BIM-based construction site robot real-time monitoring method and system

Abstract

This application relates to a BIM-based real-time monitoring method and system for construction site robots, belonging to the field of building engineering monitoring technology. The method includes: acquiring a real-time BIM model of the target construction body, planning monitoring points, and obtaining a monitoring point sequence and anomaly rate sequence; monitoring and controlling the robot according to the monitoring point sequence, extracting BIM data from the monitoring points and verifying consistency with inspection data to obtain an initial consistency rate; predicting the movement anomaly rate based on the BIM data, configuring verification parameters based on the initial consistency rate and anomaly rate, and obtaining a verification consistency rate after verification; calculating the fused anomaly rate based on the anomaly rate, movement anomaly rate, and verification consistency rate of the monitoring points, marking the monitoring points, and obtaining the monitoring results. This invention solves the problem that traditional construction site monitoring only uses a single monitoring method, without considering related errors and probability issues, resulting in inaccurate judgments on the rationality of construction progress management and insufficient reliability of results, failing to meet the needs of efficient management and control on construction sites.

Description

Technical Field

[0001] This application relates to the field of building engineering monitoring, and in particular to a BIM-based method and system for real-time monitoring of construction site robots. Background Technology

[0002] With the continuous improvement of the level of refinement in construction management, the accuracy and reliability of real-time monitoring at construction sites have become key technical challenges.

[0003] Currently, traditional monitoring methods struggle to accurately reflect the actual construction situation and are susceptible to interference from various error factors, significantly reducing the accuracy of monitoring results. Furthermore, their reliance on a single monitoring method not only diminishes the practical value of the data but also increases the complexity and cost of subsequent construction management and problem-solving, resulting in insufficient reliability of monitoring results and failing to meet the practical needs of efficient construction site management. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a BIM-based method and system for real-time monitoring of construction site robots, which improves upon the current situation where traditional construction monitoring suffers from insufficient accuracy and poor reliability, making it difficult to meet the needs of efficient construction site management.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a BIM-based method for real-time monitoring of construction site robots, the method comprising:

[0007] Obtain the real-time BIM model of the target construction body, and based on the real-time BIM model, plan monitoring points to obtain the monitoring point sequence and anomaly rate sequence;

[0008] According to the monitoring point sequence, the robot is controlled to perform monitoring. When monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted and the consistency is verified with the inspection data to obtain the initial consistency rate.

[0009] Based on the BIM data of the monitoring points, the movement anomaly rate is predicted to obtain the movement anomaly rate. Combining the initial consistency rate and the anomaly rate of the monitoring points, the verification parameters are configured to verify the consistency of the monitoring points and obtain the verification consistency rate.

[0010] Based on the anomaly rate, movement anomaly rate, and verification consistency rate of the monitoring points, the fusion anomaly rate is calculated, the monitoring points are labeled, and the monitoring results are obtained.

[0011] Secondly, embodiments of this application provide a BIM-based real-time monitoring system for construction site robots, the system comprising:

[0012] The BIM monitoring point planning module is used to acquire the real-time BIM model of the target construction body, and to plan monitoring points based on the real-time BIM model to obtain the monitoring point sequence and the anomaly rate sequence.

[0013] The robot initial verification module is used to control the robot to perform monitoring according to the monitoring point sequence. When monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted and the consistency is verified with the inspection data to obtain the initial consistency rate.

[0014] The movement anomaly verification module is used to predict the movement anomaly rate based on the BIM data of the monitoring points, obtain the movement anomaly rate, and configure verification parameters by combining the initial consistency rate and the anomaly rate of the monitoring points to verify the consistency of the monitoring points and obtain the verification consistency rate.

[0015] The fusion anomaly labeling module is used to calculate the fusion anomaly rate based on the anomaly rate, moving anomaly rate, and verification consistency rate of the monitoring points, and then to label the monitoring points to obtain the monitoring results.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] This application proposes a BIM-based real-time monitoring method and system for construction site robots. By acquiring a real-time BIM model of the target construction site and planning monitoring points, the robot is controlled to monitor according to the monitoring point sequence and perform consistency verification. Combined with movement anomaly rate prediction and configuration of verification parameters, the fused anomaly rate is finally calculated and the monitoring results are labeled, achieving accurate real-time monitoring of the construction site. First, a real-time BIM model of the target construction site is acquired, and a monitoring point sequence and anomaly rate sequence are planned accordingly. Then, the robot is controlled to perform monitoring according to the monitoring point sequence, extracting the BIM data of the monitoring points and verifying its consistency with inspection data to obtain an initial consistency rate. Next, the movement anomaly rate is predicted based on the BIM data of the monitoring points, and verification parameters are configured based on the initial consistency rate and the anomaly rate of the monitoring points. Verification is then performed to obtain a verification consistency rate. Finally, the fused anomaly rate is calculated based on the anomaly rate of the monitoring points, the movement anomaly rate, and the verification consistency rate, and the monitoring points are labeled to obtain the monitoring results.

[0018] The technical solution of this application solves the problem of insufficient monitoring accuracy caused by BIM errors, robot recognition errors, and movement error probability in traditional construction site monitoring by integrating BIM model monitoring point planning, robot monitoring and consistency verification, dynamic configuration of movement anomaly rate prediction and verification, and multi-dimensional anomaly rate fusion analysis, thereby achieving high-precision real-time monitoring of the construction site. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0020] Figure 1 A flowchart illustrating the BIM-based real-time monitoring method for construction site robots provided in this application embodiment;

[0021] Figure 2 A schematic diagram of the structure of a BIM-based real-time monitoring system for construction site robots provided in this application embodiment.

[0022] The components represented by each number in the attached diagram are explained below:

[0023] BIM monitoring point planning module 01, robot initial verification module 02, movement anomaly verification module 03, and fusion anomaly annotation module 04. Detailed Implementation

[0024] This application provides a BIM-based method and system for real-time monitoring of construction site robots, which addresses the technical problems in existing technologies such as insufficient accuracy and reliability of construction site monitoring, difficulty in accurately judging the rationality of construction progress management, and inability to meet the needs of efficient management and control of construction sites.

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0028] Example 1, as shown in the appendix Figure 1 As shown, this application provides a BIM-based real-time monitoring method for construction site robots, the method comprising the following steps:

[0029] S110: Obtain the real-time BIM model of the target construction body, and plan monitoring points based on the real-time BIM model to obtain the monitoring point sequence and anomaly rate sequence;

[0030] In this embodiment of the application, in the scenario of real-time monitoring of construction site robots based on BIM, in order to accurately plan monitoring points and obtain the basis for initial anomaly assessment, it is necessary to divide the units through the real-time BIM model and combine it with historical construction data to conduct anomaly rate analysis in order to establish a scientific monitoring sequence basis.

[0031] Specifically, the real-time BIM model of the target construction body is retrieved through the BIM management platform at the construction site to ensure that the model data is synchronized with the on-site construction progress in real time, providing an accurate digital foundation for subsequent analysis.

