information to generate the visualized 3D model with temporal attributes.

[0049] Beneficial effects: By generating triangular patch data with material attributes through feature decoding, combining with physical rendering algorithms and dynamic LOD adjustment, high-quality visualization 3D model reconstruction is achieved, enhancing the visual realism and rendering efficiency of the 3D model, enabling the system to support real-time display of large-scale building models while ensuring rendering quality.

[0050] Furthermore, the working method of the collision detection module in step S5 includes:

[0051] Step S5.1: Construct a bounding box hierarchy tree for all objects in the scene, and the bounding boxes in the bounding box hierarchy tree adopt an axis-aligned bounding box structure;

[0052] Step S5.2: Perform preliminary collision screening using the separating axis theorem. The condition for the separating axis theorem to determine that two convex polyhedra do not intersect is that there exists an axis such that the projections of the two objects on this axis do not overlap;

[0053] Step S5.3: Perform triangle-level collision detection on the selected object pairs;

[0054] Step S5.4: Feed the detected collision information back to the multi-objective constraint condition set through the parameter feedback module to trigger re-optimization.

[0055] Beneficial effects: Using a bounding box hierarchy tree and the separating axis theorem for collision detection, and realizing optimized feedback of collision information through a parameter feedback mechanism, improving the efficiency and accuracy of collision detection, enabling the system to detect and solve possible component conflict problems during the construction process in real time, and improving the coordination of design and construction.

[0056] Furthermore, performing triangle-level collision detection on the selected object pairs in step S5.3 includes the following sub-steps:

[0057] Step S5.3.1: Organize the triangular patch data of the object into a spatial hash table structure to accelerate triangle search;

[0058] Step S5.3.2: For each pair of potentially colliding triangles, calculate the distance from the triangle vertices to the plane of the other triangle;

[0059] Step S5.3.3: Determine the collision depth by calculating the intersection position of the triangle edges and the other triangle;

[0060] Step S5.3.4: Generate a collision data structure including the collision position, collision normal vector, and collision depth.

[0061] Beneficial effects: Organize triangular patch data through a spatial hash table, calculate the distance from triangle vertices to planes and the intersection points of edges, generate a detailed collision data structure, provide accurate collision detection results, including collision position, normal vector, and depth information, and provide a detailed basis for subsequent conflict resolution and model adjustment.

[0062] In a second aspect, the present invention provides an artificial intelligence-based digital display system for construction projects, which is used to execute the artificial intelligence-based digital display method for construction projects described in the first aspect, and includes:

[0063] A data acquisition module configured to obtain the model data and construction parameter set of the building information model;

[0064] A feature extraction module including a three-dimensional convolutional neural network for generating a feature encoding matrix;

[0065] An optimization processing module configured to construct a multi-objective constraint condition set and perform parameter optimization;

[0066] A visualization rendering module including a real-time rendering engine for generating a dynamic three-dimensional model;

[0067] An interactive display module, a WebGL display interface integrating a collision detection function.

[0068] Beneficial effects: Construct a complete system architecture including a data acquisition module, a feature extraction module, an optimization processing module, a visualization rendering module, and an interactive display module, realize seamless connection and efficient operation of the entire process of digital display of construction projects, and improve performance and user experience.

[0069] Furthermore, the feature extraction module includes:

[0070] A voxelization processing unit for converting the building information model into a multi-scale voxel representation;

[0071] An attention calculation unit configured with a deformable convolution kernel and a channel attention mechanism; a feature fusion unit for integrating local features and global context information.

[0072] Beneficial effects: The feature extraction module includes a voxelization processing unit, an attention calculation unit, and a feature fusion unit, realizes efficient feature extraction of the building model, improves the system's expression ability and processing efficiency for complex building structures, and lays a foundation for high-quality model display.

[0073] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the method according to any one of the first aspect are implemented.

[0074] In summary, compared with the prior art, by integrating deep learning and building information modeling technology, the present invention realizes the intelligent processing of the entire process of digital display of construction projects, significantly improves the synchronization accuracy between the 3D model and the actual construction site, enhances the feature expression ability, realizes the collaborative optimization of structural safety, construction progress, and economic cost, can shorten the construction period and reduce costs, and greatly improves the visualization quality and interactive experience through physical rendering and collision detection, providing a systematic solution for the digital management of construction projects and solving key technical problems such as difficult model update, limited feature expression, insufficient multi-objective optimization, and lack of interactivity in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The drawings forming a part of the specification illustrate embodiments of the present invention and, together with the description, are used to explain the principles of the present invention.

