Construction progress monitoring method and system based on big data

Abstract

The invention relates to the technical field of construction progress monitoring, and discloses a construction progress monitoring method and system based on big data. The method comprises the following steps: forming a space-time alignment data set through multi-source data acquisition, filtering and quality evaluation; performing feature extraction and registration to generate a digital model; target detection classification is performed to form a completion state table; progress evaluation is achieved through component-task mapping; trend analysis and risk identification are performed to generate a prediction result; decision reference is provided for personalized information screening and augmented reality display. Through multi-source data acquisition, fusion and intelligent analysis, accurate perception, objective evaluation, scientific prediction and visual presentation of the actual state of the construction site are realized, so that a comprehensive, accurate and prospective construction progress monitoring method is provided, the construction period delay risk is effectively reduced, and the construction management efficiency is improved.

Description

Technical Field

[0001] This application relates to the technical field of construction progress monitoring, and particularly to a construction progress monitoring method and system based on big data. Background Art

[0002] With the continuous increase in the scale and complexity of construction projects, traditional construction progress monitoring methods have been difficult to meet the needs of modern construction project management. Existing construction progress monitoring technologies mainly include methods such as manual inspection records, regular photo comparison, and local measurement. These methods usually rely on the experience judgment of project managers, and the collected data is often limited to specific areas and time points. Although in recent years, BIM (Building Information Modeling) technology has been applied to a certain extent in construction management, realizing the digital expression of design information, and some projects have begun to try to use single technical means such as photogrammetry and laser scanning to assist in progress monitoring, these applications are still in the initial stage and have not formed a complete data collection, analysis, and decision-making support system.

[0003] However, the existing technology has obvious deficiencies: First, the data collection methods are single and scattered, making it difficult to obtain comprehensive and continuous construction site information; second, there is a lack of an effective multi-source data fusion mechanism, and it is impossible to integrate data from different sources and of different types into a unified analysis basis; third, the progress evaluation mainly relies on manual experience judgment, lacking objective and quantitative evaluation criteria and automated analysis methods; in addition, the existing technology is difficult to predict future progress change trends and potential risks and cannot provide forward-looking decision-making support; finally, the presentation method of progress information is single and cannot provide personalized visual decision-making references according to the needs of different users. These deficiencies seriously restrict the efficiency and quality of construction management, resulting in risks such as construction period delays, quality problems, and cost overruns being difficult to prevent and control in a timely manner. Summary of the Invention

[0004] This application provides a construction progress monitoring method and system based on big data, which is used to achieve precise perception, objective evaluation, scientific prediction, and intuitive presentation of the actual state of the construction site through multi-source data collection, fusion, and intelligent analysis, so as to provide a comprehensive, accurate, and forward-looking construction progress monitoring method, effectively reduce the risk of construction period delays, and improve the efficiency of construction management.

[0005] In a first aspect, the present application provides a construction progress monitoring method based on big data. The construction progress monitoring method based on big data includes: obtaining construction site images, videos, and point clouds through multi-source devices, filtering and quality evaluating the collected data to obtain a construction original data set with spatio-temporal alignment; according to the construction original data set, extracting features and registering the image sequence and point cloud data, spatially aligning the measured data with the design model to obtain a construction site digital model containing geometric and time information; based on the construction site digital model, performing target detection and classification on on-site components, analyzing the spatial position relationship between components to obtain a construction completion status table with component identifiers and attributes; according to the construction completion status table, establishing a mapping relationship between the actual components and the construction plan tasks, calculating the component completion degree and position deviation value to obtain a construction progress evaluation report and component-level deviation data; according to the construction progress evaluation report and component-level deviation data, analyzing the progress change trend, identifying abnormal patterns and potential risk points during the construction process to obtain a progress prediction result and a risk warning list; screening the progress prediction result and the risk warning list according to user permissions and key concerns, and superimposing and displaying the progress and deviation information in the on-site real scene to obtain a visual construction management decision-making reference.

[0006] In a second aspect, the present application provides a construction progress monitoring system based on big data. The construction progress monitoring system based on big data includes:

[0007] An evaluation module, configured to obtain construction site images, videos, and point clouds through multi-source devices, filter and quality evaluate the collected data to obtain a construction original data set with spatio-temporal alignment;

[0008] A registration module, configured to extract features and register the image sequence and point cloud data according to the construction original data set, spatially align the measured data with the design model to obtain a construction site digital model containing geometric and time information;

[0009] A classification module, configured to perform target detection and classification on on-site components based on the construction site digital model, analyze the spatial position relationship between components to obtain a construction completion status table with component identifiers and attributes;

[0010] A mapping module, configured to establish a mapping relationship between the actual components and the construction plan tasks according to the construction completion status table, calculate the component completion degree and position deviation value to obtain a construction progress evaluation report and component-level deviation data;

[0011] An analysis module, configured to analyze the progress change trend according to the construction progress evaluation report and component-level deviation data, identify abnormal patterns and potential risk points during the construction process to obtain a progress prediction result and a risk warning list;

[0012] A screening module, configured to screen the progress prediction results and the risk warning list according to user permissions and key concerns, and superimpose and display the progress and deviation information in the on-site real scene to obtain a visual construction management decision-making reference.

[0013] In a third aspect, a computer device is provided, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory to cause the computer device to execute the above-mentioned big data-based construction progress monitoring method.

[0014] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, and when the instructions are run on a computer, the computer is caused to execute the above-mentioned big data-based construction progress monitoring method.

[0015] In the technical solution provided by this application, by obtaining construction site images, videos, and point cloud data through multi-source devices, a comprehensive and continuous perception of the construction site is achieved, breaking through the limitations of traditional single data sources and improving the integrity and diversity of the original data; by filtering and quality evaluating the collected data, the accuracy and reliability of the data are ensured, laying a solid foundation for subsequent analysis; the spatio-temporal aligned construction original data set solves the problem of multi-source heterogeneous data fusion and realizes the unified representation and processing of different types of data. Based on the feature extraction and registration technology of the original data set, the measured data can be accurately corresponded to the design model, forming a digital model of the construction site containing geometric and time information, creating conditions for the comparison between the actual situation and the plan; compared with the traditional BIM model, this digital model not only contains static design information, but also incorporates the measured data of the dynamic construction process, truly reflecting the actual state and changes of the construction site. Through deep learning algorithms, object detection and classification are performed on on-site components, automatically identifying the types and attributes of components, greatly improving the accuracy and efficiency of component identification. Especially for component identification in complex environments, the feature extraction ability of deep neural networks is significantly superior to traditional computer vision methods; the spatial relationship analysis algorithm makes the position constraints and connection relationships between components clearly visible, forming a construction completion status table with component identifiers and attributes, providing accurate basic data for progress evaluation. The method of establishing a mapping relationship between actual components and construction plan tasks realizes the conversion from the physical completion status to the planned completion ratio, solving the problems of strong subjectivity and poor consistency in traditional progress evaluation; by quantifying the component installation accuracy through the bounding box occupancy rate and point-plane distance evaluation methods, the construction quality evaluation is transformed from qualitative description to quantitative indicators, and the progress evaluation report and component-level deviation data provide an objective basis for management decisions. The progress change trend analysis method based on time series analysis and machine learning can mine potential laws from historical data, identify abnormal patterns and risk points; the risk propagation network established by combining historical project experience data significantly enhances the accuracy and forward-looking of risk prediction, and the progress prediction results and risk warning list provide an opportunity for managers to plan ahead. Finally, through augmented reality technology, the progress and deviation information are intuitively superimposed and displayed on the on-site real scene. Combined with the personalized information screening mechanism of user permissions and focus of attention, the complex progress data becomes intuitive and easy to understand. Decision-makers can directly obtain the required information on-site and make responses, greatly improving the decision-making efficiency and accuracy. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of an embodiment of the construction progress monitoring method based on big data in the embodiments of the present application;

[0018] Figure 2 It is a schematic diagram of an embodiment of the construction progress monitoring system based on big data in the embodiments of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of a computer device in the embodiments of the present invention. Specific embodiments

[0020] The embodiments of the present application provide a construction progress monitoring method and system based on big data. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that shown or described here. In addition, the terms "comprising" or "having" and any variation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 An embodiment of the construction progress monitoring method based on big data in the embodiments of the present application includes:

[0022] Step S101: Obtain construction site images, videos and point clouds through multi-source devices, filter and quality evaluate the collected data, and obtain a spatio-temporally aligned construction raw data set;

[0023] Step S102: According to the construction raw data set, extract features and register the image sequence and point cloud data, spatially align the measured data with the design model, and obtain a construction site digital model containing geometric and time information;

[0024] Step S103: Based on the construction site digital model, perform object detection and classification on the on-site components, analyze the spatial position relationship between the components, and obtain a construction completion status table with component identifiers and attributes;

[0025] Step S104: According to the construction completion status table, establish a mapping relationship between the actual components and the construction plan tasks, calculate the component completion degree and position deviation value, and obtain a construction progress evaluation report and component-level deviation data;

[0026] Step S105: Analyze the trend of progress changes based on the construction progress evaluation report and component-level deviation data, identify abnormal patterns and potential risk points during the construction process, and obtain the progress prediction results and a list of risk warnings.

