BIM construction progress dynamic tracking method based on multi-modal data fusion

Abstract

The invention discloses a BIM construction progress dynamic tracking method based on multi-modal data fusion, and the method comprises the steps: obtaining BIM data, component progress data and visual data of each building component in a construction region in real time, and carrying out the unified space-time reference and data cleaning processing of the component progress data and the visual data; performing feature extraction on the processed data and the BIM data to obtain a standardized spatial-temporal feature data set, constructing a construction process digital twin map according to the data set, performing spatial-temporal convolutional network processing to obtain a component construction progress state reasoning result of each building component, setting a threshold value, and obtaining a construction progress state reasoning result of each building component; and comparing the construction progress state reasoning result to obtain a component risk assessment result, taking measures according to the risk assessment result, executing the measures, and simultaneously obtaining the executed multi-modal data to update the digital twin map. According to the invention, through multi-modal data fusion, dynamic tracking and effective management of the construction progress are realized, the construction risk can be found and processed in time, and the construction management efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of building information modeling technology, and in particular to a BIM construction progress dynamic tracking method based on multimodal data fusion. Background Art

[0002] In the field of construction project management, dynamic tracking and precise control of construction progress are core requirements for ensuring project schedules and optimizing resource allocation. In existing technologies, construction progress management mainly relies on manual records combined with static analysis of BIM models, which makes it difficult to integrate multi-source heterogeneous data in real time and comprehensively, resulting in problems in construction progress tracking.

[0003] Traditional methods lack efficient calibration mechanisms for the spatial coordinates of the BIM model, the local coordinates of UWB positioning, and the geodetic coordinates of drone vision data. This leads to insufficient time synchronization accuracy, resulting in large errors in data association and matching, making it difficult to form standardized spatiotemporal feature representations, which seriously affects the accuracy of subsequent analysis. Furthermore, existing technologies for assessing component construction progress status often rely on a single data modality or simple statistical analysis. This fails to effectively capture the coupling relationship between process logic dependencies and time series data, making it difficult to proactively predict risks such as progress deviations and resource conflicts. Furthermore, after risk assessment, existing methods lack precise adjustment strategies driven by multidimensional data, provide slow feedback on the effects of implemented measures, and are unable to dynamically update the model, resulting in insufficient real-time and adaptability in progress control. Therefore, a construction progress tracking method is urgently needed that can unify the spatiotemporal benchmarks of multimodal data, intelligently infer dynamic features, and enable closed-loop management of progress adjustments to enhance intelligent management in complex construction scenarios. To this end, a BIM construction progress dynamic tracking method based on multimodal data fusion is proposed. Summary of the Invention

[0004] The present invention aims to provide a BIM construction progress dynamic tracking method based on multimodal data fusion to solve the problems raised in the above background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for dynamically tracking BIM construction progress based on multimodal data fusion includes the following steps:

[0007] S1. Acquire BIM data, component progress data, and visual data of each building component in the construction area in real time, and unify the spatiotemporal benchmarks for the component progress data and visual data of each building component;

[0008] S2. Performing data cleaning on the component progress data and visual data of each building component that has been subjected to the unified spatiotemporal reference, obtaining cleaned component progress data and visual data for each building component, performing feature extraction on the BIM data of each building component and the corresponding cleaned component progress data and visual data, obtaining a standardized spatiotemporal feature dataset for each building component, and storing the dataset in a database;

[0009] S3. Based on the standardized spatiotemporal feature dataset for each building component, a digital twin of the construction process is constructed. This digital twin is processed using a spatiotemporal convolutional network to obtain inference results on the construction progress status of each building component.

[0010] S4. Setting a component construction progress status inference threshold, comparing the component construction progress status inference result of each building component with the component construction progress status inference threshold, and obtaining a component risk assessment result of each building component;

[0011] S5. Take and execute measures based on the component risk assessment results of each building component. At the same time, obtain multimodal data of each building component in the construction area after the measures are implemented, and update the digital twin diagram of the construction process.

