Construction scene modeling method for Gaussian process regression

Abstract

The invention discloses a construction scene modeling method based on Gaussian process regression, and relates to the technical field of building management, and the method comprises the steps: extracting a first-stage task from a target construction task, constructing a first task model, and carrying out the modeling of a construction scene through Gaussian process regression. And according to the construction progress, gradually activating a plurality of Gaussian models to carry out dynamic risk prediction, and generating a risk situation. And by taking the task model and the risk situation of the first stage as references, generating a second-stage incremental model through progressive modeling, and continuously activating the Gaussian model in the execution process of the second stage to perform risk prediction. And gradually updating the multi-stage construction risk situation through the construction progress, and outputting a dynamic construction situation cloud picture. The technical problem that an existing construction scene modeling method cannot accurately predict the risk change in the construction process in the complex construction environment is solved, and the technical effect of accurately predicting the risk change in the construction process in the complex construction environment is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of construction management, and in particular to a construction scene modeling method based on Gaussian process regression. Background Art

[0002] During the construction process, various factors such as construction progress, materials, and climate will affect the execution of construction tasks, and the construction site is usually faced with a complex and dynamically changing environment. Traditional construction risk prediction methods often rely on static models, lack dynamic adaptation to real-time data, and cannot effectively deal with the impact of on-site uncertainty. With the increasing demand for intelligent management of construction, the existing risk prediction models cannot meet the requirements of high precision and high real-time performance. Therefore, a new type of construction scenario modeling method is needed that can predict risks in the construction process in real time and accurately, thereby optimizing the construction process and reducing potential losses caused by risks.

[0003] In the current related technologies, there is a technical problem that the construction scenario modeling method cannot accurately predict the risk changes during the construction process in a complex construction environment. Summary of the invention

[0004] The present application provides a construction scene modeling method based on Gaussian process regression, which solves the technical problem that the existing construction scene modeling method cannot accurately predict the risk changes during the construction process in a complex construction environment.

[0005] The present application provides a construction scene modeling method using Gaussian process regression, including: The first-stage task is extracted from the target construction task, and Gaussian process regression modeling of the construction scene is performed on the first-stage task to generate a first task model, wherein the first task model includes multiple first-stage Gaussian models; during the execution of the first-stage task, the multiple first-stage Gaussian models are progressively activated in the first task model according to the construction progress, and the first-stage task is dynamically predicted in parallel to generate a first-stage risk situation; the first task model and the first-stage risk situation are used as the stage modeling benchmark, and the construction scene of the second-stage task is modeled through staged progressive modeling to generate a second incremental model; during the execution of the second-stage task, the first task model and the first-stage risk situation are used as the prediction starting point, and multiple second-stage Gaussian models in the second incremental model are progressively activated according to the construction progress to perform dynamic risk prediction, and the second-stage risk situation is output; and so on, the multi-stage construction scene risk situation of the target construction task is dynamically predicted according to the construction progress, and a construction situation cloud map is output.

[0006] The construction scene modeling method of Gaussian process regression proposed in this application first extracts the first-stage tasks from the target construction tasks, constructs the first task model, and models the construction scene through Gaussian process regression. During the execution of the first-stage tasks, multiple Gaussian models are gradually activated according to the construction progress to perform dynamic risk prediction and generate risk situations. Using the first-stage task model and risk situation as a benchmark, the second-stage incremental model is generated through progressive modeling, and the Gaussian model is continuously activated for risk prediction during the execution of the second stage. The multi-stage construction risk situation is gradually updated through the construction progress, and a dynamic construction situation cloud map is output. By combining the Gaussian process regression model and selecting appropriate kernel functions and mean functions for parameter adjustment, the technical effect of accurately predicting risk changes during the construction process in a complex construction environment is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0008] Figure 1 A schematic diagram of a construction scene modeling method using Gaussian process regression provided in an embodiment of the present application; Figure 2 A schematic diagram of the first-stage risk situation output process of a construction scenario modeling method based on Gaussian process regression provided in an embodiment of the present application. DETAILED DESCRIPTION

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0010] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0011] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0012] The present application embodiment provides a construction scene modeling method using Gaussian process regression, such as Figure 1 As shown, the method includes: Step S100, extracting the first-stage tasks from the target construction tasks, and performing Gaussian process regression modeling of the construction scene on the first-stage tasks to generate a first task model, wherein the first task model includes multiple first-stage Gaussian models. Specifically, firstly extract the first-stage tasks from the target construction tasks to ensure that the tasks are divided according to the specific time schedule and requirements in the construction plan. For example, a construction project involves multiple stages, such as earthwork, reinforced concrete pouring, and equipment installation, and each stage has different tasks and requirements. The first-stage tasks are modeled by the Gaussian process regression method, and the modeling process is trained based on historical data and construction environment factors (such as weather, construction progress, resource use, etc.), which can capture and quantify potential risks and uncertainties in the construction process. For example, in the reinforced concrete pouring stage, it is affected by factors such as weather changes, personnel flow, or equipment failure. Then, based on Gaussian process regression, a first-stage task model is generated, which includes multiple Gaussian models, each of which represents different subtasks in the first stage, such as foundation treatment, concrete pouring, formwork support, etc. The Gaussian model of each subtask makes predictions based on factors such as the actual task progress and resource consumption. For example, the Gaussian regression model is used to predict the impact of delays in earthwork operations on subsequent construction.

[0013] In a possible implementation, the first-stage task is extracted from the target construction task, and the Gaussian process regression modeling of the construction scene is performed on the first-stage task to generate a first task model, wherein the first task model includes multiple first-stage Gaussian models, and step S100 further includes step S110, performing grid scale optimization according to the construction characteristics of the first-stage task to obtain a first grid scale. Specifically, when performing grid scale optimization, it is first necessary to conduct a detailed analysis based on the construction characteristics of the first-stage task, including the geography and terrain conditions of the construction site, the complexity of construction activities, the distribution of construction materials and environmental factors. For example, deep foundation pit areas or high-altitude operation areas usually require more detailed grid division because the risks in the above areas are higher and the environmental changes are greater. When determining the grid scale, key factors include spatial location differences, construction density and environmental changes. For example, some areas may have dense construction and require higher grid resolution, while other areas can use larger grid scales. In addition, environmental changes at the construction site, such as climate or soil changes, will also affect the selection of grid scales. Through numerical optimization methods, the most suitable grid scale can be selected according to the risk prediction accuracy and computational efficiency at different grid scales. For example, smaller grid units may be required to accurately simulate the bearing capacity of the foundation during the foundation construction phase, while larger grids can be used to improve efficiency during the exterior wall construction phase. After selecting the initial grid size, it is also necessary to verify it with actual construction data to compare the deviation between the model prediction and the actual situation. If the accuracy does not meet the requirements, the grid density is adjusted to ensure that the final grid size can effectively improve the accuracy of risk prediction.

[0014] Step S120, gridding the first task model into a plurality of first task pixels using the first grid scale. Specifically, according to the selected first grid scale, the construction task area is first gridded, and each grid unit represents a small area in the construction scene. For example, if the first grid scale is 1 square meter, the entire construction area will be divided into a plurality of small grid units of 1 square meter, and each unit becomes a task pixel. Each task pixel carries the construction characteristic data of the area, such as the construction progress of the area, the building materials used, the climatic conditions, the status of the construction equipment, etc. The area represented by the task pixel may include a concrete pouring area, a steel bar binding area, a foundation treatment area, etc., and the characteristics and construction status of each area are recorded independently. In order to ensure the accuracy of the data of each pixel, the data collection of each pixel includes not only the sensor data of the construction site, but also the feedback of the operator and external factors such as weather changes. The data is associated with information such as spatial position through a Gaussian regression model, so as to generate a dynamic risk prediction for each pixel. For example, a certain area may cause a delay in the construction progress due to weather, or a safety risk caused by the complexity of the terrain. The model can predict potential problems and provide early warnings for managers. Through gridding and data association, each area in the construction process can be monitored in real time, thus achieving more efficient construction management and risk control. The dynamic data of task pixels is updated as the construction progresses, and through the iterative adjustment of the Gaussian process regression model, the prediction of construction risks always remains accurate.

