Deformation detection method and system based on constructional engineering

Abstract

The invention relates to a deformation detection method and system based on constructional engineering, and the method comprises the steps: obtaining a multi-source heterogeneous data set which comprises structure image data and sensor signal data; then spatio-temporal feature extraction processing is carried out on the data set, and an obtained multi-dimensional deformation feature set integrates change features of surface and key nodes; then deformation probability distribution is generated by using a pre-trained deformation detection model, and the possibility of deformation of different monitoring positions can be accurately judged, so that a suspicious deformation area and space-time evolution characteristics thereof are accurately positioned; generating a detection report containing monitoring position coordinates and deformation risk levels based on the space-time evolution characteristics; according to the invention, the accuracy, comprehensiveness and timeliness of building structure deformation detection can be effectively improved, and the safety of building engineering is guaranteed.

Description

Technical Field

[0001] The present application relates to the fields of construction engineering and artificial intelligence technology, and in particular to a deformation detection method and system based on construction engineering. Background Art

[0002] Building safety has always been a key concern in the construction industry. With the advancement of construction technology and the increasing scale of buildings, ensuring the structural stability of buildings throughout their lifecycle has become a critical task. During their use, building structures are subject to a variety of external factors, such as natural disasters (earthquakes, windstorms, etc.), long-term environmental degradation (such as humidity and temperature fluctuations), and structural aging.

[0003] To monitor the status of building structures, relevant technologies are constantly developing. Currently, a variety of independent monitoring technologies are being applied in construction projects. For example, sensor technology can be used to monitor physical quantities such as stress and strain in building structures. By installing various sensors, such as strain gauges and accelerometers, at key locations within the building, the stress conditions within the building structure can be determined. Image acquisition technology plays a key role in inspecting the exterior of buildings. High-definition cameras and other equipment are used to capture images of building surfaces to check for obvious signs of damage, such as surface cracks and peeling.

[0004] While these technologies each play a role, they also face numerous challenges. While sensor technology can detect internal stress and strain, its ability to monitor subtle changes in a building's surface or overall deformation is limited. Image acquisition technology, on the other hand, struggles to gain a deep understanding of the mechanical changes within a building's structure. These factors contribute to low accuracy in deformation detection for construction projects. Summary of the Invention

[0005] The main purpose of this application is to provide a deformation detection method and system based on construction projects to improve the accuracy of deformation detection of construction projects.

[0006] To achieve the above objectives, an embodiment of the present invention provides a deformation detection method based on a construction project, the method comprising: Acquire a multi-source heterogeneous data set within a building monitoring area, wherein the multi-source heterogeneous data set includes structural image data and sensor signal data at different monitoring time periods; Performing spatiotemporal feature extraction processing on the multi-source heterogeneous data set to obtain a multidimensional deformation feature set including structural surface texture change features and key node displacement trend features, wherein the structural surface texture change features reflect the grayscale value fluctuation pattern of the surface texture of the building body during the monitoring period, and the key node displacement trend features reflect the continuous change pattern of the spatial coordinates of the key nodes during the monitoring period; Input the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generate a deformation probability distribution of the building monitoring area through the deformation detection model; Determine the suspicious deformation area in the building monitoring area and the spatio-temporal evolution characteristics of the suspicious deformation area according to the deformation probability distribution. The spatio-temporal evolution characteristics include the time interval when the deformation occurs and the spatial range expansion mode; Generate a detection report including the monitoring position coordinates and the deformation risk level based on the suspicious deformation area and the spatio-temporal evolution characteristics, and send the detection report to the building safety management system to trigger an early warning assessment operation.

[0007] Correspondingly, an embodiment of the present application further provides a deformation detection system based on a building project, including: An acquisition module for acquiring a multi-source heterogeneous data set in a building monitoring area. The multi-source heterogeneous data set includes structural image data and sensor signal data in different monitoring periods; A feature extraction module for performing spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features. The structural surface texture change features reflect the gray value fluctuation law of the building main body surface texture during the monitoring period, and the key node displacement trend features reflect the continuous change mode of the key node spatial coordinates during the monitoring period; A deformation detection module for inputting the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generating a deformation probability distribution of the building monitoring area through the deformation detection model; A feature determination module for determining the suspicious deformation area in the building monitoring area and the spatio-temporal evolution characteristics of the suspicious deformation area according to the deformation probability distribution. The spatio-temporal evolution characteristics include the time interval when the deformation occurs and the spatial range expansion mode; A report generation module for generating a detection report including the monitoring position coordinates and the deformation risk level based on the suspicious deformation area and the spatio-temporal evolution characteristics, and sending the detection report to the building safety management system to trigger an early warning assessment operation.

[0008] In summary, by adopting the technical solution of the present application, by obtaining a multi-source heterogeneous data set, including structural image data and sensor signal data, information can be comprehensively collected from the exterior and key internal positions of the building; then, spatio-temporal feature extraction processing is performed on this data set, and the resulting multi-dimensional deformation feature set synthesizes the change features of the surface and key nodes, enabling the deformation situation of the building to be quantitatively described from multiple dimensions; then, a pre-trained deformation detection model is used to generate a deformation probability distribution, which can accurately judge the possibility of deformation at different monitoring positions, thereby accurately locating the suspicious deformation area and its spatio-temporal evolution characteristics; based on these, a detection report including the coordinates of the monitoring position and the deformation risk level is generated, which can enable the building safety management system to perform warning and evaluation operations in a timely and targeted manner, effectively improving the accuracy, comprehensiveness and timeliness of building structure deformation detection, and ensuring the safety of building projects. Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0010] Figure 1 It is a schematic diagram of the scenario of the deformation detection method based on building engineering in the embodiment of the present application; Figure 2 It is a flowchart of the deformation detection method based on building engineering provided by the embodiment of the present application; Figure 3 It is a schematic diagram of the process of generating a multi-source heterogeneous data set provided by the embodiment of the present application; Figure 4 It is a schematic diagram of the process of constructing a multi-dimensional deformation feature set provided by the embodiment of the present application; Figure 5 It is a schematic diagram of the process of generating deformation probability provided by the embodiment of the present application; Figure 6 It is a schematic diagram of the process of generating spatio-temporal evolution characteristics provided by the embodiment of the present application; Figure 7 It is a schematic diagram of the process of generating a detection report provided by the embodiment of the present application; Figure 8 It is a schematic diagram of the structure of the deformation detection system based on building engineering provided by the embodiment of the present application; Figure 9 It is a schematic diagram of the structure of the computer device provided by the embodiment of the present application. Detailed Embodiments

[0011] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0012] The embodiments of the present application provide a deformation detection method and system based on construction engineering, which will be described in detail below respectively.

[0013] In the embodiments of the present application, "deformation detection based on construction engineering" refers to the detection operation for the shape change condition of the construction engineering structure. During the construction and use of construction engineering, it will be affected by many factors such as external force actions (including earthquakes, winds, heavy object pressure, etc.), environmental factors (such as material expansion and contraction caused by temperature and humidity changes), the change of the material's own characteristics over time (such as aging, fatigue), and foundation settlement, etc., thus may produce deformations. To achieve deformation detection, a variety of technical means are required. On the one hand, by arranging displacement sensors, stress sensors, etc. at key positions of the building, data signals such as displacement amounts and stresses inside the structure are obtained, and these signals can reflect the microscopic deformation conditions inside the structure; on the other hand, structure image data such as surface texture images of the building main body and geometric contour images of key nodes are collected by means of image acquisition devices, and whether there are deformations such as crack development and node displacement are judged from the building appearance. Then, by synthesizing these data, the deformation degree, development trend, etc. of each part of the building are analyzed, and further provide a basis for the safety assessment and maintenance decision-making of the building, ensuring the structural safety and stability of the building throughout its life cycle.

[0014] For example, in the scenario of a large commercial building project, the building includes different functional areas such as multiple office areas, shopping areas, and underground parking lots. As Figure 1 shown, a deformation detection scenario based on construction engineering is provided, and this deformation detection scenario includes: an image acquisition device, a sensor device, a deformation analysis system, and a building safety management system; each device and system can be connected through a wired or wireless network.

[0015] Among them, various devices are distributed in each area of the building for deformation detection based on construction engineering. First of all, there are image acquisition devices. High-definition cameras are installed at positions such as the exterior facade of the building, around the main internal support columns, and at the key structural joints on each floor. These cameras can regularly collect the structural image data of the building, including the surface texture images of the building main body, such as the splicing state of exterior wall tiles and the flatness of the concrete surface, as well as the geometric contour images of key nodes, such as the shapes and positional relationships of beam-column joints and frame connection parts. During the normal use of the building, whether it is the flow of people during the day, the temperature changes brought about by sunlight irradiation, or the temperature drop at night, etc., the cameras keep working, accurately record the image information of the building appearance and key parts, and are marked with accurate time stamps, forming a complete sequence of structural images.

[0016] In this scenario, displacement sensors and stress sensors are also arranged at the key structural positions of the building, such as the bottom of the main load-bearing columns, the key stress points of the large beams, and the support structures of the underground parking lot. The displacement sensors can detect the minute displacement changes of key nodes in three-dimensional space in real time, accurate to the millimeter level. For example, during an earthquake, they can timely capture the shaking amplitude and direction of the building structure under the influence of seismic waves; the stress sensors can sense the stress changes inside the building materials, whether it is the static stress generated by the building's own weight or the dynamic stress changes caused by the movement of people and the driving of vehicles, etc., and can accurately measure them. The real-time signals output by these sensors constitute the sensor signal data, which is also marked with time, corresponding to the time stamps of the image data.

[0017] The data collected by the above-mentioned image acquisition devices and sensors will be transmitted to the deformation analysis system. The deformation analysis system first performs spatio-temporal feature extraction and processing on the multi-source heterogeneous data set within the building monitoring area obtained. For the surface texture images, the algorithm analyzes the gray value fluctuation rules of the textures in the images at different time periods to obtain the change characteristics of the structural surface texture; for the geometric contour images, the spatial coordinate changes of the key nodes at different times are calculated to obtain the displacement trend characteristics of the key nodes. The two together constitute a multi-dimensional deformation feature set.

[0018] Then, the multi-dimensional deformation feature set is input into the pre-trained deformation detection model. This model is trained based on a large amount of building structure data and historical deformation cases, and can accurately analyze the deformation probability distribution of the building monitoring area. For example, after the building has been used for many years, the foundation may have uneven settlement, and the model can judge which areas are more likely to settle and which areas are relatively stable according to the input feature data.

[0019] Based on the deformation probability distribution, the deformation analysis system determines the suspicious deformation areas within the building and their spatio-temporal evolution characteristics. For example, if a high deformation probability is found in a certain area on a certain floor, the system further analyzes the time interval during which the deformation occurred in this area, whether it suddenly appeared recently or developed slowly over a long period, as well as the spatial range expansion mode, whether it is a local small-scale deformation or has a tendency to spread to the surrounding areas.