[0012] Preferably, the method for obtaining BIM data, component progress data and visual data of each building component in the construction site in real time is:

[0013] Obtain BIM data of each building component in real time from the BIM of the construction area through the BIM platform API, including component geometry information, WBS process tree, planned construction period and resource requirements;

[0014] Three UWB base stations are deployed as anchor points at the construction site. UWB tags are installed on each building component. They have built-in MCU and wireless modules and support TOF ranging. The UWB tags periodically send pulse signals, and the UWB base stations receive the signals and record the propagation time t. The distance between the UWB base stations and the UWB tags is calculated using the TOF algorithm to obtain the UWB local coordinates of each building component. At the same time, construction workers manually fill in the component installation progress, component material quantity and reporting time of each building component through the mobile APP. The installation progress and component material quantity cannot exceed 100% to avoid invalid input. Based on this, the component progress data including the component UWB local coordinates, component completion progress, component material quantity and reporting time of each building component are obtained;

[0015] The method for calculating the distance between the UWB base station and the UWB tag by the TOF algorithm to obtain the UWB local coordinates of each building component is as follows: recording the propagation time t when receiving the signal through the UWB base station, multiplying the propagation time t by the speed of light, and obtaining the distances d1, d3 and d3 from the UWB tag to the three UWB base stations; and at the same time, obtaining the UWB local coordinates of each building component by the trilateration formula based on the base station coordinates of the three UWB base stations;

[0016] The three-sided positioning formula is:

[0017]

[0018] Among them, d1, d3 and d3 are the distances d1, d3 and d3 from the UWB tag to the three UWB base stations, respectively. i ,y i , z i ) is the base station coordinate of the UWB base station, i is 1, 2 or 3, (x u ,y u , z u ) is the UWB local coordinate of the building component.

[0019] Using drones and their onboard visible light cameras, the team captured images of each building component in the construction area, along with the image's corresponding acquisition time and geodetic coordinates. The geodetic coordinates included GPS longitude and latitude (Lat, Lon) and altitude (Alt). This data allowed them to obtain visual data for each building component, including the component image, the image's corresponding acquisition time, and the geodetic coordinates.

[0020] The component geometric information includes component BIM coordinates, component ID, component size, and component type.

[0021] Preferably, the method for unifying the spatiotemporal benchmark of the component progress data and visual data of each building component is:

[0022] Connect the UWB base station, drone, and mobile app to the NTP server through the local area network to form a time synchronization network. The NTP server sends synchronization acquisition instructions to the devices to ensure that the UWB base station, drone camera, and mobile app collect data at the same timestamp, avoiding time dislocation caused by asynchronous acquisition and achieving a unified time base for component progress data and visual data.

[0023] The component UWB local coordinates in the component progress data of each building component are converted into BIM coordinates corresponding to the component progress data through the Bursa model;

[0024] The Bursa model is:

[0025]

[0026] in, It is the BIM coordinate corresponding to the component progress data. is the component UWB local coordinate, R is the rotation matrix, k is the scale factor, is the translation parameter in the Bursa model;

[0027] The geodetic coordinates (Lat, Lon) and Alt corresponding to the component image in the visual data of each building component are converted into projection coordinates (X P , Y P ), elevation Z P Keep it as Alt, and translate the projection coordinates to obtain the visual BIM coordinates corresponding to the visual data, thereby obtaining the component progress data and visual data of each building component after the unified time and space reference;

[0028] The translation operation is:

[0029]

[0030] Among them, X g 、Y g and Z g is the visual BIM coordinate corresponding to the visual data, X P and Y P is the projection coordinate, Z P It's Alt, X u tm0、Y u tm0 and Z u tm0 is the coordinate of the global origin of the BIM model.

[0031] The UTM projection is the Universal Transverse Mercator projection, which is a conformal transverse cylindrical projection.

[0032] Preferably, the method of performing data cleaning processing on the component progress data and visual data of each building component after the unified time and space reference is:

[0033] The component progress data of each building component that has been standardized in time and space are monitored for outliers using the 3σ principle. Outliers exceeding [μ-3σ, μ+3σ] are detected and deleted. μ and σ are the mean and standard deviation of the component progress data of each building component, respectively. The component progress data after deleting outliers are filled with missing values using linear interpolation to obtain the component progress data after data cleaning.

[0034] The linear interpolation method is a mathematical method that uses a linear relationship to estimate the value of an unknown point between two known data points.

