Construction progress detection system and method based on fusion of three-dimensional model and field data

Abstract

The present invention discloses a construction progress detection system and method based on the fusion of three-dimensional model and on-site data, and relates to the technical field of power grid construction engineering. The system includes: a data acquisition module, a data preprocessing module, a data matching module, a data fusion module, a progress detection module and a feedback module. The data matching module matches the point cloud data and the panoramic image, optimizes the matching parameters by using the clone selection algorithm, and constructs the on-site three-dimensional basic data; the data fusion module deeply fuses the BIM model with the on-site three-dimensional basic data, and uses GeoBIM technology to construct a three-dimensional fusion model for construction progress detection; the progress detection module compares the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detects the construction progress, realizes the efficient collection, accurate processing and real-time monitoring of the construction site data, and significantly improves the efficiency and decision-making ability of construction management through automated progress detection and real-time feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid construction engineering, and in particular to a construction progress detection system and method based on the fusion of a three-dimensional model and on-site data. Background Art

[0002] With the expansion of the scale of tower construction and the advancement of construction technology, traditional manual monitoring methods have been unable to adapt to the requirements of modern construction management for efficiency and real-time performance. At present, the monitoring technology of construction sites mainly relies on manual inspections and manual verification, which is not only inefficient, but also unable to provide timely feedback on construction progress. At the same time, in order to meet the accuracy requirements of construction monitoring, a large amount of three-dimensional data needs to be collected in real time at the construction site. However, due to the large amount of data and limited transmission efficiency, real-time monitoring is difficult to achieve, which limits the accurate control of on-site conditions during construction management and affects the quality and efficiency of construction management. Summary of the invention

[0003] In view of the above problems, the present invention proposes a construction progress detection system and method based on the fusion of three-dimensional models and on-site data. The point cloud data and panoramic images of the initial stage of tower construction are collected by drones, matched using intelligent algorithms, and integrated with the BIM model to construct a three-dimensional fusion model for construction progress detection as a benchmark, which is compared with the panoramic images during the construction process to achieve real-time detection and feedback of the construction progress of the tower at each stage.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] In one aspect, the present invention provides a construction progress detection system based on the fusion of a three-dimensional model and field data, the system comprising:

[0006] The data acquisition module is used to collect point cloud data and panoramic images of the tower construction site through the laser radar and panoramic camera carried by the drone, and transmit them to the cloud storage and processing platform in real time through the 5G module. Specifically, it includes: collecting point cloud data and panoramic images at the beginning of construction, and collecting panoramic images of the construction site in real time during the construction monitoring stage;

[0007] The data preprocessing module is used to perform denoising, filtering, and semantic segmentation on point cloud data, and to correct and stitch panoramic images;

[0008] The data matching module is used to associate the pre-processed point cloud data with the panoramic image through RTK coordinates, select a four-point subset of geometric features in the point cloud data and the panoramic image for rough matching using the 4PCS algorithm, and then apply the clone selection algorithm to optimize the matching parameters. Through progressive optimization, global high-precision matching is completed to construct the on-site 3D basic data;

[0009] Data fusion module, which is used to input BIM model and locate the tower base coordinates from the on-site 3D basic data. The BIM model and the on-site 3D basic data are deeply integrated through the tower base coordinates, and the 3D fusion model for construction progress detection is constructed using GeoBIM technology.

[0010] A progress detection module is used to compare the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detect the construction progress, and generate a construction progress detection result and a construction progress report;

[0011] The feedback module is used to feed back the construction progress report to the construction management platform in real time for construction management personnel to make decisions.

[0012] As a preferred solution of the present invention, when the data acquisition module faces important construction nodes, detects abnormal progress or complex construction scenes during the construction monitoring stage, it supplements the collection of point cloud data for construction progress comparison. The important construction nodes and complex construction scenes are manually set by professionals.

[0013] As a preferred solution of the present invention, the data preprocessing module applies a negative selection algorithm to denoise the point cloud data, removes outliers or abnormal points in the point cloud data; downsamples by voxel grid filtering, and then further smoothes by Gaussian filtering; uses PointNet++ to perform global semantic classification, and then combines graph theory methods to refine the boundaries of regular geometric regions, and automatically performs semantic segmentation on the filtered point cloud data;

[0014] The single-frame panoramic image is subjected to distortion correction, and multiple images are stitched together using a feature point detection algorithm to generate a complete panoramic image.

[0015] As a preferred solution of the present invention, the method of applying a negative selection algorithm to denoise point cloud data includes:

[0016] Input the collected point cloud data as the original data set. Any point in the original data set By three-dimensional coordinates constitute;

[0017] Randomly extract k points from the original data set to form a self-sample set representing the normal point cloud distribution;

[0018] Randomly generate detectors, each detector represents an abnormal feature of the point cloud data, and each detector must be dissimilar to the self-points in the self-sample set, expressed as:

[0019] ;

[0020] In the formula, represents the detector set, is the jth detector in the detector set; is the i-th self-point in the self-sample set; It is used to calculate the detector Point with Self The distance function of the similarity between them is calculated according to the Euclidean distance formula; It is a preset distance threshold that can be dynamically set for different areas based on the point cloud density;

[0021] Traverse all points in the original data set and determine whether each point matches any detector in the detector set. , then point is an outlier point;

[0022] The outliers are removed from the original data set to form denoised point cloud data.