[0020] Finally, based on the suspicious deformation areas and spatio-temporal evolution characteristics, a detection report containing the monitoring position coordinates and deformation risk levels is generated. If the deformation risk level at the bottom of a certain support column is high, its coordinate position and risk degree will be clearly stated in the report. This report is sent to the building safety management system, triggering the early warning assessment operation. According to the content of the report, the safety management system can arrange for professional personnel to conduct a detailed inspection of the suspicious area, evaluate the necessity and urgency of repair, etc. The entire deformation detection system ensures the structural safety of commercial buildings in complex usage environments and during long-term operation processes.

[0021] Reference Figure 2 , Figure 2 FIG. S10. Obtain a multi-source heterogeneous data set within the building monitoring area. The multi-source heterogeneous data set includes structural image data and sensor signal data for different monitoring time periods. The structural image data includes the surface texture image of the building main body and the geometric contour image of key nodes. The sensor signal data includes the real-time output signals of displacement sensors and stress sensors arranged at key positions of the building.

[0022] The building monitoring area refers to a specific area within the building project that needs to be monitored for deformation, which is determined in advance according to factors such as building structure characteristics, functional zoning, and potential risk areas. For example, in high-rise buildings, it may include areas where load-bearing walls are located, beam-column connection areas, building facades, and structural parts of underground parking lots, etc., which are prone to deformation or have an important impact on the overall structural stability.

[0023] The multi-source heterogeneous data set is a data set composed of data from multiple different sources with different structures and properties. Among them, the structural image data is data related to the images of the building structure, while the sensor signal data is an electrical signal or other quantifiable signal collected by sensors that reflects the physical characteristics of the building structure.

[0024] Structural image data is data used to describe visual information such as the appearance of a building structure and the shape of key parts. Among them, the surface texture image of the main building body refers to an image that can reflect surface characteristics such as the material, color, and roughness of the building surface. For example, the granularity of a concrete wall surface and the splicing pattern of tiles are presented in the image. The geometric contour image of key nodes refers to the shape, edge contour, etc. of key connection points (such as beam-column joints, frame corners, etc.) in the building structure as shown in the image.

[0025] Sensor signal data is data collected by sensors arranged at key positions in the building. A displacement sensor is a sensor that can measure the position change of an object and can be used in a building to detect the displacement of key parts (such as the bottom of a load-bearing column, both ends of a large beam, etc.) relative to a reference point. A stress sensor is a sensor used to measure the internal stress of building materials. When a building structure is subjected to external forces (such as self-weight, wind force, crowd load, etc.), the stress sensor can sense the change in the internal stress of the building materials.

[0026] This step can collect various different types of data from the building monitoring area. Data is obtained by setting appropriate devices in the monitoring area. Data collection is carried out at different monitoring time periods to ensure that information on the building structure in different time states can be captured. Structural image data is obtained through image acquisition devices (such as high-definition cameras, etc.), which are installed at positions that can cover the main building body surface and key nodes. Sensor signal data is obtained through displacement sensors and stress sensors pre-installed at key positions in the building. The collected data is sorted in chronological order to form a multi-source heterogeneous data set containing structural image data and sensor signal data for different monitoring time periods.

[0027] As a possible implementation method, a distributed image acquisition system can be adopted. Specifically, multiple high-definition cameras are evenly distributed in the building monitoring area. These cameras have a timed shooting function and take pictures of the main building body surface and key nodes at fixed time intervals (such as once an hour) to obtain surface texture images and geometric contour images. For the arrangement of sensors, the key positions are determined based on the mechanical model analysis of the building structure. High-precision displacement sensors and stress sensors are arranged at the key stress points of the load-bearing structure. The displacement sensor uses the principle of laser ranging to accurately measure the displacement by emitting laser light and receiving the reflected light; the stress sensor uses a strain gauge sensor to convert the strain of the building material into an electrical signal. All the collected data is transmitted to the data storage center through wired or wireless networks for centralized storage and management, forming a multi-source heterogeneous data set. The technical effect of this method is that the distributed acquisition system can comprehensively cover the monitoring area, and sensors with different principles ensure the accuracy of the data, providing a rich data basis for subsequent deformation analysis.

[0028] S20. Perform spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features. The structural surface texture change features reflect the gray value fluctuation law of the building main body surface texture during the monitoring period, and the key node displacement trend features reflect the continuous change pattern of the key node spatial coordinates during the monitoring period.

[0029] Among them, spatio-temporal feature extraction processing is a processing operation for multi-source heterogeneous data, aiming to extract feature information related to time and space from the data. In the deformation detection of building engineering, the time feature is reflected in the change of data in different monitoring periods, and the space feature is associated with the data at different positions in the building structure.

[0030] The structural surface texture change features are features obtained by analyzing the surface texture images of the building main body in different monitoring periods, specifically manifested as the gray value fluctuation law of the surface texture during the monitoring period. The gray value is the brightness value of each pixel point in the image, and the gray value fluctuation law reflects the possible changes of the building surface texture over time. For example, the aging of surface materials and surface damage (such as crack generation, coating peeling, etc.) will cause changes in the gray value.

[0031] The key node displacement trend features are features extracted from the geometric contour images of the key nodes and the displacement sensor data, used to describe the continuous change pattern of the key node spatial coordinates during the monitoring period. The spatial coordinates refer to the positions of the key nodes in the three-dimensional space of the building, and the continuous change pattern can be linear change, periodic change or irregular change, etc. These change patterns reflect the displacement trend of the key nodes and are of great significance for judging the stability of the building structure.

[0032] The multi-dimensional deformation feature set is a set containing multiple deformation features. These features describe the deformation of the building structure from different perspectives (such as surface texture and key node displacement, etc.), and each feature is associated with time and space.

[0033] In this step, first, for the surface texture images in the structural image data, the gray value changes of the images in different monitoring periods are analyzed through specific image processing algorithms, so as to extract the structural surface texture change features. For the geometric contour images, combined with the displacement sensor data, the position changes of the key nodes at different times are obtained through the spatial coordinate calculation method, and then the key node displacement trend features are determined. These deformation-related features extracted from different data sources are combined according to certain rules to form a multi-dimensional deformation feature set.

[0034] As an example, for the extraction of the structural surface texture change features, the gray-level co-occurrence matrix algorithm can be adopted. First, convert the surface texture images of different monitoring periods into grayscale images, and then calculate the gray-level co-occurrence matrix. By analyzing the variation of the element values (such as energy, contrast, correlation, etc.) of the gray-level co-occurrence matrix over time, the gray value fluctuation law of the surface texture during the monitoring period can be obtained. For the extraction of the displacement trend features of key nodes, in the geometric contour image, use image recognition technology to identify the contour feature points of the key nodes, and combine the displacement data collected by the displacement sensors to plot the coordinate curve of the coordinates of the key nodes in the three-dimensional space over time. By performing fitting analysis on the coordinate curve (such as using the least squares method for fitting), the displacement trend features of the key nodes can be obtained. Combine these two features into a multi-dimensional deformation feature set. The technical effect of this technical implementation method is that the gray-level co-occurrence matrix algorithm can effectively extract texture features from the image, and the coordinate curve fitting analysis can accurately describe the displacement trend of the key nodes. The combination of the two provides comprehensive feature information for subsequent deformation analysis.

[0035] S30. Input the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generate a deformation probability distribution of the building monitoring area through the deformation detection model. The deformation probability distribution is used to represent the possibility of structural deformation occurring at different monitoring positions.

[0036] The pre-trained deformation detection model is a model that has been pre-trained through a large amount of building structure data (including data with known deformation conditions and normal structure data). This model has the ability to analyze and process the input multi-dimensional deformation feature set, thereby predicting the possibility of building structure deformation.

[0037] The deformation probability distribution is a probability distribution that describes the likelihood of structural deformation occurring at different monitoring positions within the building monitoring area. For example, the deformation probability of a certain column base position is 0.1, while the deformation probability of the mid-span position of a certain beam is 0.05, indicating that the possibility of deformation at the column base position is relatively higher than that at the mid-span position of the beam.

[0038] In this step, the multi-dimensional deformation feature set is provided as input data to the pre-trained deformation detection model. Based on the relationship between the building structure deformation features learned previously and the actual deformation conditions, the model analyzes and calculates the input features. The model matches and compares the input features with the feature patterns in the training data, thereby evaluating the possibility of structural deformation occurring at each monitoring position within the building monitoring area, and finally generating a deformation probability distribution representing the likelihood of structural deformation occurring at different monitoring positions.

[0039] As a possible implementation, the pre-trained deformation detection model can adopt a Support Vector Machine (SVM) model. First, in the model training stage, a large number of building structure sample data are collected, including the multi-dimensional deformation feature set data in the normal state of the structure and the deformation states of different degrees, and the actual deformation conditions corresponding to each sample are marked (such as no deformation, slight deformation, severe deformation, etc.). These data are used as the training set and input into the SVM model. By adjusting the parameters of the model (such as the type of kernel function, penalty parameter, etc.), the model can accurately distinguish samples in different deformation states. In practical applications, the multi-dimensional deformation feature set is input into the trained SVM model. The SVM model calculates the deformation probability corresponding to each monitoring position according to the position relationship of the feature data in the feature space, thereby generating a deformation probability distribution. The technical effect of this technical implementation method is that the SVM model has good generalization ability and can accurately predict the deformation probability under limited sample data.

[0040] In this embodiment, a deformation detection model based on a Deep Neural Network (DNN) can also be constructed. During the construction of the model, a DNN structure including multiple hidden layers is designed, such as a multi-layer fully connected layer structure. In the training stage, a large amount of building structure data is used for unsupervised pre-training. For example, an autoencoder is used to pre-train the data to learn the internal structural features of the data, and then supervised learning (such as fine-tuning using the marked deformation data) is used to optimize the model. The multi-dimensional deformation feature set is input into the trained DNN model. The model calculates the deformation probability of each monitoring position through forward propagation and generates a deformation probability distribution. The technical effect of this method is that the DNN model can automatically learn the complex feature representation of the data and has better prediction ability for complex building structure deformation conditions.

[0041] S40. Determine the suspicious deformation area within the building monitoring area and the spatio-temporal evolution characteristics of the suspicious deformation area according to the deformation probability distribution. The spatio-temporal evolution characteristics include the time interval when the deformation occurs and the spatial range expansion mode.

[0042] The suspicious deformation area is the area whose deformation probability exceeds a certain threshold (this threshold is set according to the building structure safety standard) selected from the building monitoring area according to the deformation probability distribution. These areas are suspected of having structural deformation risks and need further analysis.

[0043] The spatio-temporal evolution characteristics are used to describe the change characteristics of the suspicious deformation area in time and space. Among them, the time interval when the deformation occurs refers to the time period from the start of the suspicious deformation sign to the current moment; The spatial range expansion mode refers to the change in the range of the suspicious deformation area within the building space, such as the deformation gradually expanding from local points to a larger area, or a linear expansion in a certain direction, etc.