[0027] Step S106: Screen the progress prediction results and the list of risk warnings according to user permissions and key concerns, overlay and display the progress and deviation information in the on-site real scene, and obtain a visual reference for construction management decisions.

[0028] It can be understood that the execution entity of this application can be a construction progress monitoring system based on big data, or a terminal or a server. Specifically, it is not limited here. This application embodiment is described by taking the server as the execution entity as an example.

[0029] Specifically, the multi-source data acquisition technology is adopted. Image sequences are collected by deploying high-definition cameras at fixed positions on the construction site, an unmanned aerial vehicle (UAV) aerial photography system is equipped to obtain an overhead video stream, and at the same time, a laser scanner is used to collect three-dimensional point cloud data. The collected raw data has noise and redundancy and needs to be filtered. Median filtering is used for the image data to remove noise points and perform distortion correction, and the statistical outlier filtering algorithm is used for the point cloud data to remove abnormal points. Subsequently, the quality of data from different sources is evaluated, a quality scoring system including indicators such as clarity, integrity, and coverage rate is established, and high-quality data is screened. Finally, the screened various types of data are established with a unified index according to the time stamp and spatial coordinates to form a spatio-temporal aligned construction raw data set.

[0030] Feature extraction and spatial registration are performed using the construction raw data set. Scale-invariant feature transform (SIFT) feature points and descriptors are extracted from the image sequences, the matching relationship between images is identified, and a sparse three-dimensional point cloud skeleton is established. The point cloud data determines the spatial main structure by extracting geometric features such as planes and edges. These features are compared with the structural features in the building information design model, and the spatial transformation matrix between the measured data and the design model is calculated to complete the alignment transformation of the coordinate system. The data from different periods are sorted and associated through the time stamp, and the temporal change record of the construction process is constructed. Finally, a digital model of the construction site containing geometric and time information is generated.

[0031] Perform edge detection segmentation on the component areas in the digital model to determine the candidate areas of components. Calculate the feature vectors of each candidate area, including shape ratio, texture features, color distribution, etc., and match and classify them with the preset component type library, marking them as different types such as walls, columns, beams, doors and windows, pipes, etc. Measure the three-dimensional coordinates, orientations, and dimensions of the classified components to establish a geometric description of the components. Analyze the spatial position relationships between components, identify topological relationships such as support, connection, and containment, and construct a component relationship network. Finally, integrate information such as component type, position, size, installation status, etc. into a structured data table to form a construction completion status table with component identifiers and attributes. Implement construction progress assessment and deviation analysis. First, import the construction plan data in the project management system, extract the work breakdown structure and key node milestones, and construct a construction task dependency network. Establish a mapping relationship between the actual components in the construction completion status table and the planned tasks through component coding to determine the planned task items corresponding to each component. Calculate the completion status of the components involved in each task. For example, the completion rate of the wall construction task is determined according to the ratio of the actual number of installed wall components to the planned number. At the same time, measure the deviation values between the actual components and the designed positions through the bounding box occupancy rate method and the point-to-surface distance method. For example, the average distance by which the installation position of the pipe deviates from the designed position. Combine task weights and critical paths to generate an overall progress assessment report and form component-level deviation data including a heat map of deviation distribution. Analyze and predict the progress trend. Extract historical data points from the progress assessment report, construct a progress-time curve, and apply time series decomposition techniques to separate it into trend, seasonal, and random components. Analyze the correlation between component-level deviation data and schedule delays to identify construction bottlenecks. For example, areas with large installation deviations of mechanical and electrical pipes often cause delays in subsequent processes. Compare with the data in the historical project library to identify risk patterns in similar construction scenarios and establish risk propagation paths. For example, deviations in the installation of air ducts may trigger subsequent pipeline conflicts. Based on historical trends and the current status, predict the probability of future construction progress completion and generate a progress prediction curve including different confidence intervals. Form a graded risk warning list according to the impact degree and urgency of risk factors. Implement visual decision support. Filter information according to user role permissions. For example, project managers can view all information, and professional subcontractors can only view the progress and risks related to their specialties. Score and screen information based on the user's focus of attention, obtain real-time camera images and perform spatial positioning, and convert the progress and deviation information into layers and overlay them on the real scene. For example, display a red warning mark at the position of the delayed component and a green mark in the area with normal progress. Integrate the progress prediction curve and risk warning information to form an interactive interface including multiple views, and intuitively view the progress status and potential risks on the construction site through a tablet computer or a head-mounted device.

[0032] In the embodiments of the present application, by obtaining construction site images, videos, and point cloud data through multi-source devices, comprehensive and continuous perception of the construction site is achieved, breaking through the limitations of traditional single data sources and improving the integrity and diversity of the original data; by filtering and quality evaluating the collected data, the accuracy and reliability of the data are ensured, laying a solid foundation for subsequent analysis; the spatio-temporally aligned construction original data set solves the problem of multi-source heterogeneous data fusion and realizes the unified representation and processing of different types of data. Based on the feature extraction and registration techniques of the original data set, the measured data can be accurately corresponded to the design model, forming a digital model of the construction site containing geometric and time information, creating conditions for the comparison between the actual and the planned; compared with the traditional BIM model, this digital model not only contains static design information but also incorporates the measured data of the dynamic construction process, truly reflecting the actual state and changes of the construction site. Through deep learning algorithms, object detection and classification of on-site components are performed to automatically identify the types and attributes of components, significantly improving the accuracy and efficiency of component identification. Especially for component identification in complex environments, the feature extraction ability of deep neural networks is significantly superior to traditional computer vision methods; the spatial relationship analysis algorithm makes the position constraints and connection relationships between components clearly visible, forming a construction completion status table with component identifiers and attributes, providing accurate basic data for progress assessment. The method of establishing a mapping relationship between the actual components and the construction plan tasks realizes the conversion from the physical completion status to the planned completion ratio, solving the problems of strong subjectivity and poor consistency in traditional progress assessment; by quantifying the component installation accuracy through the bounding box occupancy rate and the point-plane distance evaluation method, the construction quality assessment is transformed from qualitative description to quantitative indicators, and the progress assessment report and component-level deviation data provide an objective basis for management decisions. The progress change trend analysis method based on time series analysis and machine learning can mine potential laws from historical data, identify abnormal patterns and risk points; the risk propagation network established by combining historical project experience data significantly enhances the accuracy and forward-looking of risk prediction, and the progress prediction results and risk warning list provide an opportunity for managers to take precautions. Finally, through augmented reality technology, the progress and deviation information are intuitively superimposed and displayed on the on-site real scene. Combined with the personalized information screening mechanism of user permissions and focus of attention, the complex progress data becomes intuitive and easy to understand. Decision-makers can directly obtain the required information on-site and make responses, greatly improving the decision-making efficiency and accuracy.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] Collect image sequences, video streams, and three-dimensional point cloud data of the construction site through high-definition cameras, UAV aerial photography systems, and laser scanners;

[0035] Perform distortion correction and enhancement processing on the acquired image sequence, remove image noise and redundant information, and generate a standardized image set;

[0036] Perform spatial registration and noise reduction on the collected point cloud data, remove outliers and duplicate points, and form structured point cloud data;

[0037] Dynamically schedule tasks for collection equipment through edge computing nodes, and automatically adjust the frequency and scope of data collection according to on-site environmental conditions;

[0038] Score the quality of the processed data based on multi-dimensional quality indicators and select valid data sets that meet the threshold requirements;

[0039] The filtered image sets, point cloud data and sensor information are aligned and fused according to timestamps and spatial positions, and a unified indexing mechanism is established to generate a spatiotemporally aligned construction original data set.