[0032] Next, the real-time BIM model is structurally divided into multiple independent model units. At the same time, key model features of each model unit, such as structural type, construction stage, and material properties, are extracted as the core basis for anomaly rate analysis.

[0033] Furthermore, based on these model characteristics, the most similar family model unit sets are matched in the historical construction database, and the proportion of discrepancies between the actual construction and the BIM model in each set is counted to determine the unit anomaly rate of each model unit.

[0034] Finally, the model units are sorted in descending order of their anomaly rates, and a preset number of high-risk anomaly units are selected as key monitoring targets, forming a monitoring point sequence and a corresponding anomaly rate sequence.

[0035] This step, through the detailed breakdown of the real-time BIM model and the correlation analysis of historical construction data, enables targeted planning of monitoring points, providing a priority basis for the efficient implementation of subsequent robot inspections and accurate identification of anomalies, and ensuring the accuracy and reliability of construction site monitoring.

[0036] Step S110 in the method provided in this application embodiment includes:

[0037] Obtain the real-time BIM model of the target construction body;

[0038] The real-time BIM model is divided into multiple model units, and the model features of the multiple model units are obtained.

[0039] Based on multiple model characteristics, analyze the unit anomaly rate of multiple model units;

[0040] The abnormality rates of multiple units are arranged in descending order. A preset number of monitoring points are selected before screening to obtain a monitoring point sequence and an abnormality rate sequence.

[0041] In this embodiment of the application, in order to achieve accurate monitoring of the construction site, it is necessary to determine high-risk monitoring points by disassembling and analyzing the real-time BIM model and combining it with historical construction data. This provides a clear and targeted target sequence for subsequent robot inspections, ensuring the efficiency of the monitoring process and the accuracy of anomaly identification.

[0042] Specifically, the real-time BIM model of the target construction body is first retrieved through the BIM management platform at the construction site.

[0043] The BIM model is a digital carrier that reflects the current structural status and construction progress of the construction site. It contains detailed data such as component dimensions, materials, and construction stages. Its real-time nature ensures that it is updated synchronously with the on-site construction progress, providing reliable basic data for subsequent analysis.

[0044] For example, for an office building under construction, the real-time BIM model will be updated synchronously with information such as the floor structure that has been constructed and the location of installed pipelines.

[0045] Furthermore, the acquired real-time BIM model is structurally divided into multiple independent model units.

[0046] Specifically, the division criteria can be determined based on the structural characteristics and construction procedures of the construction body, such as division by floor, component type, or construction area. Simultaneously, model features of each BIM model unit are extracted. These features include, but are not limited to, the geometric parameters of the components, material properties, construction progress nodes, and process requirements.

[0047] The geometric parameters include the length, width, height, cross-sectional dimensions, shape, and position coordinates of the component; the material properties include the material type, material strength grade, material density, and durability index; the construction progress nodes include the planned start time, planned completion time, actual start time, actual completion time, and completion percentage; and the process requirements include welding process, assembly accuracy, and casting process.

[0048] For example, taking a frame column on a certain floor as an example, its model features may include a cross-sectional dimension of 600mm×600mm, a concrete strength grade of C30, a planned pouring completion time of 30 days, and an actual completion rate of 70%, etc.

[0049] Furthermore, based on the extracted features of multiple models and the analysis of the unit anomaly rate of multiple model units, the risk level of possible deviations in each model unit during construction is quantitatively assessed.

[0050] The method provided in this application embodiment includes the step of "analyzing the unit anomaly rate of multiple model units based on multiple model features" as follows:

[0051] Based on multiple model features, multiple sets of family model units are extracted from historical construction data. Each set of family model units includes multiple family model units with the highest similarity to the corresponding model features.

[0052] The process obtains the proportion of anomalies within multiple family model unit sets, thus acquiring the unit anomaly rate for multiple model units.

[0053] In this embodiment of the application, in order to scientifically quantify the abnormal risks of each model unit, it is necessary to construct a set of family units and count the abnormal proportion by accurately matching the model features with historical construction data, so as to provide an objective numerical basis for prioritizing monitoring points and ensure that subsequent monitoring resources are tilted towards high-risk units.

[0054] Specifically, feature matching is first performed within a historical construction database based on the extracted model features. This historical construction database stores detailed feature data and corresponding construction records for each model unit in similar past projects, including both normal and abnormal construction cases.

[0055] Meanwhile, the matching process involves calculating the feature overlap between the current model unit and historical units to select the most similar historical units and form a set of family model units corresponding to that model unit.

[0056] Specifically, the feature overlap is determined by calculating the cosine similarity of the feature vectors of the current model unit and the historical units in each feature dimension of the model. First, each model feature is quantized into vector form. For example, the cross-sectional dimensions in geometric parameters are converted into two-dimensional vectors, the strength grade in material properties is converted into numerical vectors according to standard values, the planned construction period in construction progress nodes is converted into one-dimensional vectors, and the accuracy standards in process requirements are converted into vectors according to the allowable deviation range.

[0057] Furthermore, the cosine similarity between the current unit's feature vector and the feature vector of each historical unit is calculated using the formula: ", where A is the feature vector of the current unit, and B is the feature vector of the historical units, Let |A| and |B| be the vector dot product, and |A| and |B| be the vector magnitudes, respectively.

[0058] The closer the calculated similarity value is to 1, the higher the degree of feature overlap between the two units. Based on this, the top N historical units with the highest similarity (N is set according to the amount of historical data, such as 50-200) are selected to form a family of model units.

[0059] Furthermore, anomalies are identified and statistically analyzed in the selected family of model units to accurately quantify the probability of discrepancies between the actual state and BIM model data during construction.

[0060] Among them, abnormal situations specifically refer to situations in which the actual construction status is inconsistent with the BIM model data during historical construction, including but not limited to the construction progress not being updated according to the model plan, the construction dimensions deviating from the model parameters, the materials used not meeting the model requirements, and the process execution not meeting the model standards.

[0061] In the specific statistics, the construction records of each historical unit in the same family model unit set are checked one by one, model units with any of the above-mentioned abnormal situations are marked, and the number of abnormal units is counted.

[0062] For example, a certain family of model elements contains 150 historical elements. After verification, 24 elements are found to be abnormal, including 10 elements that are behind schedule, 8 elements that are out of size, and 6 elements that are not made of the correct material. Therefore, the number of abnormal elements in this set is 24.

[0063] Furthermore, the anomaly ratio is calculated to determine the unit anomaly rate. That is, the number of anomalous units in the same family model unit set is divided by the total number of units in the set, and the result is expressed as a percentage. The specific calculation formula is "Unit anomaly rate = (number of anomalous units in the same family model unit set / total number of units in the same family model unit set) × 100%".

[0064] For example, in the same example above, the formula can be calculated to be 24 / 150×100%=16%, that is, the cell anomaly rate of the current model cell is 16%.

[0065] During the statistical process, the validity of historical data must be verified simultaneously to remove invalid data caused by incomplete records or significant differences in project types, ensuring the accuracy of the anomaly ratio calculation. For example, if a historical unit only records "construction anomaly" but does not specify the specific anomaly type or parameter deviation value, then that unit will be excluded from the statistical scope.