[0044] This step can analyze the deformation probability distribution generated by the deformation detection model, and determine the monitoring positions where the deformation probability values exceed the preset threshold. Cluster the continuous areas around these positions into suspicious deformation areas. Then, for each suspicious deformation area, by querying the data records of the corresponding monitoring positions, analyze the change of its deformation probability over time to determine the time interval when the deformation occurs. At the same time, by comparing the ranges of the suspicious deformation areas in different monitoring periods (which can be determined according to the monitoring position coordinates), the spatial range expansion mode is obtained, so as to obtain the spatio-temporal evolution characteristics of the suspicious deformation area.

[0045] As an example, a clustering algorithm can be used to determine the suspicious deformation area. For example, using the DBSCAN (Density- Based Spatial Clustering of Applications with Noise) algorithm, regarding the monitoring positions in the deformation probability distribution as points in space, and according to the preset distance threshold and density threshold, clustering the high-probability points with density connection into suspicious deformation areas. For the determination of the spatio-temporal evolution characteristics, for each suspicious deformation area, extract the deformation probability data of the monitoring positions within this area in different monitoring periods from the data storage. By analyzing the time series of these data, determine the start time and end time of the deformation, so as to obtain the time interval when the deformation occurs. At the same time, through spatial analysis of the monitoring position coordinates in different periods, such as calculating the minimum bounding rectangle (MBR) of the suspicious deformation area in each period, and comparing the size and position changes of the MBRs in different periods, the spatial range expansion mode is obtained. The technical effect of this technical implementation method is that the DBSCAN algorithm can effectively process suspicious deformation areas with irregular shapes, and the analysis method of spatio-temporal evolution characteristics can accurately describe the development process of the suspicious deformation area.

[0046] S50. Generate a detection report containing the monitoring position coordinates and the deformation risk level based on the suspicious deformation area and the spatio-temporal evolution characteristics, and send the detection report to the building safety management system to trigger the warning assessment operation.

[0047] The detection report is a report containing detailed information about the suspicious deformation area, including the coordinates of the monitoring positions within the suspicious deformation area and the deformation risk level evaluated according to the spatio-temporal evolution characteristics, etc.

[0048] The deformation risk level is a level identifier obtained by grading and evaluating the risk degree of structural deformation in a suspicious deformation area based on the spatio-temporal evolution characteristics of the suspicious deformation area (such as the speed of deformation, the trend of range expansion, etc.) and the safety standards of the building structure. For example, it can be divided into low risk, medium risk, high risk levels, etc.

[0049] The building safety management system is a system for managing building safety-related matters. It receives the detection reports from the deformation detection system and performs further early warning evaluation, safety decision-making and other operations according to the report content.

[0050] In this step, based on the determined suspicious deformation area and its spatio-temporal evolution characteristics, the coordinate information of each monitoring position within the suspicious deformation area can be obtained first. Then, according to the spatio-temporal evolution characteristics (such as the length of the time interval when deformation occurs, the speed of spatial range expansion, etc.) and the safety standards of the building structure, the deformation risk level of the suspicious deformation area is evaluated. The information such as the monitoring position coordinates and the deformation risk level is combined in a certain format to generate a detection report. Finally, the detection report is sent to the building safety management system through network communication and other means. After receiving the report, the building safety management system will start the early warning evaluation operation according to the report content, such as arranging professional personnel to conduct a detailed inspection of the high-risk area, formulating a maintenance plan, etc.

[0051] As a possible implementation method, for the evaluation of the deformation risk level, a rule-based evaluation system can be established. Different evaluation rules are formulated according to factors such as the type of building structure (such as frame structure, brick-concrete structure, etc.), service life, and design bearing capacity. For example, for a newly built frame structure building, if the spatial range expansion speed of the suspicious deformation area is slow (such as less than 1 square meter per month) within a short period of time (such as within one month), it is evaluated as a low risk level; if the spatial range expansion speed is fast (such as more than 5 square meters per month), it is evaluated as a high risk level. When generating the detection report, the coordinates of the monitoring positions within the suspicious deformation area are recorded in the form of a geographic coordinate system (such as longitude and latitude), and the deformation risk level is represented in the form of numbers (1 represents low risk, 2 represents medium risk, 3 represents high risk) or words (low risk, medium risk, high risk), and a detection report is generated according to a fixed format (such as XML format). The detection report is sent to the building safety management system through a network communication protocol (such as TCP / IP protocol). The technical effect of this technical implementation method is that the rule-based evaluation system is simple and easy to implement, can accurately evaluate the risk level according to the actual situation of the building, and the standard report format and network communication protocol ensure the accurate transmission and effective reception of information.

[0052] In this embodiment, a machine learning method can also be used to evaluate the deformation risk level. A large amount of historical building deformation data is collected, including the spatio-temporal evolution feature data of the suspicious deformation area and the corresponding actual deformation results (such as whether serious structural deformation has occurred), and a classification model (such as a decision tree model) is constructed. The spatio-temporal evolution feature data of the current suspicious deformation area is used as the input, and the deformation risk level is predicted through the trained decision tree model. When generating the detection report, in addition to the monitoring position coordinates and risk level, some auxiliary information can also be added, such as the area of the suspicious deformation area, the relationship with the adjacent building structure, etc. The detection report is sent to the building safety management system through the API interface provided by the building safety management system. The technical effect of this method is that the machine learning method can automatically learn the complex relationships in the data, improving the accuracy of the risk level assessment, and the API interface facilitates the sending of the detection report and the docking with the building safety management system.

[0053] In one embodiment, referring to Figure 3 , step S10 may specifically include: S101. Through the image acquisition device arranged in the building monitoring area, collect the surface texture image of the building main body and the geometric contour image of the key nodes at a fixed monitoring period, and generate an image data sequence with monitoring timestamp marks. The surface texture image includes the color distribution information of the building surface material and the preliminary crack identification mark, and the geometric contour image includes the contour edge coordinates of the key nodes and the relative position relationship between adjacent nodes.

[0054] The image acquisition device is a device specifically used to obtain building structure images, such as high-definition cameras, 3D laser scanners, etc. These devices have certain parameters such as resolution and field of view angle to meet the requirements of collecting images of the building main body surface and key nodes.

[0055] The fixed monitoring period is a preset time interval used to determine the image acquisition frequency of the image acquisition device. For example, for some large commercial buildings, it may be set to collect images every 12 hours. The setting of this period depends on factors such as the type of building, usage environment, and importance.

[0056] The preliminary crack identification mark in the surface texture image is a kind of information that marks the possible cracks in the surface texture image. It can be the result obtained based on specific image processing algorithms. For example, the gray value mutation area identified through the edge detection algorithm is initially judged as the position of the crack, and is marked in the image with specific marks (such as color marks, line marks, etc.).

[0057] The relative position relationship between adjacent nodes in the geometric contour image refers to the position association information between key nodes and adjacent nodes in the building structure, including relationships such as distance and angle. For example, at the beam-column connection node, information such as the vertical distance and horizontal angle between the beam node and the column node.

[0058] In this step, the position and quantity of the image acquisition device can be reasonably arranged according to factors such as the scope of the building monitoring area and the complexity of the building structure. Then, start the image acquisition operation according to a fixed monitoring period. For the acquisition of the surface texture image of the building main body, the device needs to be able to accurately capture the color distribution information of the building surface material, which helps the overall evaluation of the building appearance. At the same time, preliminary identification and marking of cracks are also required, which is the preliminary investigation of potential safety hazards in the building structure. For the acquisition of the geometric contour image of the key node, the focus is on obtaining the contour edge coordinates of the key node, which is the basis for subsequent analysis of the displacement of the key node and the structural stability. At the same time, recording the relative position relationship between adjacent nodes helps to construct the overall spatial relationship model of the building structure. The acquired image data will be marked with a monitoring timestamp to form an image data sequence, and this timestamp is very important for subsequent time series analysis of the data.

[0059] In one embodiment, the image acquisition device is arranged at multiple viewing positions in the building monitoring area to ensure that all key surface areas and structural nodes of the building main body can be covered, and the installation height and angle of the device are adjusted according to the characteristics of the building structure.

[0060] In one embodiment, a high-definition camera network can be used for image acquisition. Install multiple high-definition cameras at key positions in the building monitoring area (such as the four corners of the building facade, around key beams and columns, etc.). These cameras have characteristics such as automatic focus, high resolution (such as 1080p and above), and wide viewing angle (such as the horizontal viewing angle reaching more than 90 degrees). The monitoring period of the cameras is set to acquire images once at 8 am and 8 pm every day. When acquiring the surface texture image, use an image enhancement algorithm to improve the contrast and clarity of the image to better obtain the color distribution information. For the preliminary identification and marking of cracks, use a deep learning-based semantic segmentation algorithm to segment the crack area in the image and mark it with a red line. For the geometric contour image, determine the contour edge coordinates of the key node through a feature point extraction algorithm (such as the SIFT algorithm) and calculate the relative position relationship between adjacent nodes. The acquired image data is stored in timestamp order to form an image data sequence. The technical effect of this method is that the high-definition camera network can comprehensively cover the monitoring area, and advanced image processing algorithms can accurately obtain the required image information.

[0061] S102. By means of displacement sensors and stress sensors arranged at key positions of the building, the spatial displacement signals and material stress signals of key nodes of the structure are collected in real time, generating a sensor signal sequence with synchronized timestamp marks. The spatial displacement signals include the three-dimensional component change information of the node coordinates, and the material stress signals include the amplitude fluctuation information of different stress types.

[0062] A displacement sensor is a sensor that can accurately measure the position change of an object in space and is mainly used in construction engineering to monitor the displacement of key nodes. Its measurement principle is based on principles such as optics, electricity, or mechanics, such as fiber optic sensors, capacitive sensors, etc.

[0063] A stress sensor is a sensor used to measure the magnitude and change of internal stress in building materials. According to the stress type (such as tensile stress, compressive stress, shear stress, etc.), different sensing elements and measurement methods can be adopted for stress sensors. For example, strain gauge sensors can be used to measure tensile stress and compressive stress.

[0064] The three-dimensional component change information of the node coordinates in the spatial displacement signal represents the change amount of the position coordinates of the key node in the three-dimensional space (usually with the length, width, and height directions of the building as the coordinate axes). For example, the displacement amount of the node in the x-axis direction, the displacement amount in the y-axis direction, and the displacement amount in the z-axis direction. These change information can reflect the actual movement trajectory of the key node in space.

[0065] The amplitude fluctuation information of different stress types in the material stress signal refers to the fluctuation of the stress magnitude of different stress types (such as tensile stress, compressive stress, shear stress) inside the building materials over time. For example, when the building is subjected to wind load, tensile stress may be generated in the building structure on the windward side, and compressive stress may be generated on the leeward side. The stress sensor can monitor the change of these stress amplitudes in real time.