[0015] Step S130, Gaussian process regression modeling is performed on the multiple first task pixels to generate multiple first-stage Gaussian models. Specifically, in the process of implementing Gaussian process regression modeling, the construction characteristic data of each first task pixel is first prepared, including variables such as construction progress, weather conditions, and equipment status. Each pixel also contains specific spatial location data (such as area number, coordinates, etc.). After the data is standardized, it is passed to the Gaussian process regression model as input. In the modeling process, a suitable kernel function (such as radial basis function or Mattern kernel function) is selected to establish the correlation between different task pixels. For example, the area where a certain pixel is located may be greatly affected by climate change during the construction process. The kernel function can be used to capture the impact and model it. Next, the model uses the training data to calculate the risk change trend of each pixel through regression analysis, and outputs the Gaussian model of each task pixel. For example, the model may predict that the risk of a certain area will increase in a certain period of time in the future, which may be due to the failure of the equipment that the area relies on or the change of external conditions (such as rainfall). The Gaussian model will reflect the comprehensive impact of multiple factors such as construction progress, equipment status, and environmental conditions, thereby providing dynamic risk prediction for construction management.

[0016] In a possible implementation, the grid scale is optimized according to the construction characteristics of the first stage task to obtain the first grid scale, and step S110 further includes step S111, gridding the first task model into multiple standard grid units based on the standardized grid scale. Specifically, in the process of gridding the first task model based on the standardized grid scale, it is first clarified that the goal of gridding is to decompose the complex construction scene into multiple standard grid units, each unit reflects the local characteristics of the construction area. For example, in the urban bridge construction scene, the scale of the grid unit can be determined according to the pier spacing or the foundation type to ensure that the grid division is fine enough to capture the local characteristics without increasing too much computational burden. In actual operation, the space of the construction scene is divided into uniform units according to the preset grid scale, and the units are given specific attributes, such as the spatial location of the construction area, terrain characteristics, material strength, etc. Taking underground construction as an example, a grid unit may contain foundation bearing capacity, concrete strength grade, and groundwater level change data. In multi-story building construction, different grid units can respectively describe the material characteristics of the foundation area, the load distribution of the above-ground floors, and the support strength of the underground structure. When the grid unit is initialized, it will be assigned initial attribute values ​​based on known construction data, such as the initial bearing capacity of the foundation area and the depth of the foundation pit, and will be dynamically updated during the construction progress. The grid processing allows the entire construction model to be presented in a more intuitive and detailed way, laying a solid foundation for subsequent Gaussian regression modeling and dynamic risk prediction, while improving the computational efficiency and applicability of the model.

[0017] Step S112, collect the first construction characteristic data from the first stage task, and calculate the characteristic gradient change of the multiple standard grid units according to the first construction characteristic data to obtain multiple characteristic gradient units. Specifically, during the construction process, first collect relevant construction characteristic data from the first stage task. The characteristic data includes terrain height, soil type, strength grade of construction materials, ambient temperature and humidity, foundation settlement, etc. For example, in bridge construction, the bearing capacity, soil type and settlement of the foundation are key characteristics, while in high-rise building construction, the strength grade of concrete, steel type and wind load may be important construction characteristics. The characteristic data is distributed to multiple standard grid units according to the grid scale, and each grid unit represents a small area in the construction scene, carrying its specific spatial position and construction characteristics. For example, a grid unit may correspond to the bearing capacity value of the foundation, and another grid unit may contain the compressive strength of concrete. Then, according to the construction characteristic data of the grid unit, the characteristic gradient change is calculated to quantify the characteristic difference between adjacent grid units and calculate the change rate of the difference. For example, in foundation construction, the amount of foundation settlement may vary greatly in some areas, and gradient units are formed in areas with more significant changes. Gradient units are particularly critical in construction scene modeling, especially in the identification of uneven foundation settlement or high-risk areas. Finally, according to the changes in characteristic gradients, areas with significant risks are screened out from multiple characteristic gradient units through preset thresholds, and the units are recorded and marked. The marked units represent potential risk areas and can provide accurate data support for subsequent risk prediction and Gaussian process regression modeling of construction scenarios, ensuring that possible risks in the construction process are discovered in a timely manner and corresponding measures are taken.

[0018] Step S113, preset a risk gradient threshold, and select N risk gradient units from the multiple characteristic gradient units according to the risk gradient threshold. Specifically, in order to accurately identify high-risk areas during construction, a risk gradient threshold needs to be set first. The threshold is usually based on construction characteristics, such as the rate of change of parameters such as soil settlement, concrete strength or steel bar stress. For example, when the soil settlement rate is greater than a certain threshold, it may mean the risk of uneven settlement, or the steel bar shows a strength change beyond the tolerance range during the stress process, which may lead to safety hazards in construction. When setting the threshold, it is also necessary to combine the influence of the construction environment and historical data, such as geological environment, weather changes and other factors. Then, according to the set risk gradient threshold, the system compares the calculated gradient change value of each construction area with the threshold, and selects the unit area with excessive risk. Each unit represents a part of the construction site, and characteristics such as terrain, soil type, construction technology, etc. are reflected in it. For example, if the settlement change rate of a particular area exceeds the threshold, the area will be marked as a high-risk unit. Through the screening mechanism, the system finally obtains N high-risk gradient units, and the area needs to be paid special attention. In the actual construction process, if a failure or accident occurs in an area, it will directly affect the construction progress and safety. Therefore, the selected high-risk units will serve as the basis for subsequent construction adjustments. Engineers can take reinforcement, adjust the construction sequence or other countermeasures to prevent potential construction risks and ensure the safe and smooth progress of the project.

[0019] Step S114, gridding the N risk gradient units into N groups of first updated grid units using 1 / 2 of the standardized grid scale. Specifically, by using 1 / 2 of the standardized grid scale for gridding, the originally larger risk gradient unit is subdivided into a plurality of smaller updated grid units, thereby improving the spatial resolution and being able to capture the risk changes in the construction site more finely. For example, in a construction project, the original grid unit covers a larger area, in which the risk change of soil settlement is not obvious, but by reducing the grid scale by half, the different settlement rates of the local area can be carefully revealed, helping to identify some hidden danger areas. The subdivided grid units not only reflect the higher-resolution spatial features, but also better capture the small risk changes that may occur during the construction process. By refining the grid, the risk of each small area can be more accurately predicted dynamically. For example, in areas with complex geological conditions, it will be found that the risk changes in some areas of the refined grid units are greater, so that measures can be taken earlier to avoid potential safety problems. This method can provide higher construction process prediction accuracy and ensure more scientific risk assessment in the actual construction process.

[0020] Step S115, calculate the characteristic gradient change of the N groups of first updated grid units according to the first construction characteristic data to obtain N groups of updated gradient units. Specifically, in the construction task, it is first necessary to collect the construction characteristic data related to the first stage task, and the data includes soil type, material type (such as concrete grade, steel bar type), construction process, and environmental conditions (such as temperature, humidity, etc.). The characteristic data reflects the actual situation at the construction site. Next, calculate the characteristic gradient change of the N groups of first updated grid units according to the data. The gradient calculation result of each grid unit represents the rate of change of the construction characteristics in the area, such as the change of soil density in different areas, or the difference in concrete strength at different construction locations. By calculating the gradient, the changes and potential risks in the construction process can be quantified. Subsequently, the calculated gradient is used to generate N groups of updated gradient units, and the updated unit reflects the characteristic changes at a higher resolution grid scale.

