Large building construction quality management method based on three-dimensional cloud model

Abstract

The application discloses a large building engineering construction quality management method based on a three-dimensional cloud model, relates to the technical field of building engineering, and comprises the following steps: using a collection device to collect high-precision three-dimensional point cloud data of a building construction site, and integrating the data collected by different devices through an automatic data fusion algorithm to ensure the comprehensiveness and high precision of the data. Through the multi-device synchronous collection and data fusion technology, the application realizes real-time acquisition of high-precision three-dimensional point cloud data, accurate monitoring of hidden areas, and avoidance of omissions in traditional inspection. Data preprocessing and enhancement ensure the completeness and high resolution of the point cloud, small deviations are found in time, the construction quality control precision is improved, and the risk of rework is reduced. Meanwhile, a data-driven management system dynamically captures construction changes, generates a deviation report and pushes for rectification, closed-loop feedback ensures that problems are completely solved, and the construction quality management efficiency is significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of construction engineering, and in particular to a large-scale construction engineering construction quality management method based on a three-dimensional cloud model. Background Art

[0002] Large-scale construction project quality management based on 3D cloud models utilizes 3D laser scanning technology, drone remote sensing, or other high-precision sensing equipment to collect 3D data from construction sites, generate detailed 3D point cloud models, and integrate them with Building Information Modeling (BIM) or other digital management platforms to achieve precise monitoring and quality assessment of the entire construction process. Through this model, project managers can check in real time whether the size, shape, and position of various parts of the construction site meet design requirements, and promptly identify potential quality issues such as structural deformation, construction deviations, or material defects. Furthermore, 3D cloud models can be integrated with data such as project progress and costs to help management teams make better decisions, optimize construction processes, improve construction quality, reduce rework and resource waste, and ensure that projects are completed on time and to quality standards.

[0003] The existing technology has the following deficiencies:

[0004] In the construction quality management of large-scale building projects based on 3D cloud models, the sparsity and insufficient resolution of 3D point cloud data may lead to serious consequences. Since acquisition equipment such as laser scanners or drones may be affected by environmental factors (such as lighting, weather, building obstructions) or equipment accuracy limitations in actual applications, point cloud data in certain areas may be sparse or missing. This problem of incomplete data may cause certain small but critical structural defects or deviations to be overlooked during the construction process, especially in hidden areas such as inside walls, underground pipelines or high-altitude structures. If these problems are not discovered in time, they may cause structural safety hazards in later construction or after completion, and even lead to engineering quality accidents, such as structural instability or safety accidents.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a large-scale construction project construction quality management method based on a three-dimensional cloud model. Through multi-device synchronous acquisition and data fusion technology, high-precision three-dimensional point cloud data can be obtained in real time, especially for accurate monitoring of hidden areas, to avoid omissions in traditional inspections. Combined with data preprocessing, reconstruction and enhancement, the integrity and high resolution of the point cloud are guaranteed, and small deviations and non-conformities are discovered and handled in a timely manner, thereby improving the accuracy of construction quality control and reducing rework costs and risks. At the same time, the data-driven management system dynamically captures construction changes, generates deviation reports through algorithmic analysis, and pushes rectification opinions. The closed-loop feedback mechanism ensures that rectification is thorough, significantly improving the refinement of construction quality management and decision-making efficiency, so as to solve the problems in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a large-scale construction project construction quality management method based on a three-dimensional cloud model, comprising the following steps:

[0008] Use acquisition equipment to collect high-precision 3D point cloud data at construction sites, and integrate data collected by different devices through automated data fusion algorithms to ensure comprehensiveness and high accuracy of the data;

[0009] Pre-process the collected 3D point cloud data to remove abnormal data and noise, and use multi-view reconstruction technology to fill in missing areas. Use data enhancement technology to improve the resolution and density of point cloud data to ensure that quality monitoring of key areas is not affected.

[0010] Based on the processed 3D point cloud data, a 3D digital model of the building structure is constructed and compared with the construction design drawings to identify deviations and non-conformities in construction. Key inspections are then conducted on hidden areas to ensure that potential quality risks are identified promptly.

[0011] During the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, timely capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management.

[0012] Deep learning and machine learning algorithms are used to intelligently analyze dynamically updated 3D point cloud data, automatically identifying and evaluating minor structural defects and deviations in hidden areas, and generating detailed risk reports so that construction personnel can take timely remedial measures to prevent potential structural safety hazards.

[0013] Preferably, data is collected using a collection device and the collected data is integrated to ensure the comprehensiveness and high accuracy of the data. The specific steps are as follows:

[0014] Before officially starting 3D point cloud data collection, it is necessary to first survey the construction site and assess the complexity of the site and environmental conditions;

[0015] After the equipment is arranged, the collection of high-precision 3D point cloud data begins synchronously;

[0016] After data collection is completed, the collected point cloud data is first stored and the relevant collection parameters of the equipment are saved to provide complete information for subsequent data fusion and analysis;

[0017] After data collection, an automated data fusion algorithm is used to integrate the point cloud data collected by different devices.

[0018] Preferably, the collected 3D point cloud data is pre-processed to improve the resolution and density of the point cloud data through data enhancement technology to ensure that the quality monitoring of key areas is not affected. The specific steps are as follows:

[0019] After completing the 3D point cloud data collection and integration, the data must first be cleaned to remove abnormal data and noise;

[0020] Multi-view reconstruction technology is used to fill in the problem of missing areas in the collected data due to the complexity of the construction site, equipment accuracy limitations and occlusion factors;

[0021] Improve data resolution and density to ensure high-precision quality monitoring in key areas;

[0022] After data enhancement technology improves the resolution and density of point clouds, it is necessary to ensure that quality monitoring of key areas during construction is not affected.