[0066] In the embodiment of the present application, displacement sensors and stress sensors can be reasonably arranged at key positions of the building. These key positions are usually determined based on the mechanical analysis of the building structure, such as the bottom of load-bearing columns, the mid-span of beams, the beam-column connection nodes, etc. The displacement sensor senses the position change of the key node in the three-dimensional space in real time and converts it into a spatial displacement signal containing the three-dimensional component change information of the node coordinates. The stress sensor simultaneously measures the stress inside the building materials and obtains the amplitude fluctuation information of different stress types, forming a material stress signal. These signals are all marked with synchronized timestamps to generate a sensor signal sequence. This process ensures the temporal consistency of the displacement and stress data and provides a basis for subsequent comprehensive analysis.

[0067] In one embodiment, an optical fiber displacement sensor and a strain gauge stress sensor can be adopted. The optical fiber displacement sensor is installed at key positions of the building (such as the column base, beam end, etc.). Based on the optical interference principle of the optical fiber, when the key node undergoes displacement, the optical path of the optical fiber changes, thereby accurately measuring the displacement of the node in three-dimensional space and obtaining a spatial displacement signal containing the change information of the three-dimensional components of the node coordinates. The strain gauge stress sensor is pasted on the surface of the building material. The strain gauge generates a resistance change according to the strain of the material, and through a measurement circuit, it is converted into an electrical signal, thereby obtaining the amplitude fluctuation information of different stress types (such as tensile stress, compressive stress). All sensor signals are collected by a data collector, and a synchronous timestamp is added to each signal to form a sensor signal sequence. The technical effect of this method is that the optical fiber displacement sensor has the characteristics of high precision and strong anti-interference ability, and the strain gauge stress sensor can accurately measure stress changes, and the combination of the two can provide reliable displacement and stress data.

[0068] S103. Perform time alignment processing on the structural image data and sensor signal data under the same monitoring timestamp, and arrange them in the order of monitoring time to form the multi-source heterogeneous data set.

[0069] Time alignment processing is a data processing operation, the purpose of which is to make the information from different data sources (such as structural image data and sensor signal data) but with the same monitoring time accurately correspond on the time axis. This is like placing different types of data collected at the same moment accurately at the corresponding positions on a time coordinate axis.

[0070] Monitoring time order: Arrange the data in the order of time of data collection to reflect the state changes of the building structure at different time points.

[0071] Since the structural image data and sensor signal data are collected by different devices, although each has a timestamp mark, there may be some small time deviations. This step is to eliminate this deviation and accurately associate these two types of data under the same monitoring timestamp. Through time alignment processing, it can be ensured that when analyzing the deformation of the building structure, the image information and sensor information at the same moment can be comprehensively considered. Arranging these time-aligned data in the order of monitoring time forms a multi-source heterogeneous data set. This set completely records various data information of the building at different monitoring periods, providing a comprehensive data basis for subsequent operations such as spatio-temporal feature extraction.

[0072] In one embodiment, the time stamps in the structural image data and the sensor signal data can be subtracted to find the time deviation between the two. Then, based on this deviation, one of the data (such as the structural image data) is adjusted in time so that the time stamps of the two are exactly the same. For example, if the time stamp of the sensor signal data is 5 seconds faster than the time stamp of the structural image data, then the time stamp of the structural image data is pushed back 5 seconds. Finally, the time-aligned data is arranged in the order of the monitoring time to form a multi-source heterogeneous data set. The technical effect of this method is simple and intuitive, and can effectively align the time of the two types of data.

[0073] In one embodiment, referring to Figure 4 , step S20 may specifically include: S201. Perform gray level equalization processing on the surface texture image in the structural image data, extract the gray level co-occurrence matrix features of the images in different monitoring periods, calculate the Euclidean distance between the gray level co-occurrence matrix features of adjacent periods as the surface texture change amplitude parameter, and generate the surface texture change rate parameter in combination with the monitoring time interval, and jointly constitute the structural surface texture change feature.

[0074] Gray level equalization processing is an image processing technique aimed at adjusting the gray level distribution of an image, enhancing the contrast of the image, and making the probabilities of each gray level appearing in the image more uniform. This can highlight the detailed information in the image and facilitate subsequent feature extraction.

[0075] Gray level co-occurrence matrix features are a statistical method for describing image texture features. It reflects the texture information of an image by calculating the frequencies of different gray level pixel pairs appearing at specific directions and distances. For example, in the surface texture image of a building, different building materials (such as tiles, concrete) have different textures, and these textures will have different manifestations in the gray level co-occurrence matrix.

[0076] The surface texture change amplitude parameter is obtained by calculating the Euclidean distance between the gray level co-occurrence matrix features of adjacent periods. The Euclidean distance is a common distance metric method, which is used here to measure the difference degree of the image texture features in two periods. This parameter reflects the magnitude of the change in the surface texture between different periods.

[0077] The surface texture change rate parameter is obtained by combining the monitoring time interval and the surface texture change amplitude parameter. It represents the speed of change of the surface texture per unit time and is an important indicator for measuring the change trend of the surface texture.

[0078] This step can extract parameters from the surface texture image that can reflect the characteristics of the structural surface texture changes. First, perform histogram equalization on the surface texture image in the structural image data. This step can improve the visual effect of the image and enhance the texture information. Then, for the images in different monitoring periods, extract their gray-level co-occurrence matrix features respectively. Since the gray-level co-occurrence matrix can well describe the texture characteristics of the image, by comparing the gray-level co-occurrence matrix features of adjacent periods, the changes in the surface texture can be understood. Calculate the Euclidean distance between the gray-level co-occurrence matrix features of adjacent periods. This distance value directly reflects the change range of the texture features, and it is defined as the surface texture change range parameter. Combining with the known monitoring time interval, the surface texture change rate parameter can be calculated. These two parameters together constitute the characteristics of the structural surface texture changes, thus describing the changes in the building surface texture over time from different perspectives.

[0079] S202. Perform edge detection on the geometric contour image in the structural image data, identify the contour edge coordinates of the key nodes, and calculate the position offset vector and angle deflection amount of the contour edge coordinates of the key nodes in adjacent monitoring periods as the spatial displacement parameter and direction change parameter of the key nodes.

[0080] Edge detection is an image processing technique used to detect the edge contours of objects in an image. In the building geometric contour image, the edge contours of the key nodes can be accurately found through edge detection, which is crucial for analyzing the position changes of the key nodes.

[0081] The contour edge coordinates of the key nodes refer to the coordinate values of the edge points of the key nodes determined by the edge detection algorithm in the image coordinate system in the geometric contour image. These coordinate values accurately describe the contour shape and position of the key nodes.

[0082] The position offset vector is a vector obtained by calculating the difference between the contour edge coordinates of the key nodes in adjacent monitoring periods. This vector represents the position movement direction and distance of the key node from one period to another in two-dimensional or three-dimensional space, intuitively reflecting the displacement of the key node.

[0083] The angle deflection amount refers to the angle change amount of the connection direction between the key nodes in adjacent monitoring periods. It describes the direction change of the key node relative to its adjacent nodes and is an important indicator for measuring the spatial direction change of the key node.

[0084] The purpose of this step is to obtain the spatial displacement and direction change information of key nodes. First, perform edge detection on the geometric contour image, which helps to clearly identify the contour edges of key nodes, so as to accurately obtain the contour edge coordinates. Then, compare the contour edge coordinates of key nodes in adjacent monitoring periods. Calculate the coordinate difference between them. The vector formed by this difference is the position offset vector, which can clearly represent the displacement direction and distance of key nodes in space. At the same time, select the structural connection line formed by adjacent key nodes and calculate the angle difference in the direction of this connection line in adjacent monitoring periods, that is, the angle deflection amount. The position offset vector and the angle deflection amount are used as the spatial displacement parameter and direction change parameter of key nodes respectively, providing key spatial information for analyzing the stability of building structures.

[0085] One technical implementation method is to use the Canny edge detection algorithm for edge detection processing. In Python, use the "cv2.Canny" function in the OpenCV library to perform edge detection on the geometric contour image. Then, determine the contour edge coordinates of key nodes through the coordinate calculation method in the image coordinate system. After obtaining the contour edge coordinates of key nodes for the images in adjacent monitoring periods respectively, use the vector calculation method to calculate the position offset vector. When calculating the angle deflection amount, the trigonometric function relationship can be used. Calculate the slope of the connection line according to the coordinates of adjacent key nodes, and then calculate the angle deflection amount through the change of the slope. The Canny edge detection algorithm has good edge detection effect, can accurately identify the contour edges of key nodes, and then accurately calculate the spatial displacement parameter and direction change parameter.

[0086] In one embodiment, step S202 may specifically include: A. Use a pre-trained edge detection algorithm to process the geometric contour image, identify the contour edges of building key nodes, and extract the pixel coordinates of the contour edges to generate a coordinate set containing the key points of the node contour.

[0087] The pre-trained edge detection algorithm is an algorithm trained with a large amount of image data and can effectively identify the edge information in the image. In the deformation detection of building engineering, this algorithm is specifically used to detect the contour edges of building key nodes in the geometric contour image. Based on existing knowledge and pattern recognition capabilities, it reduces the time and resource costs of re-training for specific building images. The coordinate set of key points of the node contour is a set containing the coordinates of key pixel points on the contour edges of building key nodes. The coordinates of these key points can accurately describe the contour shape of key nodes and are important basic data for subsequent analysis of node displacement and deformation.

[0088] In the embodiments of the present application, a pre-trained edge detection algorithm can be used to process the geometric contour image. The pre-trained algorithm has the ability to recognize various image edge patterns and can quickly locate the contour edges of building key nodes in the image. For example, for the geometric contour image of a beam-column joint, the algorithm can recognize the edge contour at the beam-column intersection. During the recognition process, the algorithm analyzes the pixel information of the image and determines which pixels belong to the contour edges of the key node according to the pre-learned edge feature patterns. Then, the pixel coordinates of these contour edges are accurately extracted. This extraction process is very crucial because it directly relates to the subsequent accurate description of the node position. The extracted coordinates will form a set of coordinates containing the key points of the node contour, and each coordinate point corresponds to a key position on the contour edge. Through this set of coordinate points, the contour shape of the key node can be completely depicted.

[0089] B. Establish the node matching relationship between the geometric contour images of adjacent monitoring periods, and identify the same-name key nodes through the feature point matching algorithm to ensure the correspondence of node coordinates in different periods.

[0090] The node matching relationship refers to the relationship of associating the image features representing the same building key node in the geometric contour images of adjacent monitoring periods. The establishment of this relationship is to ensure that in different periods, the analysis of the same key node is based on accurately corresponding coordinate information, so as to calculate the displacement and change of the node subsequently.

[0091] The feature point matching algorithm is an algorithm used to identify the same feature points in different images. In the deformation detection of building engineering, it is used to identify the same-name key nodes in the geometric contour images of adjacent monitoring periods. By comparing the feature descriptions of the feature points (such as corner points, edge points, etc.) in the images, points with similar features are found to determine whether they are the same key node.