[0076] Referring to the drawings, the present invention can be more clearly understood from the following detailed description, wherein:

[0077] Figure 1 is a flowchart of a method for digital display of construction projects based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] The technical solutions of the present invention will be described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific features in the embodiments of the present application and the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other. The following embodiments are only used to illustrate the technical solutions of the present invention more clearly, and cannot be used to limit the protection scope of the present invention.

[0079] As used herein, the term "and / or" is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0080] Embodiment 1

[0081] An embodiment of the present invention provides a method for digital display of construction projects based on artificial intelligence, including the following steps:

[0082] Step S1: Obtain the building information model data and construction parameter set of the target construction project;

[0083] Step S2: Input the building information model data into a pre-trained three-dimensional convolutional neural network to generate a feature encoding matrix including geometric features and semantic features. The three-dimensional convolutional neural network includes a residual connection structure and an attention mechanism module;

[0084] Step S3: Construct a multi-objective constraint condition set based on the construction parameter set, input the feature encoding matrix into a parameter optimization model for joint optimization, and generate an optimized parameter matrix that meets construction feasibility;

[0085] Step S4: Input the optimized parameter matrix into a real-time rendering engine for dynamic model reconstruction, and generate a visual three-dimensional model with temporal attributes;

[0086] Step S5: Based on the visual three-dimensional model, construct an interactive three-dimensional display interface through WebGL technology, integrate a collision detection module and a parameter feedback module. The parameter feedback module is used to feedback the interactive operation results into the multi-objective constraint condition set for iterative optimization.

[0087] Further, Step S1 specifically includes:

[0088] Step S1.1: Collect the point cloud data of the construction site through a laser scanning device, and register the point cloud data with the building information model; the point cloud data includes spatial coordinates and reflection intensity information;

[0089] Step S1.2: Use a data cleaning algorithm to eliminate the outlier points and noise data in the point cloud data. The data cleaning algorithm calculates the point cloud density threshold based on the following formula:

[0090] ρ threshold =μ ρ -α·σ ρ

[0091] where ρ threshold represents the point cloud density threshold, μ ρ represents the regional average point cloud density, σ ρ represents the standard deviation of the point cloud density, and α represents the adjustment coefficient;

[0092] Step S1.3: Conduct a difference analysis between the cleaned point cloud data and the building information model to generate model correction parameters;

[0093] Step S1.4: Update the building information model data based on the model correction parameters to make it consistent with the actual construction site status.

[0094] Specifically, the generation process of the feature encoding matrix in Step S2 includes:

[0095] Step S2.1: Perform multi-scale voxelization on the building information model to generate hierarchical features including local geometric details and global structures. The multi-scale voxelization process uses an octree recursive segmentation algorithm;

[0096] Step S2.2: Calculate the feature weight values of each voxel region through an attention mechanism. The feature weight values are calculated according to the following formula:

[0097]

[0098] where w i represents the feature weight value of the i-th voxel region; q i represents the query vector; K represents the key matrix; V represents the value matrix; d represents the feature dimension;

[0099] Step S2.3: Dynamically adjust the convolution kernel parameters in the three-dimensional convolutional neural network according to the feature weight values. The convolution kernel parameters include the convolution kernel weight matrix, bias value, and stride parameter. The adjustment method is as follows:

[0100]

[0101] where represents the adjusted convolution kernel weight; represents the original convolution kernel weight; γ represents the adjustment factor; w i represents the feature weight value;

[0102] Step S2.4: Fuse the shallow geometric features and deep semantic features through a residual connection structure to generate a feature encoding matrix.

[0103] Furthermore, the construction method of the multi-objective constraint condition set in step S3 includes:

[0104] Step S3.1: Establish a mechanical constraint equation between material strength and structural stress. The mechanical constraint equation satisfies:

[0105] σ max ≤φ·f y

[0106] where σ max represents the maximum stress; f y represents the material yield strength; φ represents the safety factor;

[0107] Step S3.2: Construct a progress constraint relationship graph of construction processes and time nodes. The relationship graph is represented by a directed acyclic graph, where:

[0108] G=(V,E,T)

[0109] Among them, G represents the relationship graph; V represents the set of construction processes; E represents the set of process dependencies; T represents the set of time constraints; for any process v i ,v j ∈V, if process v i Must be in process v j If it is completed before, there is a directed edge e ij ∈E, and satisfy the time constraint t j -t i ≥d i , where t i and t j Represents process v i and v j Start time of d i Indicates process v i duration;

[0110] Step S3.3: Setting economic constraint thresholds for project cost and material usage;

[0111] Step S3.4: Integrate the mechanical constraint equation, the progress constraint relationship map, and the economic constraint threshold into a multi-objective constraint condition set.

