Construction site comprehensive intelligent management method and system, and storage medium

Abstract

The application discloses a building construction site comprehensive intelligent management method and system and a storage medium, relates to the technical field of building construction intelligent management, and comprises the following steps: generating a spatial distribution density field by Gauss kernel density estimation based on a UWB label position and wireless signal inversion, fusing to generate a coarse-grained density, and performing smoothing processing to obtain a final density field; generating an expected density field according to a working condition plan, performing plan matching on an entity state set, calculating a plan compliance field and a plan deviation degree; and generating a comprehensive risk field based on the plan compliance field complement and the deviation degree. The application combines wireless signal inversion and UWB Gauss kernel density estimation, improves the spatial recognition accuracy of personnel and equipment distribution in the construction site, and improves the real-time performance and scene adaptability of risk identification by jointly calculating the plan compliance field and the plan deviation degree and superimposing a static risk field.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for building construction, and in particular to a comprehensive intelligent management method, system and storage medium for building construction sites. Background Technology

[0002] With the continuous advancement of digital and intelligent transformation in the construction industry, the management of construction sites is gradually evolving from manual experience-based decision-making to data-driven intelligent decision-making. The application of technologies such as the Internet of Things (IoT), Artificial Intelligence (AI), Computer Vision (CV), and wireless positioning (such as UWB and Wi-Fi probes) has provided new possibilities for real-time monitoring and dynamic management of construction sites.

[0003] Existing intelligent management technologies for construction sites still have significant shortcomings. Visual inspection results are easily affected by occlusion, lighting, and environmental interference, and there is a lack of a complementary mechanism with wireless signal data, making it difficult to simultaneously achieve both macroscopic coverage and microscopic accuracy. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a comprehensive intelligent management method, system, and storage medium for construction sites, which solves the problems that visual inspection results are easily affected by occlusion, lighting, and environmental interference, and lack a complementary mechanism with wireless signal data, making it difficult to simultaneously take into account both macroscopic coverage and microscopic accuracy.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a comprehensive intelligent management method for construction sites, which includes the following steps:

[0008] Collect multi-source heterogeneous data and work plan from the construction site and preprocess them. The multi-source heterogeneous data includes visual images, 3D point clouds, UWB tag locations, and radio frequency signal data.

[0009] The spatial distribution density field is generated by estimating the Gaussian kernel density based on the location of UWB tags and inverting the wireless signal. The density field is then fused to generate a coarse-grained density field, which is then smoothed to obtain the final density field.

[0010] Monte Carlo sampling is performed based on the final density field. The final motion field is generated by combining optical flow calculation and Gaussian smoothing. The final density field is divided into connected regions. Based on the final motion field, the centroid and average velocity of each region are calculated to form a set of entity states. The expected density field is generated according to the working condition plan. The entity state set is matched with the plan, and the plan compliance field and plan deviation are calculated.

[0011] A comprehensive risk field is generated based on the complement of the plan compliance field and the deviation, and a risk threshold is set to identify high-risk areas and trigger early warnings.

[0012] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the step of generating a spatial distribution density field based on Gaussian kernel density estimation of UWB tag location and wireless signal inversion, fusing them to generate a coarse-grained density, and then smoothing the field to obtain the final density field includes:

[0013] A construction ground coordinate system is established based on the overall construction plan, and the construction ground coordinate system is divided into a regular two-dimensional grid.

[0014] For each grid point, the Gaussian kernel density estimation method is used to convert the UWB tag location into a continuous density distribution. The inverse distance weighted approximation of the logarithmic distance path loss model is used to retrieve the spatial distribution of the signal source from the radio frequency signal. The density distribution and spatial distribution are respectively subjected to maximum-minimum normalization and fused to generate a fused coarse-grained density.

[0015] Gaussian filtering is used to smooth the fused coarse-grained density to obtain the final density field;

[0016] Set a density threshold and filter regions with a coarse-grained density greater than the density threshold, defining them as regions of interest.

[0017] Set a foreground detection threshold, extract the running foreground pixels of the visual image within the region of interest, and generate a running foreground mask.

[0018] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the method involves: Monte Carlo sampling based on the final density field, combining optical flow calculation and Gaussian smoothing to generate the final motion field, dividing the final density field into connected regions, and based on the final motion field, including:

[0019] The final density field is normalized and used as the probability density function. Monte Carlo sampling is performed on the construction plane to find the projection position of the sampling point in each visual image. The optical flow displacement of the projection position between the current frame and the previous frame is calculated. The displaced sampling point is back-projected onto the ground plane. The ground velocity of the sampling point under the camera observation is calculated. The arithmetic mean of the ground velocity is calculated and defined as the optical flow velocity.