[0092] After obtaining the geometric contour images of different monitoring periods and the sets of coordinates containing the key points of the node contour respectively in this step, it is necessary to establish the node matching relationship between the images of adjacent monitoring periods. Due to the possible slight deformation of the building structure in different periods or the influence of factors such as shooting angle and lighting, it is inaccurate to directly determine the same node according to the coordinate position. Therefore, the feature point matching algorithm is used to identify the same-name key nodes. For example, in the monitoring of the beam-column structure of a building, there may be certain differences in the beam-column joint images taken at different times. The feature point matching algorithm can judge which feature points belong to the same beam-column joint in the images of different periods by analyzing the feature information of the feature points such as corner points and edge points around the node, such as the Histogram of Oriented Gradients (HOG) feature, etc. In this way, the correspondence of node coordinates in different periods is ensured, laying a foundation for accurately analyzing the displacement and direction change of the key node.

[0093] C. For the successfully matched key nodes, calculate the coordinate differences between the contour edge coordinates of the current monitoring period and those of the previous monitoring period, and generate a position offset vector in a two-dimensional plane or a three-dimensional space. The magnitude of the position offset vector represents the displacement distance of the node, and the direction represents the spatial orientation of the displacement.

[0094] The position offset vector is a vector calculated from the differences in the contour edge coordinates of the key nodes between the current monitoring period and the previous monitoring period. This vector describes the position change of the key node from one monitoring period to the next in a two-dimensional plane or a three-dimensional space, including the distance and direction of the displacement.

[0095] The displacement distance is represented by the magnitude of the position offset vector. It is a scalar value that reflects the actual moving length of the key node in space, without considering the direction information.

[0096] After the homologous key nodes of adjacent monitoring periods are successfully identified through the feature point matching algorithm, their position offset vectors can be calculated. For each successfully matched key node, obtain the contour edge coordinates of the current monitoring period and those of the previous monitoring period. In a two-dimensional plane or a three-dimensional space, calculate the differences in the corresponding coordinates respectively. These differences form the position offset vector. The magnitude of the position offset vector can intuitively reflect the moving length of the key node in space. And the direction of the position offset vector represents the spatial orientation of the displacement, which describes the direction in which the key node has been displaced.

[0097] D. Select the structural connection line formed by adjacent key nodes as the reference direction, and calculate the angle difference between the connection line directions of the current monitoring period and the previous monitoring period as the angular deflection amount of the node connection line. The angular deflection amount is used to reflect the relative rotation trend of the structural node.

[0098] The angular deflection amount of the node connection line refers to the direction angle difference between the structural connection lines formed by adjacent key nodes between the current monitoring period and the previous monitoring period. This angle difference can reflect the rotation trend of the structural node relative to its adjacent nodes and is an important parameter for analyzing the local deformation and stability of building structures.

[0099] When analyzing the displacement changes of key nodes in the embodiments of the present application, in addition to considering the position offset vector of the nodes themselves, it is also necessary to pay attention to the rotation trend of the nodes relative to adjacent nodes. For this purpose, the structural connection line formed by adjacent key nodes is selected as the reference direction. For example, in the frame structure of a building, the connection line between beam-column nodes is a typical structural connection line. Calculate the angle difference between the directions of this connection line in the current monitoring period and the previous monitoring period. In one embodiment, first, determine the direction vector of the connection line in each monitoring period. This can be obtained by calculating the coordinate differences of adjacent key nodes. Then, use the vector angle formula to calculate the angle difference, which is the angle deflection amount of the node connection line. The angle deflection amount can intuitively reflect whether the structural nodes have relative rotation and the magnitude of the rotation, thereby providing important information for comprehensively evaluating the stability of the building structure.

[0100] E. Normalize the magnitude and direction information of the position offset vector and the angle deflection amount, and use them as the spatial displacement parameter and direction change parameter of the key node respectively.

[0101] Normalization processing is an operation that maps data to a specific interval (usually the [0,1] interval). In this step, normalize the magnitude of the position offset vector, direction information, and angle deflection amount, so that these parameters are on a unified scale, facilitating subsequent comparison and analysis in different building structures or different monitoring scenarios.

[0102] The spatial displacement parameter in this step can be the magnitude and direction information of the position offset vector after normalization processing, which is used to describe the displacement situation of the key node in space, including the relative magnitude and direction of the displacement, and can more accurately reflect the displacement differences between different nodes on a unified scale.

[0103] The direction change parameter in this step is the angle deflection amount after normalization processing, which represents the degree of direction change of the key node relative to adjacent nodes, and can better be used together with other parameters to evaluate the deformation situation of the building structure on the normalized scale.

[0104] To make the spatial displacement and direction change parameters of different key nodes comparable, it is necessary to normalize the magnitude and direction information of the position offset vector and the angle deflection obtained from the previous calculations. For the magnitude of the position offset vector, a linear normalization method can be used, such as mapping its value to the interval [0, 1]. For the direction information, it can be normalized according to the specific representation (such as the direction angle), for example, mapping the direction angle to the interval [0, 2π]. For the angle deflection, a linear normalization method can also be used to make its value within the interval [0, 1]. The normalized magnitude and direction information are used as the spatial displacement parameters of the key nodes, and the normalized angle deflection is used as the direction change parameter. In this way, in subsequent building structure deformation analysis, these parameters can be comprehensively considered on the same scale to more accurately evaluate the stability of the building structure.

[0105] S203. Extract the coordinate change sequence of the key nodes in the three-dimensional space coordinate system from the spatial displacement signal in the sensor signal data, and use the trend slope and fluctuation variance of the coordinate change sequence as the direction stability parameter and amplitude fluctuation parameter of the displacement trend.

[0106] The spatial displacement signal is a signal collected by a displacement sensor that reflects the position change of the key nodes in three-dimensional space. This signal contains the change of the position information of the key nodes in the three-dimensional space coordinate system of the building structure (such as the coordinate system established with the length, width, and height of the building as the coordinate axes) over time.

[0107] The coordinate change sequence refers to the sequence of the coordinate values of the key nodes in the three-dimensional space coordinate system extracted from the spatial displacement signal over time. For example, at a series of monitoring times, the coordinate values of the key nodes are recorded respectively to form a sequence of coordinate values changing over time.

[0108] The direction stability parameter of the displacement trend takes the trend slope of the coordinate change sequence as this parameter. The trend slope reflects the change rate of the coordinate change in a certain direction. Through this slope, it can be judged whether the displacement direction of the key nodes in three-dimensional space is stable. For example, a constant slope indicates a stable direction, and a large change in the slope indicates an unstable direction.

[0109] The amplitude fluctuation parameter of the displacement trend takes the fluctuation variance of the coordinate change sequence as this parameter. The fluctuation variance measures the fluctuation size of the coordinate values within a certain time range and reflects the stability of the displacement amplitude of the key nodes. The larger the variance, the greater the fluctuation of the displacement amplitude.

[0110] This step focuses on the spatial displacement signals in the sensor signal data, aiming to extract parameters that can reflect the displacement trend characteristics of key nodes. First, extract the coordinate change sequence of key nodes in the three-dimensional space coordinate system from the spatial displacement signals, which requires parsing and organizing the data collected by the displacement sensor. Then, analyze the coordinate change sequence and calculate the trend slope of the coordinate change sequence. This slope can reflect the change trend of the displacement direction of key nodes in the three-dimensional space, and it is defined as the direction stability parameter of the displacement trend. At the same time, calculate the fluctuation variance of the coordinate change sequence. The variance reflects the degree of dispersion of the coordinate values, that is, the fluctuation of the displacement amplitude of key nodes, and it is used as the amplitude fluctuation parameter of the displacement trend. These two parameters describe the displacement trend characteristics of key nodes from the aspects of displacement direction and amplitude.

[0111] S204. Perform spectral decomposition processing on the material stress signals in the sensor signal data to extract the frequency components and energy distribution characteristics of different stress type signals, and calculate the correlation coefficient of the energy distribution characteristics in adjacent monitoring periods as the stress change consistency parameter.

[0112] Spectral decomposition processing is a signal processing technique that converts the signal in the time domain to the frequency domain for analysis. For material stress signals, spectral decomposition can reveal the distribution of different frequency components in the signal and the energy information carried by each frequency component.

[0113] Among them, different stress types (such as tensile stress, compressive stress, etc.) are manifested as different frequency components and energy distributions in the material stress signal. The frequency component represents the frequency component of the stress signal in the frequency domain, and the energy distribution characteristic reflects the energy magnitude contained in different frequency components.

[0114] The stress change consistency parameter is obtained by calculating the correlation coefficient of the energy distribution characteristics in adjacent monitoring periods. The correlation coefficient is a statistical index that measures the linear correlation between two variables. Here, it is used to represent the similarity degree of the stress energy distribution characteristics in adjacent periods, that is, the consistency degree of stress change.

[0115] This step analyzes the material stress signal in the sensor signal data. First, perform spectral decomposition on the material stress signal, which converts the stress signal from the time domain to the frequency domain, enabling the observation of the signal characteristics at different frequencies. In the frequency domain, extract the frequency components and energy distribution characteristics of different stress type signals. Different stress types will have different manifestations in the frequency domain. For example, some stresses may be mainly concentrated in the low-frequency part, while others may have more energy distribution in the high-frequency part. Then, compare the energy distribution characteristics of adjacent monitoring periods. Calculate the correlation coefficient between them. This correlation coefficient serves as a stress change consistency parameter, which can reflect whether the stress changes between adjacent periods are consistent. If the correlation coefficient is close to 1, it indicates that the stress changes are highly consistent; if the correlation coefficient is close to 0 or negative, it indicates that the stress changes are inconsistent or have significant differences.

[0116] S205. Combine the surface texture change amplitude parameter, surface texture change rate parameter, spatial displacement parameter, direction change parameter, direction stability parameter of the displacement trend, amplitude fluctuation parameter, and stress change consistency parameter in the order of the monitoring time to form the multi-dimensional deformation feature set.

[0117] The main task of this step is to combine the various deformation-related parameters obtained in the previous steps to form a multi-dimensional deformation feature set. Combining them in the order of the monitoring time is very important because it can maintain the temporal correspondence of each parameter, thus fully reflecting the deformation state of the building structure at different time points. These parameters include the surface texture change amplitude parameter, surface texture change rate parameter, spatial displacement parameter, direction change parameter, direction stability parameter of the displacement trend, amplitude fluctuation parameter, and stress change consistency parameter. Each parameter describes the deformation of the building structure from its own specific aspect. The multi-dimensional deformation feature set formed by combining them provides a comprehensive data basis for further deformation analysis (such as input into a deformation detection model).