[0066] Similarly, after matching, anomaly statistics and anomaly rate calculation of all model units in the same family, the unit anomaly rate corresponding to each model unit is formed, realizing the quantitative characterization of the anomaly risk of multiple model units.

[0067] Furthermore, after obtaining the unit anomaly rates of all model units, these unit anomaly rates are sorted in descending order to form a clear priority sequence of the anomaly risk levels of each model unit.

[0068] The sorting process is based solely on the numerical value of the unit anomaly rate. The larger the value of the model unit, the higher the risk of discrepancies between the actual state and the BIM model data during construction, and the higher the corresponding monitoring priority.

[0069] Furthermore, based on the monitoring needs of the construction site, the robot's inspection capabilities, and resource allocation, a preset number of monitoring points (such as 50 or 100) are set. From the sorted model units, a preset number of units are selected as key monitoring targets to form a monitoring point sequence.

[0070] At the same time, the unit anomaly rates corresponding to these monitoring points are arranged in the same order to form an anomaly rate sequence that corresponds one-to-one with the monitoring point sequence.

[0071] For example, if the calculated unit anomaly rate of a certain project's model units is between 5% and 30%, after sorting them from largest to smallest, the anomaly rates of the first 30 model units are 30%, 28%, 27%...18% respectively. If the preset number is set to 30, then these 30 model units form a monitoring point sequence, and their corresponding anomaly rate sequence is [30%, 28%, 27%, ..., 18%].

[0072] This step transforms the dispersed model units into an ordered sequence of monitoring points and a corresponding sequence of anomalies based on their risk levels. This provides a clear path for the robot to conduct inspections and monitoring according to priority, ensuring that high-risk units are monitored first, thereby improving the accuracy and efficiency of on-site monitoring.

[0073] S120: Control the robot to perform monitoring according to the monitoring point sequence. When monitoring data of any monitoring point is acquired, extract the BIM data of the monitoring point and verify its consistency with the inspection data to obtain the initial consistency rate.

[0074] In this embodiment of the application, in the scenario of real-time monitoring of construction site robots based on BIM, in order to achieve accurate comparison between the actual construction status of each monitoring point and the BIM model, it is necessary to collect data in sequence by the robot and perform consistency analysis in combination with verification tools to obtain the initial matching degree evaluation results.

[0075] Specifically, based on the established sequence of monitoring points, motion control commands are first sent to the robot through the control platform to guide the robot to move to each monitoring point in sequence according to priority.

[0076] Upon reaching the target monitoring point, the robot activates its onboard sensors to collect on-site data, such as component images, dimensional parameters, and location information, to ensure that the data accurately reflects the actual construction situation at the monitoring point.

[0077] Meanwhile, during the process of the robot collecting monitoring data, the BIM data corresponding to the monitoring point is extracted from the real-time BIM model, including preset data such as component geometric parameters, material information, and position coordinates recorded in the model, to ensure that the BIM data used for comparison accurately corresponds to the current monitoring point.

[0078] Furthermore, a pre-trained BIM validator is obtained. This BIM validator is an intelligent analysis tool trained based on historical construction monitoring data, capable of quantitatively evaluating the consistency between monitoring data and BIM data.

[0079] Based on this, the monitoring data from the current monitoring points and the extracted BIM data are simultaneously input into the BIM validator. The validator analyzes the differences between the two in key features and outputs a quantitative consistency index, namely the initial consistency rate.

[0080] The closer the initial consistency rate is to 100%, the higher the degree of matching between the actual construction situation at the monitoring point and the BIM model; conversely, it indicates that there is a certain deviation.

[0081] This step, through the coordinated efforts of sequential robot monitoring, precise BIM data extraction, and BIM verifier analysis, achieves a preliminary quantitative assessment of the consistency between the construction status and the model at each monitoring point. This provides basic data support for subsequent anomaly detection and verification, ensuring the objectivity and accuracy of the monitoring process.

[0082] Step S120 in the method provided in this application embodiment includes:

[0083] According to the monitoring point sequence, the robot is controlled to perform monitoring, and when monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted.

[0084] Obtain the BIM verifier;

[0085] Input the monitoring data and BIM data from the monitoring points into the BIM verifier, and output the initial consistency rate.

[0086] In this embodiment of the application, in order to achieve accurate comparison between the actual construction status of the monitoring point and the BIM model, it is necessary to use a robot to collect on-site data in sequence, match the corresponding model data, and use a verification tool to perform consistency analysis in order to quantitatively evaluate the degree of matching between the two and provide an initial basis for subsequent anomaly judgment.

[0087] Specifically, based on the established sequence of monitoring points, the robot control platform first sends path planning instructions to the inspection robot, guiding the robot to move to each monitoring point in order of priority from high to low anomaly rate.

[0088] For example, for the frame column monitoring point with the highest anomaly rate, the platform will generate control instructions including coordinate positioning, movement path, and obstacle avoidance parameters to ensure that the robot accurately reaches the monitoring location.

[0089] Furthermore, after the robot reaches the target monitoring point, it activates the onboard sensing devices to collect on-site data.

[0090] Specifically, if the sensing device is a drone, it can capture multi-angle images of components using a high-definition industrial camera and scan the three-dimensional contours using lidar; if the sensing device is a ground robot, it can detect material properties using infrared sensors and measure installation position deviations using millimeter-wave radar.

[0091] In addition, the collected monitoring data includes, but is not limited to, the actual dimensions, appearance, location coordinates, material characteristics, and construction progress status of the components. The data must be accompanied by a timestamp and monitoring point identifier to ensure data traceability.

[0092] Simultaneously, while collecting monitoring data, the BIM data corresponding to the monitoring point is extracted from the real-time BIM model database. The extraction process accurately associates model units through the unique code of the monitoring point. The acquired data includes the pre-set component geometric parameters, material information, location information, and process standards in the BIM model, ensuring that the model data used for comparison is completely matched with the current monitoring point.

[0093] Furthermore, a pre-trained BIM validator is obtained, and the feature correlation between the monitoring data and the BIM data is mined through the validator's deep internal learning algorithm, thereby quantifying the degree of consistency between the two.

[0094] The method provided in this application embodiment includes the following steps for "obtaining a BIM verifier":

[0095] Based on historical construction monitoring data, a sample monitoring dataset of sample monitoring points is collected, and a sample BIM dataset is also collected. Consistency annotation is performed on each sample monitoring data and sample BIM data to obtain a sample consistency rate set.

[0096] A BIM validator was built based on a Siamese neural network.

[0097] Using the sample monitoring dataset, sample BIM dataset, and sample consistency rate set, the BIM validator is trained and optimized in a supervised manner until convergence, thus obtaining the BIM validator.

[0098] In this embodiment of the application, in order to build an intelligent tool that can accurately assess the consistency between construction site monitoring data and BIM model data, a systematic sample collection, annotation and model training process is required to enable the BIM verifier to deeply mine the correlation between the two types of data features and quantify the consistency rate, so as to provide a reliable basis for the comparison between construction status and model.