[0020] The projection position is determined by calibrating the internal parameters of each high-definition network camera using the Zhang Zhengyou calibration method, calibrating the external parameters of the three-dimensional point cloud using the PNP algorithm, and projecting the running foreground mask onto the ground coordinate system based on the internal and external parameters.

[0021] The instantaneous velocity of the UWB tag at each sampling point is calculated, and the optical flow velocity and the instantaneous velocity are combined to obtain a sparse velocity field. Based on the sparse velocity field, the inverse distance weighted interpolation method is used to calculate the velocity vector of each regular grid point in the construction ground coordinate system, and Gaussian smoothing is performed to obtain the final motion field.

[0022] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the following steps are included: calculating the centroid and average velocity of each area to form a set of entity states; generating a desired density field based on the work plan; performing plan matching on the set of entity states; and calculating the plan compliance field and plan deviation.

[0023] The peak finding algorithm is used to identify local density maxima in the final density field, which are defined as density peaks. The final density field is divided into connected regions, and the centroids and average velocities of the connected regions are calculated and combined into a set of entity states.

[0024] At any grid point, a two-dimensional Gaussian model is used to calculate the expected density field of each plan. The total expected density field of the plan is obtained by linearly superimposing the expected density fields of all plans.

[0025] Iterate through the set of entity states. For each entity, determine whether its centroid position falls within the work area of ​​any plan. If the condition is met, the entity is marked as an "in-plan entity"; otherwise, it is marked as an "out-of-plan entity".

[0026] The work range refers to the Euclidean distance between the centroid position and the planned work position being less than or equal to the radius of the planned work range, and the current moment being within the planned execution time window;

[0027] Based on the labeling results and the total expected density field, the plan compliance field is calculated.

[0028] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the step of generating a comprehensive risk field based on the complement and deviation of the plan compliance field includes:

[0029] Calculate the relative absolute deviation between the final density field and the expected density field of the overall plan, generate the plan deviation degree, calculate the complement of the plan compliance field, and define it as the degree of logical conflict;

[0030] For each grid point in the construction ground coordinate system, a local spatial neighborhood is defined with it as the center. The sample covariance matrix of the velocity vectors at all grid points in the local spatial neighborhood is calculated and eigenvalue decomposition is performed. The largest eigenvalue obtained is defined as the motion disorder of the grid point.

[0031] The deviation from the plan, the degree of logical conflict, and the degree of motion disorder are normalized and linearly superimposed to obtain the basic dynamic risk field.

[0032] The spatial coordinate information of all geometric elements marked as hazards is exported from the building information model of the construction organization design. The entire construction site ground is divided into a two-dimensional grid with the same coordinate system as the construction ground. A blank binary matrix is ​​created, and all grid cells covered by each hazard geometric element are marked as 1 in the binary matrix. A static hazard binary matrix aligned with the coordinate system of the construction ground is generated and defined as a static risk field.

[0033] By combining the static risk field and the basic dynamic risk field, a comprehensive risk field is generated.

[0034] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the step of generating a comprehensive risk field based on the complement and deviation of the plan compliance field, setting a risk threshold to identify high-risk areas and triggering early warnings includes:

[0035] Set a risk threshold, use the risk threshold to perform binarization segmentation on the comprehensive risk field, mark the comprehensive risk field greater than the risk threshold as 1, otherwise mark it as 0, and generate the initial risk mask;

[0036] A morphological closing operation is performed on the initial risk mask to eliminate noise and connect adjacent regions, resulting in a binary risk mask. The binary risk mask is then aggregated using a connected component analysis algorithm to generate connected components, and an early warning is issued and the region leader is notified.

[0037] As a preferred embodiment of the comprehensive intelligent management method for construction sites described in this invention, the step of collecting multi-source heterogeneous data and work plan from the construction site and performing preprocessing includes:

[0038] Smart sensors are used to collect multi-source heterogeneous data from the construction site, and API interfaces are used to collect work plan from the construction organization design, and timestamp synchronization, noise reduction and standardization are performed.

[0039] The smart sensors include high-definition network cameras, lidar, UWB positioning systems, and Wi-Fi probe sensors;

[0040] The work plan includes the coordinates of the work center, the radius of the work area, the planned execution time window, and the expected number of entities.

[0041] Secondly, this invention provides a comprehensive intelligent management system for construction sites, including:

[0042] The collection and processing module is used to collect multi-source heterogeneous data and work plan from the construction site, and to perform timestamp synchronization, noise reduction and standardization processing.

[0043] The density fusion module is used to establish a construction ground coordinate system and a regular grid. It generates a spatial density field based on UWB Gaussian kernel density estimation and Wi-Fi path loss inversion. It then forms a coarse-grained density distribution through weighted fusion and performs smoothing to obtain a refined final density field.