[0023] Preferably, a 3D digital model of the building structure is constructed based on the processed 3D point cloud data, and compared with the construction design drawings to identify deviations and non-conformities in construction, and to conduct key inspections on hidden areas to ensure that potential quality risks are identified in a timely manner. The specific steps are as follows:

[0024] After completing the collection and preprocessing of 3D point cloud data, a 3D digital model of the building structure is constructed based on the preprocessed data;

[0025] Compare and analyze the digital model built based on 3D point cloud data with the construction design drawings;

[0026] In order to ensure that the quality of hidden areas is effectively monitored, key inspections need to be performed based on the processed 3D point cloud data;

[0027] After completing comparative analysis and hidden area inspection, a deviation report on the construction process is automatically generated and fed back to construction management personnel in real time.

[0028] Preferably, during the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, timely capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management. The specific steps are as follows:

[0029] At the construction site, real-time monitoring equipment is used to continuously collect data on the construction progress. These devices collect point cloud data at regular intervals and continuously update the three-dimensional model of the construction site to capture changes and quality deviations during the construction process. The data collected each time is compared with the data at the previous moment to obtain dynamically updated three-dimensional point cloud data. The update process expression is as follows: t+1 =P t +ΔP t , where P t is the point cloud data at time point t, ΔP t is the data change, P t+1 It is the updated 3D point cloud data;

[0030] By comparing dynamically updated point cloud data with construction design drawings, deviations during the construction process can be identified. Deviation detection not only involves changes in geometric form but also needs to consider the construction progress. Real-time deviation detection is performed in conjunction with the construction progress. The calculation expression is as follows:

[0031]

[0032] , where P t，j is the updated j-th point cloud data point, D j is the jth data point in the corresponding construction design drawing, N is the total number of data points, δ t is the overall deviation at time point t;

[0033] Based on deviation detection, combined with construction progress and quality control standards, a risk assessment model is used to conduct real-time risk assessment of the construction process. If a deviation in a construction area is detected that exceeds the preset tolerance threshold, an early warning mechanism is activated to notify the construction team and remote management personnel. The risk assessment calculation expression is as follows:

[0034]

[0035] , where R t is the construction risk index at time point t, S t is the current construction progress, S max is the maximum construction progress, is the progress target in the preset construction plan, α is the dynamic adjustment coefficient, and β is the progress deviation weighting coefficient;

[0036] Finally, all deviations, risk assessments, and construction progress data will be aggregated into the remote management platform. Combining construction progress and real-time risk assessment, the platform automatically generates early warning reports, provides recommended corrective measures, and assists project managers in making subsequent decisions. The remote management decision support formula is as follows: C t =f(R t , S t , Z t ), where C t is the decision support value, Z t It is the priority parameter of project management.

[0037] Preferably, deep learning and machine learning algorithms are used to intelligently analyze the dynamically updated 3D point cloud data to generate a detailed risk report for construction personnel to take timely remedial measures to prevent potential structural safety hazards. The calculation steps are as follows:

[0038] Before conducting deep learning and machine learning analysis, it is necessary to preprocess the dynamically updated three-dimensional point cloud data and extract effective feature information. The point cloud data is denoised using statistical analysis methods, and the data is segmented using the K-means clustering algorithm to extract the features of the hidden area. The point cloud data is represented as P = {p i}={p1, p2, p3,..., p n}, where point p i ={x i 、y i 、z i} is a point in three-dimensional space, n is the total number of points in the point cloud data, and the density expression of the point cloud determined by statistical methods is as follows:

[0039]

[0040] , where ρ(p) is the density of point p, which measures the density of the point cloud around the target point, p i is the position of the i-th point in the data set, indicating p i ={ xi 、y i 、z i},δ(||p i -p||) is the weight function of distance;

[0041] We use deep learning algorithms to train pre-processed point cloud data to automatically identify and learn small structural defects and deviations in hidden areas. We use convolutional neural networks to extract spatial features from point cloud data and incorporate self-attention mechanisms to handle small structural changes in hidden areas. Deep learning model training optimizes the loss function through backpropagation to achieve optimal defect detection results. The training process is defined as:

[0042]

[0043] , where Q is the size of the dataset, y i is the true label of the i-th point in the dataset, θ is the set of all training parameters in the deep learning model, and Φ is the total loss function, which is the weighted sum of the prediction errors of all points;

[0044] After identifying potential defects and deviations, a machine learning algorithm is used to classify the detected defects and evaluate their impact on the overall structural safety. Based on the trained model, a risk score is generated by combining the severity, location, and surrounding environment of each defect. The calculation expression is as follows:

[0045]

[0046] , where R risk is the risk score, A i is the weight coefficient of the point cloud data point, the weight coefficient corresponding to the i-th point, is the existence indicator function of the point cloud data point, B i is the influence factor of the point cloud data point, indicating the influence factor of the i-th point, p design is the position of the predetermined point in the design drawing, ||p i -p design || is the deviation of the point, which represents the distance between the i-th point and its designed position;

[0047] Finally, based on the risk assessment results, a detailed risk report is generated. The report includes the defect type, location, deviation degree, and risk to the overall building structure of each hidden area. The formula for generating the risk report is as follows:

[0048]

[0049] , where Report is the risk report, is the risk score of the j-th defect, θ threshold is the preset risk threshold, is the predicted type of the jth defect, and M is the total number of defects.