[0099] Specifically, firstly, based on historical construction monitoring data, sample monitoring datasets and sample BIM datasets are collected from the set of sample monitoring points.

[0100] The sample monitoring point set was selected from representative construction units in similar past projects, covering monitoring points of different structural types, different construction stages and different anomaly types to ensure the comprehensiveness of the sample.

[0101] In addition, the sample monitoring dataset is actual construction data collected by robotic inspection equipment of the same type as those used on site, including high-definition images, measured parameters, etc., such as on-site images and measured cross-sectional dimensions of a certain frame column (595mm×598mm).

[0102] In addition, the sample BIM dataset is the preset data of the corresponding sample monitoring points in the historical BIM model, including geometric parameters, material properties, location information, process standards, etc. in the model, to ensure that the two types of sample data correspond one-to-one with the monitoring objects.

[0103] For example, for 1,000 sample monitoring points in a historical project, on-site inspection images (resolution 1920×1080), laser scanning size data (accuracy ±1mm), and component parameters in the corresponding BIM model were collected for each point, forming a sample monitoring dataset and a sample BIM dataset containing 20,000 sets of data.

[0104] Furthermore, consistency annotations are performed on the monitoring data and BIM data of each sample to obtain a sample consistency rate set.

[0105] The annotation process is jointly completed by experienced construction managers and BIM engineers, and a consistency rate of 0-100% is assigned based on the degree of deviation between the two on key features.

[0106] Specifically, when the actual monitoring data matches the BIM data perfectly, it is marked as 100%; when there is a slight deviation but within the allowable range, it is marked as 80-95%; when the deviation exceeds the allowable range, it is marked as 50-70%; and when there is a serious discrepancy, it is marked as 0-40%.

[0107] At the same time, the labeling results should be accompanied by detailed explanations of the deviations, such as "size deviation 4mm, consistency rate 92%" or "material strength grade does not match, consistency rate 60%", forming a traceable sample consistency rate set.

[0108] For example, the monitoring data of one sample is "actual floor slab thickness 118mm", and the corresponding sample BIM data is "floor slab model thickness 120mm". The deviation is 2mm and within the allowable range (±5mm), and the consistency rate of the labeled sample is 96%. The monitoring data of another sample is "wall verticality deviation 8mm", and the sample BIM data requires "verticality allowable deviation ±5mm". The deviation exceeds the range, and the consistency rate of the labeled sample is 65%.

[0109] Based on this, a BIM validator is constructed using a Siamese neural network. The Siamese neural network is a special type of deep learning model that includes two identical feature extraction subnetworks with shared weights, used to process sample monitoring data and sample BIM data, respectively.

[0110] Specifically, one sub-network is responsible for extracting edge features, size features, material features, etc. from the monitoring data; another sub-network is responsible for extracting corresponding features from the BIM data; finally, the similarity calculation layer outputs the matching degree between the two, which serves as the sample consistency rate set.

[0111] Furthermore, the BIM validator is trained and optimized in a supervised manner using a sample monitoring dataset, a sample BIM dataset, and a sample consistency rate set until convergence.

[0112] Specifically, the sample data is first preprocessed by converting the image data in the sample monitoring data into a tensor format suitable for neural network input, for example, adjusting the resolution to 224×224 pixels and normalizing it to the range of [0, 1].

[0113] Secondly, the measured parameters are standardized using the Z-score standardization method. That is, for each measured parameter, its mean and standard deviation in the sample set are calculated, and then the parameter is converted into a distribution with a mean of 0 and a standard deviation of 1 by the formula "standardized value = (original value - mean) / standard deviation" to eliminate the impact of differences in the magnitude of different parameters on model training.

[0114] Meanwhile, structured data such as geometric parameters and material properties in the sample BIM data are converted into numerical vectors, and unstructured information such as process standard descriptions is encoded to ensure that the two types of data have the same format and meet the input requirements of the twin neural network.

[0115] Furthermore, the preprocessed sample monitoring dataset and sample BIM dataset are input into the two feature extraction sub-networks of the validator, respectively. Deep features of the data are extracted through network structures such as convolutional layers and pooling layers to generate corresponding feature vectors. The initial prediction consistency rate is obtained by calculating the cosine similarity between the two feature vectors.

[0116] Furthermore, using the labeled values ​​in the sample consistency rate set as the supervision signal, the mean squared error loss function is used to calculate the deviation between the predicted consistency rate and the labeled values. The network weights and bias parameters are adjusted layer by layer through the backpropagation algorithm to continuously reduce the loss value.

[0117] Meanwhile, during the training process, the training set and the validation set are divided into a preset ratio of 8:2. After each iteration, the validation set is used to evaluate the model performance. If the loss value of the validation set does not decrease and tends to stabilize after 10 consecutive iterations, the model is judged to have converged, training is stopped, and the optimized BIM validator is finally obtained.

[0118] Based on this, the monitoring data and BIM data of the monitoring points are input into the trained BIM validator. Through the feature extraction and similarity calculation mechanism of the validator, the initial consistency rate is obtained as the output.

[0119] Specifically, in actual monitoring scenarios, once the robot reaches the target monitoring point and collects monitoring data, it simultaneously extracts the corresponding BIM data from the real-time BIM model. The same operations as sample preprocessing are then performed on these real-time acquired data.

[0120] Furthermore, the preprocessed monitoring data and BIM data are input into two feature extraction sub-networks of the BIM validator, respectively. These sub-networks, through trained and optimized convolutional layers, pooling layers, and other structures, quickly extract deep features from the real-time data, generating corresponding feature vectors. By calculating the cosine similarity between the two feature vectors, the quantitative result of the matching degree between the real-time monitoring data and the BIM data is obtained, and the initial consistency rate is output.

[0121] For example, the actual cross-sectional dimensions collected by the robot at a monitoring point of a frame beam are 398mm×596mm, which corresponds to the cross-sectional dimensions of 400mm×600mm in the BIM data. After preprocessing, the data is input into the BIM verifier. The sub-network extracts the size features, shape features, etc. of the two and calculates the similarity. The final output shows an initial consistency rate of 97%, indicating that the actual construction status of the monitoring point is highly consistent with the BIM model.

[0122] Furthermore, if the actual verticality deviation of another wall monitoring point is 7mm, and the BIM data requires a verticality allowable deviation of ±5mm, the initial consistency rate output after inputting into the verifier is 62%, indicating that there is a certain deviation between the actual construction status of the monitoring point and the BIM model.

[0123] This step uses a BIM verifier to compare and analyze real-time monitoring data with BIM data, enabling a quantitative assessment of the consistency between the actual construction status and the model data. This provides an objective initial basis for subsequent verification based on anomaly rates and movement anomaly rates, effectively improving the accuracy and efficiency of on-site construction monitoring.

[0124] S130: Based on the BIM data of the monitoring points, predict the movement anomaly rate to obtain the movement anomaly rate. Combine the initial consistency rate and the anomaly rate of the monitoring points, configure the verification parameters, and verify the consistency of the monitoring points to obtain the verification consistency rate.