[0021] Step S116, M second updated grid units are selected from the N groups of updated gradient units according to the risk gradient threshold. Specifically, during the construction process, a risk gradient threshold needs to be set first, and the threshold is used to distinguish different risk areas of the construction site. The threshold is set according to factors such as soil density change, material strength, and terrain change. For example, if the soil density of a certain area changes too much, or the strength value of some concrete columns is lower than the standard requirement, then the risk of the above area is higher and needs to be treated first. When calculating the gradient value of each grid unit, the risk change rate of each grid unit is obtained by collecting construction data and combining it with environmental data. Assuming that there are some areas on the construction site, the foundation settlement caused by soil softening due to rainfall, the calculated gradient value will reflect the degree of change in the area. Then, the calculated gradient value is compared with the set risk gradient threshold. Only when the gradient value is higher than the threshold, it is considered a high-risk area. Grid units with higher risks will be selected to form M second updated grid units, and the above areas need to be paid special attention to, reinforced, monitored or replanned. For example, if there is a high risk of settlement in an area under construction, the system will automatically screen out the area and mark it, thus ensuring that potential risk areas can be responded to and handled in a timely manner throughout the construction process.

[0022] Step S117, and so on, the grid unit screening and updating are performed until the updated grid unit meets the risk gradient threshold, and the first grid scale is output. Specifically, during the construction process, all possible high-risk areas are first screened out according to the set risk gradient threshold. The high-risk area is divided into multiple grid units in a gridding manner, where each grid unit represents an area with specific construction characteristics and risk gradient values. According to the gradient value, the system will identify the high-risk area and give priority to it. For example, if a geological area has a high risk of settlement, the system will mark the area as high risk and include the grid unit in the screening range. Next, the system will refine each screened high-risk grid unit, gradually reduce the grid scale, and ensure that the risk area can be accurately captured. If the risk gradient value of some areas is lower than the set threshold, they will be excluded in subsequent screening, and only areas with continuous high risk will be retained. This process is continuously iterated, and the accuracy of risk assessment is gradually optimized through repeated screening and updating. Finally, when the risk gradients of all grid cells meet the set threshold requirements, the final first grid scale is output, which accurately reflects the risk level of each area during the construction process, providing effective support for subsequent construction decisions, resource allocation and risk response. For example, at a construction site, the risk of ground subsidence is high. After gridding and refining the scale, the system determines the area that needs reinforcement the most and divides it into high-risk areas, so that reinforcement is given priority to ensure the safety of construction.

[0023] In a possible implementation, Gaussian process regression modeling is performed on the multiple first task pixel points to generate multiple first-stage Gaussian models, and step S130 further includes step S131, after the first task execution time is retrieved from the first-stage task, environmental data is collected according to the first task execution time to obtain the first environmental prediction data. Specifically, before the construction task starts, the system first determines the start and end time of the construction according to the execution time of the first-stage task, which is crucial for the subsequent environmental data collection. For example, if the first-stage task is foundation excavation, the determination of the execution time can provide a time window for the collection and analysis of relevant data. Then, the system collects relevant environmental data based on this time window, including climate information such as temperature, humidity, wind speed, and geographical information related to the construction site such as soil moisture, groundwater level, and terrain height. For example, in the concrete pouring stage, changes in temperature and humidity will significantly affect the hardening process of concrete, so collecting these data is crucial to the construction progress. Finally, the system processes and analyzes the collected environmental data to generate the first environmental prediction data, which not only reflects the current environmental conditions, but also predicts the environmental change trend in the future. For example, based on real-time weather and historical data, the system may predict the amount of precipitation in the next few days, prepare for construction in advance, and avoid adverse effects of rainfall on construction progress and safety. Through the collection and prediction of these data, the system can provide support for risk assessment during the construction process, such as predicting the impact of extreme weather on construction progress, or warning of landslide risks caused by excessive soil moisture.

[0024] Step S132, matching similar construction tasks according to the first construction characteristic data and the first environmental prediction data to obtain a first similar task model, wherein the first similar task model includes H sample Gaussian models of H sample pixels, and the H sample pixels have H sample pixel construction characteristic identifiers. Specifically, in the first stage of the construction task, the system first extracts construction characteristic data from the task, such as settlement control range, concrete grade, steel bar type, construction process (such as template construction, casting process) and construction location (including terrain height, spatial coordinates, etc.). At the same time, environmental prediction data (such as temperature, humidity, wind speed, etc.) will also be collected. These data will be used as a basis, and the system will find construction tasks similar to the current task by performing similarity matching with historical data. This matching is based on the similarity measurement of multiple construction characteristics, such as concrete grade, settlement control range, etc., so as to ensure that the most similar task scene is selected. In the above manner, the system generates a first similar task model, which contains multiple sample pixels, each pixel represents a construction area, and carries corresponding construction characteristic identifiers, such as design materials, construction processes, etc. Each sample pixel generates a Gaussian model through Gaussian process regression, which reflects the different characteristics and dynamic changes of construction tasks. Through the Gaussian model, the system can dynamically predict the potential risks and performance of construction tasks under various conditions, thus providing accurate data support for subsequent task execution and risk management.

[0025] Step S133, extracting multiple task construction characteristics of the multiple first task pixels from the first stage tasks. Specifically, during the execution of the first stage tasks, the system extracts rich construction characteristic data from multiple task pixels. Each pixel represents an area of ​​the construction site, including multiple construction characteristics of the area. These characteristics include settlement control range, concrete grade, steel type, construction process (such as concrete pouring process or formwork installation process), terrain height, spatial coordinates, etc. Taking the settlement control range as an example, in an earth excavation area, soil stability and settlement control measures may be involved; in a reinforced concrete structure area, concrete grade and steel type are key construction characteristics. Construction technology may include pouring speed, formwork erection method, etc. in different areas. The characteristics of each area are clearly marked by identification, such as a certain area may require a specific grade of concrete and steel type, while another area needs to consider different terrain heights and construction parameters. The extraction and identification of characteristic data provide basic data for modeling and dynamic risk prediction of subsequent construction tasks, ensuring that the construction needs and potential risks of each area can be accurately assessed throughout the construction process.

[0026] Step S134, pre-constructing a construction characteristic evaluation network. Specifically, the process of pre-constructing a construction characteristic evaluation network first extracts template pixel construction characteristics from multiple construction tasks, covering multiple factors such as terrain, height, concrete grade, steel bar type, construction process, settlement control range and design materials, reflecting the physical and technical characteristics of different construction tasks. Then, all the collected characteristic data will be standardized to ensure that they are compared under a unified standard, thereby eliminating differences in different dimensions and units. For example, changes in terrain may affect the pouring process of concrete, while the type of steel bar may affect the bearing capacity of concrete. The standardized data will be used for characteristic combination and similarity metric calculation to evaluate the similarity between different construction tasks or areas, such as by calculating Euclidean distance or cosine similarity to measure the similarity between different construction characteristics. Then, through the similarity metric channel, the construction characteristics are mapped to a unified standard space to provide support for subsequent task matching and Gaussian regression modeling. Finally, combining the characteristic similarity metric and the standardized channel, a construction characteristic evaluation network is constructed, which can analyze and evaluate the risks and adaptation of construction tasks in real time, helping construction teams to effectively respond to various challenges in complex construction environments.