[0050] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0051] The present invention uses multi-device synchronous acquisition and data fusion technology to obtain high-precision three-dimensional point cloud data in real time and comprehensively during the construction process, avoiding details that may be missed in traditional inspection methods, especially the monitoring of hidden areas (such as the inside of walls, underground pipelines, etc.). In the process of data preprocessing, reconstruction and enhancement, the integrity and high resolution of the point cloud data are guaranteed, so that small deviations and non-conformities in the building structure can be discovered and processed in a timely manner. This process can greatly improve the control accuracy of construction quality, especially in the early stages of construction, it can discover potential quality risks, avoiding engineering quality problems that were previously caused by delayed or incomplete inspections. By promptly discovering problems and making corrections, the cost and risk of rework in the later stage are effectively reduced.

[0052] The present invention establishes a data-driven construction quality management system through real-time three-dimensional point cloud data collection and comparison with design drawings, which can carry out refined management of each construction stage. By continuously collecting and updating point cloud data, the system can dynamically capture every detail change in the construction, and combined with algorithm analysis, it automatically generates a report for any deviation or non-conformity that occurs during the construction process and pushes it to the relevant personnel. This data-driven analysis and feedback mechanism greatly improves the decision-making efficiency during the construction process, reduces human errors and omissions, and enables project managers and construction personnel to respond to various changes on the construction site in a timely manner, thereby ensuring that the entire project is carried out efficiently and in compliance with the predetermined plan. In addition, the introduction of a closed-loop feedback mechanism ensures quality inspection after each problem rectification, ensures the thorough resolution of the problem, and improves the overall construction quality and management level of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0054] Figure 1 This is a flow chart of the method for large-scale construction project construction quality management method based on three-dimensional cloud model of the present invention. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0056] The present invention provides Figure 1The large-scale construction project construction quality management method based on the 3D cloud model includes the following steps:

[0057] Use acquisition equipment to collect high-precision 3D point cloud data at construction sites, and integrate data collected by different devices through automated data fusion algorithms to ensure comprehensiveness and high accuracy of the data;

[0058] Use acquisition equipment to collect data and integrate the collected data to ensure the comprehensiveness and high accuracy of the data. The specific steps are as follows:

[0059] Before officially starting 3D point cloud data collection, it is necessary to first survey the construction site and assess the complexity of the site and environmental conditions;

[0060] Plan the placement and scanning paths of each device based on the site conditions. When arranging acquisition equipment, ensure that it covers all key areas of the site and consider the overlap between devices' fields of view. This allows for effective registration of data collected by different devices during subsequent data fusion. Furthermore, the configuration and adjustment of different devices (such as laser scanners and drones) should be tailored to the specific acquisition task, including selecting appropriate parameters such as resolution, scanning angle, and scanning distance.

[0061] After the equipment is arranged, the collection of high-precision 3D point cloud data begins synchronously;

[0062] Devices such as laser scanners use laser beams to scan buildings and their surroundings, generating large amounts of point cloud data. Different devices may generate different types of point cloud data depending on their operating principles (for example, laser scanners generate precise three-dimensional coordinate data, while drones obtain point cloud data through photogrammetry). During the data collection process, it is crucial to ensure the synchronization and stability of the devices to avoid errors introduced by different collection times between devices. Synchronous collection can be achieved through the device's built-in timestamps, positioning systems (such as GPS, IMU, etc.), and external control systems to ensure that the data collected by multiple devices is comparable.

[0063] After data collection is completed, the collected point cloud data is first stored and the relevant collection parameters of the equipment are saved to provide complete information for subsequent data fusion and analysis;

[0064] During this phase, the data collected by each device can be initially calibrated. This includes checking the device's calibration status and confirming its offset and measurement accuracy. This ensures that data integration is not distorted by device errors. Furthermore, if the collection methods used by different devices vary significantly, the data source and category must be recorded to facilitate more accurate data integration.

[0065] After data collection, an automated data fusion algorithm is used to integrate the point cloud data collected by different devices;

[0066] In this process, the point cloud data collected by different devices are first preliminarily aligned and registered through algorithms to eliminate spatial differences between devices. This is usually done using feature matching-based algorithms (for example, consensus algorithms and iterative closest point (ICP) algorithms) to accurately match device data to ensure that overlapping areas between data are correctly matched. Secondly, in the process of data fusion, the algorithm also needs to weight the data from different devices, taking into account the different accuracy, scanning angles and coverage of the devices, and finally merge the data collected by all devices into a globally consistent three-dimensional point cloud model. The key at this stage is to ensure that the final integrated data is seamlessly connected, continuous and consistent in space, thereby providing accurate and reliable basic data for subsequent quality management and hidden danger detection.

[0067] Pre-process the collected 3D point cloud data to remove abnormal data and noise, and use multi-view reconstruction technology to fill in missing areas. Use data enhancement technology to improve the resolution and density of point cloud data to ensure that quality monitoring of key areas is not affected.

[0068] Pre-process the collected 3D point cloud data and use data enhancement technology to improve the resolution and density of the point cloud data to ensure that the quality monitoring of key areas is not affected. The specific steps are as follows:

[0069] After completing the 3D point cloud data collection and integration, the data must first be cleaned to remove abnormal data and noise;

[0070] Since point cloud data may be affected by various factors during the collection process, such as environmental interference, equipment errors, lighting changes, weather conditions, etc., some noise points that do not conform to the actual scene (such as isolated points or outliers) may be generated. This type of data will affect the accuracy of subsequent analysis, so it is necessary to filter it through an outlier detection algorithm. Commonly used algorithms include density-based spatial clustering (such as DBSCAN), statistical-based outlier detection (such as RANSAC algorithm), etc. These algorithms can automatically detect and delete noise data that is significantly different from other point cloud data or has an unreasonable position. In addition, abnormal data removal can also be combined with equipment parameters (such as scanning angle, distance range, etc.) for correction to further ensure the quality of point cloud data.