[0023] As a preferred solution of the present invention, the method of applying a clonal selection algorithm to optimize matching parameters and completing global high-precision matching through progressive optimization includes:

[0024] Rotation matrix in rough matching results based on 4PCS algorithm and translation vectors Initialize the antibody population, each antibody is a combination of a rotation matrix and a translation vector, expressed as:

[0025] ;

[0026] ;

[0027] In the formula, Represents the antibody population, and is the b-th rotation matrix and translation vector after random perturbation, is the population size; is the disturbance amplitude control factor, and is the random noise matrix;

[0028] The transformation applied to each antibody on the point cloud data is given by: ,in, is any point cloud data after preprocessing, is the transformed point cloud data;

[0029] Define the matching error as the fitness function, the fitness with the smallest matching error is taken as the optimal fitness, and calculate the fitness value of each antibody , the formula is:

[0030] ;

[0031] Where n is the number of matching points; It is Point cloud data in The new coordinates under the antibody transformation; is the new coordinate in the panoramic image The corresponding points;

[0032] According to the fitness value Select high fitness antibodies for cloning, clone number It is inversely proportional to the fitness and is expressed as:

[0033] ;

[0034] Where c is the cloning ratio factor; Indicates the rounding down sign to ensure that the number of clones is an integer;

[0035] Mutate each cloned antibody to generate a new antibody population ,in and is the bth rotation matrix and translation vector after mutation, ;

[0036] Select the one with the highest fitness from the mutated new antibody population The antibodies are directly entered into the next generation as elite antibodies, and the antibodies with low fitness are eliminated to avoid error accumulation;

[0037] After each round of iterative optimization, the change in the fitness of the antibody population is calculated , the formula is:

[0038] ;

[0039] In the formula, Antibodies in the T-1 round of iteration The fitness of For the antibody in the T-th round of iteration Adaptability;

[0040] If the change is less than the preset threshold, the current optimization is considered to have reached convergence and the local curvature is introduced. and the surface normal vector Update the global rotation matrix and translation vector, the formula is:

[0041] ;

[0042] In the formula, K is the number of neighboring points of the current point in the point cloud data, For the The normal vectors of the neighboring points, is the mean normal vector of the neighborhood points.

[0043] As a preferred solution of the present invention, the data fusion module imports the BIM model of the design stage into the processing platform in a standard format, and extracts the tower foundation design coordinates in the BIM model and the tower foundation actual coordinates in the on-site three-dimensional basic data;

[0044] Unify the coordinate systems of the BIM model and the on-site 3D basic data, and convert the design coordinate system of the BIM model into an RTK coordinate system that is consistent with the on-site 3D basic data;

[0045] According to the actual coordinates of the tower foundation in the on-site 3D basic data, the geometric position of the tower foundation in the BIM model is adjusted, the tower foundation design coordinates are aligned with the actual coordinates of the tower foundation, and the preliminary geometric fusion of the BIM model and the on-site 3D basic data is completed;

[0046] Deeply integrate geometric information within the tower base area, superimpose and adjust the actual geometric information of the tower base in the on-site 3D basic data with the designed geometric information in the BIM model, while retaining the geometric design objectives of the BIM model;

[0047] Combine the semantic information of the tower foundation in the BIM model with the geometric information in the on-site 3D basic data, add semantic tags to the key points in the on-site 3D basic data, describe the material characteristics and design attributes, and the key points are selected by professionals;

[0048] Use GeoBIM technology to extend the fusion results of the tower base area to the other parts of the on-site 3D basic data and BIM model, and fully integrate the geometric information and semantic information of the tower body and surrounding facilities;

[0049] Mark the site status before construction begins, and identify the initial construction status of the tower base and surrounding areas based on the comparison of the on-site 3D basic data and the BIM model;

[0050] Verify and output the generated 3D fusion model as the initial benchmark for construction progress detection.

[0051] As a preferred solution of the present invention, the progress detection module extracts feature points from the panoramic image, extracts 3D feature points matching the panoramic image from the 3D fusion model, and uses a matching algorithm to align the feature points in the panoramic image with the 3D feature points of the 3D fusion model;

[0052] Use the distance error detection algorithm to compare the panoramic image with the 3D fusion model frame by frame: For each frame of the panoramic image, calculate the spatial distance between the feature points extracted and the corresponding points in the 3D fusion model, mark the feature points that match the designed 3D fusion model as completed, the feature points that exceed the range of the 3D fusion model as abnormal, and the unmatched areas as incomplete;

[0053] Summarize the frame-by-frame comparison results, generate construction progress detection results based on the comparison results, count the completed progress, unfinished progress and abnormal area information, and generate a construction progress report.