[0027] Step S135, after the construction characteristics of the H sample pixels and the construction characteristics of multiple tasks are combined and enumerated, the construction characteristic evaluation network is used to measure the characteristic similarity of the combination results, and multiple sample mapping relationships between the construction characteristics of the H sample pixels and the construction characteristics of multiple tasks are obtained. Specifically, the characteristics of multiple construction tasks are first combined and enumerated with the construction characteristics of the H sample pixels. The construction characteristics of each task cover multiple dimensions such as terrain, construction technology, material selection, such as concrete grade, steel bar type, terrain height, etc. By enumerating all characteristic combinations, a larger set of construction characteristics can be generated to ensure the coverage of all potential construction scenes. Then, the pre-constructed construction characteristic evaluation network is used to measure the similarity of these enumerated characteristic combinations. The evaluation network measures the degree of matching by calculating the similarity between the construction characteristics of each task and the characteristics of the sample pixels, such as using Euclidean distance or cosine similarity. For example, if the concrete grade and steel bar type of a task are similar to the characteristics of a sample pixel, the network will output a higher similarity value. Based on the similarity measurement results, the mapping relationship between the construction characteristics of multiple tasks and the sample pixel characteristics is generated. The mapping relationship shows which tasks' construction characteristics are most similar to the sample pixel characteristics, and provides a basis for subsequent construction scene modeling.

[0028] Step S136, according to the multiple sample mapping relationships, the H sample Gaussian models are migrated and adjusted to generate the multiple first-stage Gaussian models adapted to the multiple first task pixel points. Specifically, firstly, multiple sample mapping relationships are obtained through the above steps, and these mapping relationships are established based on the similarity between the characteristics of the construction task and the sample pixel characteristics. Based on the mapping relationship, the H sample Gaussian models are migrated and adjusted. Specifically, the goal of the migration adjustment is to enable these Gaussian models to adapt to the new construction task characteristics, especially the construction characteristics of multiple first task pixel points. Each task pixel point represents an area in the construction scene, with different construction characteristics, such as settlement control range, concrete grade, steel bar type, etc. The migration adjustment process includes adjusting the mean and covariance matrix of the model by mathematical methods according to the existing sample Gaussian model, so that the model can reflect the new construction characteristics. The adjustment not only retains the validity of the original sample model, but also enables the model to adapt to the new construction task and provide construction risk prediction for different areas. Finally, the adjusted multiple Gaussian models can accurately predict the risks of each task pixel point in the actual construction process and effectively evaluate the construction progress of each area.

[0029] In a possible implementation, a construction characteristic evaluation network is pre-constructed, and step S134 further includes step S1341, interactively obtaining construction characteristics of multiple template pixels. Specifically, in construction scene modeling, interactively obtaining construction characteristics of multiple template pixels involves detailed division of construction areas or tasks, and each division unit is a pixel point, representing the specific construction characteristics of the area. For example, the settlement control range may be obtained through geological survey data, information such as concrete grade and steel bar type is usually extracted from design drawings or construction reports, and terrain height is obtained through field surveys or remote sensing technology. Through real-time interaction with the site, combined with automatic sensor data and construction management system interface, it is ensured that the characteristic data of each construction task area can be transmitted to the system in real time. After obtaining the characteristic data, the system standardizes the information, uses unique hot encoding for categorical data (such as steel bar type, concrete grade), and processes continuous data (such as settlement range, terrain height) through standardized methods, so that the data can be uniformly entered into the subsequent modeling stage. Points marked as template pixel construction characteristics contain a complete description of the construction task. Each template pixel construction characteristic identifies various construction conditions and requirements of the task, including but not limited to settlement control, construction technology, material selection, task parameters, etc. In addition, dynamic changes in the construction site will also affect the characteristic data, such as changes in material selection or construction technology. Therefore, these data are updated through real-time feedback to ensure that the data throughout the construction process remains up-to-date and accurate. The continuous update method ensures the accuracy of construction scene modeling and provides strong support for subsequent risk prediction and construction decisions.

[0030] Step S1342, the construction characteristics of the multiple template pixels are standardized and annotated to obtain multiple template characteristic vectors. Specifically, in the process of construction scene modeling, the construction characteristics of the multiple template pixels need to be standardized first to ensure that different types of data can be compared at the same scale. For continuous data, such as settlement range and terrain height, a standardized formula is used to convert them into values ​​with zero mean and unit standard deviation. For example, for settlement range data, its original value may be distributed in a large interval, and after standardization, each data point of the settlement range will be converted according to its mean and standard deviation, so that they fall within the same numerical range, thereby avoiding the influence between different dimensions. For categorical data, such as concrete grade and steel bar type, it is processed by one-hot encoding or embedding vectors to ensure that discrete data can be converted into a digital format that can be processed by the machine learning model. For example, concrete grade C30 can be converted into [0,1,0] by one-hot encoding, and steel bar type HRB400 is converted into a corresponding vector representation, and each category is mapped to a unique digital representation. The standardized characteristic data will be merged into a template characteristic vector, representing the construction characteristics of each task area. The vectors will be used as the basis for subsequent construction task matching and modeling, so that the construction characteristics of each task area can be compared and analyzed under a unified standard.

[0031] Step S1343, the construction characteristics of the plurality of template pixels are combined and enumerated to obtain F groups of template construction characteristics. Specifically, in the construction scene modeling, the construction characteristic data of the plurality of template pixels are first collected, each pixel represents a specific area in the construction scene, and includes key characteristics of the area, such as settlement control range, material selection (such as concrete grade and steel bar type), construction process, terrain height, spatial coordinates, etc. For example, the settlement control range may be 1-5mm, the concrete grade may have the selection of C30, C35 and C40, and the steel bar type is HRB400 or HRB500. After collecting the characteristics, the combination enumeration is then performed to combine different characteristics into multiple possible construction scenarios. For example, by combining the settlement control range, concrete grade and steel bar type, multiple construction characteristic combinations can be obtained, such as the first group combination is the settlement control range of 1-5mm, the concrete grade of C30 and the steel bar type of HRB400, and the second group combination is the settlement control range of 2-6mm, the concrete grade of C35 and the steel bar type of HRB500. This process enumerates all possible combinations and ultimately generates F groups of template construction characteristics. Each group represents a specific construction scenario, and the combination provides a basis for subsequent construction task matching and Gaussian regression modeling. For example, when a construction task needs to determine the applicable materials or processes based on these characteristics, the preset combination of construction characteristics can help optimize the construction plan and improve construction accuracy and efficiency. The F group of template construction characteristics ultimately provides accurate data support for construction tasks, which helps to dynamically adjust and optimize the risk prediction of construction scenarios.

[0032] Step S1344, perform similarity measurement on the F groups of template construction characteristics to obtain the similarities of the F template characteristics. Specifically, when performing similarity measurement on template construction characteristics, it is first necessary to perform standardization on multiple template construction characteristics to eliminate the influence of different dimensions and units. For example, for continuous data (such as settlement range, terrain height), the standardization formula is used. ,in is the data value after standardization. is the original data value, is the mean, As the standard deviation, the data is converted to a unified scale. For categorical data (such as concrete grade, steel bar type), One-Hot_Encoding or Embedding methods are used to convert them into numerical form. Next, the similarity between each group of template construction characteristics is calculated by similarity measurement methods (such as Euclidean distance, Hamming distance, etc.). For example, the settlement range and terrain height can be measured by Euclidean distance, while the concrete grade and steel bar type can be measured by Hamming distance. Through similarity calculation, the similarity value between each group of template characteristics and other groups of characteristics is obtained, and the value can provide decision support for construction tasks. Ultimately, the similarity value helps to optimize the selection of construction plans, improve construction efficiency, reduce potential risks, and optimize the risk prediction of construction scenarios through further Gaussian regression models to cope with unforeseen complex situations during construction.