[0125] In this embodiment of the application, in the scenario of real-time monitoring of construction site robots based on BIM, in order to avoid the impact of robot movement deviation on monitoring results and improve the reliability of consistency verification, it is necessary to perform secondary verification on the initial monitoring results by predicting movement anomaly risks and dynamically configuring verification parameters to obtain a more accurate consistency rate assessment.

[0126] First, based on the BIM data of the monitoring points, the movement anomaly rate is predicted. The movement anomaly rate refers to the probability that the actual monitoring position does not match the target monitoring point during the robot's movement to the monitoring point due to factors such as positioning errors and path deviations. To achieve this prediction, a pre-trained movement anomaly analyzer needs to be obtained.

[0127] Specifically, the motion anomaly analyzer is an intelligent model trained on a sample BIM data set and a sample motion anomaly rate set. The sample BIM data set contains BIM data of each monitoring point in historical projects, while the sample motion anomaly rate set records the probability of the robot's actual motion deviation in the corresponding scenario.

[0128] Furthermore, after inputting the BIM data of the current monitoring point into the movement anomaly analyzer, the analyzer outputs the movement anomaly rate of the monitoring point by mining the correlation between the environmental features and movement deviations hidden in the BIM data.

[0129] Furthermore, the initial anomaly rate is calculated by combining the initial consistency rate. Specifically, the initial anomaly rate is the difference between "1 minus the initial consistency rate", which is used to quantify the degree of deviation between the actual construction and the BIM model during initial monitoring.

[0130] Furthermore, the fusion anomaly rate is calculated based on the anomaly rate of the monitoring points, the initial anomaly rate, and the moving anomaly rate.

[0131] Among them, the fusion anomaly rate comprehensively assesses the overall deviation probability of monitoring points by integrating three types of anomaly risks. That is, it is calculated by weighted summation to aggregate multi-dimensional anomaly risk values ​​and form a comprehensive quantitative representation of the overall deviation probability of monitoring points.

[0132] Based on this, review parameters are configured according to the fusion anomaly rate. The review parameters mainly include the number of reviews, and the fusion anomaly rate is positively correlated with the number of reviews. Specifically, it can be calculated by "fusion anomaly rate × preset review base" and rounded down to ensure that high-risk points are reviewed more thoroughly.

[0133] Furthermore, according to the determined verification parameters, the robot is controlled to perform multiple verifications of the monitoring points to ensure consistency. That is, the robot moves back to the monitoring point according to the planned path, collects new monitoring data, extracts BIM data simultaneously and inputs it into the BIM verifier to obtain the consistency rate of each verification; then the average of the multiple verification consistency rates is calculated as the verification consistency rate of the monitoring point.

[0134] This step effectively reduces the impact of single monitoring errors by introducing a mobile anomaly rate assessment to evaluate robot positioning risks and combining it with a multi-dimensional anomaly rate dynamic configuration verification strategy. This makes the consistency rate assessment more closely reflect the actual construction status, provides more reliable data support for subsequent anomaly judgment, and further improves the accuracy and reliability of construction site monitoring.

[0135] Step S130 in the method provided in this application embodiment includes:

[0136] A motion anomaly analyzer is obtained, wherein the motion anomaly analyzer is trained using a sample BIM data set and a sample motion anomaly rate set, and the sample motion anomaly rate includes the probability of robot motion deviation.

[0137] The BIM data is input into the movement anomaly analyzer, and the movement anomaly rate is output.

[0138] The initial anomaly rate is calculated based on the initial consistency rate.

[0139] The fusion anomaly rate is calculated based on the anomaly rate, initial anomaly rate, and moving anomaly rate of the monitoring points.

[0140] Based on the fusion anomaly rate and preset review parameters, the review parameters are calculated, including the number of reviews.

[0141] According to the verification parameters, the monitoring points are verified for consistency, and the average of the verification consistency rate is calculated to obtain the verification consistency rate.

[0142] In this embodiment of the application, in order to eliminate the interference of robot movement deviation on monitoring results and improve the reliability of consistency verification, it is necessary to dynamically configure the verification parameters by predicting movement anomaly risks and integrating multi-dimensional anomaly indicators, and to perform secondary verification on the initial monitoring results in order to obtain a more accurate consistency rate assessment.

[0143] Specifically, a movement anomaly analyzer is first acquired to accurately predict the probability of robot movement deviations. This movement anomaly analyzer is an intelligent prediction model trained based on a sample BIM data set and a sample movement anomaly rate set.

[0144] The sample BIM data set includes model features such as the 3D coordinates of each monitoring point in historical projects, the distribution of surrounding components, and path complexity. The sample movement anomaly rate set records the probability of positioning deviation caused by the offset probability due to construction environment occlusion and uneven ground when the robot actually moves in the corresponding scenario.

[0145] Specifically, the BIM data of the current monitoring point is input into the movement anomaly analyzer. The gradient boosting decision tree (GBDT) algorithm built into the analyzer is used to mine the correlation between features and anomaly rate, so as to output the movement anomaly rate of the monitoring point.

[0146] First, the sample BIM data set is standardized by converting the three-dimensional coordinates into relative position vectors, quantifying the distribution of surrounding components into density indices, and constructing a comprehensive index for path complexity using parameters such as path curvature and the number of obstacles, so that all model features are converted into quantifiable numerical vectors.

[0147] The transformation of three-dimensional coordinates is achieved by establishing a local coordinate system with the construction body reference point as the origin, and subtracting the reference point coordinates from the absolute coordinates of each monitoring point to obtain the relative position vector; the quantification of the distribution of surrounding components is achieved by counting the number of components within a preset radius around the monitoring point and dividing by the spatial volume within that range to obtain the density index.

[0148] In addition, the path curvature is obtained by averaging the curvature values ​​of each curve segment on the robot's movement path. The larger the curvature value, the more curved the path. The number of obstacles is obtained by identifying and counting the obstructions on the robot's preset path.

[0149] At the same time, the probability values ​​in the sample movement anomaly rate set are used as labels and mapped one-to-one with the feature vectors to form a training dataset.

[0150] Furthermore, a moving anomaly analyzer is built based on the gradient boosting decision tree (GBDT) framework.

[0151] The framework is composed of multiple decision trees, each trained based on the residuals of its predecessor trees. Specifically, the first tree initially fits the distribution of the movement anomaly rate by splitting the sample features (e.g., splitting preferentially when the path curvature is >0.8); subsequent trees gradually correct the prediction error until the model's prediction deviation on the training set is lower than a preset threshold (e.g., mean squared error ≤0.001).

[0152] Meanwhile, during training, 5-fold cross-validation is used to optimize hyperparameters such as the number and depth of trees to avoid overfitting.

[0153] Ultimately, this enables the motion anomaly analyzer to accurately capture the correlation between BIM features and robot motion anomaly rates, ensuring high accuracy when predicting new monitoring point BIM data.