[0071] Multi-view reconstruction technology is used to fill in the problem of missing areas in the collected data due to the complexity of the construction site, equipment accuracy limitations and occlusion factors;

[0072] By fusing point cloud data collected from different perspectives and devices, and utilizing image processing-based depth inference and multi-perspective 3D reconstruction algorithms (such as stereo matching and disparity map generation), point cloud information for missing areas can be estimated. This process not only relies on traditional point cloud registration methods, but also requires the use of point cloud data from adjacent areas for inference and restoration. By integrating data from different perspectives, multi-perspective reconstruction technology can more accurately restore the three-dimensional form of the construction site from various angles, thereby filling data gaps caused by occlusion or other reasons and ensuring the integrity of the 3D data of the entire building.

[0073] Improve data resolution and density to ensure high-precision quality monitoring in key areas;

[0074] Data enhancement techniques are applied to increase the density and resolution of point clouds. These techniques include spatially refining point cloud data through interpolation algorithms (such as interpolation based on the least squares method), multi-scale processing methods (such as multi-level reconstruction using data of different accuracy levels), and super-resolution reconstruction algorithms (such as depth map enhancement technology generated by convolutional neural networks), thereby improving accuracy and density. In particular, in areas with complex details, improving resolution through enhancement technology not only provides more detailed data but also ensures the accuracy of construction quality monitoring, preventing minor structural deviations from being detected due to insufficient resolution.

[0075] After data augmentation technology improves the resolution and density of point clouds, it is necessary to ensure that quality monitoring of key areas during construction is not affected;

[0076] By utilizing enhanced point cloud data, precise quality analysis and deviation detection are performed on the construction site, with a particular focus on hidden structural areas (such as underground pipelines, wall interiors, and elevated structures). By comparing with construction drawings, any minor construction deviations or quality risks can be accurately analyzed, allowing potential quality issues to be discovered promptly. For example, subtle structural issues within a wall can be accurately detected using enhanced high-resolution point cloud data, avoiding potential safety hazards that could impact the structure. This series of pre-processing steps ensures that every critical area during the construction process is effectively monitored, ultimately guaranteeing the overall quality and safety of the project.

[0077] Based on the processed 3D point cloud data, a 3D digital model of the building structure is constructed and compared with the construction design drawings to identify deviations and non-conformities in construction. Key inspections are then conducted on hidden areas to ensure that potential quality risks are identified promptly.

[0078] Based on the processed 3D point cloud data, a 3D digital model of the building structure is constructed and compared with the construction design drawings to identify deviations and non-conformities in construction. Hidden areas are inspected to ensure that potential quality risks are identified in a timely manner. The specific steps are as follows:

[0079] After completing the collection and preprocessing of 3D point cloud data, a 3D digital model of the building structure is constructed based on the preprocessed data;

[0080] Through point cloud to mesh model conversion, the collected point cloud data is converted into a three-dimensional mesh model suitable for analysis and operation. This process is usually completed through surface reconstruction algorithms (such as Poisson reconstruction algorithm or triangular mesh generation algorithm) to form accurate building geometry. Especially in complex structural areas (such as roofs, pipes, interior walls, etc.), more sophisticated reconstruction techniques are used to ensure the accuracy of details. Using these digital models, the spatial layout of the building and the geometric characteristics of its various parts can be accurately described, providing a solid data foundation for subsequent construction quality analysis, deviation detection and hidden area inspection.

[0081] Compare and analyze the digital model built based on 3D point cloud data with the construction design drawings;

[0082] Using computer-aided design (CAD) software or a building information modeling (BIM) system, construction design drawings are converted into three-dimensional models that match the point cloud data. Model comparison and analysis tools then automatically detect deviations between the design drawings and the actual construction, identifying areas that don't meet design standards or construction specifications. This process involves comparing each structural component of the building (such as walls, beams, columns, and floor slabs) to identify dimensional deviations, positional discrepancies, or construction errors. This comparative analysis helps identify potential problems during construction and prevent quality incidents or potential problems later in the process.

[0083] To ensure that the quality of hidden areas (such as the interior of walls, underground pipelines, ceilings and other high structures) is effectively monitored, it is necessary to conduct key inspections based on the processed 3D point cloud data;

[0084] Utilizing digital models, these hidden areas can be sliced ​​through virtual cutting and cross-section analysis techniques, revealing information such as the internal structure of the wall, the layout of underground pipelines, and the accuracy of overhead structures. Furthermore, through local refinement, the resolution of these areas can be improved, ensuring that even minor deviations can be promptly detected and corrected. Inspections of hidden areas should not only include testing structural accuracy but also the layout of hidden facilities such as pipelines and wiring, preventing construction quality issues from being overlooked and ensuring that the quality of hidden projects meets standards.

[0085] After completing comparative analysis and hidden area inspection, deviation reports during construction are automatically generated and provided to construction management personnel in real time;

[0086] These deviation reports detail the error values, potential quality risks, and areas requiring rectification for each area, enabling construction personnel to take timely corrective measures. For example, if a wall position deviation or pipeline layout in a certain area is found to not meet design requirements, the system will issue an early warning, prompting the engineer or project manager to conduct an on-site inspection and repair. This real-time feedback mechanism not only helps the construction team identify problems promptly but also provides data support for subsequent project quality monitoring. By comparing with design drawings and combining precise point cloud data, construction quality can be effectively ensured, preventing potential hazards from affecting the long-term stability and safety of the building.