[0112] Specifically, the dynamic model reconstruction in step S4 includes:

[0113] Step S4.1, feature decoding is performed on the optimized parameter matrix to generate triangular facet data with material attributes;

[0114] Step S4.2: Calculate lighting effects and shadow information based on a physical rendering algorithm, where the physical rendering algorithm includes physically based rendering path tracing and a global illumination model;

[0115] Step S4.3: Dynamically adjust the model level of detail (LOD) based on the viewing distance; the LOD level is determined by the following formula:

[0116]

[0117] Among them, LOD level represents the level of detail, d represents the distance from the viewpoint to the object, and d0 represents the reference distance;

[0118] Step S4.4: Combine the triangular facet data with material attributes with the lighting effect and shadow information to generate a visual three-dimensional model with time series attributes.

[0119] Specifically, the working method of the collision detection module in step S5 includes:

[0120] Step S5.1: construct a bounding box hierarchy tree for all objects in the scene, where the bounding boxes in the bounding box hierarchy tree adopt an axially aligned bounding box structure;

[0121] Step S5.2: Perform preliminary collision screening using the Separating Axis Theorem. The condition for the Separating Axis Theorem to determine that two convex polyhedra do not intersect is that there exists an axis such that the projections of the two objects on this axis do not overlap.

[0122] Step S5.3: Perform triangle-level collision detection on the filtered object pairs.

[0123] Step S5.4: Feed the detected collision information back to the multi-objective constraint condition set through the parameter feedback module to trigger re-optimization.

[0124] Preferably, performing triangle-level collision detection on the filtered object pairs in Step S5.3 includes the following sub-steps:

[0125] Step S5.3.1: Organize the triangular facet data of the object into a spatial hash table structure to accelerate triangle lookup.

[0126] Step S5.3.2: For each pair of potentially colliding triangles, calculate the distance from the triangle vertices to the plane of the other triangle.

[0127] Step S5.3.3: Determine the collision depth by calculating the intersection position of the triangle edges with the other triangle.

[0128] Step S5.3.4: Generate a collision data structure including the collision position, collision normal vector, and collision depth.

[0129] As an embodiment, a digital display system for construction projects based on artificial intelligence, used to execute the digital display method for construction projects based on artificial intelligence, includes:

[0130] A data acquisition module, configured to obtain the model data and construction parameter set of the building information model.

[0131] A feature extraction module, including a three-dimensional convolutional neural network, used to generate a feature encoding matrix.

[0132] An optimization processing module, configured to construct a multi-objective constraint condition set and perform parameter optimization.

[0133] A visualization rendering module, including a real-time rendering engine, used to generate a dynamic three-dimensional model.

[0134] An interactive display module, a WebGL display interface integrating a collision detection function.

[0135] Furthermore, for the digital display system for construction projects based on artificial intelligence, the feature extraction module includes:

[0136] A voxelization processing unit for converting a building information model into a multi-scale voxel representation;

[0137] An attention calculation unit configured with a deformable convolution kernel and a channel attention mechanism; A feature fusion unit for integrating local features and global context information.

[0138] Further, the visualization rendering module includes:

[0139] A material processing unit for generating physically based material property parameters;

[0140] A lighting calculation unit configured to calculate the lighting distribution and shadow effects in real time;

[0141] A view optimization unit for adjusting the model rendering details according to the viewing perspective;

[0142] A timing control unit configured to manage the display of the construction sequence of building components.

[0143] Embodiment 2

[0144] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the following method are implemented:

[0145] Step S1, obtaining the building model data and construction parameters of the target building project, including:

[0146] Step S1.1, collecting the point cloud data of the construction site through a laser scanning device, and registering the point cloud data with the building information model;

[0147] Step S1.2, using a data cleaning algorithm to eliminate the outlier points and noise data in the point cloud data;

[0148] Step S1.3, performing a difference analysis on the cleaned point cloud data and the building information model to generate model correction parameters.