[0033] Step S1345, using the construction characteristics of the multiple template pixels and the multiple template characteristic vectors to train a characteristic standardization channel. Specifically, in order to effectively process the construction characteristic data of the multiple template pixels, firstly, the construction characteristic data related to the template is collected, and the data includes the settlement control range, concrete grade, steel bar type, construction process, terrain height and spatial coordinates. The characteristics represent multiple key factors in the construction process, such as construction methods, material selection and site environment. Next, the collected construction characteristic data is standardized. For continuous data, such as the settlement control range and terrain height, the standardized formula To convert the data into dimensionless data so that it can be compared on the same scale. For categorical data, such as concrete grade and steel bar type, one-hot encoding or embedding vector representation is used to convert the categorical data into numerical form for easy processing by the model. Then, based on the standardized characteristic data, a characteristic standardization channel is constructed. The goal of this channel is to learn how to uniformly convert different construction characteristic data into a standardized format through training. Through the training process, the standardized construction characteristic data and the corresponding characteristic vectors are input into the network. The system adjusts the weights through back propagation so that when faced with new construction characteristic data, it can automatically convert it into a standardized result. After the training is completed, new construction data is used for testing to evaluate the effect of the standardization channel and check whether it can accurately handle different types of construction characteristic data in actual construction and improve the accuracy of prediction.

[0034] Step S1346, using the F groups of template construction characteristics and the F template characteristic similarities to train a similarity measurement channel. Specifically, in order to train the similarity measurement channel, first prepare F groups of template construction characteristic data, each group of template construction characteristics includes different data points related to construction, such as terrain height, concrete grade, steel bar type, construction process, etc. The similarity between each group of template characteristics is calculated to form F template characteristic similarities, which represent the degree of matching between different template construction characteristics. Then, based on the above data, a similarity measurement channel is constructed, which is trained by a deep learning or machine learning algorithm to learn how to quantify the similarity between construction task characteristics. By inputting template construction characteristics and similarity data, the network weights are adjusted using a back propagation algorithm to optimize the accuracy of similarity prediction. During the training process, the system continuously adjusts the model parameters to improve the accuracy of similarity calculation, so that when new construction task data is input, the similarity measurement channel can evaluate the similarity between the new task and the existing template according to the training results. For example, if a new construction task has similar terrain and material selection as the template task, the system can predict the risks and challenges that the task may face during the construction process through the similarity measurement channel, thereby helping project managers make more reasonable decisions. The well-trained similarity metric channel can provide accurate similarity assessment for new construction tasks, thereby effectively supporting risk prediction, resource allocation and task planning during the construction process.

[0035] Step S1347, cascade the characteristic standardization channel and the similarity measurement channel to complete the construction of the construction characteristic evaluation network. Specifically, in the process of constructing the construction characteristic evaluation network, the input construction characteristic data is first standardized through the characteristic standardization channel, including standardization of continuous data such as settlement range and terrain height, and using unique hot encoding or embedded vector representation for categorical data such as concrete grade and steel bar type. The standardized data helps to eliminate the deviation between data of different dimensions, making subsequent processing more consistent and accurate. Then, the standardized data is passed to the similarity measurement channel, in which the system evaluates the matching degree between tasks by calculating the characteristic similarity between different tasks. The similarity measurement channel uses deep learning or machine learning algorithms to calculate the similarity between tasks for construction characteristics such as material selection, construction process, etc. For example, for an ongoing concrete pouring task, the system can evaluate the potential risks in its construction process based on the similarity of historical construction tasks and provide real-time feedback. Then, the standardization channel and the similarity measurement channel are cascaded so that the two can work together in a unified framework. The standardization channel first processes all input data and converts it into a consistent standardized format. The similarity measurement channel performs task matching and similarity measurement based on standardized data to achieve the accuracy of task evaluation. Ultimately, through training and optimization, the network can provide real-time support for construction task decision-making, help construction teams identify potential problems, optimize resource allocation, and predict risks during construction. For example, when excavating deep foundation pits, by evaluating construction characteristics, the system can identify possible geological changes in advance, thereby providing effective risk warnings for the construction team.

[0036] In a possible implementation, multiple template pixel construction characteristics are interactively obtained, and step S1341 further includes step S13411, wherein the template pixel construction characteristics include settlement control range, concrete grade, steel type, construction process, terrain height and construction materials. Specifically, during the construction process, multiple template pixel construction characteristics directly affect the building quality and construction progress, and the key characteristics involved include settlement control range, concrete grade, steel type, construction process, terrain height and construction materials. First, the settlement control range involves the settlement changes that may occur in the soil or infrastructure during the construction process, which is crucial to ensure the stability of the building. For example, in high-rise buildings, the settlement control range of the foundation needs special attention to prevent the building from tilting or cracking due to excessive settlement. Secondly, the concrete grade refers to the strength grade of the concrete, which directly determines the compressive performance of the building structure. For example, in key structures such as high-rise buildings or bridges, high-strength concrete is usually required, which not only affects the time arrangement of the construction process, but also puts forward requirements for the material ratio and maintenance methods. The type of steel bar is an important part of the building structure, which determines the type, specification and arrangement of the steel bar, and thus affects the stability of the load-bearing system. For example, complex structures or load-bearing parts may require high-strength steel bars, while ordinary walls or non-load-bearing parts use ordinary steel bars. Construction technology involves various technical methods from material selection to the actual construction process, which directly determines the quality and safety of construction. For example, different mixing and pouring processes used in the concrete pouring process will directly affect the strength and durability of the concrete. The terrain height reflects the geographical conditions of the construction site. The undulation of the terrain may affect the design and construction process of the foundation. For example, when building in mountainous or hilly areas, special support structures or drainage system designs may be required. Finally, construction materials cover all raw materials used in the construction process. Different material types and qualities have an important impact on construction time, cost and post-maintenance. For example, the use of fire-resistant and heat-insulating materials can improve the safety and energy efficiency of buildings. All characteristics comprehensively affect the decision-making and prediction models in the construction process. Dynamic risk prediction through methods such as Gaussian regression can ensure the smooth progress of construction tasks.

[0037] Step S200, during the execution of the first stage task, the multiple first stage Gaussian models are progressively activated in the first task model according to the construction progress, and the dynamic risk parallel prediction of the first stage task is performed to generate the first stage risk situation. Specifically, during the execution of the first stage task, according to the construction progress, the system will gradually activate multiple Gaussian models in the first task model. These Gaussian models represent different risk factors that may occur during the construction process, such as environmental changes, material quality, equipment failure, etc. As the construction progresses, the activation state of each model will change with the different stages of the task, ensuring that these risks can be tracked and evaluated in real time. By processing these activated Gaussian models in parallel, the system can predict multiple potential risks at the same time to ensure a comprehensive risk assessment. In this process, the system generates a risk situation map of the first stage based on the output of the model, and the risk situation map can show the risk level at different time points or regions during the construction process. For example, when the construction progress approaches a specific node, the system will update the risk map of that stage, which may show that the risk of certain areas is increased, such as abnormal changes in the settlement control range or substandard quality of construction materials in a specific area. This real-time update and dynamic adjustment of risk situations enables construction managers to take timely countermeasures based on real-time data to avoid possible delays or construction accidents. By gradually activating models and dynamically updating risk assessments, the system can effectively optimize the construction process to ensure that construction progress and quality are maximized.