[0044] The motion entity module is used to generate a sparse velocity field through optical flow analysis and UWB velocity calculation, interpolate to form a continuous motion field, and use the DBSCAN algorithm to identify entities and their states at the construction site.

[0045] The risk warning module is used to match the entity status with the planned expected density field, calculate the plan compliance, deviation and motion disorder, generate a basic dynamic risk field, and combine the static hazard source information exported from BIM with the dynamic risk field to generate a comprehensive risk field and perform risk threshold detection and regional warning.

[0046] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any step of the integrated intelligent management method for construction sites as described in the first aspect of the present invention.

[0047] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the integrated intelligent management method for construction sites as described in the first aspect of the present invention.

[0048] The beneficial effects of this invention are as follows: By combining Gaussian kernel density estimation of UWB tag locations with wireless signal inversion, this invention enhances the integrity and stability of the spatial distribution modeling of personnel and equipment at construction sites. Even under conditions of occlusion, lighting changes, and complex environmental interference, it can still maintain the ability to continuously perceive the distribution of entities, thereby improving the comprehensive balance between macroscopic coverage and local accuracy of spatial identification results. By combining optical flow calculation with Monte Carlo sampling, the continuity and stability of the movement field at the construction site are improved, enabling a fine characterization of the dynamic behavior of work entities. By combining the joint calculation of the plan compliance field and the plan deviation field with the superposition of the static risk field, the real-time performance and scene adaptability of risk identification are improved, enabling the system to promptly detect unplanned operations, personnel gatherings, and high-risk movement areas. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the operation of the integrated intelligent management method for building construction sites in Example 1.

[0051] Figure 2 This is a schematic diagram of the integrated intelligent management method for building construction sites in Example 1.

[0052] Figure 3 This is a schematic diagram illustrating the relationship between the density field and the risk field in Example 1.

[0053] Figure 4 This is a schematic diagram of the integrated intelligent management system for building construction sites in Example 2. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0057] Example 1, referring to Figures 1 to 3 This is the first embodiment of the present invention, which provides a comprehensive intelligent management method for construction sites, including the following steps:

[0058] S1. Collect multi-source heterogeneous data and work plan from the construction site and perform preprocessing.

[0059] Specifically, it involves collecting multi-source heterogeneous data and work plan from construction sites and preprocessing them, including:

[0060] Smart sensors are used to collect multi-source heterogeneous data from the construction site, and API interfaces are used to collect work plan from the construction organization design, and timestamp synchronization, noise reduction and standardization are performed.

[0061] The intelligent sensors include high-definition network cameras, lidar, UWB (ultra-wideband) positioning systems, and Wi-Fi probe sensors;

[0062] The multi-source heterogeneous data includes visual images (converted to grayscale), 3D point clouds, UWB tag locations, and radio frequency signal data;

[0063] The work plan includes the coordinates of the work center, the radius of the work area, the planned execution time window (start time and end time), and the expected number of entities.

[0064] S2. Based on the Gaussian kernel density estimation of UWB tag location and the wireless signal inversion, a spatial distribution density field is generated, which is then fused to generate a coarse-grained density and smoothed to obtain the final density field.

[0065] Specifically, a spatial distribution density field is generated based on Gaussian kernel density estimation of UWB tag locations and wireless signal inversion. This density field is then fused to generate a coarse-grained density, which is then smoothed to obtain the final density field, including:

[0066] Establish a construction ground coordinate system with the design origin of the overall construction plan (usually the southwest corner of the site) as the coordinate origin, the east direction as the X-axis, the north direction as the Y-axis, and the vertical upward direction as the Z-axis;

[0067] The construction ground coordinate system is divided into a regular (0.5m × 0.5m) two-dimensional grid;

[0068] For each grid point, the Gaussian kernel density estimation method is used to transform the UWB tag location into a continuous density distribution, as shown in the formula:

[0069] ,

[0070] ,

[0071] ,

[0072] in For in position UWB density estimates at location and time t The coordinates of the points on the two-dimensional grid. Let be the number of active UWB tags at time t, and h be the bandwidth parameter, set using the Silverman rule of thumb. This represents the average of the standard deviations of the UWB tag positions in the x and y directions. For Gaussian kernel function, Let be the two-dimensional position coordinates of the j-th UWB tag at time t;

[0073] By leveraging the characteristics of wireless positioning signals—that they are unaffected by visual obstruction and have a wide coverage range—a macroscopic density field of personnel and equipment distribution within the construction site can be quickly constructed, providing spatial priors for subsequent detailed analysis.

[0074] Using the path loss model of Wi-Fi signal propagation, and employing the inverse distance weighted approximation of the logarithmic distance path loss model, the spatial distribution of the signal source is inverted from the wireless radio frequency signal (received signal strength), as shown in the formula:

[0075] ,

[0076] in For in position Density estimates at location and time t, where M is the total number of Wi-Fi probes. Let be the average signal strength received by the c-th Wi-Fi probe at time t. Let c be the installation location of the c-th Wi-Fi probe. For minimum signal strength, It is a very small positive number;

[0077] To simplify calculations, an inverse distance model with linearly weighted signal strength is used to construct the radio frequency density field as an efficient approximation of the ideal path loss model.