[0154] Based on this, the standardized BIM data of the current monitoring point is input into the trained motion anomaly analyzer. The analyzer uses a gradient boosting decision tree algorithm to split and combine the input features layer by layer, calls the learned feature association rules, and outputs the motion anomaly rate of the monitoring point.

[0155] For example, the relative position vector of a monitoring point shows that it is close to the construction elevator passage, with a density index of 12 (dense components) and a path curvature of 0.9 (multiple turns). The movement anomaly analyzer combines these features to output a movement anomaly rate of 14%.

[0156] Furthermore, the initial anomaly rate is calculated based on the initial consistency rate. Specifically, the initial anomaly rate can be expressed as "initial anomaly rate = 1 - initial consistency rate". This formula transforms the initial consistency rate, an indicator reflecting the degree of consistency, into a quantitative value that directly characterizes the risk of deviation.

[0157] For example, if the initial consistency rate of a certain monitoring point is 92%, then the initial anomaly rate = 1 - 92% = 8%. This value intuitively reflects the likelihood of a deviation between the actual construction status and the BIM model in the initial monitoring.

[0158] Furthermore, the fusion anomaly rate is calculated based on the anomaly rate of the monitoring points, the initial anomaly rate, and the moving anomaly rate.

[0159] Specifically, the fusion anomaly rate is calculated by comprehensively assessing the overall deviation probability of monitoring points by integrating three types of anomaly risks. It needs to be obtained by weighted summation, and the specific calculation formula can be expressed as "fusion anomaly rate = anomaly rate of monitoring points × anomaly weight of monitoring points + initial anomaly rate × initial anomaly weight + moving anomaly rate × moving anomaly weight".

[0160] The allocation of each weight needs to take into account the inherent abnormal risks of the monitoring points themselves, as well as the deviations generated during the initial monitoring process and the errors that may be brought about by the robot's movement, so as to make the evaluation results more comprehensive.

[0161] Therefore, the anomaly rate of the monitoring point is set to the highest weight (0.4), as it reflects the inherent risk of anomalies in the construction process of the monitoring point itself, and is the core basis of the assessment. The initial anomaly rate and the moving anomaly rate supplement the assessment of deviation risk from the two aspects of monitoring process and robot movement, respectively. Therefore, each is assigned a weight of 0.3, so as to scientifically aggregate multi-dimensional risks and form a comprehensive quantitative representation of the overall anomaly situation of the monitoring point.

[0162] For example, if the anomaly rate of a certain monitoring point is 18%, the initial anomaly rate is 8%, and the moving anomaly rate is 14%, then the fusion anomaly rate = 18%×0.4 + 8%×0.3 + 14%×0.3 = 7.2% + 2.4% + 4.2% = 13.8%.

[0163] Furthermore, based on the obtained fusion anomaly rate and preset verification parameters, verification parameters are calculated. These verification parameters include the number of verifications. The preset verification parameters include a base number of verifications and an amplification factor.

[0164] Specifically, the basic number of reviews is the minimum number of reviews required to ensure basic monitoring accuracy, usually taken as 1 time; the amplification factor is the proportional coefficient used to convert the fusion anomaly rate into additional review times, usually taken as 10; the specific formula for calculating the number of reviews is "review times = basic number of reviews + round (fusion anomaly rate × amplification factor)", where "round" is a rounding function to achieve quantitative determination of the number of reviews.

[0165] The calculation method of multiplying the fusion anomaly rate by the amplification factor, rounding it, and then adding it to the basic number of verifications ensures that the monitoring point with the higher fusion anomaly rate receives more verifications, thereby reducing the error impact of high-risk points by increasing the number of verifications.

[0166] For example, when the fusion anomaly rate is 13.8%, the number of reviews = 1 + round (13.8% × 10) = 1 + 1 = 2 times; if the fusion anomaly rate is 25%, then the number of reviews = 1 + round (25% × 10) = 1 + 3 = 4 times.

[0167] Based on this, the monitoring points are verified for consistency according to the obtained verification parameters, and the average of the verification consistency rate is calculated to obtain the verification consistency rate, so as to reduce the possible errors in a single monitoring and improve the reliability of the monitoring results.

[0168] Specifically, based on the determined number of checks, the robot replans the optimal path multiple times to move to the monitoring point. Each time it reaches the monitoring point, it restarts the onboard sensors to collect on-site monitoring data, including the actual size, position coordinates, and appearance integrity of the components, to ensure that the data collected each time can truly reflect the real-time construction status of the monitoring point.

[0169] Meanwhile, during each review and collection of monitoring data, the latest BIM data corresponding to the monitoring point is accurately extracted from the real-time BIM model database, including the updated component geometric parameters, material properties, process standards and other preset data in the model, to ensure that the BIM data used for comparison is completely matched with the current monitoring point.

[0170] Furthermore, the monitoring data collected for each review, along with the extracted BIM data, is input into a pre-trained BIM validator. The validator analyzes the degree of matching between the two in key features and outputs the consistency rate for each review. After all reviews have been completed, the arithmetic mean of these consistency rates is calculated, which is the consistency rate for that monitoring point.

[0171] For example, a certain monitoring point is verified twice. During the first verification, the monitoring data collected by the robot and the BIM data are analyzed by the verifier and the output consistency rate is 89%. During the second verification, the output consistency rate is 91%. Therefore, the verification consistency rate of the monitoring point is (89% + 91%) / 2 = 90%.

[0172] Ultimately, by conducting multiple reviews and taking the average, the random errors that may exist in a single monitoring session, such as instantaneous robot positioning deviations and temporary sensor interference, were effectively reduced. This made the final review consistency rate more objectively reflect the degree of consistency between the actual construction status of the monitoring points and the BIM model, providing a reliable basis for subsequent fusion anomaly assessment.

[0173] S140: Calculate the fusion anomaly rate based on the anomaly rate, movement anomaly rate, and verification consistency rate of the monitoring points, mark the monitoring points, and obtain the monitoring results.

[0174] In this embodiment of the application, in the scenario of real-time monitoring of construction site robots based on BIM, in order to obtain the final anomaly assessment result by integrating various anomaly indicators from the previous period, it is necessary to accurately label the anomaly status of the monitoring points in order to form a comprehensive and reliable monitoring result.

[0175] Specifically, the review anomaly rate is first calculated based on the review consistency rate. The formula for calculating the review anomaly rate can be expressed as "Review Anomaly Rate = 1 - Review Consistency Rate". This formula transforms the review consistency rate, which reflects the degree of review consistency, into a quantitative value that characterizes the deviation risk in the review stage.

[0176] Furthermore, based on the anomaly rate, movement anomaly rate, and verification anomaly rate of the monitoring points, a weighted summation is used to calculate the fusion anomaly rate, so as to comprehensively assess the overall anomaly probability of the monitoring points.

[0177] Based on this, the monitoring points are labeled according to the fusion anomaly rate, and the monitoring results are obtained by analyzing the numerical range of the fusion anomaly rate.

[0178] The labeling rules can be divided into different levels based on preset thresholds, and each level comes with specific processing suggestions.