[0087] During the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, timely capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management.

[0088] During the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, promptly capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management. The specific steps are as follows:

[0089] At the construction site, real-time monitoring equipment is used to continuously collect data on the construction progress. These devices collect point cloud data at regular intervals and continuously update the three-dimensional model of the construction site to capture changes and quality deviations during the construction process. The data collected each time is compared with the data at the previous moment to obtain dynamically updated three-dimensional point cloud data. The update process expression is as follows: t+1 =P t +ΔP t , where P t It is the point cloud data representing the time point t, which usually contains a large amount of spatial coordinate information and additional data such as reflection intensity. These data constitute the three-dimensional model of the building or construction site, including the precise location and attributes of each measurement point, ΔP t is the data change, which represents the change of the 3D point cloud data at the construction site from time point t to time point t+1. t+1 It is the updated 3D point cloud data, which means the updated 3D point cloud data at time t+1, including all changes and new measurement information between time point t and time point t+1;

[0090] This process ensures that the 3D point cloud data of the construction site can reflect the actual situation at each moment, providing the latest data support for subsequent analysis.

[0091] By comparing dynamically updated point cloud data with construction design drawings, deviations during the construction process can be identified. Deviation detection not only involves changes in geometric form but also needs to consider the construction progress. Real-time deviation detection is performed in conjunction with the construction progress. The calculation expression is as follows:

[0092]

[0093] , where P t，j is the updated j-th point cloud data point, D j is the jth data point in the corresponding construction design drawing, N is the total number of data points, δ t is the overall deviation at time point t;

[0094] Through the above formula, the deviation value can be calculated in real time during the construction process to determine whether the construction is carried out according to the design plan, and provide a basis for the next warning.

[0095] Based on deviation detection, combined with construction progress and quality control standards, a risk assessment model is used to conduct real-time risk assessment of the construction process. If a deviation in a construction area is detected that exceeds the preset tolerance threshold, an early warning mechanism is activated to notify the construction team and remote management personnel. The risk assessment calculation expression is as follows:

[0096]

[0097] , where R t is the construction risk index at time point t, S t is the current construction progress, S max is the maximum construction progress, which is the progress target in the preset construction plan. α is the dynamic adjustment coefficient, which is used to adjust the impact of construction quality deviation on the final result, reflecting the influence of quality problems over time. β is the progress deviation weighting coefficient, which is used to express the construction progress deviation (i.e. ) on the quality monitoring results, measuring the indirect impact of progress deviation on construction quality, that is, the potential impact of delay or advance in construction progress on quality deviation;

[0098] By calculating the construction risk index R t ,If the risk value exceeds the set threshold, the system will trigger the early warning mechanism, notifying on-site managers to deal with potential problems in a timely manner.

[0099] Finally, all deviations, risk assessments, and construction progress data will be aggregated into the remote management platform. Combining construction progress and real-time risk assessment, the platform automatically generates early warning reports, provides recommended corrective measures, and assists project managers in making subsequent decisions. The remote management decision support formula is as follows: Ct =f(R t , S t , Z t ), where C t is the decision support value, Z t It is the priority parameter of project management;

[0100] Based on this function, the platform automatically generates management decision recommendations, such as adjusting construction plans, strengthening quality control measures, or initiating on-site remediation work. Through this decision-support model, remote managers can quickly identify potential risks during construction and make targeted decisions, effectively improving project quality control and management efficiency.

[0101] Using deep learning and machine learning algorithms, the system intelligently analyzes dynamically updated 3D point cloud data, automatically identifying and evaluating minor structural defects and deviations in hidden areas. It also generates detailed risk reports, enabling construction personnel to take timely remedial measures to prevent potential structural safety hazards.

[0102] Deep learning and machine learning algorithms are used to intelligently analyze dynamically updated 3D point cloud data to generate detailed risk reports, allowing construction personnel to take timely remedial measures to prevent potential structural safety hazards. The calculation steps are as follows:

[0103] Before deep learning and machine learning analysis, it is necessary to preprocess the dynamically updated three-dimensional point cloud data and extract effective feature information. Statistical analysis methods (such as point cloud density analysis based on Gaussian distribution) are used to denoise the point cloud data, and the K-means clustering algorithm is used to segment the data and extract the features of the hidden area. The point cloud data is represented as P = {p i}={p1, p2, p3,..., p n}, where point p i ={x i 、y i 、z i} is a point in three-dimensional space, n is the total number of points in the point cloud data, and the density expression of the point cloud determined by statistical methods is as follows:

[0104]

[0105] , where ρ(p) is the density of point p, which measures the density of the point cloud around the target point, p i is the position of the i-th point in the data set, indicating p i ={x i 、y i 、z i},δ(||p i-p||) is the distance weight function, which controls the influence of each point on the target point density, and is usually attenuated according to the distance between the point and the target point;

[0106] We use deep learning algorithms to train pre-processed point cloud data to automatically identify and learn small structural defects and deviations in hidden areas. We use convolutional neural networks (CNNs) to extract spatial features from point cloud data and incorporate a self-attention mechanism to handle small structural changes in hidden areas. Deep learning model training optimizes the loss function through backpropagation to achieve optimal defect detection results. The training process is defined as:

[0107]

[0108] , where Q is the size of the dataset, y i is the true label of the i-th point in the dataset, θ is the set of all training parameters in the deep learning model, which is continuously adjusted by the optimization algorithm during the model training process, ultimately enabling the model to accurately process the input data and make predictions, and Φ is the total loss function, which is the weighted sum of the prediction errors of all points;