[0078] The density distribution and spatial distribution are respectively subjected to minimization normalization, and then fused using a weighted fusion method to generate a fused coarse-grained density, as shown in the formula:

[0079] ,

[0080] in To generate a coarse-grained fusion density, The standard deviation of Wi-Fi positioning error is determined based on the measured calibration method. The standard deviation of the positioning error for the UWB system is obtained from the equipment specifications. and These are the normalized density distribution and spatial distribution, respectively;

[0081] The percentile method is used to calculate 80% of the historical coarse-grained density. A density threshold is set, and regions with coarse-grained density greater than the density threshold are selected and defined as regions of interest.

[0082] Set a foreground detection threshold, extract the moving foreground pixels from the visual image within the region of interest, and generate a moving foreground mask. The formula is as follows:

[0083] ,

[0084] in Let be the foreground binary mask value of the c-th camera at time t and image pixel coordinates (u,v). A value of 1 indicates that the pixel is identified as moving foreground, and a value of 0 indicates that the pixel is identified as background. Let be the pixel intensity value (e.g., grayscale value) of the c-th camera at time t and image pixel coordinates (u,v). This represents the pixel intensity value at pixel coordinates (u,v) of the background model established using a Gaussian mixture model. The foreground detection threshold is set by calculating three times the standard deviation of pixel intensity changes in historical visual image sequences.

[0085] By using Gaussian kernel density estimation, continuous distribution modeling of UWB tag positioning data was achieved, making the spatial aggregation pattern of personnel and equipment at the construction site more realistic and smooth. This overcomes the shortcomings of traditional discrete point-based statistics which are susceptible to outliers. Wireless signal inversion compensates for the occlusion blind spots of visual and UWB systems in complex construction environments, improves the spatial coverage integrity and robustness of the overall monitoring system, and provides macroscopic density constraints for subsequent multi-source fusion.

[0086] S3. Based on the final density field, Monte Carlo sampling is performed. The final motion field is generated by combining optical flow calculation and Gaussian smoothing. The final density field is divided into connected regions. Based on the final motion field, the centroid and average velocity of each region are calculated to form a set of entity states. The expected density field is generated according to the working condition plan. The entity state set is matched with the plan, and the plan compliance field and plan deviation are calculated.

[0087] Specifically, Monte Carlo sampling is performed based on the final density field, and optical flow calculation and Gaussian smoothing are combined to generate the final motion field. The final density field is divided into connected regions, and based on the final motion field, the following are included:

[0088] The final density field is normalized and used as the probability density function. Monte Carlo sampling is performed on the construction plane (all sampling points are located on the ground plane (Z=0). The projection coordinates of each sampling point on the image plane of each camera are calculated using the perspective projection model of each camera) to find the projection position of the sampling point in each visual image.

[0089] During the construction deployment phase, a checkerboard calibration board was used to calibrate the internal parameters of each fixed high-definition network camera using the Zhang Zhengyou calibration method, obtaining its intrinsic parameter matrix and distortion coefficient vector. Based on the 3D point cloud of the construction site, feature points (such as corner points of buildings and specific parts of permanent equipment) corresponding to the visual images were selected from the point cloud. The external parameters were calibrated using the PNP (Perspective-n-Point) algorithm to determine the rotation matrix and translation vector of each high-definition network camera relative to the construction ground coordinate system.

[0090] Distortion removal processing is performed on each frame of visual image from the high-definition network camera. The Lucas-Kanade optical flow algorithm is used to obtain the instantaneous motion vector of each pixel in the image optical flow field of two adjacent frames. The running foreground pixels in the image are traversed, and the coordinates of the running foreground pixels are back-projected to the camera coordinate system using the intrinsic parameter matrix. The calculated ground intersection coordinates are mapped to a regular grid with the same resolution as the coarse-grained density field.

[0091] The Lucas-Kanade optical flow algorithm is used to calculate the optical flow displacement of the projection position between the current frame and the previous frame. The displaced sampling points are then back-projected onto the ground plane (back-projection is accomplished by solving a system of linear equations formed by the camera projection model). The ground velocity of the sampling points as observed by the camera is then calculated, using the following formula:

[0092] ,

[0093] in Let the ground velocity of the k-th sampling point be the value of the c-th camera. and Let be the two-dimensional position coordinates of the new sampling point and the k-th sampling point in the construction ground coordinate system. The time interval for visual images is determined by the camera frame rate;

[0094] The arithmetic mean of ground velocities is defined as the optical flow velocity.