[0149] Step S2, inputting the building model data into a feature extraction network to generate a feature encoding matrix, and the generation process of the feature encoding matrix includes:

[0150] Step S2.1, performing multi-scale voxelization processing on the building information model to generate hierarchical features including local geometric details and global structures;

[0151] Step S2.2, calculating the feature weight values of each voxel region through an attention mechanism;

[0152] Step S2.3, dynamically adjusting the convolution kernel parameters according to the feature weight values.

[0153] Step S3. Based on the construction parameters, construct a set of constraint conditions, and input the feature encoding matrix into the optimization model for parameter optimization. The method for constructing the multi-objective constraint condition set includes:

[0154] Step S3.1. Establish a mechanical constraint equation between material strength and structural stress;

[0155] Step S3.2. Construct a progress constraint relationship graph of construction processes and time nodes;

[0156] Step S3.3. Set an economic constraint threshold for project cost and material consumption.

[0157] Step S4. Input the optimized feature encoding matrix into the rendering engine to generate a dynamic visualization model. The specific dynamic model reconstruction in Step S4 includes:

[0158] Step S4.1. Decode the features of the optimized parameter matrix to generate triangular patch data with material attributes;

[0159] Step S4.2. Calculate the lighting effect and shadow information based on the physically based rendering algorithm;

[0160] Step S4.3. Dynamically adjust the level of detail (LOD) of the model according to the viewing distance.

[0161] Step S5. Output an interactive three-dimensional display interface including a collision detection function, specifically including:

[0162] Step S5.1. Construct an axis-aligned bounding box (AABB) hierarchy of all objects in the scene;

[0163] Step S5.2. Use the separating axis theorem for preliminary collision screening;

[0164] Step S5.3. Perform precise triangle-level collision detection on the selected object pairs.

[0165] A computer-readable storage medium provided by an embodiment of the present invention stores a computer program that can execute a method for digital display of a construction project based on artificial intelligence provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0166] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and deformations can be made, and these improvements and deformations should also be regarded as the protection scope of the present invention.

Claims

1. A digital display method for construction projects based on artificial intelligence, characterized in that, It includes the following steps: Step S1: Obtain the building information model data and construction parameter set of the target construction project; Step S2: Input the building information model data into a pre-trained three-dimensional convolutional neural network to generate a feature encoding matrix including geometric features and semantic features. The three-dimensional convolutional neural network includes a residual connection structure and an attention mechanism module; Step S3: Construct a multi-objective constraint condition set based on the construction parameter set, input the feature encoding matrix into a parameter optimization model for joint optimization, and generate an optimized parameter matrix that meets construction feasibility; Step S4: Input the optimized parameter matrix into a real-time rendering engine for dynamic model reconstruction to generate a visualized three-dimensional model with temporal attributes; Step S5: Based on the visualized three-dimensional model, construct an interactive three-dimensional display interface through WebGL technology, integrating a collision detection module and a parameter feedback module. The parameter feedback module is used to feedback the interactive operation results into the multi-objective constraint condition set for iterative optimization.

2. The method for digital display of construction projects based on artificial intelligence according to claim 1, wherein Step S1 specifically includes: Step S1.1: Collect the point cloud data of the construction site through a laser scanning device, and register the point cloud data with the building information model; the point cloud data includes spatial coordinates and reflection intensity information; Step S1.2: Use a data cleaning algorithm to eliminate the outlier points and noise data in the point cloud data. The data cleaning algorithm calculates the point cloud density threshold based on the following formula: ρ threshold = μ ρ - α·σ ρ Among them, ρ threshold Indicates the point cloud density threshold, μ ρ represents the average point cloud density of the region, σ ρ represents the standard deviation of point cloud density, and α represents the adjustment coefficient; Step S1.3: Perform a difference analysis on the cleaned point cloud data and the building information model to generate model correction parameters; Step S1.4: Update the building information model data based on the model correction parameters to make it consistent with the actual construction site state.

3. The digital display method for construction projects based on artificial intelligence according to claim 1, wherein The generation process of the feature encoding matrix in Step S2 includes: Step S2.1: Perform multi-scale voxelization processing on the building information model to generate hierarchical features including local geometric details and global structures. The multi-scale voxelization processing uses an octree recursive segmentation algorithm; Step S2.2: Calculate the feature weight values of each voxel region through the attention mechanism. The feature weight values are calculated according to the following formula: where, w i represents the feature weight value of the i-th voxel region; q i represents the query vector; K represents the key matrix; V represents the value matrix; d represents the feature dimension; Step S2.3: Dynamically adjust the convolution kernel parameters in the three-dimensional convolutional neural network according to the feature weight values. The convolution kernel parameters include the convolution kernel weight matrix, bias value, and stride parameter. The adjustment method is: Among them, represents the adjusted convolutional kernel weight; represents the original convolutional kernel weight; γ represents the adjustment factor; w i represents the feature weight value; Step S2.4: Fuse the shallow geometric features and deep semantic features through the residual connection structure to generate the feature encoding matrix.