[0109] After identifying potential defects and deviations, machine learning algorithms (such as support vector machines (SVMs) or decision tree algorithms) are used to classify the detected defects and evaluate their impact on the overall structural safety. Based on the trained model, a risk score is generated by combining the severity, location, and surrounding environment of each defect. The calculation expression is as follows:

[0110]

[0111] , where R risk is the risk score, A i is the weight coefficient of the point cloud data point, the weight coefficient corresponding to the i-th point (i.e. a point in the three-dimensional point cloud data), It is the existence indicator function of the point cloud data point. When the i-th point exists in the point cloud data and is identified as a potential defect by the intelligent algorithm, it returns 1, otherwise it returns 0. i is the influence factor of the point cloud data point, which represents the influence factor of the i-th point and quantifies the possible impact of the deviation of this point on the safety of the entire structure. design is the position of the predetermined point in the design drawing, ||p i -p design || is the deviation of the point, which represents the distance between the i-th point and its designed position, and represents the deviation of the point relative to the design drawing;

[0112] Finally, based on the risk assessment results, a detailed risk report is generated. The report includes the defect type, location, deviation degree, and risk to the overall building structure of each hidden area. The formula for generating the risk report is as follows:

[0113]

[0114] , where Report is the risk report, is the risk score of the jth defect, which indicates the degree of threat to the safety of the building structure. threshold is the preset risk threshold, is the predicted type of the j-th defect, identified by the deep learning model through analysis of point cloud data, such as "crack" and "dislocation", and M is the total number of defects.

[0115] The generated report can visualize structural risks and provide them to the construction team, helping them to promptly address potential hidden dangers during the construction process and prevent serious quality problems in the building structure.

[0116] Implementation Method 1: Ensuring comprehensive and high-precision 3D point cloud data is crucial at construction sites, especially large-scale projects. Traditional single-device acquisition methods may have limitations, such as limited field of view, low accuracy, and slow acquisition speed. Therefore, this implementation method utilizes a multi-device synchronous acquisition and data fusion strategy to ensure comprehensive and high-precision data collection.

[0117] First, a variety of equipment, including high-precision laser scanners, drones (drone-mounted lidar), and mobile laser scanning devices, are used to simultaneously collect all-round, multi-angle three-dimensional point cloud data at the construction site. These devices are characterized by their ability to cover construction areas at different heights and orientations. For example, laser scanners can accurately measure point cloud data of the ground and building facades, while drones can cover high-rise structures and hard-to-reach areas, ensuring that point cloud data in all key areas is accurately captured. At the same time, mobile laser scanning equipment can be used to dynamically scan changing structures during construction, ensuring that data can be updated at all times during the construction phase.

[0118] Secondly, the point cloud data collected from different devices will have errors and deviations due to device differences. At this time, it is necessary to integrate this data through an automated data fusion algorithm to ensure the uniformity and consistency of data from different sources. Common point cloud data fusion algorithms include registration methods based on coordinate transformation and ICP (iterative closest point) algorithms. These algorithms can eliminate the spatial displacement caused by position differences between devices during the collection process, align all data through precise mathematical models, and generate a unified point cloud model. In this way, data from different devices can be perfectly fused in the same coordinate system, ensuring the comprehensiveness and high precision of the data.

[0119] The fused 3D point cloud data serves as the foundation for conversion into a 3D digital model of the building through surface reconstruction technology. During this process, the point cloud data is converted into a surface mesh through algorithms, forming a digital model that captures the building's geometry. In construction projects, these digital models not only accurately reflect the geometric characteristics of the building structure but also capture detailed changes during construction, providing a visual foundation for subsequent construction deviation analysis.

[0120] The completed digital model must also be compared with the construction design drawings. Through the computer-aided design (CAD) system or building information modeling (BIM) technology, the construction drawings are converted into three-dimensional digital models, and compared and analyzed with the model constructed by point cloud data. This process can automatically identify deviations or non-conformities, especially subtle deviations in size, position and shape. Where the design drawings do not match the actual construction situation, the system will automatically mark the areas that need correction and generate actionable rectification suggestions. The core advantage of this implementation method is that through three-dimensional model comparison, deviations and problems in the construction process can be discovered at the first time, ensuring that the quality of the project is effectively controlled at every link.

[0121] Implementation 2: Although 3D point cloud data acquired by devices like laser scanners and drones can cover the entire construction site, in practice, point cloud data often exhibits sparseness, excessive noise, and missing areas. To ensure high accuracy of this data, especially providing sufficient resolution and density in critical areas (such as hidden structures and detailed areas), this implementation utilizes multi-view reconstruction and data enhancement techniques to optimize details and ensure data quality and accuracy.

[0122] First, multi-view reconstruction technology is used to fill in missing areas in point cloud data. Due to the complexity of construction sites, some areas, such as wall interiors, complex pipeline layouts, or elevated structures, may be missing data due to occlusion or limited acquisition angles. To address this, multi-view reconstruction technology uses data reconstruction algorithms from multiple acquisition angles to effectively fuse point cloud data from different perspectives, restoring previously invisible areas. This process is typically accomplished through methods such as stereo matching and depth map reconstruction to ensure that the geometry of all areas is accurately captured.

[0123] On this basis, the density of point cloud data in key areas is enhanced. Using deep learning technology, a model is trained using large amounts of point cloud data to predict and generate point cloud distribution in low-density areas. This process significantly improves the resolution of point cloud data, especially in detailed areas such as pipes, window frames, and wall joints. The enhanced algorithm increases the point cloud density, providing more refined reference data for subsequent deviation detection. This enhanced point cloud data enables millimeter-level accuracy in analyzing minor structural deviations, ensuring that no potential quality risks are missed.