[0095] The instantaneous velocity of the UWB tag at each sampling point is calculated using the central difference method. The optical flow velocity and the instantaneous velocity are then combined (the two types of observations exist independently) to obtain a sparse velocity field.

[0096] Based on the sparse velocity field, the inverse distance weighted interpolation method is used to calculate the velocity vector of each regular grid point in the construction ground coordinate system, and Gaussian smoothing is applied to obtain the final motion field, as shown in the formula:

[0097] ,

[0098] ,

[0099] in For grid points The overall velocity obtained by interpolation is... and Let be the two-dimensional coordinates of the a-th row, o-th column, and i-th grid point, respectively. Let M be the velocity vector of the i-th observation point, and M be the total number of sampling points in the sparse velocity field. To prevent division by zero by small constants (only distance is used) The 10 most recent observation points are used in the calculation to improve efficiency.

[0100] By replacing traditional uniform grid sampling with probability-driven spatial sampling, redundant computation is significantly reduced while maintaining high-resolution representation of high-activity areas. This improves the response sensitivity to dynamic clustering areas, making subsequent motion field interpolation more consistent with the actual construction behavior distribution. It also has good scalability, enabling rapid adaptive generation of motion samples at construction sites of different sizes. The combination of optical flow and 3D backprojection constructs a bidirectional mapping from visual space to physical space, allowing the system to recover motion trajectories on the actual construction plane. This avoids the limitations of traditional motion extraction based on target detection, which is susceptible to occlusion. It achieves continuous estimation of the overall motion of crowds and equipment. Backprojection calculation combined with time interval parameters can calibrate the visual frame rate and physical velocity, enhancing the temporal consistency of velocity estimation. The combined application of interpolation and smoothing can effectively suppress oscillations caused by local abnormal velocities, improving the spatiotemporal continuity of the motion field. The resulting "final motion field" retains both macroscopic directionality and refines to local dynamics, providing high-quality motion features for subsequent connected region identification. It avoids boundary artifacts caused by direct splicing of optical flow fields, making the field's topological structure natural and continuous, facilitating risk flow analysis.

[0101] Furthermore, the centroid and average velocity of each region are calculated to form a set of entity states. Based on the work plan, a desired density field is generated. The entity state set is then matched against the plan, and the plan compliance field and plan deviation field are calculated, including:

[0102] The peak finding algorithm is used to identify local density maxima in the final density field (for a grid, if its density value is greater than the density values ​​of all other grids in its eight neighborhoods, it is determined to be a density peak point). Each density peak point is used as a seed point to perform region growing in the final density field to segment the connected regions occupied by each entity (each region corresponds to a potential entity, such as a single worker or piece of equipment).

[0103] The region growth (growth criteria) includes starting from the seed point, checking its four neighbors, and if the density value of the neighboring grid is greater than or equal to a global threshold (10% of the maximum density), and the grid has not been marked by other regions, then the grid is incorporated into the region.

[0104] The newly incorporated mesh is used as the new boundary, and the region growth is repeated until the boundaries of all regions can no longer be expanded (i.e. no unoccupied neighboring meshes satisfy the growth conditions), thus obtaining the connected region occupied by each entity.

[0105] Calculate the centroid of the connected region (obtained by the density-weighted average of the coordinates of all grid points in the region) and the average velocity (obtained by the density-weighted average of the final motion field in the region), and combine them into a set of entity states;

[0106] At any grid point, using a two-dimensional Gaussian model, the expected density field for each plan is calculated, as follows:

[0107] ,

[0108] in For the m-th plan at the grid point location The expected density at time t For the entity to be invested in the m-th plan, Let be the coordinates of the center of the work area for the m-th plan. The standard deviation of the Gaussian distribution is used to determine the relationship between the standard deviation and the planned work radius, based on the "3" of the normal distribution. The principle is established that, based on one-third of the planned operation radius, the operation radius is used for entity matching, while the Gaussian standard deviation is used for density modeling, which is logically consistent. This is a time window indicator function. The current time t is within the planned execution time window; the value is 1 if the current time t is within the window, and 0 otherwise. and The start and end times;

[0109] The total expected density field of the plan is obtained by linearly superimposing the expected density fields of all plans (i.e. plans within the time window);

[0110] Traverse the entity state set. For each entity, determine whether its centroid location falls within the work area of ​​any plan. If the condition is met, the entity is marked as an "entity within the plan" (if the entity's resource type attribute belongs to the list of any plan in the area (e.g., earthwork operation plan is associated with excavators, and the identified entity is also an excavator), then it is marked as an "entity within the plan" and assigned to that plan (if it is an overlapping area, then it is assigned to the plan that is closest and matches the type). If the entity's resource type attribute does not conform to the list of any plan in the area (e.g., a crane enters an area that only contains earthwork excavation plans), then even if it overlaps in space, it is still forcibly marked as an "entity outside the plan"); otherwise, it is marked as an "entity outside the plan".