[0040] Specifically, construction site data is collected through multi-source devices. High-definition cameras are fixedly installed at key positions on the construction site to collect image sequence data. The typical collection frequency is one frame every 10 - 30 seconds, forming a continuous record of the construction process. The UAV aerial photography system flies along a preset route to obtain an overhead video stream of the construction site from different angles and heights, usually cruising and collecting at a frequency of 1 - 2 times a day. The laser scanner measures the three-dimensional coordinates of spatial points by emitting laser beams and receiving reflected signals, generating point cloud data. The scanning density is usually set to thousands of points per square meter. These three types of devices work together to ensure comprehensive data coverage of the construction site. The collected image sequences need to be subjected to distortion correction and enhancement processing. First, apply a lens distortion correction algorithm to the images to eliminate the barrel or pincushion distortion caused by wide-angle lenses and restore the geometric accuracy of the images. Then, perform image enhancement, including contrast adjustment, brightness equalization, and sharpening, to improve the clarity of the images. For noise processing, use median filtering or Gaussian filtering algorithms to remove random noise and interference in the images. Identify and remove redundant background regions through image segmentation technology, only retaining the effective regions containing construction activities. The processed images are organized according to the shooting time and spatial location to form a standardized image set. Point cloud data processing includes two key steps: spatial registration and noise reduction. Spatial registration is to integrate multiple groups of point cloud data obtained by scanning at different positions into a unified coordinate system. The iterative closest point (ICP) algorithm is used to find the optimal rigid body transformation relationship between two groups of point clouds, calculate the rotation matrix and translation vector, and splice the point clouds scanned multiple times into a complete three-dimensional scene. Noise reduction processing first identifies outliers through statistical outlier analysis, calculates the average distance of each point to its neighboring points, and determines and removes the points that significantly deviate from the average value. Then, screen for duplicate points. When the distance between two points is less than a preset threshold, merge them into one point to reduce data redundancy. Organize the point cloud data through spatial index structures such as octrees or KD-trees to form structured point cloud data with hierarchical query capabilities. Edge computing nodes play a key scheduling role in the data collection process. This node is implemented through an embedded computing device and is deployed near the on-site data collection devices to monitor changes in environmental conditions in real-time and adjust the collection strategy. Specifically, the edge node analyzes the lighting conditions, weather conditions, and construction activity intensity, and accordingly dynamically adjusts the exposure parameters, collection frequency, and image resolution of the camera. When detecting insufficient light, automatically reduce the camera collection frequency and increase the exposure time; when finding areas with intensive construction activities, increase the data collection frequency and resolution in that area; during construction pauses, reduce the collection frequency to save storage space. At the same time, the edge node adjusts the data compression ratio and upload priority according to the data transmission bandwidth status to ensure the priority transmission of important data.

[0041] Multi-dimensional quality assessment is a key link to ensure data availability. A comprehensive scoring system including clarity, integrity, coverage, accuracy, and timeliness is established for the processed data. The clarity of image data is calculated through edge sharpness and contrast; the integrity of point cloud data is evaluated through spatial distribution density to check for large areas of data holes; coverage is calculated by comparing the spatial range of the collected data with the target monitoring area; accuracy is assessed by the deviation value from the reference point; and timeliness is judged based on the time interval between data collection time and the current time. The comprehensive quality score is calculated according to the weights of each dimension index, and the data with scores higher than the preset threshold is selected as the effective data set, while the low-quality data is excluded. The data from different sources are time-aligned according to the timestamp to establish a time series relationship. Then, spatial alignment is performed through three-dimensional spatial coordinates to establish a unified spatial reference system for image, point cloud, and sensor data. The world coordinates corresponding to the image are calculated through the internal and external parameters of the camera; the point cloud data itself contains spatial coordinate information; and the spatial position of the sensor data is calculated through the device installation position and measurement parameters. A spatio-temporal unified indexing mechanism is established, using time and spatial coordinates as multi-dimensional index keys to support rapid retrieval of data by time period, spatial region, or comprehensive conditions. Through the above steps, a spatio-temporal aligned construction raw data set is formed, laying a foundation for subsequent analysis and processing.

[0042] Taking a high-rise building construction project as an example, during the concrete pouring process of the third-floor slab, 4 high-definition cameras are deployed at fixed positions around. The drone conducts two aerial surveys at 10 am and 3 pm every day, and the laser scanner conducts a comprehensive scan once after the end of each day's work. After distortion correction and enhancement processing of the 4500 original images collected by the cameras, 3800 effective images are selected; the original point cloud data obtained by the laser scanner is about 800 million points, and 240 million effective points are retained after noise reduction and duplicate removal processing. The edge computing node detected that the sunlight directly irradiated from 2 pm to 4 pm caused overexposure of the images, and automatically adjusted the camera exposure parameters downward and increased the acquisition frequency; in rainy days, the point cloud acquisition density was automatically increased to cope with the reduced visibility. In the quality assessment, due to the obstruction of the construction tower crane, the data quality in the northeast corner area was relatively low, and the system supplemented the acquisition of this area. The finally generated spatio-temporal aligned construction raw data set includes complete coverage of all construction areas and establishes an index structure that supports multi-dimensional queries by floor, area, component type, and time period.

[0043] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0044] Extract the feature points and feature descriptors of the image sequence from the construction raw data set, establish the matching relationship between images, and generate sparse reconstructed point clouds;

[0045] Densify the sparse reconstructed point cloud, fill the missing areas in the space, and construct continuous surface geometric information;

[0046] Extract main geometric features such as planes, line segments, and corner points from the point cloud data to form a structural feature set of the construction scene;

[0047] Pairwise compare the structural feature set with the geometric features of the building information design model, calculate the transformation matrix, and correct the spatial positions of the measured data;

[0048] Sort and organize the data collected at different times through timestamps to establish a temporal relationship network of the construction progress;

[0049] Integrate the corrected geometric data with the temporal relationship network, attach time attribute tags to each spatial point, and generate a digital model of the construction site containing geometric and time information.

[0050] Specifically, extract feature points and feature descriptors from the image sequence. Feature points are points with unique properties in the image, such as corner points, edge points, etc., which can maintain stable recognition under different perspectives. The extraction process uses the SIFT (Scale-Invariant Feature Transform) or SURF (Speeded-Up Robust Features) algorithm. The SIFT algorithm constructs a Gaussian difference pyramid, detects local extreme points in different scale spaces, calculates their main directions, and generates descriptors. Descriptors are vectors representing the gray-scale distribution of pixels around feature points, usually composed of 128-dimensional or 64-dimensional numerical values. Calculate the Euclidean distance between feature points for each pair of images, find the matching point pairs with the smallest distance, and then eliminate the incorrect matches through the RANSAC (Random Sample Consensus) algorithm. Based on the spatial position relationship of the matching point pairs, calculate the relative pose between the cameras, and restore the three-dimensional coordinates of the feature points through triangulation to form a sparse reconstructed point cloud. Although the number of points in this point cloud is limited, it accurately represents the scene structure. The sparse reconstructed point cloud needs to be densified to fill the missing areas in the space. This process relies on PMVS (Patch-based Multi-View Stereo) or MVS (Multi-View Stereo Vision) technology. First, expand dense sampling points around the sparse point cloud, construct normal vectors and local plane assumptions for each sampling point, and then verify this assumption among multiple images, retaining the points with high consistency. For the missing areas, use the Poisson surface reconstruction algorithm to convert the point cloud into an implicit function representation, solve the Poisson equation to fill the gaps, and then extract the isosurface to generate a continuous mesh surface. In severely occluded areas, use plane constraints and symmetry assumptions for completion. For example, if a wall plane is identified, extend along this plane to fill the missing part. Through these steps, the sparse point cloud is transformed into a dense model with continuous surface geometric information.

[0051] Extracting geometric features from point cloud data is the key to constructing a structural feature set. First, plane extraction is carried out. The RANSAC algorithm is used to randomly select three points in the point cloud to determine the plane equation, calculate the distances from other points to this plane, count the number of points that meet the threshold, and through multiple iterations, the optimal plane parameters are found. For line segment extraction, the Hough transform is applied to detect line features at the intersections of the identified planes, or edge lines are directly extracted at locations where the point cloud density gradient changes significantly. Corners are identified at the intersections of line segments, and the intersection coordinates and included angles are calculated. The extracted planes, line segments, and corners are organized into a topological relationship graph, which records the connection relationships between various elements, such as which line segments form a plane and which corners connect which line segments, thus forming a structural feature set that characterizes the main structure of the construction scene. The structural feature set is paired and compared with the Building Information Design Model (BIM) to achieve the alignment of the measured data and the design model. First, the BIM model is converted into a geometric representation, and the plane, line segment, and corner features in it are extracted to establish a feature library corresponding to the measured structural feature set. Then, a feature matching algorithm is used to calculate the similarity matrix between the structural features and find the best matching pairs. Based on the matching pairs, a set of control points is determined, and the rigid body transformation matrix is solved, including the rotation matrix R and the translation vector T. This transformation converts the coordinate system of the measured data to the coordinate system of the BIM model. The calculation of the transformation matrix uses the SVD (Singular Value Decomposition) or ICP (Iterative Closest Point) algorithm, and the optimal solution is obtained by minimizing the sum of the Euclidean distances between the control point pairs. The transformation matrix is applied to perform coordinate transformation on all the measured data to complete the spatial position correction.