[0054] As a preferred solution of the present invention, the construction progress report is automatically generated by a data visualization tool combined with the semantic data of GeoBIM, and includes:

[0055] Overall progress chart: Displays the proportion of completed, uncompleted and abnormal items through bar charts or pie charts;

[0056] Abnormal area annotation: Visually annotate abnormal areas on 3D fusion models and panoramic images;

[0057] Details of unfinished construction: List the unfinished construction parts and locations;

[0058] Comparison of completed parts: Show the actual status of the completed parts and whether they are consistent with the design goals.

[0059] On the other hand, the present invention provides a construction progress detection method based on the fusion of three-dimensional model and field data, based on the construction progress detection system based on the fusion of three-dimensional model and field data as described above, the method comprises:

[0060] The laser radar and panoramic camera carried by the drone collect point cloud data and panoramic images of the tower construction site, and transmit them to the cloud storage and processing platform in real time through the 5G module. Specifically, point cloud data and panoramic images are collected at the beginning of construction, and panoramic images of the construction site are collected in real time during the construction monitoring stage.

[0061] Perform denoising, filtering, and semantic segmentation preprocessing operations on point cloud data, and perform correction and stitching on panoramic images;

[0062] The pre-processed point cloud data is associated with the panoramic image through RTK coordinates. The 4PCS algorithm is used to select a four-point subset of geometric features in the point cloud data and the panoramic image for rough matching. The clonal selection algorithm is then applied to optimize the matching parameters. Global high-precision matching is completed through progressive optimization to construct the on-site 3D basic data.

[0063] Input the BIM model and locate the tower base coordinates from the on-site 3D basic data. Through the tower base coordinates, deeply integrate the BIM model with the on-site 3D basic data, and use GeoBIM technology to build a 3D fusion model for construction progress detection.

[0064] Compare the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detect the construction progress, and generate a construction progress detection result and a construction progress report;

[0065] The construction progress report is fed back to the construction management platform in real time for construction management personnel to make decisions.

[0066] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the construction progress detection system based on the fusion of three-dimensional model and field data as described above is implemented.

[0067] The beneficial effects of the present invention are as follows: by combining drones with lidar and panoramic cameras, the time and cost of manual data collection can be greatly reduced. The real-time transmission capability of the 5G module further eliminates the disadvantages of data delay in traditional methods, greatly improving the visibility and collection efficiency of construction site information. Data denoising, filtering and semantic segmentation ensure the clarity and accuracy of point cloud data. The correction and stitching of panoramic images eliminate the geometric errors that may occur during the collection process, providing high-quality basic data input for subsequent data processing. Data matching uses RTK coordinates and algorithm optimization to ensure the accurate association between point cloud data and panoramic images, which not only improves the matching speed, but also significantly reduces the matching error. It lays the foundation for high-precision 3D modeling; by deeply integrating the BIM model with the construction site data, it realizes the real-time association and dynamic visualization of the site and the design model, solves the disconnection problem between the design and the actual construction, and improves the construction management's control over complex scenes; the frame-by-frame comparison progress detection mechanism enables the system to quickly identify the completed parts, unfinished parts and abnormal areas in the construction, and the generated construction progress report simplifies the workflow of construction management and provides accurate construction completion status; the real-time feedback mechanism of the construction progress report provides construction managers with clear real-time data support, which can quickly respond to abnormal situations, optimize the allocation of construction resources, and ensure the timely completion of the construction plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. 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 creative labor. Among them:

[0069] Figure 1 It is a system modular structure diagram of the present invention;

[0070] Figure 2 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0072] like Figure 1 FIG. 1 is an embodiment of the present invention, which provides a construction progress detection system based on the fusion of a three-dimensional model and field data, including:

[0073] (1) Data acquisition module

[0074] The data acquisition module is used to collect point cloud data and panoramic images of the tower construction site through the laser radar (LiDAR) and panoramic camera carried by the drone, and transmit them to the cloud storage and processing platform in real time through the 5G module;

[0075] Specifically, point cloud data and panoramic images are collected at the beginning of construction (before construction officially starts) to construct the three-dimensional basic data of the site and comprehensively record the initial state of the construction site, including the terrain, environment, and key geometric features of the area to be constructed;

[0076] During the construction monitoring phase, the latest panoramic images of the construction site are collected in real time to record the construction progress and compare them with the 3D model before construction to determine the progress difference;

[0077] Panoramic images are fast to acquire and low in cost, making them suitable for high-frequency data acquisition. Point cloud data is highly accurate, but the acquisition and processing costs are high. Therefore, this embodiment only needs to acquire point cloud data and panoramic images once before the construction officially starts. During the monitoring phase, panoramic images are used as the main method for real-time high-frequency monitoring to quickly record the construction progress. At important construction nodes (such as the completion of the tower base or important steel structure parts), when abnormal progress is detected and further analysis is required, or when complex construction scenes (such as the installation of complex components) require more accurate three-dimensional data, additional point cloud data is acquired for detailed comparison. Important construction nodes and complex construction scenes are manually set by professionals. This strategy can reduce the cost of data acquisition and processing while ensuring monitoring accuracy.