[0111] The work range refers to the Euclidean distance between the centroid position and the planned work position being less than or equal to the radius of the planned work range, and the current moment being within the planned execution time window;

[0112] Based on the labeling results and the total expected density field, the plan compliance field is calculated using the following formula:

[0113] ,

[0114] in For the grid point location The plan compliance at time t is defined as follows: 1 indicates that the location is completely occupied by planned entities, and 0 indicates that there are no planned entities. L is the total number of identified entities, set based on the number of entities marked "planned entities". This is an indicator function; it takes a value of 1 when the grid point is within the entity's region, and 0 otherwise. This is an indicator function that takes the value 1 when the entity is identified as an "in-plan entity" (and 0 when it is an "out-of-plan entity").

[0115] It avoids the problem of traditional clustering algorithms being sensitive to initial parameters. The region division is formed naturally with the density distribution. The joint representation of the centroid of the entity and the average velocity makes the spatial location, motion trend and behavior state of each entity quantifiable. It provides a visualized spatial matching result for the plan compliance field, and makes the deviation between the construction progress and the site status quantifiable in real time. The entity-level matching mechanism can identify local unplanned clusters caused by scheduling errors and operation conflicts. This field can serve as the logical basis for subsequent risk field inference, transforming the static construction plan into a computable dynamic matching index.

[0116] S4. Generate a comprehensive risk field based on the complement and deviation of the plan compliance field, set risk thresholds to identify high-risk areas and trigger early warnings;

[0117] Specifically, a comprehensive risk field is generated based on the complement of the plan compliance field and the deviation, including:

[0118] Calculate the relative absolute deviation between the final density field and the expected density field of the overall plan to generate the plan deviation degree. Calculate the complement of the plan compliance field (1 minus the plan compliance field, where the plan compliance field represents the proportion of "within the plan" entities, and its complement naturally represents the proportion of "outside the plan" or "non-compliant" entities), and define it as the degree of logical conflict.

[0119] For each grid point in the construction ground coordinate system, a local spatial neighborhood (e.g., a 3×3 grid window) is defined centered on that point. The sample covariance matrix of the velocity vectors at all grid points within the local spatial neighborhood is calculated, and eigenvalue decomposition is performed. The largest eigenvalue obtained is defined as the motion disorder of the grid point, as shown in the formula:

[0120] ,

[0121] in Let be the 2x2 covariance matrix of the neighborhood velocity. For grid points The local spatial neighborhood centered on the center For the average velocity field in the neighborhood grid points The value, It is the arithmetic mean of all velocity vectors in the neighborhood, and T is the transpose operation;

[0122] The deviation from the plan, the degree of logical conflict, and the degree of motion disorder are normalized and linearly superimposed to obtain the basic dynamic risk field.

[0123] The spatial coordinate information of all geometric primitives (surfaces, lines and their buffer areas) marked as hazards are exported from the building information model (BIM) of the construction organization design. The entire construction site ground is divided into a two-dimensional grid with the same coordinate system as the construction ground. A blank binary matrix (with all initial values ​​of 0) is created. Using the ray casting method, all grid cells covered by each hazard geometric primitive (surface region or polygon generated by buffer) are marked as 1 in the binary matrix. A static hazard binary matrix aligned with the construction ground coordinate system is generated and defined as a static risk field.

[0124] Before construction, a safety engineer or BIM engineer shall, in accordance with safety standards such as the "Technical Specification for Safety of High-Altitude Operations in Building Construction" (JGJ 80), perform geometric modeling and attribute marking of various static hazard sources in the construction drawing design model of the project.

[0125] By combining the static risk field and the basic dynamic risk field, a comprehensive risk field is generated, as shown in the formula:

[0126] ,

[0127] in As a comprehensive risk field, Based on the dynamic risk field, This is a static risk field with a value of 0 or 1.

[0128] The multimodal motion computation method based on Monte Carlo-optical flow-density fusion improves the accuracy and stability of dynamic behavior modeling. By using Gaussianized plan expression and entity matching, a quantifiable "plan compliance space" is constructed, realizing the computability of construction plans and real-time risk assessment. The generated motion field and matching field can be directly used for safety early warning, resource scheduling and intelligent supervision systems, providing a highly robust data foundation for intelligent construction.

[0129] Furthermore, a comprehensive risk field is generated based on the complement and deviation of the plan compliance field. Risk thresholds are set to identify high-risk areas and trigger early warnings, including:

[0130] The 90% of historical risk values ​​are calculated using the percentile method. A risk threshold is set, and the comprehensive risk field is binarized using the risk threshold. Comprehensive risk fields that are greater than the risk threshold are marked as 1, and those that are not marked as 0, thus generating an initial risk mask.