[0035] For each pixel point of the component image in the visual data of each building component in the unified spatiotemporal benchmark, three rows and three columns centered on each pixel point are selected as the pixel area of the pixel point. The pixel area of each pixel point is calculated using the Gaussian weighted sum formula to obtain the pixel value of each pixel point after Gaussian filtering and denoising. Based on this, denoising processing is achieved for the visual data of each building component in the unified spatiotemporal benchmark. The grayscale histogram of the denoised component image is statistically analyzed to obtain the pixel frequency of each grayscale level. The original grayscale of the component image is mapped to the new grayscale using the cumulative distribution function to obtain the component progress data after data cleaning.

[0036] The Gaussian weighted sum formula is:

[0037]

[0038] Where G(x,y) is the pixel value of the component image at coordinate (x,y) after Gaussian filtering and denoising, I(i,j) is the pixel value of the component image at coordinate (i,j), and σ is 1.5;

[0039] The cumulative distribution function is:

[0040]

[0041] Where L is the total number of gray levels, p(r i ) is the gray value r in the component image i The frequency of pixel appearance, s k After histogram equalization processing, the gray value in the component image is r k The pixel points are mapped to the grayscale values in the new image, and k is the grayscale range.

[0042] Preferably, the method of performing feature extraction processing on the BIM data of each building component, and the component progress data and visual data after corresponding data cleaning to obtain a standardized spatiotemporal feature data set for each building component is:

[0043] The component images in the visual data of each building component after data cleaning are monitored through the YOLO algorithm to obtain the component completion and component type. At the same time, the visual BIM coordinates and acquisition time are extracted from the visual data of each building component after data cleaning. Based on this, the visual features of each building component including acquisition time, component completion, component type and visual BIM coordinates are obtained;

[0044] The YOLO algorithm is an efficient target detection model;

[0045] Extract the component installation progress, component material quantity, and reporting time from the component progress data of each building component after data cleaning. Calculate the component installation progress and reporting time using the construction progress speed formula to obtain the component construction speed. Simultaneously, calculate the component material quantity and component installation progress using the unit progress material consumption formula to obtain the component unit progress material consumption. Extract the BIM coordinates from the component progress data of each building component after data cleaning, and thereby obtain the component progress characteristics of each building component, including reporting time, component unit progress material consumption, component construction speed, and BIM coordinates.

[0046] The construction progress speed formula is:

[0047]

[0048] Among them, R is the component construction speed, P2 and P1 are the component installation progress corresponding to different filling times, and t2 and t1 are the filling times;

[0049] The formula for unit progress material consumption is:

[0050]

[0051] Among them, P2 and P1 are the component installation progress corresponding to different reporting times, M2 and M1 are the component material consumption corresponding to the component installation progress at different reporting times, and C is the component unit progress material consumption;

[0052] Directly extract component BIM coordinates, component ID, component type, WBS process tree, planned construction period and resource requirements from the BIM data of each building component. At the same time, according to the WBS process tree, analyze the logical dependencies between the processes to which the building components belong, and obtain the logical dependencies between components. Based on this, the BIM features of each building construction, including component BIM coordinates, component ID, planned construction period, logical dependencies between components and resource requirements, are obtained.

[0053] Based on the visual BIM coordinates and acquisition time in the visual features, the BIM coordinates and reporting time in the component progress features, and the component BIM coordinates in the BIM features, the component BIM coordinates are used as a benchmark to ensure that the visual BIM coordinates of the visual features, the BIM coordinates of the component progress features, and the component BIM coordinates of the BIM features are consistent. At the same time, based on the acquisition time or reporting time as the time benchmark, the BIM features, visual features, and component progress features of each building component are matched and merged to obtain a standardized spatiotemporal feature dataset for each building component.

[0054] The standardized spatiotemporal feature dataset includes timestamp, component completion, component type, component unit progress material consumption, component construction speed, component BIM coordinates, component ID, planned construction period, logical dependencies between components, and resource requirements;

[0055] The timestamp is the acquisition time.