[0078] (2) Data preprocessing module

[0079] The data preprocessing module is used to perform preprocessing operations such as denoising, filtering, and semantic segmentation on point cloud data, and to correct and stitch panoramic images;

[0080] Specifically, the negative selection algorithm is used to denoise the point cloud data, and outliers or abnormal points in the point cloud data (such as isolated points caused by equipment noise) are removed to ensure the basic quality of the data; the voxel grid filtering is used to downsample, and then the Gaussian filter is used for further smoothing to eliminate the boundary irregularities that may be caused by denoising, while retaining important geometric details; PointNet++ is used for global semantic classification, and then the graph theory method is combined to further refine the boundaries of regular geometric areas (such as the plane part of the tower base), and the filtered point cloud data is automatically classified into categories such as buildings, tower bases, ground, and vegetation;

[0081] Correct the distortion of a single-frame panoramic image and use feature point detection algorithms such as SIFT (Scale-invariant feature transform) to stitch multiple images together to generate a complete panoramic image.

[0082] The negative selection algorithm originates from the artificial immune system and simulates the "self-non-self" discrimination mechanism of the biological immune system. It can be used to detect abnormal data. In point cloud processing, the algorithm identifies "abnormal" or "abnormal" points (such as outliers) in the point cloud data and removes them, thereby ensuring the overall quality of the data.

[0083] In one preferred embodiment, a negative selection algorithm is applied to denoise the point cloud data, and the method includes:

[0084] Input the collected point cloud data as the original data set. Any point in the original data set By three-dimensional coordinates The distribution of noise points is generally sparse, and the main body of the point cloud is continuous;

[0085] Randomly extract k points from the original data set to form a self-sample set representing the normal point cloud distribution. Its features are considered to represent the main body of the point cloud, that is, the point set in the point cloud data that conforms to the distribution law of most point clouds;

[0086] Randomly generate detectors, each of which represents a possible abnormal feature of the point cloud data (such as a large density difference with surrounding points or far away from other points), which must be dissimilar to the self-points in the self-sample set (for example, mismatch is judged by distance threshold), expressed as:

[0087] ;

[0088] In the formula, represents the detector set, is the jth detector in the detector set; is the i-th self-point in the self-sample set; It is used to calculate the detector Point with Self The distance function of the similarity between them is calculated according to the Euclidean distance formula; It is a preset distance threshold, which indicates the maximum matching range of normal points in the point cloud data. The matching thresholds of different areas are dynamically set according to the point cloud density to improve the detection accuracy.

[0089] Traverse all points in the original data set and determine whether each store matches any detector in the detector set. , then point is an outlier point;

[0090] The outliers are removed from the original data set to form denoised point cloud data; the outliers can be stored separately for subsequent quality analysis.

[0091] (3) Data matching module

[0092] The data matching module is used to associate the pre-processed point cloud data with the panoramic image through RTK (Real-time Kinematic) coordinates, and use the 4PCS (Four Point Congruent Set, a matching algorithm in computer vision) algorithm to select four point subsets of geometric features in the point cloud data and the panoramic image (such as corner points, plane intersections, etc.) for rough matching, and then use the Clone Selection Algorithm (CSA) in AIS (Artificial Immune Systems) to optimize the matching parameters, complete global high-precision matching through progressive optimization, and construct the on-site 3D basic data;

[0093] RTK coordinates are generated by the GNSS module carried by the drone, providing centimeter-level global geographic reference for point cloud data, preliminarily aligning point cloud data with panoramic images, and providing a basis for subsequent precise matching. Four point subsets with obvious geometric features are selected in the point cloud and panoramic images for rough matching. The 4PCS algorithm selects four points based on the principle of "maximum coplanarity" to reduce computational complexity and achieve fast alignment on low-overlapping data or large data sets.

[0094] The artificial immune algorithm (AIS) was originally borrowed from the field of biology. Its clone selection algorithm was introduced into point cloud and image matching for parameter optimization. By introducing mutation and fitness evaluation, it can effectively find the global optimal matching parameters and make up for the shortcomings of the 4PCS algorithm in complex scenes. Traditional progressive optimization is mostly based on simple geometric features (such as points and lines), while this solution further combines high-level features such as point cloud curvature and surface normal vectors, introduces more complex surface features based on traditional geometric features, and improves the point cloud data's ability to describe complex geometric structures.