[0118] In one embodiment, the value of each parameter and the corresponding monitoring timestamp can be stored in a corresponding table in the database. For example, a table named "DeformationFeatures" can be created, containing fields such as "MonitoringTime" (monitoring time), "SurfaceTextureChangeAmplitude" (surface texture change amplitude parameter), and "SurfaceTextureChangeRate" (surface texture change rate parameter). Then, data is extracted from the table in the order of monitoring time through a database query statement, and this data is combined into a multidimensional deformation feature set. The technical effect of this method is that the database management system can be used to conveniently store, manage, and query data, ensuring the integrity and accuracy of the data.

[0119] In one embodiment, the pre-trained deformation detection model may include a time series feature processing layer and a spatial feature association layer, wherein the time series feature processing layer is used to process a multi-dimensional deformation feature set having a corresponding relationship with monitoring timestamps, and the spatial feature association layer is used to analyze the spatial correlation of feature parameters at different monitoring locations. Figure 5 , step S30 may specifically include: S301. Input the multidimensional deformation feature set into the time series feature processing layer in the order of monitoring time, perform time series dependency modeling on the feature parameters of adjacent monitoring periods through the long short-term memory network structure, and extract the time series feature vector containing the feature correlation relationship between the historical period and the current period. The time series feature vector contains the weight distribution information of different feature parameters in the time dimension.

[0120] The time series feature processing layer is a specific layer in the deformation detection model, specifically designed to process data with time series characteristics—a collection of multidimensional deformation features arranged in chronological order. The primary function of this layer is to mine data for features and relationships along the time dimension.

[0121] The Long Short-Term Memory (LSTM) network structure is a special recurrent neural network (RNN) architecture designed to address the vanishing or exploding gradient problems of traditional RNNs when processing long sequences of data. LSTM uses a special gating mechanism (input gate, forget gate, and output gate) to control the flow of information, effectively modeling the temporal dependencies in long sequences of data.

[0122] Temporal dependency modeling is the process of modeling the interdependencies of data over time. In construction deformation monitoring, the goal is to identify correlations between deformation features across different monitoring periods. For example, how changes in structural surface texture or displacement of key nodes at a previous moment affect corresponding features at the current moment.

[0123] The time-series feature vector is a vector representation obtained after time-series dependence modeling. It synthesizes the feature correlation relationships between historical periods and the current period, and includes the weight distribution information of different feature parameters in the time dimension. These weights represent the relative importance of each feature parameter to the overall deformation situation at different time points.

[0124] In one embodiment, the multi-dimensional deformation feature set arranged in the order of monitoring time can be input into the time-series feature processing layer. This step ensures the chronological order of the data, meeting the requirements of time-series analysis. Then, a long short-term memory network structure is used for time-series dependence modeling. The input of the LSTM network is the feature parameters of adjacent monitoring periods. For example, in the monitoring of a certain building structure, it may be the surface texture change rate parameters, key node displacement trends, etc. of the previous hour and the current hour. The LSTM network processes these inputs through its internal gating mechanism. The forget gate determines which historical information needs to be forgotten, the input gate controls the inflow of new information, and the output gate determines the final output information. Through such a mechanism, the LSTM can capture the complex relationships between features in different periods, thereby extracting the time-series feature vector that includes the feature correlation relationships between historical periods and the current period. Different elements in this vector correspond to different feature parameters, and the weight of each element represents the importance of the feature parameter in the time dimension. For example, for a building area where the surface texture changes relatively stably in the long term, the weight of the feature parameter related to the surface texture change in the time-series feature vector may be relatively low; while for an area where key node displacements often fluctuate, the weight of the displacement-related feature parameter may be relatively high.

[0125] S302. Input the time-series feature vector into the spatial feature correlation layer, construct a spatial node correlation graph of the building monitoring area through a graph neural network structure, where the nodes represent the monitoring positions, and the edges represent the spatial distances and structural connection relationships between the monitoring positions, and calculate the spatial influence factors of the feature parameters at different monitoring positions based on the spatial node correlation graph.

[0126] The spatial feature correlation layer is a component of the deformation detection model, and its main function is to analyze the spatial correlation of the feature parameters at different monitoring positions. It realizes the integration and analysis of the spatial information of the building structure by constructing a spatial node correlation graph.

[0127] The graph neural network structure (GNN) is a neural network model specifically used to process graph-structured data. In the deformation detection of building engineering, the graph-structured data is represented as a spatial node correlation graph of the building monitoring area. The GNN can learn the feature information of the nodes and edges in the graph and model the relationships between the nodes.

[0128] The spatial node association graph is a graphical structure representation, where nodes represent monitoring locations in the building monitoring area, and edges represent the spatial distances and structural connection relationships between the monitoring locations. For example, in a building frame structure, beam-column joints can be used as nodes, and the connections between beams and columns form the edges. The attributes of the edges can include information such as the length of the beam (spatial distance) and the force transfer relationship (structural connection relationship).

[0129] The spatial influence factor is a parameter obtained through the analysis and calculation of the spatial node association graph, which is used to measure the influence degree of the characteristic parameters of different monitoring locations on the calculation of the deformation probability of other monitoring locations. It reflects the role of spatial position relationships and structural connection relationships in deformation propagation and mutual influence.

[0130] After obtaining the time-series feature vectors, they are input into the spatial feature association layer. In this layer, first, a spatial node association graph of the building monitoring area is constructed using the graph neural network structure. During the construction process, the attributes of nodes and edges are determined according to the actual layout and mechanical relationships of the building structure. Nodes represent specific monitoring locations, and each node carries the corresponding characteristic parameter information obtained from the time-series feature vectors. Edges accurately describe the spatial distances and structural connection relationships between the monitoring locations, which is very important for analyzing the overall stability of the building structure and the deformation propagation path. For example, in a multi-story building, there may be strong structural connection relationships between the column foot nodes on the same floor due to the integrity of the foundation, and this relationship is reflected through the attributes of the edges in the spatial node association graph. Then, based on the constructed spatial node association graph, the graph neural network calculates the spatial influence factors of the characteristic parameters of different monitoring locations by aggregating the feature information of adjacent nodes. For example, a key column node located in the center of the building may have a relatively large spatial influence factor on the calculation of the deformation probability of surrounding nodes due to its close structural connection relationships with multiple surrounding beam and column nodes; while some non-key nodes located at the edge of the building may have relatively small spatial influence factors.

[0131] In this embodiment, the DGL (Deep Graph Library) framework can be used to construct the graph neural network. In DGL, the graph structure is defined and the attributes of nodes and edges are set. The node feature data obtained from the time-series feature vectors is associated with the graph structure. A graph neural network model based on DGL is constructed, such as using algorithms like GraphSAGE. The model is trained to calculate the spatial influence factor by learning from the actual deformation data of the building structure. DGL has high-efficient graph computing performance and flexible model construction capabilities, and is suitable for handling spatial association analysis tasks in complex building structures.

[0132] S303. Generate a comprehensive feature representation based on the temporal feature vector and the spatial influence factor, and perform a non-linear transformation on the comprehensive feature representation through a fully-connected neural network structure to generate a deformation probability distribution including deformation probability values at different monitoring positions, where the deformation probability value represents the likelihood of structural deformation occurring at the corresponding monitoring position during the current monitoring period.

[0133] The comprehensive feature representation is a form of feature representation that combines the temporal feature vector and the spatial influence factor. It integrates the feature correlation relationship in the time dimension (represented by the temporal feature vector) and the feature correlation in the spatial dimension (represented by the spatial influence factor), providing comprehensive feature information for subsequent calculation of the deformation probability value.

[0134] The fully-connected neural network structure is one of the most basic neural network structures, where each neuron in one layer is connected to all neurons in the next layer. In this step, the fully-connected neural network is used to perform a non-linear transformation on the comprehensive feature representation to adapt to the complex task of calculating the deformation probability of the building structure.

[0135] In the embodiment of the present application, a comprehensive feature representation can be generated based on the previously obtained temporal feature vector and spatial influence factor. This process is to fuse the feature information in the two dimensions of time and space, so that the generated comprehensive feature representation can comprehensively reflect the deformation of the building structure. For example, when considering the column grid structure of a large commercial building, the comprehensive feature representation of a certain column node should include both its own feature changes in the time series (such as the displacement trend in historical periods) and the spatial influence of surrounding nodes on it (such as the influence brought by the structural connection relationship of adjacent columns and beam nodes). Then, the comprehensive feature representation is input into the fully-connected neural network structure. The fully-connected neural network performs a non-linear transformation on the input through multiple hidden layers. This non-linear transformation can capture the complex non-linear relationships between features. For example, some minor deformations in the building structure may lead to a greater deformation risk under the non-linear action of multiple factors. Finally, after being processed by the fully-connected neural network, a deformation probability distribution including deformation probability values at different monitoring positions is output. Each deformation probability value represents the likelihood of structural deformation occurring at the corresponding monitoring position during the current monitoring period. For example, in a high-rise building, the calculated deformation probability value of a certain column foot position on a certain floor is 0.1, while the deformation probability value of another column foot position is 0.05, which indicates that the former has a relatively higher likelihood of structural deformation.

[0136] In one embodiment, referring to Figure 6 , step S40 may specifically include: S401. Analyze the deformation probability values of each monitoring position in the deformation probability distribution, define the monitoring positions with deformation probability values exceeding the preset probability threshold as high-probability value monitoring positions, and cluster the continuously adjacent high-probability value monitoring positions to form a suspicious deformation area, where the preset probability threshold is preset according to the building structure safety standard.

[0137] The preset probability threshold is a value determined in advance according to the building structure safety standard. It is a boundary for judging whether a deformation may occur at a monitoring position. When the deformation probability value of a monitoring position exceeds this threshold, it is considered to have a relatively high possibility of deformation.

[0138] The suspicious deformation area is an area formed by clustering continuously adjacent monitoring positions with deformation probability values exceeding the preset probability threshold. This area is suspected of having a structural deformation risk and requires further analysis of its spatio-temporal evolution characteristics.

[0139] In the embodiment of the present application, first, the deformation probability distribution is analyzed to obtain the deformation probability value corresponding to each monitoring position. The deformation probability distribution is an important data containing the deformation information of each monitoring position in the entire building monitoring area. Then, screening is carried out according to the preset probability threshold. This probability threshold is set based on the building structure safety standard. For example, it is determined comprehensively according to factors such as the type of building (such as high-rise building, industrial factory building, etc.), service life, building materials, etc. When the deformation probability value of a monitoring position exceeds this threshold, it means that the possibility of structural deformation at this position is relatively large, and this monitoring position can be defined as a high-probability value monitoring position. Finally, these continuously adjacent high-probability value monitoring positions are clustered together to form a suspicious deformation area. For example, in the monitoring of the frame structure of a large building, if the deformation probability values of the monitoring positions around some beam-column joints are all relatively high and adjacent to each other, then the area where these joints are located will be clustered as a suspicious deformation area. This clustering method can initially divide the area where deformation risks may exist and provide a target area for subsequent in-depth analysis.