[0056] Preferably, the method for constructing a digital twin diagram of the construction process based on the standardized spatiotemporal feature dataset of each building component is:

[0057] The component ID in the standardized spatiotemporal feature dataset of each building component is used as the unique node identifier, and the logical dependency relationship between components in the standardized spatiotemporal feature dataset of each building component is used as the directed edge connecting nodes. At the same time, according to the component ID, the timestamp, component completion degree, component type, component unit progress material consumption, component construction speed, planned construction period and resource requirements in the standardized spatiotemporal feature dataset of each building component are used as node attributes to be associated with the node corresponding to the component ID, thereby constructing a digital twin diagram of the construction process.

[0058] Preferably, the method of processing the digital twin graph of the construction process through a spatiotemporal convolutional network to obtain the inference result of the component construction progress status of each building component is:

[0059] Based on the digital twin graph of the construction process, the node feature matrix X and adjacency matrix A of the digital twin graph of the construction process are obtained. The node feature matrix and adjacency matrix A are input into the spatiotemporal convolutional network. The spatiotemporal convolutional network integrates the data processing capabilities of spatial and temporal dimensions. The spatial dimension uses the graph convolution layer to extract the dependency relationship features between nodes. The temporal dimension uses the time convolution layer to process the timestamp sequence to capture the changing trend of component status over time. After multi-layer spatiotemporal feature fusion, the fully connected layer outputs the component construction progress status inference results of each building component, including component completion deviation ΔC, process-level time deviation ΔT, and resource conflict index RCI.

[0060] Preferably, the method for obtaining the component risk assessment results of each building component is:

[0061] If ΔC ≥ -5%, the component completion deviation ΔC of the building component is normal;

[0062] If -15% ≤ ΔC < -5%, the component completion deviation ΔC of the building component is low risk;

[0063] If -30% ≤ ΔC < -15%, the component completion deviation ΔC of the building component is medium risk;

[0064] If ΔC<-30%, the component completion deviation ΔC of the building component is high risk;

[0065] If ΔT < 3, the process-level time deviation ΔT of the building component is normal.

[0066] If 3≤ΔT<5, the process-level time deviation ΔT of the building component is low risk;

[0067] If 5≤ΔT<7, the process-level time deviation ΔT of the building component is medium risk;

[0068] If ΔT ≥ 7, the process-level time deviation ΔT of the building component is high risk;

[0069] If RCI<04, the resource conflict index RCI of the building component is normal.

[0070] If 0.4≤RCI<0.6, the resource conflict index RCI of the building component is low risk;

[0071] If 0.6≤RCI<0.8, the resource conflict index RCI of the building component is medium risk;

[0072] If RCI ≥ 0.8, the resource conflict index RCI of the building component is high risk;

[0073] Based on this, the component completion deviation ΔC risk assessment result, process-level time deviation ΔT risk assessment result and resource conflict index RCI risk assessment result of each building component are obtained. The highest risk assessment result among the component completion deviation ΔC risk assessment result, process-level time deviation ΔT risk assessment result and resource conflict index RCI risk assessment result is taken as the component risk assessment result of each building component.

[0074] Preferably, the method of taking measures according to the component risk assessment results of each building component and obtaining multimodal data of each building component in the construction area after the measures are implemented to update the digital twin diagram of the construction process is as follows:

[0075] Based on the component risk assessment results of each building component, select the building components with high risk assessment results. For these high-risk building components, prioritize resource allocation for construction according to the resource requirements in the building component's BIM data until the planned construction period in the building component's BIM data is met;

[0076] For building components with medium risk, low risk, or normal risk as assessed, no measures will be taken. The construction progress will be deemed to be within a controllable range, with no need for additional resource allocation or process adjustments. Only routine monitoring will be maintained.

[0077] Obtain multimodal data for each building component in the construction area after execution, process the multimodal data of each building component accordingly, and use it to update the digital twin diagram of the construction process. Synchronize the processed multimodal data to the corresponding node of the digital twin diagram. If the execution of measures for high-risk building components leads to changes in the logic of subsequent processes, update the directed edges between nodes.