[0052] The processing in the time dimension organizes the data by sorting it according to timestamps. Each data point carries the acquisition time information, and the data is arranged in chronological order to form a time series. For time periods with sparse data, time interpolation techniques are used to fill in the gaps and maintain data continuity. According to the intervals of the timestamps, the data is divided into time windows of different granularities such as days and weeks, and the change amounts within each window are calculated, such as newly added components and position changes. By comparing the data differences between adjacent time windows, construction activities and progress changes are identified, and a temporal relationship network representing the construction progress is constructed. This network records the appearance times, installation sequences, and interdependencies of different components. The corrected geometric data is integrated with the temporal relationship network, and time attributes are attached to each point or component in space. Specifically, time tags are added to each geometric element, including the first appearance time, change history, and status update records. The resulting digital model of the construction site not only contains accurate three-dimensional spatial information but also records the temporal evolution of the construction process, becoming a four-dimensional model that supports querying and analysis in both time and space dimensions. Such a digital model can visually display the construction status at any moment, track the installation progress of specific components, and analyze whether the construction sequence conforms to the plan.

[0053] Taking an office building construction project as an example, 15,000 SIFT feature points are extracted from the construction raw dataset. Through matching and triangulation, a sparse point cloud of approximately 8,000 spatial points is generated. This sparse point cloud is expanded to a dense point cloud of 2 million points through the PMVS algorithm, and the main structural features are identified, including 120 planes (such as walls, floors, etc.), 350 line segments (such as edges, beams, etc.), and 180 corner points (such as building corners, column intersections, etc.). These structural features are matched and compared with the BIM model, and the calculated transformation matrix corrects the measured data to the design coordinate system, with the average alignment error controlled within 3 centimeters. The data is organized according to the time stamp, and a time-series network for a 4-month construction period is constructed, recording the construction sequence from the foundation structure to the enclosure system and then to the interior decoration. The finally formed digital model of the construction site contains approximately 500 main components, each component with geometric information and time stamps, and this model accurately reflects the construction progress and actual status.

[0054] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0055] Extract the feature map of the component area from the digital model of the construction site, outline the component contour through edge detection, and form a candidate area for component segmentation;

[0056] Calculate the feature vectors of the candidate areas, determine the component types according to the preset component classification criteria, and generate a set of components with type labels;

[0057] Based on the set of components with type labels, calculate the three-dimensional coordinates, orientations, and dimension parameters of each component, and establish a geometric description document for the components;

[0058] Conduct spatial association analysis on the components in the geometric description document, identify the connection relationships and relative position constraints between components, and form a component topology network diagram;

[0059] Extract the component installation standards and quality requirements from the design specification library, compare them with the geometric parameters of the actual components, and calculate the installation deviation values and completion status indicators;

[0060] Integrate the type labels of the components, the geometric description document, the component topology network diagram, the installation deviation values, and the completion status indicators into a structured data record, and organize it into a construction completion status table with component identifiers and attributes.

[0061] Specifically, extracting the feature map of the component area from the digital model of the construction site is a multi-step and delicate process. First, the image data in the digital model of the construction site is processed by edge detection technology. The edge detection uses the improved Canny algorithm, which first performs Gaussian filtering on the image to reduce noise, then calculates the image gradient amplitude and direction, then obtains the refined edge through non-maximum suppression, and finally applies the double threshold method to determine the edge pixels. In this way, the structural feature lines such as column-beam nodes and wall panel joints in the construction site are outlined to form a preliminary component outline. For point cloud data, the region growing method and RANSAC algorithm are used for plane segmentation, clustering the point cloud into different plane regions, and further refining the boundaries of the component. These processed contour information is integrated into candidate regions for component segmentation, each of which represents a possible building component.

[0062] The calculation of feature vectors for candidate regions is the core step of component classification. Feature vector calculation uses deep convolutional neural networks to extract high-dimensional features and combines them with geometric feature analysis. It can be expressed as:

[0063]

[0064] in, The geometric shape feature vector representing the component, including parameters such as aspect ratio, area-to-volume ratio, and curvature distribution; Represents the texture density feature vector, describing the surface detail features of the component; Represents the material feature vector, characterized by parameters such as reflectivity and color distribution; Represents the contextual feature vector, describing the spatial correlation between the candidate area and the surrounding environment. The calculated feature vector is matched with the preset component classification standard library for similarity, and the support vector machine (SVM) or random forest algorithm is used for classification judgment. The component classification judgment can be expressed as:

[0065]

[0066] Among them, C t represents the classification result of the component, T k represents the kth component type (such as column, beam, wall, plate, etc.), K represents the preset component type set, Represents the feature vector Of type T k In this way, the candidate areas are divided into different types such as columns, beams, walls, plates, doors, windows, pipes, etc., forming a set of components with type labels.

[0067] In the three-dimensional parameter calculation stage of components, for the identified components, accurate spatial information is extracted. The three-dimensional coordinate point set of a component can be expressed as:

[0068] P c =(x v ,y v ,z v )|v∈[1,N p

[0069] where P c represents the point set of the component, and (x v ,y v ,z v ) represents the spatial coordinates of the v-th feature point on the component, and N p represents the total number of feature points. The orientation of the component is calculated by the principal component analysis (PCA), and three principal direction vectors of the point cloud of the component are extracted:

[0070]

[0071] where represents the orientation matrix of the component, and respectively represent the unit vectors of the component in the three principal axis directions. The dimensional parameters of the component are calculated by the minimum circumscribed rectangle method:

[0072] S c =[L x ,L y ,L z

[0073] where S c represents the dimensional vector of the component, and L x , L y , L z respectively represent the lengths of the component in the three principal axis directions. These calculated geometric parameters form the geometric description document of the component, and a detailed geometric model is established for each component.

[0074] In the spatial association analysis stage, the topological relationships of the components in the geometric description document are identified. By calculating the shortest distance, overlapping area, and connection point position between components, a spatial relationship matrix between components is established. The types of spatial relationships include support relationship, connection relationship, inclusion relationship, and adjacent relationship, etc. For example, by judging the spatial position relationship between columns and beams, it can be determined that there is a support connection between them; by analyzing the relative position of walls and doors and windows, the inclusion relationship can be determined. These relationship information is integrated into a component topological network diagram to form a logical model of the overall building structure.

[0075] ​​The installation deviation calculation process extracts the standard parameters of corresponding components from the design specification library and compares them with the geometric parameters measured in reality. Deviation calculation involves three aspects: position deviation, angle deviation, and dimension deviation. The position deviation is obtained by calculating the Euclidean distance between the center point of the actual component and the designed position; the angle deviation is calculated by the included angle between the actual orientation vector and the designed orientation vector; the dimension deviation is calculated by the difference between the actual dimension and the designed dimension. According to the calculated deviation values compared with the preset allowable error range, the installation status of the component is determined and marked with different levels such as "qualified", "slight deviation", or "severe deviation", generating a completion status identifier.

[0076] Finally, the type label of the component, the geometric description document, the component topology network diagram, the installation deviation value, and the completion status identifier are integrated according to a unified data structure format to form a complete component data record. These records are organized in multiple indexing methods such as component ID, spatial position, and component type to form a construction completion status table with component identifiers and attributes. This status table details the identification results, location information, connection relationships, and installation quality assessments of each component on the construction site, providing an accurate data basis for subsequent construction progress assessments.

[0077] Taking a concrete frame structure building as an example, this method first extracts the edge features of components such as walls, columns, beams, and slabs from the digital model to form hundreds of candidate regions. Feature vectors are calculated for each candidate region. For example, the geometric shape features of a certain beam component show that its aspect ratio is 15:1, the surface texture presents the unique roughness pattern of concrete, and the material features show a uniform gray tone distribution. By comprehensively comparing these feature vectors with the preset standard of beam components, the system determines it as the type of "concrete beam". Further analyzing the spatial parameters of this beam, calculating its center line coordinates, main axis direction vector, and three-dimensional dimensions (length 6000mm, width 300mm, height 500mm), a geometric description document is generated. Through spatial correlation analysis, it is identified that this beam has a support connection relationship with the column components at both ends and a load-bearing relationship with the upper slab component, forming a local structure network. Comparing with the standard parameters of this beam in the design specification, the calculated position deviation is 22mm, the inclination angle is 1.2°, and the width deviation is -5mm, all within the allowable range. Therefore, its completion status is marked as "qualified". Finally, this information is integrated into a structured record and added to the construction completion status table to provide accurate data for subsequent progress assessments and decision-making support.

[0078] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0079] Import the construction plan data from the project management system, extract the work breakdown structure and key node milestones, and generate a construction task dependency network;

[0080] By matching component codes, establish a corresponding relationship between the component records in the construction completion status table and the planned items in the construction task dependency network to form a component-task mapping table;

[0081] Based on the component-task mapping table, count the completion status of the components involved in each task item, and calculate the task completion rate and progress weight value;

[0082] Calculate the difference between the geometric parameters and design parameters of the components in the construction completion status table, and quantify the component installation accuracy through the boundary box occupancy rate and point-plane distance evaluation methods to generate a component position deviation matrix;

[0083] According to the task completion rate and progress weight value, combined with the critical path of the project, comprehensively analyze the overall progress status and the achievement of critical nodes, and compile a construction progress evaluation report;

[0084] Classify and summarize the component position deviation matrix according to component type, area, and deviation degree to generate component-level deviation data including deviation distribution maps and hot spots.