[0140] S402. Extract the monitoring timestamps and spatial coordinates of each monitoring position in the suspicious deformation area, analyze the continuous change trend of the deformation probability value in the time dimension, determine the start monitoring time and end monitoring time of the suspicious deformation area, and use the time difference between the start monitoring time and the end monitoring time as the deformation duration parameter.

[0141] The monitoring timestamp is the time information marked for the data of each monitoring position during the data acquisition process. It records the moment of data acquisition and is used to analyze the changes in the data in the time series.

[0142] Spatial coordinates: Represent the specific position information of the monitoring location in the building space, usually represented in the form of three-dimensional coordinates (such as x, y, z coordinates), which helps to determine the specific position of the suspicious deformation area in the building structure.

[0143] The deformation duration parameter is obtained by calculating the time difference between the start monitoring time and the end monitoring time of the suspicious deformation area. This parameter describes the time length experienced by the suspicious deformation area from the start of being monitored to the current moment, and is an important indicator for analyzing the spatio-temporal evolution characteristics of the suspicious deformation area.

[0144] In the embodiment of the present application, for the already determined suspicious deformation area, first, the monitoring timestamps and spatial coordinates of each monitoring location therein need to be extracted. The monitoring timestamp can reflect the time sequence of the data at each monitoring location, and the spatial coordinates determine the specific position of each monitoring location in the building structure. Then, analyze the continuous change trend of the deformation probability value in the time dimension within the suspicious deformation area. This can be achieved by performing data fitting or trend analysis on the deformation probability values at different timestamps. For example, a linear regression analysis method can be used to observe whether the deformation probability value rises, falls, or fluctuates over time. According to this change trend, determine the start monitoring time of the suspicious deformation area, that is, the earliest time point when the deformation probability value starts to deviate significantly from the normal range or shows a specific change trend; and the end monitoring time, that is, the current monitoring time. Finally, calculate the time difference between these two times to obtain the deformation duration parameter. This parameter is of great significance for evaluating the development speed and risk level of the suspicious deformation area. For example, a longer deformation duration may mean more serious structural problems or a slower deformation development process.

[0145] S403. Obtain the minimum bounding rectangle coordinates and centroid coordinates of the suspicious deformation area in the spatial dimension, analyze the position offset and size change of the minimum bounding rectangle coordinates in adjacent monitoring periods, and generate the spatial range expansion direction and expansion rate parameter of the suspicious deformation area according to the position offset and size change.

[0146] The minimum bounding rectangle is the smallest rectangle that can completely contain the suspicious deformation area, and its coordinates can determine the position of this rectangle in the building space. The minimum bounding rectangle is a simple and effective way to describe the spatial range of an area.

[0147] The centroid coordinates are the coordinate positions of the center of gravity of the suspicious deformation area in the building space, which reflect the central position of the suspicious deformation area in space.

[0148] The spatial range expansion direction is determined by analyzing the position offset of the minimum bounding rectangle coordinates in adjacent monitoring periods. It indicates the direction in which the suspicious deformation area expands in space and is important direction information for describing the spatial evolution of the suspicious deformation area.

[0149] The expansion rate parameter is calculated based on the size change of the minimum bounding rectangle in adjacent monitoring periods. It describes the expansion speed of the spatial range of the suspicious deformation area per unit time and is an important indicator for measuring the spatial evolution speed of the suspicious deformation area.

[0150] In this step, for the suspicious deformation area, first calculate its minimum bounding rectangle coordinates and centroid coordinates in the spatial dimension. The minimum bounding rectangle coordinates can be determined by finding the maximum and minimum coordinate values of the suspicious deformation area in each direction. The centroid coordinates can be calculated as the weighted average of the coordinates of all monitoring positions within the area. Then, analyze the position offset and size change of the minimum bounding rectangle coordinates in adjacent monitoring periods. The position offset can be obtained by calculating the difference in the center coordinates of the minimum bounding rectangles in different periods, and this difference reflects the translation of the suspicious deformation area in space. The size change is to compare the changes in the length, width, and height of the minimum bounding rectangles in adjacent periods. Determine the spatial range expansion direction based on the position offset. For example, if the coordinate difference of the minimum bounding rectangle in the x-axis direction is positive, it indicates that the suspicious deformation area has an expansion trend in the positive x-axis direction. Calculate the expansion rate parameter based on the size change. For example, calculate the volume change rate or area change rate (in the two-dimensional case) of the minimum bounding rectangle in adjacent periods as the expansion rate parameter. These parameters can comprehensively describe the spatial evolution of the suspicious deformation area.

[0151] S404. Based on the deformation duration parameter, the spatial range expansion direction, and the expansion rate parameter, construct the spatio-temporal evolution characteristics of the suspicious deformation area. The spatio-temporal evolution characteristics are used to describe the temporal evolution law and spatial expansion pattern of the suspicious deformation area during the monitoring period.

[0152] The spatio-temporal evolution characteristics are a characteristic representation that comprehensively describes the evolution of the suspicious deformation area in time and space. It integrates information such as the deformation duration parameter, the spatial range expansion direction, and the expansion rate parameter, and can comprehensively reflect the development process of the suspicious deformation area in the building structure from the start of monitoring to the current moment.

[0153] When constructing the spatio-temporal evolution characteristics of the suspicious deformation area in the embodiments of the present application, it is necessary to comprehensively consider the previously obtained deformation duration parameter, the spatial range expansion direction, and the expansion rate parameter. First, the deformation duration parameter gives the development span of the suspicious deformation area in the time dimension. The spatial range expansion direction describes the movement or expansion trend of the suspicious deformation area in space, for example, whether it expands in a certain specific direction (such as the vertical direction, the horizontal direction, etc.), or presents an irregular expansion pattern. The expansion rate parameter quantifies the expansion speed of the suspicious deformation area in space. Integrating this information together, the spatio-temporal evolution characteristics of the suspicious deformation area can be constructed. For example, if a suspicious deformation area has a long deformation duration, the spatial range expansion direction is towards a certain weak structure direction of the building, and the expansion rate is fast, then this spatio-temporal evolution characteristic indicates that the suspicious deformation area has a high structural risk and it is necessary to take timely measures for further detection and repair.

[0154] In one embodiment, referring to Figure 7 , step S50 may specifically include: S501. According to the suspicious deformation area and the corresponding spatio-temporal evolution characteristics, obtain the spatial coordinates of all monitoring positions within the suspicious deformation area, and arrange them in the order of monitoring time to form a deformation position coordinate sequence.

[0155] The deformation position coordinate sequence is a sequence of spatial coordinates of monitoring positions within the suspicious deformation area arranged in the order of monitoring time, and this sequence can reflect the spatial changes of each position within the suspicious deformation area over time.

[0156] S502. According to the deformation probability value in the deformation probability distribution, the expansion rate parameter in the spatio-temporal evolution characteristics, the design standard of the building structure, and the historical monitoring data, establish a deformation risk level assessment rule, and the deformation risk level assessment rule maps the deformation probability value and the expansion rate parameter to a preset risk.

[0157] The expansion rate parameter in the spatio-temporal evolution characteristics describes the expansion speed of the suspicious deformation area in space, reflects the development trend of the deformation, and is used to evaluate the risk level.

[0158] The design standard of the building structure is the specification and requirement regarding the structural safety, stability, etc. determined during the design stage of the building, such as the bearing capacity, the allowable deformation range, etc., and is the basic reference for evaluating the deformation risk.

[0159] The historical monitoring data is the data accumulated from the past monitoring of the building, which contains the state information of the building structure at different times and can provide a reference and comparison basis for the current risk assessment.

[0160] Deformation risk level assessment rule: A rule that establishes a mapping relationship between deformation probability values and expansion rate parameters and a preset risk level interval, used to determine the risk level of a suspicious deformation area.

[0161] S503. Calculate the deformation risk level of each suspicious deformation area according to the deformation risk level assessment rule, where the deformation risk level includes a level identifier corresponding to the risk degree.

[0162] The deformation risk level assessment rule is an established rule for determining the risk level of a suspicious deformation area. It comprehensively considers multiple factors and maps relevant parameters to different risk level intervals.

[0163] Deformation risk level: A classification identifier for the risk degree of a suspicious deformation area, such as low risk, medium risk, high risk, etc. Different levels represent different degrees of structural deformation risk.

[0164] S504. Integrate the deformation position coordinate sequence, spatio-temporal evolution characteristics, and deformation risk level to generate a detection report including monitoring position coordinates, time interval, spatial expansion mode, and deformation risk level.

[0165] This step is to integrate the key information obtained in the previous steps to generate a detection report. First, the deformation position coordinate sequence provides the spatial information of the monitoring positions within the suspicious deformation area. These coordinates are arranged in the order of monitoring time and can reflect the spatial states of each position at different times. The spatio-temporal evolution characteristics include information such as the time interval and the spatial expansion mode. The time interval describes the time period from the appearance of the suspicious deformation in the suspicious deformation area to the current time, and the spatial expansion mode illustrates the development trend of the suspicious deformation area in space. The deformation risk level intuitively represents the risk degree of the suspicious deformation area. Integrate these information together and generate a detection report in a specific format. For example, a table form or a specific data structure (such as JSON format) can be used to organize the report content to ensure that the report includes necessary information such as monitoring position coordinates, time interval, spatial expansion mode, and deformation risk level, so that the building safety management system can clearly understand the overall situation of the suspicious deformation area.

[0166] Correspondingly, in order to better implement the above method, an embodiment of the present application further provides a deformation detection system based on a construction project. As Figure 8 shown, the deformation detection system 80 based on a construction project includes: An acquisition module 801, configured to acquire a multi-source heterogeneous data set within a building monitoring area, where the multi-source heterogeneous data set includes structural image data and sensor signal data for different monitoring time periods, the structural image data includes surface texture images of the building main body and geometric contour images of key nodes, and the sensor signal data includes real-time output signals of displacement sensors and stress sensors arranged at key positions of the building; A feature extraction module 802, configured to perform spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features, where the structural surface texture change features reflect the gray value fluctuation law of the building main body surface texture during the monitoring time period, and the key node displacement trend features reflect the continuous change pattern of the key node spatial coordinates during the monitoring time period; A deformation detection module 803, configured to input the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generate a deformation probability distribution of the building monitoring area through the deformation detection model, where the deformation probability distribution is used to represent the possibility of structural deformation occurring at different monitoring positions; A feature determination module 804, configured to determine a suspicious deformation area within the building monitoring area and the spatio-temporal evolution features of the suspicious deformation area according to the deformation probability distribution, where the spatio-temporal evolution features include the time interval when the deformation occurs and the spatial range expansion mode; A report generation module 805, configured to generate a detection report including monitoring position coordinates and deformation risk levels based on the suspicious deformation area and the spatio-temporal evolution features, and send the detection report to a building safety management system to trigger an early warning assessment operation.