4. The method for digital display of construction projects based on artificial intelligence according to claim 1, characterized in that, The construction method of the multi-objective constraint condition set in Step S3 includes: Step S3.1: Establish a mechanical constraint equation between material strength and structural stress. The mechanical constraint equation satisfies: σ max ≤φ·f y Among them, σ max represents the maximum stress; f y represents the yield strength of the material; φ represents the safety factor; Step S3.2: Construct a progress constraint relationship graph of construction processes and time nodes. The relationship graph is represented by a directed acyclic graph, where: G=(V,E,T) Among them, G represents the relational graph; V represents the set of construction processes; E represents the set of process dependencies; T represents the set of time constraints; for any process v i , v j ∈V, if process v i must be completed before process v j , then there is a directed edge e ij ∈E, and it satisfies the time constraint t j -t i ≥d i , where t i and t j respectively represent the start times of processes v i and v j ; d i represents the duration of process v i . Step S3.3: Set an economic constraint threshold for project cost and material usage; Step S3.4: Integrate the mechanical constraint equation, the progress constraint relationship graph, and the economic constraint threshold into a multi-objective constraint condition set.

5. The method for digital display of construction projects based on artificial intelligence according to claim 1, characterized in that, The specific dynamic model reconstruction described in step S4 includes: Step S4.1: Decode the features of the optimization parameter matrix to generate triangular patch data with material attributes; Step S4.2: Calculate the lighting effect and shadow information based on the physically based rendering algorithm, which includes physically based rendering path tracing and global illumination models; Step S4.3: Dynamically adjust the model detail level LOD according to the viewing distance; Step S4.4: Combine the triangular patch data with material attributes with the lighting effect and shadow information to generate the visualized 3D model with temporal attributes.

6. The method for digital display of construction projects based on artificial intelligence according to claim 1, characterized in that: The working method of the collision detection module described in step S5 includes: Step S5.1: Construct a bounding box hierarchy tree for all objects in the scene, and the bounding boxes in the bounding box hierarchy tree adopt an axis-aligned bounding box structure; Step S5.2: Use the separating axis theorem for preliminary collision screening. The condition for the separating axis theorem to determine that two convex polyhedra do not intersect is that there exists an axis such that the projections of the two objects on this axis do not overlap; Step S5.3: Perform triangle-level collision detection on the selected object pairs; Step S5.4: Feed the detected collision information back to the multi-objective constraint condition set through the parameter feedback module to trigger re-optimization.

7. The method for digital display of construction projects based on artificial intelligence according to claim 6, characterized in that, Performing triangle-level collision detection on the selected object pairs in step S5.3 includes the following sub-steps: Step S5.3.1: Organize the triangular patch data of the object into a spatial hash table structure to accelerate triangle search; Step S5.3.2: For each pair of potentially colliding triangles, calculate the distance from the triangle vertices to the plane of the other triangle; Step S5.3.3: Determine the collision depth by calculating the intersection position of the triangle edge and the other triangle; Step S5.3.4: Generate a collision data structure including the collision position, collision normal vector, and collision depth.

8. An artificial intelligence-based digital display system for construction projects, which is used to execute the artificial intelligence-based digital display method for construction projects described in claim 1, characterized in that, Including: A data acquisition module configured to obtain the model data and construction parameter set of the building information model; A feature extraction module including a 3D convolutional neural network for generating a feature encoding matrix; An optimization processing module configured to construct a multi-objective constraint condition set and perform parameter optimization; A visualization rendering module including a real-time rendering engine for generating a dynamic 3D model; An interactive display module, a WebGL display interface integrating collision detection functions.

9. The digital display system for construction projects based on artificial intelligence according to claim 8, wherein The feature extraction module includes: A voxelization processing unit for converting the building information model into a multi-scale voxel representation; An attention calculation unit configured with a deformable convolution kernel and a channel attention mechanism; a feature fusion unit for integrating local features and global context information.

10. A computer-readable storage medium, characterized in that, Stores a computer program, which when executed by a processor, implements the method steps described in any one of claims 1-7.

Images