[0085] Specifically, importing construction plan data from the project management system is a key link in the construction progress monitoring method based on big data. The specific implementation process first connects to the project management system through a data interface, and uses API calls or direct database queries to extract the original plan data containing information such as task name, planned start time, planned completion time, predecessor tasks, and resource allocation. After extraction, the original data is structured. The work breakdown structure (WBS) is identified through a hierarchical decomposition algorithm, and the overall project tasks are decomposed into task units at different levels such as subsystems and sub-projects. At the same time, critical node milestones are identified through time node analysis. These milestones are important time points for project progress control, such as foundation completion, main body topping out, equipment installation completion, etc. On this basis, the topological sorting algorithm is applied to analyze the dependency relationships between tasks, and a construction task dependency network in the form of a directed acyclic graph (DAG) is constructed. This network clearly shows the sequence and parallel relationships between tasks, providing a logical basis for progress evaluation.

[0086] Associating the construction completion status table with the construction task dependency network through component code matching is an important step in realizing the mapping between physical entities and planned tasks. First, analyze the coding rules of each component in the construction completion status table. These codes usually contain information such as area identification, component type, floor location, etc. At the same time, analyze the composition rules of task codes in the construction plan to find the corresponding relationship between the two coding systems. Then design a matching algorithm that establishes coding mapping rules by extracting common elements in the codes (such as area codes, floor numbers, component type codes, etc.). For cases where the coding rules are not unified, use fuzzy matching and semantic analysis methods to infer possible corresponding relationships by calculating the similarity between task descriptions and component attributes. In this way, establish a one-to-one or one-to-many mapping relationship between each component record in the construction completion status table and the corresponding planned item in the construction task dependency network, forming a component-task mapping table. This table contains fields such as component ID, task ID, corresponding weight, etc., and becomes the basic data structure for subsequent progress calculation.

[0087] Statistical analysis of task completion based on the component-task mapping table is the core link in evaluating the actual progress. For each task item, first screen out all component records associated with this task from the component-task mapping table, and then extract the completion status identifiers of these components from the construction completion status table. Assign different completion degree weight values according to different levels of component completion status (such as not started, in progress, completed, etc.), and then combine the importance weight of the component in the task to calculate the weighted average completion degree. When calculating specifically, multiply the completion degree of each component by its weight and then divide by the total weight value to obtain the overall completion rate of the task. At the same time, determine the progress weight value of the task according to factors such as the workload proportion and key degree of the task in the whole project. This weight value reflects the impact degree of the task on the overall project progress. In this way, quantitatively convert the component-level completion status into task-level progress indicators and establish the quantitative relationship between physical completion and planned completion.

[0088] Calculating the differences between the geometric parameters and design parameters of components in the construction completion status table is an important step in evaluating construction quality. First, extract the actual geometric parameters of each component from the construction completion status table, including spatial coordinates, dimensions, orientations, etc., and at the same time extract the standard parameters of the corresponding components from the design model. Then calculate the differences between the two sets of parameters, including position deviation (the Euclidean distance between the actual coordinates and the design coordinates), dimension deviation (the difference between the actual dimension and the design dimension), and angle deviation (the included angle between the actual orientation and the design orientation). On this basis, use the bounding box occupancy rate evaluation method to calculate the overlap degree between the actual component bounding box and the design bounding box. The higher the overlap rate, the more accurate the positioning. At the same time, use the point-to-plane distance evaluation method to calculate the average distance from the actual component point cloud to the surface of the design model. The smaller the distance, the higher the shape matching degree. Through the comprehensive application of these two evaluation methods, quantify the installation accuracy of the components and organize the results into a component position deviation matrix, which is indexed by component ID and contains various deviation values and accuracy scores.

[0089] Comprehensive analysis based on the task completion rate and progress weight value is the basis for forming the construction progress evaluation report. First, multiply the completion rate of each task by its progress weight value to obtain the weighted completion degree, and then sum up the weighted completion degrees of all tasks to calculate the overall project progress completion percentage. At the same time, apply the Critical Path Method (CPM) to analyze the task dependency network and identify the task chain that has the greatest impact on the total project duration, that is, the critical path. Pay special attention to the completion of tasks on the critical path and calculate the progress deviation value of the critical path. In addition, for key node milestones, compare the difference between the actual completion time and the planned time to form a milestone achievement evaluation. On this basis, combine the overall progress completion degree, the critical path status, and the milestone achievement situation, and use the Analytic Hierarchy Process to form a comprehensive evaluation result and compile a construction progress evaluation report including progress status, deviation analysis, risk warning, and trend prediction.

[0090] Hierarchical aggregation of the component position deviation matrix is a key step in forming component-level deviation data. First, group and statistically analyze the deviation matrix according to component types (such as columns, beams, walls, slabs, etc.), calculate the average deviation value and standard deviation of various components, and analyze the installation accuracy differences of different types of components. Then group by spatial area (such as floors, functional areas, etc.) to analyze the construction quality distribution in different areas. At the same time, set thresholds according to the deviation degree for grading. For example, divide the position deviation into levels such as the normal range (<10mm), slight deviation (10 - 20mm), and severe deviation (>20mm). Through this multi-dimensional hierarchical aggregation, generate a visual report containing statistical charts and deviation distribution heat maps, intuitively showing the spatial distribution and concentrated areas of construction deviations, and helping managers quickly identify the hot spots of construction quality problems.

[0091] Taking a high-rise building project as an example, this method first imports the construction plan data containing 523 task items from the project management system. Through the hierarchical decomposition algorithm, it identifies three major systems: the main structure, mechanical and electrical installation, and decoration, as well as 15 sub-projects such as the foundation and foundation, main construction, and exterior enclosure. It extracts 8 key milestone nodes such as the completion of the foundation and the topping out of the main body, and constructs a complete construction task dependency network. Then, through the coding matching algorithm, a corresponding relationship is established between the 3,850 component records in the construction completion status table and the 523 planned tasks, forming a component-task mapping table. For example, a mapping relationship is established between the component with the code "F3-C-042" (indicating column No. 42 in Area C on the 3rd floor) and the task "T-157" (construction of the frame column on the 3rd floor). Based on the mapping table, statistics show that task "T-157" is associated with 24 frame column components, of which 22 are completed and 2 are in progress. Calculated by weight, the task completion rate is 91.7%. At the same time, through geometric parameter analysis, the average position deviation of these column components is calculated to be 8.3 mm, and the average angle deviation is 0.7°, both within the allowable range. By comprehensively analyzing the completion of all tasks and combining the critical path and milestone achievement status, a detailed progress evaluation report is formed, showing that the overall progress completion rate of the project is 78.2%, and the critical path is delayed by 3 days. At the same time, through the hierarchical summary of the component position deviation, a deviation distribution heat map is generated, showing that there is a large concentration of installation deviation in the northeast corner area on the 5th floor, providing an accurate quality control basis for construction management.

[0092] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0093] Extract historical progress data points from the construction progress evaluation report, construct a time series data set, and draw a progress change curve graph;

[0094] Perform trend decomposition on the progress change curve graph, separate seasonal factors, long-term trends, and random fluctuation components, and obtain progress fluctuation characteristics;

[0095] Combine component-level deviation data, establish a deviation-progress correlation matrix, identify the correlation between deviation accumulation areas and progress delays, and form a construction bottleneck analysis report;

[0096] Compare the construction bottleneck analysis report with the data in the historical project library, extract the risk evolution paths of similar scenarios, and establish a risk propagation network;

[0097] Based on the progress fluctuation characteristics and the risk propagation network, calculate the progress completion probability distribution within the future time window, and generate a progress prediction result including multiple confidence intervals;

[0098] Risk factors and their impact weights are extracted from the risk propagation network, ranked and sorted according to urgency and scope of impact, and a risk warning list is formed.

[0099] Specifically, extracting historical progress data points from the construction progress assessment report is the first step in progress forecasting. This process uses data parsing technology to extract the progress completion rate value at each assessment time point from the progress assessment report and convert it into a binary structure containing a timestamp and a progress value. Data extraction uses a structured text parsing algorithm to locate and extract the date-progress correspondence in the report to form preliminary time series data. Subsequently, the extracted data is cleaned and normalized, including removing outliers, supplementing missing values, and standardizing the time format to ensure the integrity and consistency of the data set. The cleaned data is arranged in chronological order to form a complete time series data set, reflecting the progress change trajectory of the project from the beginning to the current time. Based on this data set, a progress change curve is drawn using time series visualization technology to intuitively display the time evolution characteristics of the project progress.