[0131] A morphological closing operation is performed on the initial risk mask to eliminate noise and connect adjacent regions, resulting in a binary risk mask. The binary risk mask is then aggregated using the Connected-Component Analysis algorithm to generate connected components, and an early warning is issued and the region leader is notified.

[0132] Example 2, refer to Figure 4 As a second embodiment of the present invention, a comprehensive intelligent management system for construction sites includes:

[0133] The collection and processing module is used to collect multi-source heterogeneous data and work plan from the construction site, and to perform timestamp synchronization, noise reduction and standardization processing.

[0134] The density fusion module is used to establish a construction ground coordinate system and a regular grid. It generates a spatial density field based on UWB Gaussian kernel density estimation and Wi-Fi path loss inversion. It then forms a coarse-grained density distribution through weighted fusion and performs smoothing to obtain a refined final density field.

[0135] The motion entity module is used to generate a sparse velocity field through optical flow analysis and UWB velocity calculation, interpolate to form a continuous motion field, and use the DBSCAN algorithm to identify entities and their states at the construction site.

[0136] The risk warning module is used to match the entity status with the planned expected density field, calculate the plan compliance, deviation and motion disorder, generate a basic dynamic risk field, and combine the static hazard source information exported from BIM with the dynamic risk field to generate a comprehensive risk field and perform risk threshold detection and regional warning.

[0137] This embodiment also provides a computer device applicable to the integrated intelligent management method for construction sites, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the integrated intelligent management method for construction sites as proposed in the above embodiment.

[0138] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0139] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the comprehensive intelligent management method for construction sites as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0140] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive intelligent management method for construction sites, characterized by: Includes the following steps: Collect multi-source heterogeneous data and work plan from the construction site and preprocess them. The multi-source heterogeneous data includes visual images, 3D point clouds, UWB tag locations, and radio frequency signal data. The spatial distribution density field is generated by estimating the Gaussian kernel density based on the location of UWB tags and inverting the wireless signal. The density field is then fused to generate a coarse-grained density field, which is then smoothed to obtain the final density field. Monte Carlo sampling is performed based on the final density field. The final motion field is generated by combining optical flow calculation and Gaussian smoothing. The final density field is divided into connected regions. Based on the final motion field, the centroid and average velocity of each region are calculated to form a set of entity states. The expected density field is generated according to the working condition plan. The entity state set is matched with the plan, and the plan compliance field and plan deviation are calculated. A comprehensive risk field is generated based on the complement and deviation of the plan compliance field. A risk threshold is set to identify high-risk areas and trigger early warnings. The calculation of the centroid and average velocity of each region forms a set of entity states. Based on the work plan, a desired density field is generated. The entity state set is then matched against the plan, and the plan compliance field and plan deviation field are calculated, including: The peak finding algorithm is used to identify local density maxima in the final density field, which are defined as density peaks. The final density field is divided into connected regions, and the centroids and average velocities of the connected regions are calculated and combined into a set of entity states. At any grid point, a two-dimensional Gaussian model is used to calculate the expected density field of each plan. The total expected density field of the plan is obtained by linearly superimposing the expected density fields of all plans. Traverse the set of entity states. For each entity, determine whether its centroid position falls within the work area of ​​any plan. If the condition is met, the entity is marked as an "in-plan entity"; otherwise, it is marked as an "out-of-plan entity". The work range refers to the Euclidean distance between the centroid position and the planned work position being less than or equal to the radius of the planned work range, and the current moment being within the planned execution time window; Based on the labeling results and the total expected density field, calculate the plan compliance field; The generation of a comprehensive risk field based on the complement of the plan compliance field and the deviation includes: Calculate the relative absolute deviation between the final density field and the expected density field of the overall plan, generate the plan deviation degree, calculate the complement of the plan compliance field, and define it as the degree of logical conflict; For each grid point in the construction ground coordinate system, a local spatial neighborhood is defined with it as the center. The sample covariance matrix of the velocity vectors at all grid points in the local spatial neighborhood is calculated and eigenvalue decomposition is performed. The largest eigenvalue obtained is defined as the motion disorder of the grid point. The deviation from the plan, the degree of logical conflict, and the degree of motion disorder are normalized and linearly superimposed to obtain the basic dynamic risk field. The spatial coordinate information of all geometric elements marked as hazards is exported from the building information model of the construction organization design. The entire construction site ground is divided into a two-dimensional grid with the same coordinate system as the construction ground. A blank binary matrix is ​​created, and all grid cells covered by each hazard geometric element are marked as 1 in the binary matrix. A static hazard binary matrix aligned with the coordinate system of the construction ground is generated and defined as a static risk field. By combining the static risk field and the basic dynamic risk field, a comprehensive risk field is generated.