[0038] In one possible implementation, Figure 2 As shown, during the execution of the first-stage task, the multiple first-stage Gaussian models are progressively activated in the first task model according to the construction progress, and the first-stage risk is predicted in parallel for the first-stage task, and the first-stage risk situation is generated. Step S200 further includes step S210, which presets the adjacent pixel activation radius. Specifically, presetting the adjacent pixel activation radius is a crucial step in construction scene modeling. The activation radius defines the influence range of each construction task pixel point, which in turn affects the risk assessment of the surrounding area. In different construction tasks, the size of the activation radius needs to be adjusted according to the complexity of the task. For example, infrastructure construction tasks may require a larger activation radius in order to analyze the potential risks of the surrounding environment and other construction tasks, while small residential construction projects can set a smaller radius to reduce the amount of calculation and focus on evaluating the main risk areas. During the construction process, as the project progresses and the scope of construction expands, the activation radius should be dynamically adjusted to ensure that a wider range of potential risk areas are gradually covered, thereby improving the accuracy of risk prediction during the construction process and helping the project team to more effectively control the dynamic risk situation at each stage of construction.

[0039] Step S220, according to the construction progress of the first stage task, in the first task model, the plurality of first stage Gaussian models are progressively activated with the adjacent pixel activation radius as a constraint. Specifically, during the construction process, as the task progresses, the construction progress gradually advances, and the progress directly affects the risk prediction model of the construction area. In order to accurately capture the dynamic changes in the construction process, an adjacent pixel activation radius is set, and multiple first stage Gaussian models are progressively activated with this as a constraint. When the progress of the construction task reaches a predetermined node, the plurality of Gaussian models in the first task model will be gradually activated according to the construction progress, which not only takes into account the expansion of the construction area, but also reflects the gradual advancement of the construction task in real time. For example, in the early stage of construction, the Gaussian model may only cover the core construction area, but as the task progresses, the model will gradually expand its scope of influence until it covers the entire construction site. By dynamically adjusting the activation radius and the model range, the regional risk changes in the construction process can be more accurately reflected. When the progress of the construction task reaches a certain stage, the activated Gaussian model will further load the construction characteristics based on the construction data collected in real time, ensuring that the task model can accurately assess the risk situation, thereby providing real-time and accurate support for construction decisions.

[0040] Step S230, loading the real-time collected construction data into the multiple first-stage Gaussian models that are progressively activated, and obtaining multiple superimposed construction characteristic risk values ​​of the multiple first-task pixel points. Specifically, during the construction process, the real-time collected construction data includes various environmental factors and construction dynamics, such as settlement data, equipment usage, ambient temperature, humidity changes, material properties, etc. The data is acquired in real time through sensors and monitoring systems, and loaded into multiple first-stage Gaussian models that have been gradually activated. Each Gaussian model represents the construction characteristics of a small area (pixel point) in the construction site, such as foundation settlement, concrete strength, etc. During the data loading process, the construction characteristics and risk values ​​of each pixel point are updated according to the changes in real-time construction data. Through the above method, not only the risk value of each construction area can be obtained, but also the problems that may occur during the construction process can be dynamically reflected, such as incomplete hardening of concrete due to temperature fluctuations, or material expansion due to excessive humidity. Finally, the construction characteristics and risk values ​​of multiple task pixels are superimposed to generate a comprehensive construction risk situation, which helps the construction team to monitor the risk changes of each area during the construction process in real time and adjust the construction strategy in time according to the prediction results. The above method ensures the safety and accuracy of the construction process and improves the efficiency of risk management.

[0041] Step S240, spatial difference is used to integrate the multiple superimposed construction characteristic risk values, and the risk situation of the first stage is output. Specifically, in the risk prediction of the construction process, spatial difference is used to integrate the construction characteristic risk values ​​of different regions. When the corresponding risk values ​​are generated for each task pixel point of the construction task, the scattered data often have discontinuities or blanks between regions. Especially in the actual construction environment, due to different equipment distribution, acquisition conditions or construction characteristics, the risk data of some areas may be missing or incomplete. Therefore, the spatial difference method is used to interpolate these risk data, which can smoothly transition the risk information of the surrounding adjacent areas, so that the originally isolated data points can be seamlessly connected to the overall model. The interpolation method not only fills the data gaps, but also obtains a more accurate and continuous construction risk situation map by calculating the risk value changes between adjacent areas. For example, in a construction area, some construction processes have higher risks (such as deep foundation pit operations). After the risk value is spatially differed, it can be merged with the risk values ​​of other surrounding processes to form a continuous risk distribution map. Ultimately, by further analyzing and processing the integrated risk values, the construction team can monitor the risk situation at the construction site in real time and accurately identify potential high-risk areas, such as settlement problems caused by soil unevenness during construction in a certain area, and take timely measures to avoid the expansion of risks.

[0042] Step S300, using the first task model and the first stage risk situation as the stage modeling benchmark, the second stage task is modeled through staged progressive modeling to generate a second incremental model. Specifically, in the process of construction scene modeling, the first stage task model and risk situation are used as the benchmark to provide detailed data on the construction progress and potential risks of the first stage. The data reflects the factors such as process, materials and geological conditions in the construction process of the first stage tasks such as foundation treatment and underground pipeline installation, and generates the corresponding risk situation. After completing the first stage task, the information is used as the benchmark for modeling the second stage task. The second stage task may include structural construction or masonry work of the superstructure. Therefore, when modeling, in addition to inheriting the task model of the first stage, it is also necessary to analyze new processes, materials and technical requirements. On this basis, through the staged progressive modeling method, the second incremental model is generated to ensure that the construction scene of the second stage can accurately reflect the characteristics of the new task and supplement the deficiencies in the first stage model. The modeling process is not only based on the risk situation of the first stage, but also can timely identify and respond to potential risks in the second stage to ensure the efficiency and safety of the construction process.

[0043] In a possible implementation, the first task model and the first stage risk situation are used as the stage modeling benchmark, and the second stage task is modeled through staged progressive modeling to generate a second incremental model. Step S300 further includes step S310, extracting the second stage task from the target construction task according to the task execution topological sequence. Specifically, in the construction project, the task execution topological sequence is used to clarify the order of different construction tasks to ensure that each task is carried out after its dependency is completed. In this process, first, the system needs to establish a topological sequence according to the dependency between the construction tasks, which means that the execution of each task must be carried out in a specific order. For example, the earth excavation task must be completed before the infrastructure laying task to ensure the smooth progress of subsequent work. Next, the system extracts the second stage task from the target construction task. The task can usually be started after the first stage is completed. For example, after the earth excavation is completed, it will involve infrastructure construction, underground facility installation, etc. The extraction of the second stage task is not only the selection of the order, but also takes into account the dependency relationship between the task and the first stage, ensuring that the second stage task can only be started after the results of the previous stage are completed. For example, the construction of the structural frame can only begin after the underground pipeline is laid. This process also requires the system to monitor the progress of task execution to ensure that each task is started at the appropriate time to avoid construction delays due to errors in the dependencies between tasks.

[0044] Step S320, taking the first task model as the stage modeling benchmark, and performing construction scene modeling according to the second stage task to generate the second task model. Specifically, in the construction scene modeling, taking the first task model as the stage modeling benchmark, the construction characteristics, risk assessment and prediction results generated by the first stage task can be used to provide support and guidance for the second stage task. Specifically, the first task model contains the construction environment information, resource allocation and risk situation of the first stage, and the above information provides a solid foundation for the implementation of the second stage task. For example, in a certain construction project, the first stage task involves foundation excavation and foundation treatment. The task model of this stage records the soil settlement characteristics, construction equipment layout and material usage in detail. When entering the second stage task, such as structural frame construction, the data of the first task model will be used as input to provide an important reference for the modeling of the second stage. At the same time, according to the specific construction characteristics of the second stage task, such as the concrete grade, steel bar layout method and construction sequence required for the frame structure, combined with the output results of the first task model, the construction scene of the second stage is further refined. For example, the first-stage model may predict that a certain area has a high risk of settlement, and the second-stage model will adjust the amount of concrete poured or optimize the steel bar layout for that area to reduce the risk. In this process, the first-task model not only provides a basic reference, but also combines it with the characteristics of the second-stage task through dynamic modeling to generate a targeted and forward-looking second-task model, thereby achieving a phased progression of construction scenario modeling. This modeling method not only ensures the close connection between the first-stage and second-stage tasks, but also lays the foundation for resource optimization and risk control in construction.