[0100] Trend decomposition of the progress change curve is a key step in understanding the law of progress fluctuations. Trend decomposition uses a time series decomposition algorithm to split the original progress data into three basic components: long-term trends, seasonal factors, and random fluctuations. Long-term trends reflect the main development direction of the project progress and are extracted through methods such as moving average or polynomial fitting; seasonal factors reflect periodic change patterns, such as regular fluctuations caused by alternating working days / rest days, and differences in resource allocation at the beginning / end of the month, which are extracted through Fourier analysis or seasonal index decomposition; random fluctuations represent irregular changes, including disturbances caused by various temporary factors. Through this decomposition, we can gain a deep understanding of the inherent laws of progress fluctuations and provide a basis for predicting future progress.

[0101] Establishing a deviation-schedule correlation matrix in combination with component-level deviation data is an important means to discover the relationship between quality problems and schedule delays. The process first aggregates the component-level deviation data according to spatial regions and time periods, and calculates the average deviation value and deviation density of each region in different time periods. At the same time, the schedule delays of each region are extracted from the schedule data, including the difference and change rate between the plan and the actual. Then, through the correlation analysis method, the correlation coefficient between the deviation index and the schedule delay index is calculated to form a two-dimensional correlation matrix. Each element in the matrix represents the correlation strength between a specific type of deviation and a specific schedule problem. Through matrix data mining, highly correlated areas, that is, areas and processes where deviation accumulation is highly correlated with schedule delays, are identified to form a construction bottleneck analysis report, detailing the bottleneck location, severity, and causes.

[0102] Comparing the construction bottleneck analysis report with the data in the historical project library is an important step in making full use of empirical knowledge. This process first parametrically describes the construction bottleneck characteristics of the current project, including attributes such as bottleneck type, location characteristics, and occurrence stage. Then, it accesses the historical project library and filters out historical cases similar to the current bottleneck characteristics through similarity calculation, and uses a case retrieval algorithm to quickly locate relevant records. It deeply analyzes the retrieved historical cases and extracts their risk evolution paths, including precursor signals of problem occurrence, diffusion processes, influence scopes, and final results. Based on the common characteristics of multiple similar cases, it constructs a risk propagation network model. In this network, nodes represent risk states, edges represent risk transfer paths, and weights represent transfer probabilities, comprehensively describing the evolution law of risks in time and space.

[0103] Calculating the progress completion probability distribution based on the progress fluctuation characteristics and the risk propagation network is the core of achieving scientific prediction. This calculation process comprehensively applies time series prediction and Monte Carlo simulation methods. First, based on the fluctuation characteristics of historical progress data, it establishes a benchmark prediction model for progress changes to predict the expected progress value at future time points. Then, it combines the risk factors and their probability distributions in the risk propagation network to generate a large number of possible risk scenarios and evaluates the progress impact under each scenario. Through multiple simulation calculations, it obtains the progress completion probability distribution function at different future time points:

[0104]

[0105] where P(C t ≤α) represents the probability that the completion degree at time point t does not exceed α, represents the probability density function of the progress completion degree at time point t. Based on the probability distribution function, it generates progress prediction results for different confidence intervals, such as 50%, 75%, and 90% confidence intervals, providing decision-making references under different risk preferences for project management.

[0106] Extracting risk factors and their influence weights from the risk propagation network is an important part of risk management. This process first conducts a topological analysis of the risk propagation network, calculates the centrality indicators of each node, including degree centrality, betweenness centrality, and eigenvector centrality, etc., to identify key risk nodes in the network. Then, through sensitivity analysis, it evaluates the influence degree of each risk factor on schedule delay and calculates the influence weight. According to the urgency of the risk (proximity of occurrence time) and the influence scope (affected workload), it conducts two-dimensional grading of risk factors to form a risk matrix. Finally, it ranks the risk factors according to the comprehensive risk score and generates a structured risk prompt list, containing detailed information such as risk descriptions, potential impacts, early warning signals, and countermeasure suggestions.

[0107] Taking a large commercial complex project as an example, this method first extracts progress data points from the progress evaluation reports for 90 consecutive days to construct a complete time series data set. Through trend decomposition, it is found that there is an obvious phenomenon of slower progress on weekends (seasonal factor) in this project, as well as a long-term trend of decelerating growth in the overall progress recently. Combining the analysis of component-level deviation data, it is found that there are high-density pipeline installation deviations in the commercial podium part, and this area is also the area with the most serious progress delays. By calculating the correlation coefficient of 0.87, it is confirmed that the two are highly correlated. Querying the historical project database, it is found that bottleneck problems in pipeline construction areas similar to the commercial podium have occurred in the past 5 similar projects. The typical evolution path starts from pipeline positioning deviations, leading to pipeline intersections and collisions, and then causing rework, ultimately resulting in overall delays in mechanical and electrical installation. Based on these data, the Monte Carlo simulation is used to calculate the probability distribution of the project achieving the target progress within the next 30 days. The results show that there is a 90% confidence level to reach 85% of the plan, but only a 50% confidence level to reach 92% of the plan. The key risk factors extracted from the risk network include "insufficient installation accuracy of pipeline supports", "frequent design changes", etc. After sorting by urgency and impact scope, a targeted risk prompt list is formed.

[0108] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0109] Read the user role permission configuration table, determine the information access level according to the user identity identifier, filter the sensitive data in the progress prediction results and the risk prompt list, and obtain a targeted information set;

[0110] Based on the targeted information set, combined with the focus parameters set by the user, perform importance scoring on the progress and risk data, and screen out the key information elements;

[0111] Obtain the on-site real-time image stream, perform spatial positioning and attitude correction on the images, and establish a transformation matrix between the image coordinate system and the construction coordinate system;

[0112] According to the transformation matrix, convert the key information elements into layer data, and perform visual arrangement according to the transparency and color coding rules;

[0113] Overlay the layer data onto the real-time image stream, construct an augmented reality scene, and display the progress status and deviation marks at the positions of key components;

[0114] Integrate the augmented reality scene with the decision-making suggestion text, generate a multi-view interaction interface according to the information presentation template, and form a visual construction management decision-making reference.

[0115] Specifically, reading the user role permission configuration table is the primary step in realizing personalized information presentation. In this process, user permission configuration data containing fields such as user ID, role type, department affiliation, scope of responsibilities, and permission level is first obtained through database query or configuration file parsing. When a user logs in to the system and provides identity credentials, the authenticity of the user's identity is confirmed through the authentication module, and the user's unique identifier is extracted. Then, a matching item is searched in the permission configuration table to obtain the user's role type (such as project manager, construction supervisor, quality supervision, safety management, etc.) and permission level (such as levels 1-5). According to the obtained permission level, corresponding data filtering rules are executed to screen the data items in the progress prediction results and risk warning list. Data filtering adopts a label-based access control mechanism, that is, each piece of data is pre-labeled with a sensitivity level and the affiliated department, and by comparing the matching relationship between the user's permissions and the data labels, it is determined which data items are visible to the current user. For example, for the financial risk data in the risk warning list, it is only displayed to users with permissions above the financial supervisor level; for the detailed probability distribution data in the progress prediction, it is only shown to the project management level. Through this refined data filtering process, a targeted information set that conforms to the user's permission scope is generated.

[0116] Screening key information based on the targeted information set is an important step in achieving information focus. This process first reads the focus parameter settings made by the user through the interaction interface. These parameters include the focus area range (such as a specific floor, functional area), project type (such as structure, mechanical and electrical, decoration), time window (such as near term, medium term, long term), and focus dimension (such as schedule delay, quality deviation, safety risk), etc. Then, the matching degree between these parameters and the data item attributes in the targeted information set is calculated. The matching degree calculation uses a multi-dimensional similarity algorithm, comprehensively considering the matching degrees in the spatial dimension, time dimension, and content dimension. Next, the importance of each information item is scored. The scoring formula comprehensively considers four factors: the urgency of the information, the scope of influence, the degree of abnormality, and the user's attention. Weight coefficients are assigned to each factor, and the weighted total score is calculated. Finally, according to the scoring results, the information items are sorted, and the information ranked at the top is selected as the key information elements to ensure that the displayed content not only conforms to the user's focus but also has practical importance.

[0117] Obtaining real-time on-site image streams and performing spatial positioning are the fundamental steps in realizing augmented reality. This process first captures the real-time video stream of the construction site through the camera of a mobile device (such as a tablet computer or a smartphone) or a dedicated augmented reality glasses. To ensure the accuracy of image positioning, multi-source positioning technologies are adopted, including GPS positioning (to determine the geographical coordinates of the device), inertial measurement unit (to record the attitude angles of the device), and computer vision positioning (to determine the relative position by identifying feature points in the scene). Among them, computer vision positioning uses the Simultaneous Localization and Mapping (SLAM) technology, which identifies feature points in video frames, tracks the changes of these feature points in consecutive frames, and calculates the movement trajectory of the camera and the three-dimensional structure of the scene. After identifying a sufficient number of feature points, they are compared with the pre-established three-dimensional model of the construction site through a feature matching algorithm to determine the positional relationship between the current perspective and the standard coordinate system. Finally, the transformation matrix from the image coordinate system (pixel coordinates) to the construction global coordinate system (actual physical coordinates) is calculated. This matrix contains a translation vector and a rotation matrix, which completely describes the mapping relationship between the two coordinate systems.