[0095] Specifically, a clonal selection algorithm is applied to optimize matching parameters, and global high-precision matching is achieved through progressive optimization. The method includes:

[0096] Rotation matrix in rough matching results based on 4PCS algorithm and translation vectors Initialize the antibody population, each antibody is a combination of a rotation matrix and a translation vector, expressed as:

[0097] ;

[0098] ;

[0099] In the formula, Represents the antibody population, and is the b-th rotation matrix and translation vector after random perturbation, is the population size; is the disturbance amplitude control factor, and is the random noise matrix;

[0100] The transformation applied to each antibody on the point cloud data is given by: , where p is any point cloud data after preprocessing, is the transformed point cloud data;

[0101] Define the matching error as the fitness function, the fitness with the smallest matching error is taken as the optimal fitness, and calculate the fitness value of each antibody , the formula is:

[0102] ;

[0103] Where n is the number of matching points; It is Point cloud data in The new coordinates under the antibody transformation; is the new coordinate in the panoramic image The corresponding points;

[0104] According to the fitness value Select high fitness antibodies for cloning, clone number It is inversely proportional to the fitness and is expressed as:

[0105] ;

[0106] Where c is the cloning ratio factor; Indicates the rounding down sign to ensure that the number of clones is an integer;

[0107] Mutate each cloned antibody to generate a new antibody population ,in and is the bth rotation matrix and translation vector after mutation, ;

[0108] Select the one with the highest fitness from the mutated new antibody population The antibodies are directly entered into the next generation as elite antibodies, and the antibodies with low fitness are eliminated to avoid error accumulation;

[0109] After each round of iterative optimization, the change in the fitness of the antibody population is calculated , the formula is:

[0110] ;

[0111] In the formula, Antibodies in the T-1 round of iteration The fitness of For the antibody in the T-th round of iteration Adaptability;

[0112] If the change is less than the preset threshold, the current optimization is considered to have reached convergence and the local curvature is introduced. and the surface normal vector Update the global rotation matrix and translation vector, the formula is:

[0113] ;

[0114] In the formula, K is the number of neighboring points of the current point in the point cloud data, For the The normal vectors of the neighboring points, is the mean normal vector of the neighborhood points.

[0115] (4) Data fusion module

[0116] The data fusion module is used to input the BIM model (Building Information Modeling) and locate the tower base coordinates from the on-site 3D basic data. The BIM model and the on-site 3D basic data are deeply integrated through the tower base coordinates, and the GeoBIM technology is used to build a 3D fusion model for construction progress detection.

[0117] GeoBIM technology is a technology that combines geographic information system (GIS) and building information model (BIM). It uses geographic information system to describe geographic spatial information and building information model to describe building information, thereby achieving integrated management of geographic space and building information.

[0118] In one embodiment, the data fusion module imports the BIM model of the design phase into the processing platform in a standard format (such as IFC or OBJ), and the BIM model includes geometric information and semantic information generated in the design phase, such as tower foundation design coordinates, geometric shapes, and attribute data;

[0119] Extract the tower foundation design coordinates in the BIM model and clarify the geometric information of the tower foundation design location;

[0120] Extract the actual coordinates of the tower foundation from the on-site 3D basic data to obtain the geometric center point and spatial boundary of the tower foundation;

[0121] Unify the coordinate systems of the BIM model and the on-site 3D basic data, and convert the design coordinate system of the BIM model into an RTK coordinate system that is consistent with the on-site 3D basic data;

[0122] According to the actual coordinates of the tower foundation in the on-site 3D basic data, the geometric position of the tower foundation in the BIM model is adjusted, the tower foundation design coordinates are aligned with the actual coordinates of the tower foundation, and the preliminary geometric fusion of the BIM model and the on-site 3D basic data is completed;

[0123] Deeply integrate geometric information within the tower base area, superimpose and adjust the actual geometric information of the tower base in the on-site 3D basic data (such as the actual boundary and shape reflected by the point cloud) with the design geometric information in the BIM model (such as the ideal shape of the tower base), to ensure that the constructed 3D fusion model can combine the characteristics of the actual on-site status while retaining the geometric design objectives of the BIM model;

[0124] Combine the semantic information of the tower foundation in the BIM model (such as material properties and construction plan) with the geometric information in the on-site 3D basic data, add semantic tags to the key points in the on-site 3D basic data to describe the material characteristics and design properties. The key points are selected by professionals.

[0125] Use GeoBIM technology to extend the fusion results of the tower base area to the other parts of the on-site 3D basic data and BIM model, and fully integrate the geometric information and semantic information of the tower body and surrounding facilities to ensure geometric accuracy and semantic consistency;

[0126] Mark the site status before construction begins. Based on the comparison between the on-site 3D basic data and the BIM model, identify the initial construction status of the tower base and surrounding areas, provide a benchmark for subsequent construction progress detection, and identify unconstructed areas and completed site preparation parts;

[0127] Verify the generated 3D fusion model to ensure the geometric alignment accuracy and semantic integrity of the model, and check the key fusion results of the tower base area and surrounding sites to confirm that they are consistent with the actual situation on site.