[0045] Step S330, performing evolution difference analysis on the first task model and the second task model to obtain the second incremental model. Specifically, the process of evolution difference analysis begins with clarifying and refining the boundaries of the first task model and the second task model. The first task model usually describes the completed construction area, such as the completed foundation pouring and the preliminary support structure, while the second task model covers all construction areas at the current stage, including the possible expansion of wall construction and frame erection parts. By comparing the spatial distribution and construction characteristics of the two, the newly added area in the second task model can be clearly determined. The newly added area not only includes the unfinished construction part, but also may involve new construction characteristics, such as the need to use high-strength concrete, special steel bar types, or introduce complex construction processes to adapt to the increased load requirements. Subsequently, for the newly added area, its construction characteristic data, such as material selection, terrain height and design parameters, are extracted, and updated in real time in combination with the dynamic construction progress. For example, when the newly added area enters the frame welding stage, the model will dynamically reflect the changes in construction characteristics such as welding strength and weld position. Finally, through the integrated analysis of the newly added areas, the second incremental model was generated. The above model accurately describes the construction characteristics and risk situation of the newly added areas, providing reliable data support and risk prediction guarantee for the next stage of construction, such as identifying high-risk areas or priority resource allocation points. The whole process ensures the combination of model dynamics and real-time performance. Especially in multi-stage construction tasks, this analysis method effectively improves the accuracy of modeling and the scientific nature of decision-making.

[0046] Step S340, optimizing the second grid scale according to the construction task characteristics of the second stage task. Specifically, optimizing the second grid scale according to the construction task characteristics of the second stage task first requires clarifying the scope of collecting feature data from the construction task. For example, in high-rise building construction, feature data may include foundation load capacity, concrete pouring requirements, steel bar diameter distribution, and terrain height changes. Next, the entire construction area is divided into initial standardized grid units, for example, each grid unit size is set to 10 meters × 10 meters, which is used to quickly evaluate the basic characteristic distribution of the entire construction site. Subsequently, the gradient changes of each grid unit are calculated based on the collected construction characteristic data, such as the settlement rate of different areas during concrete pouring or the change range of steel bar density. In order to ensure accuracy, grid units with gradient changes higher than a preset threshold are screened out, such as identifying areas with significant foundation settlement or obvious differences in steel bar layout density. These high-gradient grid units are further refined and reduced to a size of 5 meters × 5 meters or smaller to improve feature resolution. For example, in bridge construction, high-gradient areas may be concentrated at the junction of bridge piers and beams, and refining these areas can ensure modeling accuracy. After that, the characteristic gradient calculation and screening process is repeated until the refined grid unit meets the optimization conditions, such as the characteristic gradient change is lower than the set threshold or reaches the calculated resolution limit. Finally, an optimized grid scale, such as 1 meter × 1 meter, is generated to accurately capture the characteristic distribution in the construction area. This optimized grid scale provides a solid foundation for construction scenario modeling and subsequent dynamic risk prediction, and is particularly suitable for the refined management of complex construction areas, such as high-risk scenarios such as large bridge foundations or subway tunnel excavation.

[0047] Step S350, taking the second grid scale as a constraint, construct the multiple second-stage Gaussian models of the multiple second task pixels in the second incremental model. Specifically, the construction area is divided into multiple small grid units by optimizing the second grid scale, each unit corresponds to a pixel point, and the pixel point constitutes the basic unit of the construction scene. For example, in road construction, the construction area can be divided into a grid of 2 meters × 2 meters, and each grid unit records the material properties, terrain height and progress information of the area. Then, construction characteristic data such as concrete strength grade, foundation bearing capacity, etc. are extracted for each pixel point, and these data will be used as input for subsequent Gaussian process regression modeling. In the initialization stage, specific characteristic data, such as foundation settlement rate, real-time status of construction progress, and initial risk value, are assigned to each pixel point in the construction area to ensure that complete input data is provided for the Gaussian model. Subsequently, a Gaussian process regression model is constructed separately for each pixel point based on the construction characteristic data of the pixel point. For example, in the construction of building foundations, the Gaussian model may predict the risk value of foundation settlement in a certain area. In order to capture the spatial correlation between pixels, the radial basis function (RBF) kernel function is used for modeling, so that the characteristics of adjacent pixels affect each other, ensuring that the model can reflect the continuity and consistency in the actual construction scene. In addition, by adjusting the parameters of the kernel function (such as length scale) and the mean function, the model is verified and optimized so that it can accurately capture dynamic construction risks. For example, in the construction of high-rise building foundations, adjusting the model parameters can more accurately predict the settlement of each foundation column. Finally, all Gaussian models are combined to form a second incremental model, which comprehensively describes the dynamic characteristics and risk situation of the second-stage construction tasks. The incremental model can be used to show the risk changes in key support areas in bridge construction and help construction personnel adjust construction strategies in real time.

[0048] Step S400, during the execution of the second-stage task, the first task model and the first-stage risk situation are used as the prediction starting point, and multiple second-stage Gaussian models in the second incremental model are gradually activated according to the construction progress to perform dynamic risk prediction, and the second-stage risk situation is output. Specifically, during the execution of the second-stage task, the first task model and the first-stage risk situation are used as the prediction starting point, and multiple second-stage Gaussian models in the second incremental model are gradually activated in combination with the construction progress to achieve dynamic risk prediction. Specifically, real-time monitoring of the construction progress can be achieved through the task completion ratio or time node. For example, in bridge construction, when the concrete pouring of the pier reaches a certain ratio, the system will activate the Gaussian model corresponding to the pier and associate the state of the surrounding pixels. Then, the real-time collected construction data, such as material status, weather changes, equipment operating parameters, etc., will be loaded into the corresponding Gaussian model for prediction and analysis. For example, in the construction of high-rise buildings, the wind speed and the stability data of high-altitude construction equipment directly affect the risk value prediction, thereby helping construction personnel to identify potential hidden dangers in advance. Subsequently, by integrating the output results of multiple Gaussian models, a comprehensive risk situation map is generated, such as marking the risk levels of different areas of the site: unstable foundations may be high risk, and large support stress may be medium risk. Finally, the risk situation is output in the form of a dynamic cloud map, which clearly presents the risk changes and distribution of various areas of the construction site. For example, in the construction of subway tunnels, the cloud map can show the risk evolution of different soil layers during the advancement of the shield machine in real time, providing a decision-making basis for adjusting the construction plan and optimizing resource allocation.

[0049] Step S500, and so on, according to the construction progress, the multi-stage construction scene risk situation dynamic prediction of the target construction task is performed, and the construction situation cloud map is output. Specifically, with the generation and dynamic prediction of the construction situation cloud map as the core, the whole process is gradually realized through the dynamic risk assessment of the multi-stage task. First, according to the construction progress, the Gaussian model of the corresponding stage is dynamically activated, based on the real-time collected construction data, such as equipment operation data, material properties and weather information. In bridge construction, after the pile foundation is completed, the Gaussian model of the pier construction is activated to evaluate the risk area, such as the settlement risk of the pier foundation or the concrete cracking risk. Then, the risk situation of each stage is integrated into a global perspective through progressive analysis. For example, in tunnel construction, the stability of the soil layer advanced by the shield machine is used as the benchmark for the next stage of track laying. The construction situation cloud map integrates all risk prediction results through a spatial difference algorithm, and displays the risk distribution of the construction area in the form of an intuitive cloud map. For example, in the construction of a large dam, the cloud map intuitively represents the pressure distribution change of the dam structure by color partitioning, with high-risk areas marked in red and low-risk areas in green. In construction, the cloud map dynamically displays the wind load risk of each layer of the high-rise building structure, helping the construction party optimize resource allocation and construction strategy. Through the construction situation cloud map, the construction team can grasp the risk distribution and dynamic evolution of the whole stage in real time. For example, in water conservancy projects, the potential risk of flood impact on the dam under construction can be timely identified, and the protection measures can be quickly adjusted to ensure construction safety and efficiency.