[0118] Performing spatial mapping of information according to the transformation matrix is the core step in achieving precise visualization. This process first classifies key information elements according to their spatial attributes, including point information (such as component nodes and equipment positions), line information (such as pipeline paths and structural edges), and surface information (such as area progress and wall quality). For each type of information, according to its position data in the construction coordinate system, through the previously calculated transformation matrix, it is mapped to the image coordinate system of the current perspective to determine the exact display position on the screen. At the same time, the field of view range is detected, and only the information elements within the current field of view are processed to reduce the computational burden. Then, according to the type and importance of the information, preset visualization rules are applied to generate corresponding layer data. These rules include color coding (such as using red to represent risks, green to represent normal, and yellow to represent warnings), transparency settings (important information has a high opacity, and secondary information is semi-transparent), icon selection (different types of problems use different icons), and text annotation (key values and short descriptions), etc. Through these rules, abstract data is transformed into intuitive visual elements to form multiple information layers.

[0119] Overlaying layer data onto a real-time image stream is an implementation step in constructing an augmented reality scene. This process uses image synthesis technology to layer information elements one by one according to the layer priority while maintaining the integrity of the original image. The overlay process takes into account perspective effects and occlusion relationships to ensure the visual consistency between the augmented reality elements and the actual scene. For component information that requires precise positioning, such as installation deviation marks, edge adsorption technology is used to make the marks accurately attach to the edges of the components; for regional information, such as progress completion, a semi-transparent heat map is used to cover the corresponding area, and the display effect is adjusted according to the actual lighting conditions to ensure readability in different environments. To handle occlusion problems in real-time images, a depth estimation algorithm is applied to calculate the relative depth of objects in the scene, ensuring that virtual information elements can be correctly occluded by foreground objects and enhancing the sense of reality. At the same time, in response to changes in the user's perspective, the layer position and perspective effect are updated in real-time to ensure that the augmented reality elements always maintain the correct spatial relationship with the actual scene and can maintain a stable display effect even when the user is moving.

[0120] Integrating the augmented reality scene with decision-making advice text is the final step in forming a complete decision-making reference. This process first selects a suitable information presentation template from a preset interface template library according to the user's role characteristics and habitual preferences. The template defines the information layout, interaction methods, and navigation structure. For different types of decision-making needs, multiple view modes are designed, including a global overview view (showing the overall progress and key indicators), a detailed analysis view (showing in-depth analysis of specific problems), a time evolution view (showing past and predicted future trends), and a comparison and evaluation view (showing the differences between the plan and the actual situation) etc. In each view, the augmented reality visualization content is integrated with the corresponding text explanations, data charts, and decision-making advice to form an information-rich and hierarchical interactive interface. To enhance the user experience, intuitive interaction mechanisms, such as touch gestures (zooming in, zooming out, rotating, clicking), voice commands, and gaze tracking, are designed to enable users to easily switch between different views and explore the information they are interested in in-depth. Finally, a visually intuitive and content-rich visualization construction management decision-making reference is formed to provide the required decision-making support for managers at different levels.

[0121] Taking a large public building construction project as an example, when the project supervisor logs in to the progress monitoring system through a mobile tablet device, the system first reads the user permission configuration information, confirms that the supervisor has the role of a supervision engineer with a permission level of 3, responsible for structural and mechanical and electrical supervision work. Based on this identity information, the system filters out 1,200 data records related to the structure and mechanical and electrical engineering from the progress prediction results and risk warning list containing more than 2,500 data items, and at the same time eliminates sensitive information such as contract disputes and cost overruns that require a higher permission level to view, forming a targeted information set. The system reads the key focus parameters set by the supervision engineer, including "Focus area: Central Hall", "Project type: Mechanical and Electrical", "Time window: The recent two weeks", and "Focus dimension: Installation quality". According to these parameters, the system calculates the matching degree and importance score for the data items in the targeted information set, and screens out the top 20 key information elements, including specific problem points such as "High risk of collision of air-conditioning pipes in the central hall". When the supervision engineer walks to the construction area of the central hall with the tablet device, the system captures the real-time image stream through the device camera, and at the same time uses SLAM technology to identify more than 50 feature points in the environment, matches them with the pre-established BIM model, and accurately calculates the conversion relationship between the image coordinate system and the construction coordinate system under the current perspective. Based on this transformation matrix, the system converts the identified key information elements (such as pipe collision points, bracket deviations, etc.) into layer data, and emphasizes them with red highlights and flashing icons according to their importance. These layer data are superimposed on the real-time image stream, and deviation and risk marks are displayed above the actual pipe installation position, and detailed text explanations and processing suggestions are provided at the edge of the screen. Through this intuitive visual presentation, the supervision engineer immediately discovers the installation deviation problems at two pipe intersection points, and according to the decision-making suggestions provided by the system, guides the construction team to make adjustments on site, effectively avoiding possible subsequent collision conflicts and rework.

[0122] The above describes the construction progress monitoring method based on big data in the embodiments of the present application. Next, the construction progress monitoring system based on big data in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the construction progress monitoring system based on big data in the embodiments of the present application includes:

[0123] An evaluation module, configured to obtain construction site images, videos, and point clouds through multi-source devices, filter and quality-evaluate the collected data, and obtain a spatio-temporally aligned original construction dataset;

[0124] A registration module, configured to perform feature extraction and registration on the image sequence and point cloud data according to the original construction dataset, spatially align the measured data with the design model, and obtain a digital construction site model containing geometric and time information;

[0125] A classification module for performing object detection and classification on on-site components based on the digital model of the construction site, analyzing the spatial position relationships between components, and obtaining a construction completion status table with component identifiers and attributes;

[0126] A mapping module for establishing a mapping relationship between actual components and construction plan tasks according to the construction completion status table, calculating the completion degree of components and the position deviation value, and obtaining a construction progress assessment report and component-level deviation data;

[0127] An analysis module for analyzing the progress change trend according to the construction progress assessment report and component-level deviation data, identifying abnormal patterns and potential risk points during the construction process, and obtaining a progress prediction result and a risk warning list;

[0128] A screening module for screening the progress prediction result and the risk warning list according to user permissions and key concerns, superimposing and displaying the progress and deviation information in the on-site real scene, and obtaining a visual construction management decision reference.

[0129] Through the collaborative cooperation of the above-mentioned various components, images, videos, and point cloud data of the construction site are obtained through multi-source devices, achieving comprehensive and continuous perception of the construction site, breaking through the limitations of traditional single data sources, and improving the integrity and diversity of the original data; by filtering and quality evaluating the collected data, the accuracy and reliability of the data are ensured, laying a solid foundation for subsequent analysis; the spatio-temporal aligned construction original data set solves the problem of multi-source heterogeneous data fusion, realizing the unified representation and processing of different types of data. Based on the feature extraction and registration technology of the original data set, the measured data can be accurately corresponded to the design model, forming a digital model of the construction site containing geometric and time information, creating conditions for the comparison between the actual and the plan; compared with the traditional BIM model, this digital model not only contains static design information, but also incorporates the measured data of the dynamic construction process, truly reflecting the actual state and changes of the construction site. Through deep learning algorithms, object detection and classification of on-site components are carried out, automatically identifying the types and attributes of components, greatly improving the accuracy and efficiency of component recognition. Especially for component recognition in complex environments, the feature extraction ability of deep neural networks is significantly superior to traditional computer vision methods; the spatial relationship analysis algorithm makes the position constraints and connection relationships between components clearly visible, forming a construction completion status table with component identifiers and attributes, providing accurate basic data for progress assessment. The method of establishing a mapping relationship between the actual components and the construction plan tasks realizes the conversion from the physical completion state to the planned completion ratio, solving the problems of strong subjectivity and poor consistency in traditional progress assessment; by quantifying the component installation accuracy through the bounding box occupancy rate and point-plane distance evaluation methods, the construction quality assessment is transformed from qualitative description to quantitative indicators, and the progress assessment report and component-level deviation data provide an objective basis for management decisions. The progress change trend analysis method based on time series analysis and machine learning can mine potential laws from historical data, identify abnormal patterns and risk points; the risk propagation network established by combining historical project experience data significantly enhances the accuracy and forward-looking of risk prediction. The progress prediction results and risk warning list provide managers with the opportunity to plan ahead. Finally, through augmented reality technology, the progress and deviation information are intuitively superimposed and displayed on the on-site real scene. Combined with the personalized information screening mechanism of user permissions and focus of attention, the complex progress data becomes intuitive and easy to understand. Decision-makers can directly obtain the required information on-site and make responses, greatly improving the decision-making efficiency and accuracy.