[0128] The output 3D fusion model includes the fusion geometric information of the on-site 3D basic data and the BIM model before construction begins, the initial construction status annotation and semantic information. This model serves as the initial benchmark for construction progress detection and provides support for subsequent real-time monitoring and construction progress analysis.

[0129] (5) Progress detection module

[0130] The progress detection module is used to compare the latest panoramic images of the construction site collected in real time with the 3D fusion model frame by frame, automatically detect the construction progress, and generate construction progress detection results and construction progress reports;

[0131] In one specific embodiment, the progress detection module extracts feature points from the panoramic image, extracts 3D feature points matching the panoramic image from the 3D fusion model, and uses a matching algorithm (such as 4-point set method 4PCS or iterative closest point ICP) to align the feature points in the panoramic image with the 3D feature points of the 3D fusion model;

[0132] Use a 3D space comparison algorithm (such as a distance error detection algorithm) to compare the panoramic image with the 3D fusion model frame by frame: For each frame of the panoramic image, calculate the spatial distance between the feature points extracted and the corresponding points in the 3D fusion model, mark the feature points that meet the design 3D fusion model as "completed", the feature points that exceed the range of the 3D fusion model as "abnormal", and the unmatched areas as "incomplete";

[0133] Summarize the frame-by-frame comparison results and count the following indicators:

[0134] Completed progress: volume or proportion of construction completed;

[0135] Unfinished progress: volume or area of ​​unfinished parts;

[0136] Abnormal area information: the location of each abnormal point and its distance error;

[0137] The construction progress detection results are generated based on the comparison results, including the classification of construction status and a detailed data list.

[0138] Generate construction progress report, including:

[0139] Overall progress chart: such as a bar chart or pie chart showing the percentage of completed, uncompleted and abnormal;

[0140] Abnormal area annotation: Visually annotate abnormal areas on 3D models and panoramic images;

[0141] Details of unfinished construction: List the unfinished construction parts and their locations;

[0142] Comparison of completed parts: Show the actual status of the completed parts and whether they are consistent with the design goals.

[0143] Use data visualization tools (such as Python's Matplotlib or Power BI) combined with GeoBIM's semantic data to automatically generate construction progress reports in the following output formats: PDF, HTML, or interactive visualization tools.

[0144] (6) Feedback module

[0145] The feedback module is used to feed back the construction progress report to the construction management platform in real time for construction management personnel to make decisions.

[0146] like Figure 2 As shown, another embodiment of the present invention provides a construction progress detection method based on the fusion of three-dimensional model and field data. Based on the construction progress detection system based on the fusion of three-dimensional model and field data as described above, the method includes the following steps:

[0147] S1: The laser radar and panoramic camera carried by the drone are used to collect point cloud data and panoramic images of the tower construction site, and the data is transmitted to the cloud storage and processing platform in real time through the 5G module. Specifically, point cloud data and panoramic images are collected at the beginning of construction, and panoramic images of the construction site are collected in real time during the construction monitoring stage.

[0148] S2: Perform pre-processing operations such as denoising, filtering, and semantic segmentation on the point cloud data, and correct and stitch the panoramic image;

[0149] S3: The pre-processed point cloud data is associated with the panoramic image through RTK coordinates, and the 4PCS algorithm is used to select a four-point subset of geometric features in the point cloud data and the panoramic image for rough matching. The clonal selection algorithm is then applied to optimize the matching parameters. Global high-precision matching is completed through progressive optimization to construct the on-site 3D basic data.

[0150] S4: Input the BIM model and locate the tower base coordinates from the on-site 3D basic data. Through the tower base coordinates, deeply integrate the BIM model with the on-site 3D basic data, and use GeoBIM technology to build a 3D fusion model for construction progress detection.

[0151] S5: comparing the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detecting the construction progress, and generating a construction progress detection result and a construction progress report;

[0152] S6: Feedback the construction progress report to the construction management platform in real time for construction management personnel to make decisions.

[0153] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, it can be implemented in whole or in part in the form of a computer program product, which includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. Therefore, an embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the construction progress detection system based on the fusion of three-dimensional model and field data as described above is implemented.

[0154] In summary, the present invention can greatly reduce the time and cost of manual data collection by combining drones with lidar and panoramic cameras. The real-time transmission capability of 5G modules further eliminates the disadvantages of data delay in traditional methods, greatly improving the visibility and collection efficiency of construction site information. Data denoising, filtering and semantic segmentation ensure the clarity and accuracy of point cloud data. The correction and stitching of panoramic images eliminate the geometric errors that may occur during the collection process, providing high-quality basic data input for subsequent data processing. Data matching uses RTK coordinates and algorithm optimization to ensure the accurate association between point cloud data and panoramic images, which not only improves the matching speed, but also significantly reduces the matching error. High-precision 3D modeling lays the foundation; by deeply integrating the BIM model with the construction site data, real-time correlation and dynamic visualization of the site and design model are achieved, solving the disconnection problem between design and actual construction and improving the construction management's control over complex scenes; the frame-by-frame comparison progress detection mechanism enables the system to quickly identify completed parts, unfinished parts and abnormal areas in construction, and the generated construction progress report simplifies the workflow of construction management and provides accurate construction completion status; the real-time feedback mechanism of the construction progress report provides construction managers with clear real-time data support, which can quickly respond to abnormal situations, optimize the allocation of construction resources, and ensure the timely completion of the construction plan.