[0179] This step integrates multi-dimensional anomaly indicators to form a comprehensive judgment on the abnormal state of the monitoring points, and presents the monitoring results in a clear and labeled format. This provides a direct and effective basis for decision-making in quality management and risk control at the construction site, and further enhances the completeness and practicality of the monitoring process.

[0180] Step S140 in the method provided in this application embodiment includes:

[0181] The review anomaly rate is calculated based on the review consistency rate.

[0182] The fusion anomaly rate is calculated based on the anomaly rate, movement anomaly rate, and verification anomaly rate of the monitoring points. The monitoring points are then labeled to obtain the monitoring results.

[0183] In this embodiment of the application, in order to integrate multi-dimensional monitoring data to obtain the final anomaly assessment result, it is necessary to accurately label the abnormal status of the monitoring points by integrating the deviation risk of the review stage with various anomaly rate indicators in the early stage, so as to form a comprehensive and reliable monitoring result.

[0184] First, the review anomaly rate is calculated based on the review consistency rate. Specifically, the review anomaly rate can be calculated using the formula "Review Anomaly Rate = 1 - Review Consistency Rate". This transformation quantifies the indicator reflecting the consistency level of the review stage into a numerical value representing the risk of deviation at that stage.

[0185] For example, if the consistency rate of a certain monitoring point is 90%, then the abnormality rate is 1-90%=10%. This value intuitively reflects the possibility that there are still deviations between the actual construction and the BIM model after multiple reviews.

[0186] Furthermore, the fusion anomaly rate is calculated based on the anomaly rate of the monitoring points, the movement anomaly rate, and the verification anomaly rate.

[0187] Specifically, the fusion anomaly rate is obtained by weighted summation, and the specific calculation formula can be expressed as "fusion anomaly rate = anomaly rate of monitoring points × anomaly weight of monitoring points + moving anomaly rate × moving anomaly weight + review anomaly rate × review anomaly weight".

[0188] The weighting is determined by comprehensively considering the actual impact of each anomaly rate. For example, the weight of anomalies at monitoring points is set to 0.3, the weight of moving anomalies is set to 0.2, and the weight of re-verified anomalies is set to 0.5, in order to highlight the importance of the re-verified anomaly rate in the overall assessment, since it is a more reliable quantitative value of deviation risk obtained after multiple re-verifications.

[0189] For example, if the monitoring point anomaly rate of a certain monitoring point is 20%, the movement anomaly rate is 15%, and the verification anomaly rate is 10%, then the fusion anomaly rate = 20% × 0.3 + 15% × 0.2 + 10% × 0.5 = 6% + 3% + 5% = 14%.

[0190] Based on this, the monitoring points are labeled according to the obtained fusion anomaly rate. By defining different anomaly threshold ranges, the monitoring points are divided into different anomaly levels to obtain monitoring results.

[0191] Specifically, the labeling rules are set by preset fusion anomaly rate thresholds. For example, when the fusion anomaly rate is ≤5%, it is labeled as “normal”; when the fusion anomaly rate is 5% < fusion anomaly rate ≤15%, it is labeled as “minor anomaly”; and when the fusion anomaly rate is >15%, it is labeled as “serious anomaly”. Each level corresponds to different processing suggestions.

[0192] For example, when the fusion anomaly rate of a certain monitoring point is 14%, since 5% < 14% < 15%, it is marked as "minor anomaly" according to the labeling rules, with the accompanying handling suggestion "It is recommended to strengthen the daily inspection of this monitoring point and closely monitor its subsequent changes" in order to achieve accurate description and targeted response to the abnormal state of the monitoring point.

[0193] This step, by comparing the fusion anomaly rate with a preset threshold and labeling it in a tiered manner, combined with specific handling suggestions, makes the monitoring results more practical and instructive, providing a clear action basis for risk management and quality assurance at the construction site.

[0194] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0195] This application proposes a BIM-based real-time monitoring method for construction site robots. First, a real-time BIM model of the target construction body is acquired, divided into model units, and model features are extracted. The unit anomaly rate is determined by combining historical construction data, and high-risk units are selected to form a monitoring point sequence and anomaly rate sequence. Next, the robot is controlled to perform monitoring according to the monitoring point sequence, extracting BIM data and inspection data from the monitoring points. Consistency verification is performed using a BIM verifier to obtain an initial consistency rate. Then, the movement anomaly rate is predicted based on the monitoring point BIM data. The fused anomaly rate is calculated by combining the initial consistency rate and the monitoring point anomaly rate. Based on this, verification parameters are configured and verification is performed to obtain the verification consistency rate. Finally, the verification anomaly rate is calculated based on the verification consistency rate. The monitoring point anomaly rate, movement anomaly rate, and verification anomaly rate are merged to obtain the final fused anomaly rate. Monitoring points are then labeled to obtain the monitoring results.

[0196] The method provided in this application adopts a technical solution of "real-time BIM model monitoring point planning - robot initial monitoring and consistency verification - movement anomaly rate prediction and verification parameter configuration - multi-dimensional anomaly rate fusion and result labeling". It integrates anomaly rate analysis based on historical construction data, BIM verifier constructed by Siamese neural network, movement anomaly prediction by gradient boosting decision tree, and multi-dimensional anomaly rate weighted fusion. It solves the problem of insufficient monitoring accuracy caused by BIM error, robot recognition error and movement error probability in traditional construction site monitoring, and realizes high-precision real-time monitoring of construction site.

[0197] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of the BIM-based real-time monitoring method for construction site robots provided in Embodiment 1, this application also provides a BIM-based real-time monitoring system for construction site robots, specifically including:

[0198] BIM monitoring point planning module 01 is used to obtain the real-time BIM model of the target construction body, and to plan monitoring points based on the real-time BIM model to obtain the monitoring point sequence and the anomaly rate sequence.

[0199] The robot initial verification module 02 is used to control the robot to perform monitoring according to the monitoring point sequence. When monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted and the consistency is verified with the inspection data to obtain the initial consistency rate.

[0200] The movement anomaly verification module 03 is used to predict the movement anomaly rate based on the BIM data of the monitoring points, obtain the movement anomaly rate, and configure verification parameters in combination with the initial consistency rate and the anomaly rate of the monitoring points to verify the consistency of the monitoring points and obtain the verification consistency rate.

[0201] The fusion anomaly labeling module 04 is used to calculate the fusion anomaly rate based on the anomaly rate, moving anomaly rate and verification consistency rate of the monitoring points, and to label the monitoring points to obtain the monitoring results.

[0202] In one embodiment, the BIM monitoring point planning module 01 is further configured to:

[0203] Obtain a real-time BIM model of the target construction body; divide the real-time BIM model to obtain multiple model units and obtain the model features of the multiple model units; analyze the unit anomaly rate of the multiple model units based on the multiple model features; arrange the multiple unit anomaly rates in descending order, filter a preset number of monitoring points, and obtain a monitoring point sequence and anomaly rate sequence.