[0078] The multimodal data includes BIM data, component progress data and visual data.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. The present invention is based on multimodal data fusion and spatiotemporal benchmark unification technology to achieve high-precision dynamic tracking of construction progress. By integrating BIM data, component progress data located by UWB, and drone vision data, a multi-source heterogeneous data fusion system is constructed, which solves the problems of single data source and insufficient real-time performance of traditional methods. The NTP server is used to achieve multi-device time synchronization. Combined with the Bursa model and UTM projection technology, different modal data are unified to the BIM global spatiotemporal benchmark, eliminating analysis deviations caused by coordinate misalignment and time asynchrony. The YOLO algorithm is used to identify component completion and UWB real-time positioning of component three-dimensional coordinates, forming a double verification with BIM design data, greatly improving the accuracy of construction progress tracking and providing high-precision real-time data support for complex construction scenarios.

[0081] 2. The present invention improves the construction risk prediction capability and resource scheduling efficiency, constructs a dynamic digital twin graph based on a standardized spatiotemporal feature data set, and integrates process logic dependencies and time series dynamic features through a spatiotemporal convolutional network to achieve intelligent reasoning of component completion deviation, process-level time deviation, and resource conflict index, solving the problem of complex construction logic being difficult to model in the existing technology, thereby improving the accuracy of risk identification. At the same time, through multi-dimensional risk threshold setting, the BIM data-driven resource allocation strategy is automatically triggered for high-risk components, and the digital twin model is updated through real-time data collection after execution, forming a closed-loop management of monitoring, evaluation, adjustment, and feedback, thereby reducing the delay rate of high-risk components and improving resource utilization, effectively shortening the construction period and reducing management costs, and realizing intelligent control and dynamic optimization of construction progress.

[0082] Figures in the specification

[0083] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0084] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0085] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0086] Examples, such as Figure 1 As shown, a method for dynamic tracking of BIM construction progress based on multimodal data fusion includes the following steps:

[0087] S1. Acquire BIM data, component progress data, and visual data of each building component in the construction area in real time, and unify the spatiotemporal benchmarks for the component progress data and visual data of each building component;

[0088] S2. Performing data cleaning on the component progress data and visual data of each building component that has been subjected to the unified spatiotemporal reference, obtaining cleaned component progress data and visual data for each building component, performing feature extraction on the BIM data of each building component and the corresponding cleaned component progress data and visual data, obtaining a standardized spatiotemporal feature dataset for each building component, and storing the dataset in a database;

[0089] S3. Based on the standardized spatiotemporal feature dataset for each building component, a digital twin of the construction process is constructed. This digital twin is processed using a spatiotemporal convolutional network to obtain inference results on the construction progress status of each building component.

[0090] S4. Setting a component construction progress status inference threshold, comparing the component construction progress status inference result of each building component with the component construction progress status inference threshold, and obtaining a component risk assessment result of each building component;

[0091] S5. Take and execute measures based on the component risk assessment results of each building component. At the same time, obtain multimodal data of each building component in the construction area after the measures are implemented, and update the digital twin diagram of the construction process.

[0092] Furthermore, the working principle of the present invention is described below by way of examples:

[0093] Taking the steel column hoisting construction link of a certain prefabricated high-rise building project as an example, an H-shaped steel column component numbered COL-001 is selected, with a designed height of 8m and a planned construction period of 8 hours, to demonstrate the dynamic tracking process of the construction progress of the present invention.

[0094] The BIM data of COL-001 is obtained in real time through the Revit API. The BIM data includes geometric information and progress information. The geometric information includes BIM coordinates (10.5, 15.2, 0), component ID (COL-001), component type (H-shaped steel column), and progress information includes WBS process tree (main structure → steel column hoisting), planned construction period (2025-04-28T08:00:00~16:00:00), and resource requirements (2 50-ton truck cranes and 10 installation workers). For component progress data, 3 UWB base stations are deployed and the steel column installation is carried out. UWB tag, the coordinates of the three UWB base stations are (0, 0, 0), (50, 0, 0) and (0, 50, 0), respectively. The distances from the tag to the base station are measured by the TOF algorithm to be 12.3m, 42.8m and 37.5m respectively. The UWB local coordinates are solved by the three-sided positioning method to be (10.2, 15.5, 0.1). At the same time, the construction personnel filled in the following information through the mobile APP at 12:00: the installation progress is 60%, the material quantity is 80 sets of high-strength bolts and the filling time is 2025-04-28T12:00:00. For the visual data, DJI A Mavic 3 drone captured an image of the steel column at 12:05 PM, obtaining the geodetic coordinates (Lat = 30.0001°, Lon = 120.0002°, Alt = 10 m) and the acquisition time (2025-04-28T12:05:00). The UWB base station, drone, and mobile app were connected to the NTP server, using 2025-04-28T12:00:00 as the unified time base. Using the Bursa model, the UWB local coordinates were converted to BIM coordinates (10.5, 15.2, 0), which are consistent with the design coordinates. The geodetic coordinates were then converted to plane coordinates (333500, 3320000) using the UTM projection. After translation based on the BIM global origin coordinates (333490, 3319985, 0), the visual BIM coordinates were (10, 15, 10), which matched the actual installation area of the steel column.