[0050] The embodiment of the present application extracts the first-stage tasks from the target construction tasks, constructs the first task model, and models the construction scene through Gaussian process regression. During the execution of the first-stage tasks, multiple Gaussian models are gradually activated according to the construction progress to perform dynamic risk prediction and generate risk situations. Using the first-stage task model and risk situation as a benchmark, the second-stage incremental model is generated through progressive modeling, and the Gaussian model is continuously activated for risk prediction during the execution of the second stage. The multi-stage construction risk situation is gradually updated through the construction progress, and a dynamic construction situation cloud map is output, which achieves the technical effect of accurately predicting the risk changes in the construction process in a complex construction environment by combining the Gaussian process regression model and selecting suitable kernel functions and mean functions for parameter adjustment.

[0051] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A construction scene modeling method based on Gaussian process regression, characterized in that: The method comprises: Extracting the first-stage task from the target construction task, and performing Gaussian process regression modeling of the construction scene on the first-stage task to generate a first task model, wherein the first task model includes multiple first-stage Gaussian models; During the execution of the first-stage tasks, the multiple first-stage Gaussian models are progressively activated in the first task model according to the construction progress, dynamic risk parallel prediction is performed on the first-stage tasks, and the first-stage risk situation is generated; The first task model and the first stage risk situation are used as the stage modeling benchmark, and the construction scenario modeling of the second stage task is performed through staged progressive modeling to generate a second incremental model; During the execution of the second-stage task, the first task model and the first-stage risk situation are used as the prediction starting point, and multiple second-stage Gaussian models in the second incremental model are gradually activated according to the construction progress to perform dynamic risk prediction and output the second-stage risk situation; By analogy, the risk situation of multi-stage construction scenarios of the target construction task is dynamically predicted according to the construction progress, and a construction situation cloud map is output.

2. A construction scene modeling method based on Gaussian process regression as claimed in claim 1, characterized in that: Gaussian process regression modeling of the construction scene is performed on the first-stage task to generate a first task model. The method includes: Optimize the grid scale according to the construction characteristics of the first-stage task to obtain the first grid scale; Using the first grid scale to grid the first task model into a plurality of first task pixel points; Gaussian process regression modeling is performed on the multiple first task pixel points to generate multiple first-stage Gaussian models.

3. A construction scene modeling method based on Gaussian process regression as claimed in claim 1, characterized in that: The first task model and the first stage risk situation are used as stage modeling benchmarks, and construction scenario modeling is performed on the second stage task through staged progressive modeling to generate a second incremental model. The method includes: extracting the second-stage tasks from the target construction tasks according to the task execution topology sequence; Using the first task model as a stage modeling benchmark, performing construction scene modeling according to the second stage task, and generating a second task model; Performing evolution difference analysis on the first task model and the second task model to obtain the second incremental model; The second grid scale is obtained by optimizing the construction task characteristics of the second stage task; Taking the second grid scale as a constraint, construct the multiple second-stage Gaussian models of the multiple second task pixel points in the second incremental model.

4. A construction scene modeling method using Gaussian process regression as claimed in claim 2, characterized in that: According to the construction characteristics of the first-stage task, a grid scale optimization is performed to obtain a first grid scale, and the method includes: Gridding the first task model into a plurality of standard grid cells based on a standardized grid scale; Collecting first construction characteristic data from the first stage task, and calculating characteristic gradient changes of the plurality of standard grid units according to the first construction characteristic data to obtain a plurality of characteristic gradient units; Presetting a risk gradient threshold, and selecting N risk gradient units from the plurality of characteristic gradient units according to the risk gradient threshold; Using 1 / 2 of the standardized grid scale to grid the N risk gradient units into N groups of first updated grid units; Calculate characteristic gradient changes of the N groups of first updated grid units according to the first construction characteristic data to obtain N groups of updated gradient units; Selecting M second update grid units from the N groups of update gradient units according to the risk gradient threshold; In this way, the grid unit screening and updating are performed until the obtained updated grid unit meets the risk gradient threshold, and the first grid scale is output.

5. A construction scene modeling method using Gaussian process regression as claimed in claim 4, characterized in that: Performing Gaussian process regression modeling on the plurality of first task pixel points to generate a plurality of first-stage Gaussian models, the method comprising: After retrieving the first task execution time from the first stage task, collecting environmental data according to the first task execution time to obtain first environmental prediction data; Matching similar construction tasks according to the first construction characteristic data and the first environment prediction data to obtain a first similar task model, wherein the first similar task model includes H sample Gaussian models of H sample pixel points, and the H sample pixel points have H sample pixel construction characteristic identifiers; Extracting a plurality of task construction characteristics of the plurality of first task pixels from the first stage tasks; Pre-construction construction characteristics evaluation network; After combining and enumerating the H sample pixel construction characteristics and the multiple task construction characteristics, the construction characteristic evaluation network is used to measure the characteristic similarity of the combination results to obtain multiple sample mapping relationships between the H sample pixel construction characteristics and the multiple task construction characteristics; The H sample Gaussian models are migrated and adjusted according to the multiple sample mapping relationships to generate the multiple first-stage Gaussian models adapted to the multiple first task pixel points.

6. A construction scene modeling method using Gaussian process regression as claimed in claim 5, characterized in that: A construction characteristic evaluation network is pre-built, the method comprising: Interactively obtain multiple template pixel construction characteristics; Performing characteristic standardization annotation on the construction characteristics of the plurality of template pixels to obtain a plurality of template characteristic vectors; Combining and enumerating the plurality of template pixel construction characteristics to obtain F groups of template construction characteristics; Performing similarity measurement on the F groups of template construction characteristics to obtain F template characteristic similarities; Using the plurality of template pixel construction characteristics and the plurality of template characteristic vectors to train a characteristic standardization channel; The similarity measurement channel is trained using the F sets of template construction characteristics and the similarity of the F template characteristics; The characteristic standardization channel and the similarity measurement channel are cascaded to complete the construction of the construction characteristic evaluation network.

7. A construction scene modeling method using Gaussian process regression as claimed in claim 6, characterized in that: The template pixel construction characteristics include settlement control range, concrete grade, steel bar type, construction technology, terrain height and construction materials.

8. A construction scene modeling method using Gaussian process regression as claimed in claim 2, characterized in that: During the execution of the first-stage task, the plurality of first-stage Gaussian models are progressively activated in the first task model according to the construction progress, dynamic risk parallel prediction is performed on the first-stage task, and the first-stage risk situation is generated. The method includes: Preset adjacent pixel activation radius; According to the construction progress of the first-stage task, in the first task model, progressively activate the plurality of first-stage Gaussian models with the adjacent pixel activation radius as a constraint; Loading the real-time collected construction data into the multiple first-stage Gaussian models that are progressively activated to obtain multiple superimposed construction characteristic risk values ​​of the multiple first task pixel points; The plurality of superimposed construction characteristic risk values ​​are integrated by using spatial difference to output the risk situation of the first stage.

Images