[0130] Referring to Figure 3 , in the embodiment of the present invention, a computer device is further provided. This computer device can be a server, and its internal structure can be as Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected by a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0131] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0132] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0133] The above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A construction progress monitoring method based on big data, characterized in that The construction progress monitoring method based on big data includes: Obtaining construction site images, videos, and point clouds through multi-source devices, filtering and quality evaluating the collected data to obtain a construction original data set with spatio-temporal alignment; According to the construction original data set, extracting features and registering the image sequence and point cloud data, spatially aligning the measured data with the design model to obtain a construction site digital model containing geometric and time information; Based on the construction site digital model, performing object detection and classification on on-site components, analyzing the spatial position relationship between components to obtain a construction completion status table with component identifiers and attributes; According to the construction completion status table, establishing a mapping relationship between the actual components and the construction plan tasks, calculating the component completion degree and position deviation value to obtain a construction progress evaluation report and component-level deviation data; According to the construction progress evaluation report and component-level deviation data, analyzing the progress change trend, identifying abnormal patterns and potential risk points during the construction process to obtain a progress prediction result and a risk warning list; Screening the progress prediction result and the risk warning list according to user permissions and key concerns, overlaying and displaying the progress and deviation information in the on-site real scene to obtain a visual construction management decision-making reference.

2. The construction progress monitoring method based on big data according to claim 1, wherein, The step of obtaining construction site images, videos, and point clouds through multi-source devices, filtering and quality evaluating the collected data to obtain a construction original data set with spatio-temporal alignment includes: Collecting image sequences, video streams, and three-dimensional point cloud data of the construction site through high-definition cameras, unmanned aerial vehicle aerial photography systems, and laser scanners; Performing distortion correction and enhancement processing on the collected image sequence, removing image noise and redundant information to generate a standardized image set; Performing spatial registration and noise reduction processing on the collected point cloud data, removing outlier points and duplicate points to form structured point cloud data; Performing dynamic task scheduling on the collection devices through edge computing nodes, automatically adjusting the data collection frequency and range according to the on-site environmental conditions; Performing quality scoring on the processed data based on multi-dimensional quality indicators, screening out valid data sets that meet the threshold requirements; Aligning and fusing the screened image set, point cloud data, and sensor information according to time stamps and spatial positions, establishing a unified indexing mechanism to generate a construction original data set with spatio-temporal alignment.

3. The construction progress monitoring method based on big data according to claim 1, characterized in that, The step of according to the construction original data set, extracting features and registering the image sequence and point cloud data, spatially aligning the measured data with the design model to obtain a construction site digital model containing geometric and time information includes: Extracting feature points and feature descriptors of the image sequence from the construction original data set, establishing a matching relationship between images to generate a sparse reconstructed point cloud; Performing densification processing on the sparse reconstructed point cloud, filling in the missing spatial regions to construct continuous surface geometric information; Extracting main geometric features such as planes, line segments, and corner points from the point cloud data to form a structural feature set of the construction scene; Pairing and comparing the structural feature set with the geometric features of the building information design model, calculating the transformation matrix, and correcting the spatial position of the measured data; Sort and organize the data collected at different times through timestamps to establish a temporal relationship network of the construction progress. Integrate the corrected geometric data with the temporal relationship network, attach time attribute tags to each spatial point, and generate a digital model of the construction site containing geometric and time information.

4. The construction progress monitoring method based on big data according to claim 1, characterized in that Based on the digital model of the construction site, perform object detection and classification on the on-site components, analyze the spatial position relationships between the components, and obtain a construction completion status table with component identifiers and attributes, including: Extract the feature maps of the component regions from the digital model of the construction site, outline the component contours through edge detection, and form candidate regions for component segmentation. Calculate the feature vectors of the candidate regions, determine the component types according to the preset component classification criteria, and generate a set of components with type tags. Based on the set of components with type tags, calculate the three-dimensional coordinates, orientations, and dimension parameters of each component, and establish a geometric description document for the components. Perform spatial association analysis on the components in the geometric description document, identify the connection relationships and relative position constraints between the components, and form a component topology network diagram. Extract the component installation standards and quality requirements from the design specification library, compare them with the geometric parameters of the actual components, and calculate the installation deviation values and completion status identifiers. Integrate the type tags of the components, the geometric description document, the component topology network diagram, the installation deviation values, and the completion status identifiers into structured data records, and organize them into a construction completion status table with component identifiers and attributes.

5. The construction progress monitoring method based on big data according to claim 1, wherein, Based on the construction completion status table, establish a mapping relationship between the actual components and the construction plan tasks, calculate the component completion degrees and position deviation values, and obtain a construction progress evaluation report and component-level deviation data, including: Import the construction plan data from the project management system, extract the work breakdown structure and key node milestones, and generate a construction task dependency network. Establish a corresponding relationship between the component records in the construction completion status table and the plan items in the construction task dependency network through component code matching to form a component-task mapping table. Based on the component-task mapping table, count the completion status of the components involved in each task item, and calculate the task completion rate and progress weight values. Calculate the difference between the geometric parameters and design parameters of the components in the construction completion status table, quantify the component installation accuracy through the boundary box occupancy rate and point-plane distance evaluation methods, and generate a component position deviation matrix. According to the task completion rate and progress weight values, combined with the project critical path, comprehensively analyze the overall progress status and the achievement status of the key nodes, and compile a construction progress evaluation report. Classify and summarize the component position deviation matrix according to the component type, region, and deviation degree, and generate component-level deviation data including a deviation distribution map and hot spots.

6. The construction progress monitoring method based on big data according to claim 1, wherein Based on the construction progress evaluation report and the component-level deviation data, analyze the progress change trend, identify abnormal patterns and potential risk points during the construction process, and obtain a progress prediction result and a risk warning list, including: Extract the historical progress data points from the construction progress evaluation report, construct a time series data set, and draw a progress change curve graph. Perform trend decomposition on the progress change curve graph, separate seasonal factors, long-term trends, and random fluctuation components, and obtain progress fluctuation characteristics; Combine the component-level deviation data to establish a deviation-progress correlation matrix, identify the correlation between deviation accumulation areas and progress delays, and form a construction bottleneck analysis report; Compare the construction bottleneck analysis report with the data in the historical project database, extract the risk evolution paths of similar scenarios, and establish a risk propagation network; Based on the progress fluctuation characteristics and the risk propagation network, calculate the progress completion probability distribution within the future time window, and generate a progress prediction result containing multiple confidence intervals; Extract risk factors and their influence weights from the risk propagation network, rank them according to urgency and influence scope, and form a risk warning list.

7. The construction progress monitoring method based on big data according to claim 1, characterized in that, Screen the progress prediction result and the risk warning list according to user permissions and key concerns, overlay and display the progress and deviation information in the on-site real scene, and obtain a visual construction management decision-making reference, including: Read the user role permission configuration table, determine the information access level according to the user identity identifier, filter the sensitive data in the progress prediction result and the risk warning list, and obtain a targeted information set; Based on the targeted information set, combined with the key concern parameters set by the user, perform importance scoring on the progress and risk data, and screen out key information elements; Obtain the on-site real-time image stream, perform spatial positioning and attitude correction on the images, and establish a conversion matrix between the image coordinate system and the construction coordinate system; According to the conversion matrix, convert the key information elements into layer data, and perform visual arrangement according to the transparency and color coding rules; Overlay the layer data onto the real-time image stream, construct an augmented reality scene, and display the progress status and deviation marks at the positions of key components; Integrate the augmented reality scene and the decision-making advice text, generate a multi-view interaction interface according to the information presentation template, and form a visual construction management decision-making reference.

8. A construction progress monitoring system based on big data, which is used to implement the construction progress monitoring method based on big data according to any one of claims 1-7, and is characterized in that, The construction progress monitoring system based on big data includes: An evaluation module for obtaining construction site images, videos, and point clouds through multi-source devices, filtering and quality evaluating the collected data, and obtaining a spatially and temporally aligned construction original data set; A registration module for extracting features and registering the image sequence and point cloud data according to the construction original data set, spatially aligning the measured data with the design model, and obtaining a construction site digital model containing geometric and time information; A classification module for performing target detection and classification on on-site components based on the construction site digital model, analyzing the spatial position relationship between components, and obtaining a construction completion status table with component identifiers and attributes; A mapping module for establishing a mapping relationship between actual components and construction plan tasks according to the construction completion status table, calculating the component completion degree and position deviation value, and obtaining a construction progress evaluation report and component-level deviation data; An analysis module for analyzing the progress change trend according to the construction progress evaluation report and component-level deviation data, identifying abnormal patterns and potential risk points during the construction process, and obtaining a progress prediction result and a risk warning list. A screening module for screening the progress prediction results and risk warning lists according to user permissions and key concerns, superimposing and displaying progress and deviation information in the on-site real scene, and obtaining a visual construction management decision-making reference.

9. A computer device, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the big data-based construction progress monitoring method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon. When the computer program is run by the processor, the processor is caused to execute the big data-based construction progress monitoring method described in any one of claims 1 to 7.

Images