2. The integrated intelligent management method for construction sites as described in claim 1, characterized in that: The Gaussian kernel density estimation based on UWB tag location and the wireless signal inversion generate a spatial distribution density field, which is then fused to generate a coarse-grained density and smoothed to obtain the final density field, including: A construction ground coordinate system is established based on the overall construction plan, and the construction ground coordinate system is divided into a regular two-dimensional grid. For each grid point, the Gaussian kernel density estimation method is used to convert the UWB tag location into a continuous density distribution. The inverse distance weighted approximation of the logarithmic distance path loss model is used to retrieve the spatial distribution of the signal source from the radio frequency signal. The density distribution and spatial distribution are respectively subjected to maximum-minimum normalization and fused to generate a fused coarse-grained density. Gaussian filtering is used to smooth the fused coarse-grained density to obtain the final density field; Set a density threshold and filter regions with a coarse-grained density greater than the density threshold, defining them as regions of interest. Set a foreground detection threshold, extract the running foreground pixels of the visual image within the region of interest, and generate a running foreground mask.

3. The integrated intelligent management method for construction sites as described in claim 2, characterized in that: The process of generating the final motion field by performing Monte Carlo sampling based on the final density field, combined with optical flow calculation and Gaussian smoothing, includes: The final density field is normalized and used as the probability density function. Monte Carlo sampling is performed on the construction plane to find the projection position of the sampling point in each visual image. The optical flow displacement of the projection position between the current frame and the previous frame is calculated. The displaced sampling point is back-projected onto the ground plane. The ground velocity of the sampling point under the camera observation is calculated. The arithmetic mean of the ground velocity is calculated and defined as the optical flow velocity. The projection position is determined by calibrating the internal parameters of each high-definition network camera using the Zhang Zhengyou calibration method, calibrating the external parameters of the three-dimensional point cloud using the PNP algorithm, and projecting the running foreground mask onto the ground coordinate system based on the internal and external parameters. The instantaneous velocity of the UWB tag at each sampling point is calculated, and the optical flow velocity and the instantaneous velocity are combined to obtain a sparse velocity field. Based on the sparse velocity field, the inverse distance weighted interpolation method is used to calculate the velocity vector of each regular grid point in the construction ground coordinate system, and Gaussian smoothing is performed to obtain the final motion field.

4. The integrated intelligent management method for construction sites as described in claim 1, characterized in that: The process of generating a comprehensive risk field based on the complement and deviation of the plan compliance field, setting risk thresholds to identify high-risk areas and triggering early warnings includes: Set a risk threshold, use the risk threshold to perform binarization segmentation on the comprehensive risk field, mark the comprehensive risk field greater than the risk threshold as 1, otherwise mark it as 0, and generate the initial risk mask; A morphological closing operation is performed on the initial risk mask to eliminate noise and connect adjacent regions, resulting in a binary risk mask. The binary risk mask is then aggregated using a connected component analysis algorithm to generate connected components, and an early warning is issued and the region leader is notified.

5. The integrated intelligent management method for construction sites as described in claim 1, characterized in that: The collection and preprocessing of multi-source heterogeneous data and work plan from the construction site includes: Smart sensors are used to collect multi-source heterogeneous data from the construction site, and API interfaces are used to collect work plan from the construction organization design, and timestamp synchronization, noise reduction and standardization are performed. The smart sensors include high-definition network cameras, lidar, UWB positioning systems, and Wi-Fi probe sensors; The work plan includes the coordinates of the work center, the radius of the work area, the planned execution time window, and the expected number of entities.

6. A comprehensive intelligent management system for construction sites, used to implement the comprehensive intelligent management method for construction sites as described in any one of claims 1 to 5, characterized in that: include: The collection and processing module is used to collect multi-source heterogeneous data and work plan from the construction site, and to perform timestamp synchronization, noise reduction and standardization processing. The density fusion module is used to establish a construction ground coordinate system and a regular grid. It generates a spatial density field based on UWB Gaussian kernel density estimation and Wi-Fi path loss inversion. It then forms a coarse-grained density distribution through weighted fusion and performs smoothing to obtain a refined final density field. The motion entity module is used to generate a sparse velocity field through optical flow analysis and UWB velocity calculation, interpolate to form a continuous motion field, and use the watershed algorithm to identify the entities and their states at the construction site. The risk warning module is used to match the entity status with the planned expected density field, calculate the plan compliance, deviation and motion disorder, generate a basic dynamic risk field, and combine the static hazard source information exported from BIM with the dynamic risk field to generate a comprehensive risk field and perform risk threshold detection and regional warning.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the integrated intelligent management method for construction sites as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the integrated intelligent management method for construction sites as described in any one of claims 1 to 5.

Images