[0124] The enhanced point cloud data is used to generate a high-precision 3D model using surface reconstruction technology. This is then combined with automated analysis tools and BIM to detect deviations. This allows for automatic identification of minor deviations and design discrepancies during construction, providing immediate feedback to construction personnel, enabling more refined and comprehensive quality monitoring.

[0125] Implementation Method 3: To ensure that all construction quality issues are promptly identified and addressed throughout the construction process, this implementation method proposes a closed-loop feedback system for real-time construction monitoring and deviation analysis. By collecting point cloud data from the construction site in real time and comparing it with the design drawings and construction progress, deviations can be identified and corrective measures can be proposed, ensuring that project quality remains within controllable limits.

[0126] First, during the construction process, mobile laser scanners and drones continuously collect real-time data. These devices continuously update the 3D point cloud data of the construction site, capturing changes and updates to the on-site structure. This process is integrated with the construction progress management system to ensure that each data collection is synchronized with the construction progress. Through real-time data streaming, construction managers can obtain the latest data for comparison with design drawings and building models, ensuring that every detail of the construction process is reviewed.

[0127] The collected real-time data is then analyzed for deviations from the design drawings. Automated analysis tools compare the 3D point cloud data of each construction step with the 3D model of the design drawings, automatically detecting any deviations and non-conformities. Particularly in hidden areas (such as underground pipelines, wall interiors, and elevated structures), the system uses virtual sections and local magnification technology for focused monitoring to ensure that the construction quality in these areas is fully verified. If deviations are detected, the system generates a real-time report and sends it to the construction team and project manager, ensuring timely feedback on construction issues.

[0128] Finally, a closed-loop feedback mechanism enables continuous improvement in construction quality management. After each deviation is detected, the system updates the model data based on the corrective actions taken, providing real-time feedback to the construction site. Once the issue is rectified and fixed, new point cloud data is collected again and compared with the previous data to verify the repair. This closed-loop system of continuous monitoring, feedback, and rectification ensures real-time control of construction quality, preventing quality issues from being overlooked or delayed.

[0129] The present invention uses multi-device synchronous acquisition and data fusion technology to obtain high-precision three-dimensional point cloud data in real time and comprehensively during the construction process, avoiding details that may be missed in traditional inspection methods, especially the monitoring of hidden areas (such as the inside of walls, underground pipelines, etc.). In the process of data preprocessing, reconstruction and enhancement, the integrity and high resolution of the point cloud data are guaranteed, so that small deviations and non-conformities in the building structure can be discovered and processed in a timely manner. This process can greatly improve the control accuracy of construction quality, especially in the early stages of construction, it can discover potential quality risks, avoiding engineering quality problems that were previously caused by delayed or incomplete inspections. By promptly discovering problems and making corrections, the cost and risk of rework in the later stage are effectively reduced.

[0130] The present invention establishes a data-driven construction quality management system through real-time three-dimensional point cloud data collection and comparison with design drawings, which can carry out refined management of each construction stage. By continuously collecting and updating point cloud data, the system can dynamically capture every detail change in the construction, and combined with algorithm analysis, it automatically generates a report for any deviation or non-conformity that occurs during the construction process and pushes it to the relevant personnel. This data-driven analysis and feedback mechanism greatly improves the decision-making efficiency during the construction process, reduces human errors and omissions, and enables project managers and construction personnel to respond to various changes on the construction site in a timely manner, thereby ensuring that the entire project is carried out efficiently and in compliance with the predetermined plan. In addition, the introduction of a closed-loop feedback mechanism ensures quality inspection after each problem rectification, ensures the thorough resolution of the problem, and improves the overall construction quality and management level of the project.