[0095] The 3σ principle was used to detect some abnormal distance data of the component progress data, which deviated from the mean by 3.5σ. After deletion, the missing material quantity data was filled by linear interpolation to obtain the cleaned data: installation progress 60%, material quantity 80 sets and BIM coordinates (10.5, 15.2, 0). The steel column image in the visual data was denoised by Gaussian filtering, and the contrast was enhanced by histogram equalization to clearly show the bolt connection status. For visual features, the YOLO algorithm was used to identify that the completion degree of the steel column image was 60% and the component type was H-shaped steel column. At the same time, the visual BIM coordinates (10.5, 15.2, 0) were extracted. The collection time is 12:05; for the component progress characteristics, based on the 30% progress and 40 sets of materials reported at 10:00, the construction speed R is calculated to be 15% / hour, and the unit progress material consumption C is 133.3 sets / 100 using the construction progress speed formula and the unit progress material consumption formula. At the same time, the BIM coordinates (10.5, 15.2, 0) and the reporting time 12:00 are extracted; for the BIM features, the component ID, planned construction period, logical dependency (previous process foundation leveling completed), and resource requirements are extracted. Based on the BIM coordinates and collection time, the three types of features are merged to form a standardized spatiotemporal feature dataset.

[0096] A digital twin graph of the local construction process was constructed using the component ID (COL-001) as a node and its temporal and spatial attributes. Logical dependencies (from foundation leveling to steel column hoisting) were used as directed edges. The node feature matrix X and the adjacency matrix A were obtained from this digital twin graph. These matrixes were then input into the ST-GCN model. The resulting inference results were: a completion deviation of -15%, a process-level time deviation of 2 hours, meaning the actual progress was 2 hours behind schedule, and a resource conflict index (RCI) of 0.5, indicating that only one mobile crane was currently in use, posing a risk of insufficient resources.

[0097] According to the comparison of the set threshold settings, the risk assessment result of the completion deviation of the steel column component is low risk, the risk assessment result of the process-level time deviation of the steel column component is normal, and the risk assessment result of the resource conflict index of the steel column component is low risk. The highest risk level is taken as low risk, and the component risk assessment result of the steel column component is low risk.

[0098] According to the low risk assessment result of the steel column component, the routine monitoring strategy was triggered, and the team was reminded through the mobile APP to add a mobile crane. At 13:00, the construction party dispatched a mobile crane to the steel column lifting area, and the total resources invested reached 2 units. The construction speed increased to 20% / hour. At 14:00, the component progress data and visual data of the new steel column component were obtained. The component progress data showed a completion rate of 80%, 110 sets of materials, and no deviation in the BIM coordinates. The visual data showed a completion rate of 80%, and there was no abnormality in the bolt connection. After time and space calibration, the node attributes of the digital twin diagram were updated, and the risk level was reduced to normal.

[0099] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Modifications to the technical solutions described in the aforementioned embodiments, or equivalent replacements of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A BIM construction progress dynamic tracking method based on multimodal data fusion, characterized in that: The following steps are involved: S1. Acquire BIM data, component progress data, and visual data of each building component in the construction area in real time, and unify the spatiotemporal benchmarks for the component progress data and visual data of each building component; S2. Performing data cleaning on the component progress data and visual data of each building component that has been subjected to the unified spatiotemporal reference, obtaining cleaned component progress data and visual data for each building component, performing feature extraction on the BIM data of each building component and the corresponding cleaned component progress data and visual data, obtaining a standardized spatiotemporal feature dataset for each building component, and storing the dataset in a database; S3. Based on the standardized spatiotemporal feature dataset for each building component, a digital twin of the construction process is constructed. This digital twin is processed using a spatiotemporal convolutional network to obtain inference results on the construction progress status of each building component. S4. Setting a component construction progress status inference threshold, comparing the component construction progress status inference result of each building component with the component construction progress status inference threshold, and obtaining a component risk assessment result of each building component; S5. Take and execute measures based on the component risk assessment results of each building component. At the same time, obtain multimodal data of each building component in the construction area after the measures are implemented, and update the digital twin diagram of the construction process.

2. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 1, characterized in that: The method for obtaining BIM data, component progress data and visual data of each building component in the construction site in real time: Obtain BIM data of each building component in real time from the BIM of the construction area through the BIM platform API, including component geometry information, WBS process tree, planned construction period and resource requirements; Deploy UWB base stations at the construction site, install UWB tags on each building component, calculate the distance between the UWB base station and the UWB tag using the TOF algorithm, and obtain the UWB local coordinates of each building component. At the same time, construction workers manually fill in the component installation progress, component material quantity, and reporting time of each building component through the mobile app, thereby obtaining component progress data for each building component, including component UWB local coordinates, component completion progress, component material quantity, and reporting time; Using a drone and a visible light camera mounted on the drone, the component image of each building component in the construction area, as well as the acquisition time and geodetic coordinates corresponding to the component image, is acquired, thereby obtaining visual data of each building component, including the component image, the acquisition time and geodetic coordinates corresponding to the component image; The TOF algorithm is a technology that calculates distance by measuring the propagation time of a signal between a transmitter and a receiver; The component geometric information includes component BIM coordinates, component ID, component size, and component type.

3. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 2 is characterized in that: The method for unifying the spatiotemporal benchmark of the component progress data and visual data of each building component: Connect the UWB base station, drone, and mobile app to the NTP server, and use the NTP server to control the UWB base station, drone, and mobile app to collect data simultaneously, which is used to unify the time base for component progress data and visual data; The component UWB local coordinates in the component progress data of each building component are converted into BIM coordinates corresponding to the component progress data through the Bursa model; The geodetic coordinates corresponding to the component image in the visual data of each building component are converted into projection coordinates through UTM projection, and the projection coordinates are translated to obtain the visual BIM coordinates corresponding to the visual data. Based on this, the component progress data and visual data of each building component after the unified time and space reference are obtained; The NTP server is a server used to achieve high-precision time synchronization in a computer network; The UTM projection is the Universal Transverse Mercator projection, which is a conformal transverse cylindrical projection.

4. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 3 is characterized in that: The method of performing data cleaning processing on the component progress data and visual data of each building component after the unified time and space reference: The component progress data of each building component that has been unified in time and space is monitored for outliers using the 3σ principle, and outliers are deleted. The component progress data after outliers are deleted is filled with missing values using linear interpolation to obtain component progress data after data cleaning. The visual data of each building component that has been unified in time and space is subjected to denoising by the Gaussian filtering algorithm. The denoised visual data is then enhanced by histogram equalization to obtain component progress data after data cleaning. The histogram equalization is a technique used in the field of image processing.

5. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 4 is characterized in that: The method of performing feature extraction processing on the BIM data of each building component, and the component progress data and visual data after corresponding data cleaning to obtain a standardized spatiotemporal feature data set for each building component: The component images in the visual data of each building component after data cleaning are monitored through the YOLO algorithm to obtain the component completion and component type. At the same time, the visual BIM coordinates and acquisition time are extracted from the visual data of each building component after data cleaning. Based on this, the visual features of each building component including acquisition time, component completion, component type and visual BIM coordinates are obtained; Extract the component installation progress, component material quantity, and reporting time from the component progress data of each building component after data cleaning. Calculate the component installation progress and reporting time using the construction progress speed formula to obtain the component construction speed. Simultaneously, calculate the component material quantity and component installation progress using the unit progress material consumption formula to obtain the component unit progress material consumption. Extract the BIM coordinates from the component progress data of each building component after data cleaning, and thereby obtain the component progress characteristics of each building component, including reporting time, component unit progress material consumption, component construction speed, and BIM coordinates. Directly extract component BIM coordinates, component ID, component type, WBS process tree, planned construction period and resource requirements from the BIM data of each building component. At the same time, according to the WBS process tree, the logical dependency relationship between components is obtained. Based on this, the BIM characteristics of each building construction, including component BIM coordinates, component ID, planned construction period, logical dependency relationship between components and resource requirements, are obtained. Based on the visual BIM coordinates and acquisition time in the visual features, the BIM coordinates and reporting time in the component progress features, and the component BIM coordinates in the BIM features, the BIM features, visual features, and component progress features of each building component are matched and merged to obtain a standardized spatiotemporal feature dataset for each building component. The standardized spatiotemporal feature dataset includes timestamp, component completion, component type, component unit progress material consumption, component construction speed, component BIM coordinates, component ID, planned construction period, logical dependencies between components, and resource requirements; The timestamp is the acquisition time.

6. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 5 is characterized in that: The method for constructing a digital twin diagram of the construction process based on the standardized spatiotemporal feature dataset of each building component: The component ID in the standardized spatiotemporal feature dataset of each building component is used as the unique node identifier, and the logical dependency relationship between components in the standardized spatiotemporal feature dataset of each building component is used as the directed edge connecting nodes. At the same time, according to the component ID, the timestamp, component completion degree, component type, component unit progress material consumption, component construction speed, planned construction period and resource requirements in the standardized spatiotemporal feature dataset of each building component are used as node attributes to be associated with the node corresponding to the component ID, thereby constructing a digital twin diagram of the construction process.

7. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 6 is characterized in that: The method of processing the digital twin graph of the construction process through a spatiotemporal convolutional network to obtain the inference results of the construction progress status of each building component: Based on the digital twin graph of the construction process, the node feature matrix and adjacency matrix of the digital twin graph of the construction process are obtained. The node feature matrix and adjacency matrix are input into the spatiotemporal convolutional network to obtain the component construction progress status inference results of each building component, including component completion deviation ΔC, process-level time deviation ΔT, and resource conflict index RCI; The spatiotemporal convolutional network is a deep learning model that integrates spatial and temporal dimension data processing.

8. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 7 is characterized in that: The method for obtaining the component risk assessment results of each building component: If ΔC ≥ -5%, the component completion deviation ΔC of the building component is normal; If -15% ≤ ΔC < -5%, the component completion deviation ΔC of the building component is low risk; If -30% ≤ ΔC < -15%, the component completion deviation ΔC of the building component is medium risk; If ΔC<-30%, the component completion deviation ΔC of the building component is high risk; If ΔT < 3, the process-level time deviation ΔT of the building component is normal. If 3≤ΔT<5, the process-level time deviation ΔT of the building component is low risk; If 5≤ΔT<7, the process-level time deviation ΔT of the building component is medium risk; If ΔT ≥ 7, the process-level time deviation ΔT of the building component is high risk; If RCI<04, the resource conflict index RCI of the building component is normal. If 0.4≤RCI<0.6, the resource conflict index RCI of the building component is low risk; If 0.6≤RCI<0.8, the resource conflict index RCI of the building component is medium risk; If RCI ≥ 0.8, the resource conflict index RCI of the building component is high risk; Based on this, the component completion deviation ΔC risk assessment result, process-level time deviation ΔT risk assessment result and resource conflict index RCI risk assessment result of each building component are obtained. The highest risk assessment result among the component completion deviation ΔC risk assessment result, process-level time deviation ΔT risk assessment result and resource conflict index RCI risk assessment result is taken as the component risk assessment result of each building component.

9. The method for dynamic tracking of BIM construction progress based on multimodal data fusion according to claim 8, characterized in that: The method for updating the digital twin of the construction process by taking measures based on the component risk assessment results of each building component and obtaining multimodal data of each building component in the construction area after the measures are implemented: Based on the component risk assessment results of each building component, select the building components with high risk assessment results. For these high-risk building components, prioritize resource allocation for construction according to the resource requirements in the building component's BIM data until the planned construction period in the building component's BIM data is met; No measures will be taken for building components with medium risk, low risk and normal risk according to the risk assessment results; Obtain multimodal data of each building component in the construction area after execution, process the multimodal data of each building component accordingly, and use it to update the digital twin of the construction process; The multimodal data includes BIM data, component progress data and visual data.

Images