[0204] In one embodiment, the robot initial verification module 02 is further configured to:

[0205] According to the monitoring point sequence, control the robot to perform monitoring. When monitoring data of any monitoring point is acquired, extract the BIM data of the monitoring point; and obtain the BIM verifier.

[0206] Input the monitoring data and BIM data from the monitoring points into the BIM verifier, and output the initial consistency rate.

[0207] In one embodiment, the movement anomaly verification module 03 is further configured to:

[0208] A motion anomaly analyzer is obtained, wherein the motion anomaly analyzer is trained using a sample BIM data set and a sample motion anomaly rate set, and the sample motion anomaly rate includes the probability of robot movement deviation; the BIM data is input into the motion anomaly analyzer, and the motion anomaly rate is output.

[0209] Based on the initial consistency rate, the initial anomaly rate is calculated; based on the anomaly rate of the monitoring points, the initial anomaly rate, and the moving anomaly rate, the fusion anomaly rate is calculated; based on the fusion anomaly rate and preset verification parameters, verification parameters are calculated, wherein the verification parameters include the number of verifications; according to the verification parameters, the monitoring points are verified for consistency and the average of the verification consistency rates is calculated to obtain the verification consistency rate.

[0210] In one embodiment, the fusion anomaly annotation module 04 is further configured to:

[0211] Based on the consistency rate of the verification, the verification anomaly rate is calculated; based on the anomaly rate, movement anomaly rate, and verification anomaly rate of the monitoring points, the fusion anomaly rate is calculated, the monitoring points are marked, and the monitoring results are obtained.

[0212] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0213] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0214] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A BIM-based real-time monitoring method for construction site robots, characterized in that, The method includes: Obtain the real-time BIM model of the target construction body, and based on the real-time BIM model, plan monitoring points to obtain the monitoring point sequence and anomaly rate sequence; According to the monitoring point sequence, the robot is controlled to perform monitoring. When monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted and the consistency is verified with the inspection data to obtain the initial consistency rate. Based on the BIM data of the monitoring points, the movement anomaly rate is predicted to obtain the movement anomaly rate. Combining the initial consistency rate and the anomaly rate of the monitoring points, the verification parameters are configured to verify the consistency of the monitoring points and obtain the verification consistency rate. The fusion anomaly rate is calculated based on the anomaly rate, movement anomaly rate, and verification consistency rate of the monitoring points. The monitoring points are then labeled to obtain the monitoring results. Obtain the real-time BIM model of the target construction body, and based on the real-time BIM model, plan monitoring points to obtain the inspection point sequence and anomaly rate sequence, including: Obtain the real-time BIM model of the target construction body; The real-time BIM model is divided into multiple model units, and the model features of the multiple model units are obtained. Based on multiple model characteristics, analyze the unit anomaly rate of multiple model units; The abnormality rates of multiple units are arranged in descending order. A preset number of monitoring points are selected before filtering to obtain a monitoring point sequence and an abnormality rate sequence. Based on multiple model characteristics, the unit anomaly rate of multiple model units is analyzed, including: Based on multiple model features, multiple sets of family model units are extracted from historical construction data. Each set of family model units includes multiple family model units with the highest similarity to the corresponding model features. The proportion of anomalies occurring within multiple family model unit sets is processed to obtain the unit anomaly rate of multiple model units; According to the monitoring point sequence, the robot is controlled to perform monitoring. When monitoring data is acquired at any monitoring point, the BIM data of the monitoring point is extracted and its consistency is verified with the inspection data to obtain an initial consistency rate, including: According to the monitoring point sequence, the robot is controlled to perform monitoring, and when monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted. Obtain the BIM verifier; Input the monitoring data and BIM data from the monitoring points into the BIM verifier, and output the initial consistency rate.

2. The BIM-based real-time monitoring method for construction site robots according to claim 1, characterized in that, Obtain a BIM validator, including: Based on historical construction monitoring data, a sample monitoring dataset of sample monitoring points is collected, and a sample BIM dataset is also collected. Consistency annotation is performed on each sample monitoring data and sample BIM data to obtain a sample consistency rate set. A BIM validator was built based on a Siamese neural network. Using the sample monitoring dataset, sample BIM dataset, and sample consistency rate set, the BIM validator is trained and optimized in a supervised manner until convergence, thus obtaining the BIM validator.

3. The BIM-based real-time monitoring method for construction site robots according to claim 1, characterized in that, Based on the BIM data of the monitoring points, the movement anomaly rate is predicted, and the movement anomaly rate is obtained, including: A motion anomaly analyzer is obtained, wherein the motion anomaly analyzer is trained using a sample BIM data set and a sample motion anomaly rate set, and the sample motion anomaly rate includes the probability of robot motion deviation. The BIM data is input into the movement anomaly analyzer, and the movement anomaly rate is output.

4. The BIM-based real-time monitoring method for construction site robots according to claim 1, characterized in that, Based on the initial consistency rate and the anomaly rate of the monitoring points, verification parameters are configured to perform consistency verification on the monitoring points, thereby obtaining the verification consistency rate, including: The initial anomaly rate is calculated based on the initial consistency rate. The fusion anomaly rate is calculated based on the anomaly rate, initial anomaly rate, and moving anomaly rate of the monitoring points. Based on the fusion anomaly rate and preset review parameters, the review parameters are calculated, including the number of reviews. According to the verification parameters, the monitoring points are verified for consistency, and the average of the verification consistency rate is calculated to obtain the verification consistency rate.

5. The BIM-based real-time monitoring method for construction site robots according to claim 1, characterized in that, Based on the anomaly rate, movement anomaly rate, and verification consistency rate of the monitoring points, the fusion anomaly rate is calculated. The monitoring points are then labeled to obtain the monitoring results, including: The review anomaly rate is calculated based on the review consistency rate. The fusion anomaly rate is calculated based on the anomaly rate, movement anomaly rate, and verification anomaly rate of the monitoring points. The monitoring points are then labeled to obtain the monitoring results.

6. A BIM-based real-time monitoring system for construction site robots, characterized in that, The system is used to execute the BIM-based real-time monitoring method for construction site robots according to any one of claims 1-5, and the system includes: The BIM monitoring point planning module is used to acquire the real-time BIM model of the target construction body, and to plan monitoring points based on the real-time BIM model to obtain the monitoring point sequence and the anomaly rate sequence. The robot initial verification module is used to control the robot to perform monitoring according to the monitoring point sequence. When monitoring data of any monitoring point is acquired, the BIM data of the monitoring point is extracted and the consistency is verified with the inspection data to obtain the initial consistency rate. The movement anomaly verification module is used to predict the movement anomaly rate based on the BIM data of the monitoring points, obtain the movement anomaly rate, and configure verification parameters by combining the initial consistency rate and the anomaly rate of the monitoring points to verify the consistency of the monitoring points and obtain the verification consistency rate. The fusion anomaly labeling module is used to calculate the fusion anomaly rate based on the anomaly rate, moving anomaly rate, and verification consistency rate of the monitoring points, and then to label the monitoring points to obtain the monitoring results.

Images