[0131] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A large-scale construction project construction quality management method based on a three-dimensional cloud model, characterized in that: The following steps are involved: Use acquisition equipment to collect high-precision 3D point cloud data at construction sites, and integrate data collected by different devices through automated data fusion algorithms to ensure comprehensiveness and high accuracy of the data; Pre-process the collected 3D point cloud data to remove abnormal data and noise, and use multi-view reconstruction technology to fill in missing areas. Use data enhancement technology to improve the resolution and density of point cloud data to ensure that quality monitoring of key areas is not affected. Based on the processed 3D point cloud data, a 3D digital model of the building structure is constructed and compared with the construction design drawings to identify deviations and non-conformities in construction. Key inspections are then conducted on hidden areas to ensure that potential quality risks are identified promptly. During the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, timely capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management. Using deep learning and machine learning algorithms, the system intelligently analyzes dynamically updated 3D point cloud data, automatically identifying and evaluating minor structural defects and deviations in hidden areas. It also generates detailed risk reports, enabling construction personnel to take timely remedial measures to prevent potential structural safety hazards. During the construction process, real-time monitoring and data analysis algorithms are used to dynamically update the 3D point cloud data of the construction site, promptly capture quality deviations during the construction process, and conduct comparative analysis based on the construction progress to achieve real-time early warning and remote management. The specific steps are as follows: At the construction site, real-time monitoring equipment is used to continuously collect data on the construction progress. These devices collect point cloud data at regular intervals and continuously update the 3D model of the construction site to capture changes and quality deviations during the construction process. The data collected at each time is compared with the data at the previous moment to obtain dynamically updated 3D point cloud data. The update process is expressed as follows: , where It indicates a point in time Point cloud data at each moment, is the amount of data change, It is the updated 3D point cloud data; By comparing dynamically updated point cloud data with construction design drawings, deviations during the construction process can be identified. Deviation detection not only involves changes in geometric form but also needs to consider the construction progress. Real-time deviation detection is performed in conjunction with the construction progress. The calculation expression is as follows: , where It is the updated Point cloud data points, It corresponds to the first data points, is the total number of data points, It's time Overall deviation of the moment; Based on deviation detection, combined with construction progress and quality control standards, a risk assessment model is used to conduct real-time risk assessment of the construction process. If a deviation in a construction area is detected that exceeds the preset tolerance threshold, an early warning mechanism is activated to notify the construction team and remote management personnel. The risk assessment calculation expression is as follows: , where It's time The construction risk index at the moment, The current construction progress. It is the maximum construction progress, which is the progress target in the preset construction plan. is the dynamic adjustment coefficient, is the weighting coefficient of schedule deviation; Finally, all deviations, risk assessments, and construction progress data will be aggregated into the remote management platform. Combining construction progress and real-time risk assessments, the platform automatically generates early warning reports, provides recommended corrective measures, and assists project managers in making subsequent decisions. The remote management decision support formula is as follows: , where is the decision support value, It is the priority parameter of project management; Deep learning and machine learning algorithms are used to intelligently analyze dynamically updated 3D point cloud data to generate detailed risk reports, allowing construction personnel to take timely remedial measures to prevent potential structural safety hazards. The calculation steps are as follows: Before conducting deep learning and machine learning analysis, it is necessary to preprocess the dynamically updated 3D point cloud data and extract effective feature information. The point cloud data is denoised using statistical analysis methods and the data is segmented using the K-means clustering algorithm to extract the features of hidden areas. The point cloud data is represented as , where point is a point in three-dimensional space. is the total number of points in the point cloud data. The density expression of the point cloud is determined by statistical methods as follows: , where Yes The density of the point cloud around the target point is measured. It is the first The position of the point, indicating , is the weight function of distance; The deep learning algorithm is used to train pre-processed point cloud data to automatically identify and learn small structural defects and deviations in hidden areas. A convolutional neural network is used to extract spatial features from the point cloud data, and a self-attention mechanism is combined to process small structural changes in hidden areas. The deep learning model is trained by optimizing the loss function through backpropagation to achieve the best defect detection effect. The training process is defined as: , where is the dataset size, It is the first The true label of each point, is the set of all training parameters in the deep learning model, is the total loss function, which is the weighted sum of all point prediction errors; After identifying potential defects and deviations, a machine learning algorithm is used to classify the detected defects and evaluate their impact on the overall structural safety. Based on the trained model, a risk score is generated by combining the severity, location, and surrounding environment of each defect. The calculation expression is as follows: , where is the risk score, is the weight coefficient of the point cloud data point, The weight coefficient corresponding to each point, is the existence indicator function of the point cloud data point, is the influence factor of the point cloud data point, indicating the The impact factor of each point, It is the position of the predetermined point in the design drawing. is the deviation of the point, indicating the The distance between a point and its designed position; Finally, based on the risk assessment results, a detailed risk report is generated. The report includes the defect type, location, deviation degree, and risk to the overall building structure of each hidden area. The formula for generating the risk report is as follows: , where It is a risk report. It is Risk score for each defect, is the preset risk threshold, It is The predicted type of defect, is the total number of defects.

2. The large-scale construction project construction quality management method based on a three-dimensional cloud model according to claim 1 is characterized in that: Use acquisition equipment to collect data and integrate the collected data to ensure the comprehensiveness and high accuracy of the data. The specific steps are as follows: Before officially starting 3D point cloud data collection, it is necessary to first survey the construction site and assess the complexity of the site and environmental conditions; After the equipment is arranged, the collection of high-precision 3D point cloud data begins synchronously; After data collection is completed, the collected point cloud data is first stored and the relevant collection parameters of the equipment are saved to provide complete information for subsequent data fusion and analysis; After data collection, an automated data fusion algorithm is used to integrate the point cloud data collected by different devices.

3. The large-scale construction project construction quality management method based on a three-dimensional cloud model according to claim 1 is characterized in that: Pre-process the collected 3D point cloud data and use data enhancement technology to improve the resolution and density of the point cloud data to ensure that the quality monitoring of key areas is not affected. The specific steps are as follows: After completing the 3D point cloud data collection and integration, the data must first be cleaned to remove abnormal data and noise; Multi-view reconstruction technology is used to fill in the problem of missing areas in the collected data due to the complexity of the construction site, equipment accuracy limitations and occlusion factors; Improve data resolution and density to ensure high-precision quality monitoring in key areas; After data enhancement technology improves the resolution and density of point clouds, it is necessary to ensure that quality monitoring of key areas during construction is not affected.

4. The large-scale construction project construction quality management method based on a three-dimensional cloud model according to claim 1 is characterized in that: Based on the processed 3D point cloud data, a 3D digital model of the building structure is constructed and compared with the construction design drawings to identify deviations and non-conformities in construction. Hidden areas are inspected to ensure that potential quality risks are identified in a timely manner. The specific steps are as follows: After completing the collection and preprocessing of 3D point cloud data, a 3D digital model of the building structure is constructed based on the preprocessed data; Compare and analyze the digital model built based on 3D point cloud data with the construction design drawings; In order to ensure that the quality of hidden areas is effectively monitored, key inspections need to be performed based on the processed 3D point cloud data; After completing comparative analysis and hidden area inspection, a deviation report on the construction process is automatically generated and fed back to construction management personnel in real time.

Images