[0155] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The construction progress detection system based on the fusion of three-dimensional model and field data is characterized by: The system comprises: The data acquisition module is used to collect point cloud data and panoramic images of the tower construction site through the laser radar and panoramic camera carried by the drone, and transmit them to the cloud storage and processing platform in real time through the 5G module. Specifically, it includes: collecting point cloud data and panoramic images at the beginning of construction, and collecting panoramic images of the construction site in real time during the construction monitoring stage; The data preprocessing module is used to perform denoising, filtering, and semantic segmentation on point cloud data, and to correct and stitch panoramic images; The data matching module is used to associate the pre-processed point cloud data with the panoramic image through RTK coordinates, select a four-point subset of geometric features in the point cloud data and the panoramic image for rough matching using the 4PCS algorithm, and then apply the clone selection algorithm to optimize the matching parameters. Through progressive optimization, global high-precision matching is completed to construct the on-site 3D basic data; The method of applying a clone selection algorithm to optimize matching parameters and completing global high-precision matching through progressive optimization includes: Rotation matrix in rough matching results based on 4PCS algorithm and translation vectors Initialize the antibody population, each antibody is a combination of a rotation matrix and a translation vector, expressed as: ； ； In the formula, Represents the antibody population, and is the b-th rotation matrix and translation vector after random perturbation, is the population size; is the disturbance amplitude control factor, and is the random noise matrix; The transformation applied to each antibody on the point cloud data is given by: ,in, is any point cloud data after preprocessing, is the transformed point cloud data; Define the matching error as the fitness function, the fitness with the smallest matching error is taken as the optimal fitness, and calculate the fitness value of each antibody , the formula is: ； Where n is the number of matching points; It is Point cloud data in The new coordinates under the antibody transformation; is the new coordinate in the panoramic image The corresponding points; According to the fitness value Select high fitness antibodies for cloning, clone number It is inversely proportional to the fitness and is expressed as: ； Where c is the cloning ratio factor; Indicates the rounding down sign to ensure that the number of clones is an integer; Mutate each cloned antibody to generate a new antibody population ,in and is the bth rotation matrix and translation vector after mutation, ; Select the one with the highest fitness from the mutated new antibody population The antibodies are directly entered into the next generation as elite antibodies, and the antibodies with low fitness are eliminated to avoid error accumulation; After each round of iterative optimization, the change in the fitness of the antibody population is calculated. , the formula is: ； In the formula, Antibodies in the T-1 round of iteration The fitness of For the antibody in the T-th round of iteration Adaptability; If the change is less than the preset threshold, the current optimization is considered to have reached convergence and the local curvature is introduced. and the surface normal vector Update the global rotation matrix and translation vector, the formula is: ； Where K is the number of neighboring points of the current point in the point cloud data, For the The normal vectors of the neighboring points, is the mean normal vector of the neighborhood points; Data fusion module, which is used to input BIM model and locate the tower base coordinates from the on-site 3D basic data. The BIM model and the on-site 3D basic data are deeply integrated through the tower base coordinates, and the 3D fusion model for construction progress detection is constructed using GeoBIM technology. A progress detection module is used to compare the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detect the construction progress, and generate a construction progress detection result and a construction progress report; The feedback module is used to feed back the construction progress report to the construction management platform in real time for construction management personnel to make decisions.

2. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 1 is characterized in that: During the construction monitoring phase, when the data acquisition module faces important construction nodes, detects abnormal progress or complex construction scenes, it will collect additional point cloud data for construction progress comparison. The important construction nodes and complex construction scenes are manually set by professionals.

3. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 1 is characterized in that: The data preprocessing module uses a negative selection algorithm to denoise the point cloud data and remove outliers or abnormal points in the point cloud data; downsamples through voxel grid filtering and then further smoothes using Gaussian filtering; uses PointNet++ to perform global semantic classification, and then combines graph theory methods to refine the boundaries of regular geometric regions, and automatically performs semantic segmentation on the filtered point cloud data; The single-frame panoramic image is subjected to distortion correction, and multiple images are stitched together using a feature point detection algorithm to generate a complete panoramic image.

4. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 3 is characterized in that: The method of applying a negative selection algorithm to denoise point cloud data includes: Input the collected point cloud data as the original data set. Any point in the original data set By three-dimensional coordinates constitute; Randomly extract k points from the original data set to form a self-sample set representing the normal point cloud distribution; Randomly generate detectors, each detector represents an abnormal feature of the point cloud data, and each detector must be dissimilar to the self-points in the self-sample set, expressed as: ； In the formula, represents the detector set, is the jth detector in the detector set; is the i-th self-point in the self-sample set; It is used to calculate the detector Point with Self The distance function of the similarity between them is calculated according to the Euclidean distance formula; It is a preset distance threshold that can be dynamically set for different areas based on the point cloud density; Traverse all points in the original data set and determine whether each point matches any detector in the detector set. , then point is an outlier point; The outliers are removed from the original data set to form denoised point cloud data.

5. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 1 is characterized in that: The data fusion module imports the BIM model of the design stage into the processing platform in a standard format, and extracts the tower foundation design coordinates in the BIM model and the tower foundation actual coordinates in the on-site three-dimensional basic data; Unify the coordinate systems of the BIM model and the on-site 3D basic data, and convert the design coordinate system of the BIM model into an RTK coordinate system that is consistent with the on-site 3D basic data; According to the actual coordinates of the tower foundation in the on-site 3D basic data, the geometric position of the tower foundation in the BIM model is adjusted, the tower foundation design coordinates are aligned with the actual coordinates of the tower foundation, and the preliminary geometric fusion of the BIM model and the on-site 3D basic data is completed; Deeply integrate geometric information within the tower base area, superimpose and adjust the actual geometric information of the tower base in the on-site 3D basic data with the designed geometric information in the BIM model, while retaining the geometric design objectives of the BIM model; Combine the semantic information of the tower foundation in the BIM model with the geometric information in the on-site 3D basic data, add semantic tags to the key points in the on-site 3D basic data, describe the material characteristics and design attributes, and the key points are selected by professionals; Use GeoBIM technology to extend the fusion results of the tower base area to the other parts of the on-site 3D basic data and BIM model, and fully integrate the geometric information and semantic information of the tower and surrounding facilities; Mark the site status before construction begins, and identify the initial construction status of the tower base and surrounding areas based on the comparison of the on-site 3D basic data and the BIM model; Verify and output the generated 3D fusion model as the initial benchmark for construction progress detection.

6. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 1 is characterized in that: The progress detection module extracts feature points from the panoramic image, extracts 3D feature points matching the panoramic image from the 3D fusion model, and uses a matching algorithm to align the feature points in the panoramic image with the 3D feature points of the 3D fusion model; Use the distance error detection algorithm to compare the panoramic image with the 3D fusion model frame by frame: For each frame of the panoramic image, calculate the spatial distance between the feature points extracted and the corresponding points in the 3D fusion model, mark the feature points that match the designed 3D basic model as completed, the feature points that exceed the range of the 3D fusion model as abnormal, and the unmatched areas as incomplete; Summarize the frame-by-frame comparison results, generate construction progress detection results based on the comparison results, count the completed progress, unfinished progress and abnormal area information, and generate a construction progress report.

7. The construction progress detection system based on the fusion of three-dimensional model and field data according to claim 6 is characterized in that: The construction progress report is automatically generated by the data visualization tool combined with the semantic data of GeoBIM, and includes: Overall progress chart: Displays the proportion of completed, uncompleted and abnormal items through bar charts or pie charts; Abnormal area annotation: Visually annotate abnormal areas on 3D fusion models and panoramic images; Details of unfinished construction: List the unfinished construction parts and locations; Comparison of completed parts: Show the actual status of the completed parts and whether they are consistent with the design goals.

8. The detection method of the construction progress detection system based on the fusion of three-dimensional model and field data according to any one of claims 1 to 7, characterized in that: The method comprises: The laser radar and panoramic camera carried by the drone collect point cloud data and panoramic images of the tower construction site, and transmit them to the cloud storage and processing platform in real time through the 5G module. Specifically, point cloud data and panoramic images are collected at the beginning of construction, and panoramic images of the construction site are collected in real time during the construction monitoring stage. Perform denoising, filtering, and semantic segmentation preprocessing operations on point cloud data, and perform correction and stitching on panoramic images; The pre-processed point cloud data is associated with the panoramic image through RTK coordinates. The 4PCS algorithm is used to select a four-point subset of geometric features in the point cloud data and the panoramic image for rough matching. The clonal selection algorithm is then applied to optimize the matching parameters. Global high-precision matching is completed through progressive optimization to construct the on-site 3D basic data. Input the BIM model and locate the tower base coordinates from the on-site 3D basic data. Through the tower base coordinates, deeply integrate the BIM model with the on-site 3D basic data, and use GeoBIM technology to build a 3D fusion model for construction progress detection. Compare the panoramic image of the construction site collected in real time with the three-dimensional fusion model frame by frame, automatically detect the construction progress, and generate a construction progress detection result and a construction progress report; The construction progress report is fed back to the construction management platform in real time for construction management personnel to make decisions.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a construction progress detection system based on the fusion of three-dimensional model and field data as described in any one of claims 1 to 7 is implemented.

Images