[0167] For the implementation of each of the above modules, reference may specifically be made to the foregoing method embodiments, which will not be elaborated herein. The technical effects achieved by each module and device are described with reference to the foregoing method embodiments.

[0168] It should be noted that, in specific implementation, the above modules can be combined arbitrarily, integrated into one or several modules, or implemented as independent entities. In addition, the above modules can be implemented in the form of hardware or in the form of software function modules. When the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc.

[0169] Such as Figure 9As shown in the figure, an embodiment of the present application further provides a computer device 90, which includes a processor 901 and a memory 902. Among them, the memory 902 stores a computer program. When the computer program is executed by the processor 901, the processor 901 is caused to execute the steps of any of the above methods.

[0170] In several embodiments provided by the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0171] The above-disclosed are only the preferred embodiments of the present application. Of course, the scope of the rights of the present application cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. A deformation detection method based on construction engineering, characterized in that The method includes: Obtaining a multi-source heterogeneous data set within the building monitoring area, where the multi-source heterogeneous data set includes structural image data and sensor signal data for different monitoring periods; Performing spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features. The structural surface texture change features reflect the gray value fluctuation law of the building main body surface texture during the monitoring period, and the key node displacement trend features reflect the continuous change mode of the key node spatial coordinates during the monitoring period; Inputting the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generating a deformation probability distribution of the building monitoring area through the deformation detection model; Determining the suspicious deformation area within the building monitoring area and the spatio-temporal evolution characteristics of the suspicious deformation area according to the deformation probability distribution. The spatio-temporal evolution characteristics include the time interval when the deformation occurs and the spatial range expansion mode; Generating a detection report including monitoring position coordinates and deformation risk levels based on the suspicious deformation area and the spatio-temporal evolution characteristics, and sending the detection report to the building safety management system to trigger an early warning assessment operation.

2. The deformation detection method based on construction engineering according to claim 1, characterized in that The obtaining of the multi-source heterogeneous data set within the building monitoring area, where the multi-source heterogeneous data set includes structural image data and sensor signal data for different monitoring periods, includes: Through the image acquisition devices arranged in the building monitoring area, collecting the surface texture images of the building main body and the geometric contour images of the key nodes according to a fixed monitoring period, and generating an image data sequence with monitoring timestamp marks; Through the displacement sensors and stress sensors arranged at key positions of the building, collecting the spatial displacement signals and material stress signals of the structural key nodes in real time, and generating a sensor signal sequence with synchronous timestamp marks; Performing time alignment processing on the structural image data and sensor signal data under the same monitoring timestamp, and arranging them in the order of monitoring time to form the multi-source heterogeneous data set.

3. The deformation detection method based on construction engineering according to claim 2, wherein The image acquisition devices are arranged at multiple viewing angle positions in the building monitoring area to ensure that all key surface areas and structural nodes of the building main body can be covered, and the installation height and angle of the devices are adjusted according to the building structure characteristics.

4. The deformation detection method based on construction engineering according to claim 1, characterized in that, The structural surface texture change features include: surface texture change amplitude parameters and surface texture change rate parameters; the key node displacement trend features include: key node spatial displacement parameters, direction change parameters, and direction stability parameters and amplitude fluctuation parameters of the key node spatial displacement trend; The performing of spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features includes: Performing gray level equalization processing on the surface texture images in the structural image data, and extracting the gray level co-occurrence matrix features of the images for different monitoring periods; calculating the Euclidean distance between the gray level co-occurrence matrix features of adjacent periods as the surface texture change amplitude parameter, and generating the surface texture change rate parameter in combination with the monitoring time interval; Perform edge detection processing on the geometric contour image in the structural image data to identify the contour edge coordinates of the key nodes; calculate the position offset vector and angle deflection amount of the contour edge coordinates of the key nodes in adjacent monitoring periods as the spatial displacement parameter and direction change parameter of the key nodes; Extract the coordinate change sequence of the key nodes in the three-dimensional space coordinate system from the spatial displacement signal in the sensor signal data, and use the trend slope and fluctuation variance of the coordinate change sequence as the direction stability parameter and amplitude fluctuation parameter of the spatial displacement trend; Perform spectral decomposition processing on the material stress signal in the sensor signal data to extract the frequency components and energy distribution characteristics of different stress type signals, and calculate the correlation coefficient of the energy distribution characteristics in adjacent monitoring periods as the stress change consistency parameter; Combine the surface texture change amplitude parameter, surface texture change rate parameter, spatial displacement parameter, direction change parameter, direction stability parameter of the displacement trend, amplitude fluctuation parameter, and stress change consistency parameter in the order of monitoring time to form the multi-dimensional deformation feature set.

5. The deformation detection method based on construction engineering according to claim 4, wherein The performing edge detection processing on the geometric contour image in the structural image data to identify the contour edge coordinates of the key nodes, calculating the position offset vector and angle deflection amount of the contour edge coordinates of the key nodes in adjacent monitoring periods as the spatial displacement parameter and direction change parameter of the key nodes includes: Process the geometric contour image using a pre-trained edge detection algorithm to identify the contour edges of the building key nodes, extract the pixel coordinates of the contour edges, and generate a coordinate set containing the key points of the node contour; Establish the node matching relationship of the geometric contour images in adjacent monitoring periods, identify the same-name key nodes through the feature point matching algorithm, and ensure the correspondence of the node coordinates in different periods; For the successfully matched key nodes, calculate the coordinate difference between the contour edge coordinates in the current monitoring period and the contour edge coordinates in the previous monitoring period to generate a position offset vector in a two-dimensional plane or three-dimensional space. The magnitude of the position offset vector represents the displacement distance of the node, and the direction represents the spatial direction of the displacement; Select the structural connection line formed by adjacent key nodes as the reference direction, calculate the angle difference between the connection line directions in the current monitoring period and the previous monitoring period as the angle deflection amount of the node connection line, and the angle deflection amount is used to reflect the relative rotation trend of the structural node; Normalize the magnitude and direction information of the position offset vector and the angle deflection amount, and use them as the spatial displacement parameter and direction change parameter of the key nodes respectively.

6. The deformation detection method based on construction engineering according to claim 1, characterized in that, The pre-trained deformation detection model includes a time series feature processing layer and a spatial feature correlation layer. The time series feature processing layer is used to process the multi-dimensional deformation feature set with monitoring timestamp correspondence, and the spatial feature correlation layer is used to analyze the spatial correlation of the feature parameters at different monitoring positions; The inputting the multi-dimensional deformation feature set into the pre-trained deformation detection model to generate the deformation probability distribution of the building monitoring area by the deformation detection model includes: Input the multi-dimensional deformation feature set into the time series feature processing layer in the order of monitoring time. Use the long short-term memory network structure to model the temporal dependence of the feature parameters in adjacent monitoring periods, and extract the temporal feature vector containing the feature correlation relationship between the historical period and the current period; Input the temporal feature vector into the spatial feature correlation layer. Use the graph neural network structure to construct the spatial node correlation graph of the building monitoring area, and calculate the spatial influence factors of the feature parameters at different monitoring positions based on the spatial node correlation graph; Generate the comprehensive feature representation according to the temporal feature vector and the spatial influence factor, and perform non-linear transformation processing on the comprehensive feature representation through the fully connected neural network structure to generate the deformation probability distribution containing the deformation probability values at different monitoring positions.

7. The deformation detection method based on construction engineering according to claim 1, characterized in that, The step of determining the suspicious deformation area in the building monitoring area according to the deformation probability distribution includes: Analyze the deformation probability values of each monitoring position in the deformation probability distribution, define the monitoring positions with deformation probability values exceeding the preset probability threshold as the high-probability value monitoring positions, and cluster the continuously adjacent high-probability value monitoring positions to form the suspicious deformation area. The preset probability threshold is preset according to the building structure safety standard.

8. The deformation detection method based on construction engineering according to claim 7, wherein The step of determining the spatio-temporal evolution characteristics of the suspicious deformation area in the building monitoring area according to the deformation probability distribution includes: Extract the monitoring timestamps and spatial coordinates of each monitoring position in the suspicious deformation area, and analyze the continuous change trend of the deformation probability value in the time dimension to determine the start monitoring time and end monitoring time of the suspicious deformation area. Take the time difference between the start monitoring time and the end monitoring time as the deformation duration parameter; Obtain the minimum bounding rectangle coordinates and centroid coordinates of the suspicious deformation area in the spatial dimension, and analyze the position offset and size change of the minimum bounding rectangle coordinates in adjacent monitoring periods. Generate the spatial range expansion direction and expansion rate parameters of the suspicious deformation area according to the position offset and size change; Based on the deformation duration parameter, the spatial range expansion direction and the expansion rate parameter, construct the spatio-temporal evolution characteristics of the suspicious deformation area.

9. The deformation detection method based on construction engineering according to claim 8, wherein, Generating a detection report including the monitoring position coordinates and the deformation risk level based on the suspicious deformation area and the spatio-temporal evolution characteristics includes: According to the suspicious deformation area and the corresponding spatio-temporal evolution characteristics, obtain the spatial coordinates of all monitoring positions in the suspicious deformation area, and arrange them in the order of monitoring time to form a deformation position coordinate sequence; Establish a deformation risk level assessment rule according to the deformation probability value in the deformation probability distribution, the expansion rate parameter in the spatio-temporal evolution characteristics, the design standard of the building structure and the historical monitoring data; Calculate the deformation risk level of each suspicious deformation area according to the deformation risk level assessment rule; Integrate the deformation position coordinate sequence, the spatio-temporal evolution characteristics and the deformation risk level to generate a detection report including the monitoring position coordinates, the time interval, the spatial expansion mode and the deformation risk level.

10. A deformation detection system based on construction engineering, characterized in that, The system includes: An acquisition module, configured to acquire a multi-source heterogeneous data set within a building monitoring area, where the multi-source heterogeneous data set includes structural image data and sensor signal data for different monitoring periods; A feature extraction module, configured to perform spatio-temporal feature extraction processing on the multi-source heterogeneous data set to obtain a multi-dimensional deformation feature set including structural surface texture change features and key node displacement trend features. The structural surface texture change features reflect the gray value fluctuation law of the building main body surface texture during the monitoring period, and the key node displacement trend features reflect the continuous change pattern of the key node spatial coordinates during the monitoring period; A deformation detection module, configured to input the multi-dimensional deformation feature set into a pre-trained deformation detection model, and generate a deformation probability distribution of the building monitoring area through the deformation detection model; A feature determination module, configured to determine a suspicious deformation area within the building monitoring area and the spatio-temporal evolution features of the suspicious deformation area according to the deformation probability distribution. The spatio-temporal evolution features include the time interval when the deformation occurs and the spatial range expansion mode; A report generation module, configured to generate a detection report including monitoring position coordinates and deformation risk levels based on the suspicious deformation area and the spatio-temporal evolution features, and send the detection report to a building safety management system to trigger a warning assessment operation.

Images