Building curtain wall deformation degree detection method and system

Abstract

The invention provides a building curtain wall deformation degree detection method and system, and relates to the technical field of intelligent building monitoring, and the method comprises the steps: obtaining multi-modal curtain wall data of a target building curtain wall based on a multi-modal data collection network; constructing a curtain wall deformation detection model, inputting the multi-modal curtain wall data into the curtain wall deformation detection model, and generating a curtain wall deformation detection result; based on a three-level alarm response mechanism, triggering an alarm response of a corresponding level according to a curtain wall deformation detection result; based on a model closed-loop optimization mechanism, parameters of the curtain wall deformation detection model are correspondingly adjusted, so that comprehensive, accurate and real-time detection and early warning can be performed on the deformation degree of the building curtain wall, the requirement of manual intervention is greatly reduced, the curtain wall detection efficiency is improved, meanwhile, the continuous operation capability and the immediate response characteristic of a detection system are guaranteed, and the detection accuracy is improved. And the safety and the reliability of the building curtain wall are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent building monitoring, and particularly to a method and system for detecting the deformation degree of a building curtain wall. Background Art

[0002] In modern architecture, building curtain walls are widely used. With the increase in building height and the growth of service life, the safety problems of curtain walls have become increasingly prominent.

[0003] Traditional means for detecting the deformation of building curtain walls, such as manual inspections and simple sensor monitoring, have defects such as low efficiency, poor accuracy, and inability to perform real-time monitoring. Manual inspections rely on manual observation, which not only consumes a large amount of manpower and time but is also easily affected by subjective factors, resulting in missed inspections and misjudgments. The existing method of combining fixedly installed cameras with manual image analysis also faces problems such as low efficiency, reliance on manpower, and inability to perform real-time continuous monitoring, making it difficult to meet the requirements for efficient, accurate, and real-time building curtain wall detection, seriously affecting the timeliness of building safety management, and increasing operation and maintenance costs.

[0004] Therefore, it is necessary to provide a method and system for detecting the deformation degree of a building curtain wall to solve the above technical problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a method and system for detecting the deformation degree of a building curtain wall, which are used to solve the problems of low detection accuracy, inability to perform real-time continuous monitoring, high operation and maintenance costs, and lack of efficient and automated detection means existing in the prior art.

[0006] A method for detecting the deformation degree of a building curtain wall provided by the present invention, the detection method comprising: Based on a multi-modal data acquisition network, acquiring multi-modal curtain wall data of a target building curtain wall; Constructing a curtain wall deformation detection model, inputting the multi-modal curtain wall data into the curtain wall deformation detection model, and generating a curtain wall deformation detection result; Based on a three-level alarm response mechanism, triggering an alarm response at the corresponding level according to the curtain wall deformation detection result; Based on a model closed-loop optimization mechanism, correspondingly adjusting the parameters of the curtain wall deformation detection model.

[0007] Preferably, the multi-modal data acquisition network includes a multi-view camera array, an infrared thermal image acquisition device, and an acceleration sensor; The multi-view camera array, the infrared thermal image acquisition device, and the acceleration sensor are respectively used to acquire visible light curtain wall images, infrared thermal imaging data, and deformation vibration signals of the target building curtain wall; Among them, the multi-modal curtain wall data includes the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal.

[0008] Preferably, obtaining the visible light curtain wall image of the target building curtain wall based on the multi-view camera array specifically includes: Collecting point cloud image data on the surface of the target building curtain wall by using a laser scanning device according to a preset scanning pitch; Constructing a three-dimensional digital twin model corresponding to the target building curtain wall based on the point cloud image data by using a surface fitting algorithm; Taking the lowest overlap rate of the field of view angles of adjacent cameras and the largest curtain wall coverage as the optimization goal, calculating the fitness function of the particle swarm in the three-dimensional digital twin model based on the particle swarm optimization algorithm, and iteratively updating the particle velocity and position until the optimization goal is satisfied to obtain the camera positions on the surface of the three-dimensional digital twin model; For the camera positions that need to perform global detection, install a fish-eye lens to obtain the global curtain wall image of the target building curtain wall. For the camera positions that need to perform local detection, install a telephoto lens to obtain the local curtain wall image of the target building curtain wall; Summarizing the global curtain wall image and the local curtain wall image to generate the visible light curtain wall image.

[0009] Preferably, before inputting the multi-modal curtain wall data into the curtain wall deformation recognition model, preprocessing operations are performed on the multi-modal curtain wall data, specifically including: Performing grayscale and normalization processing on the visible light curtain wall image; Performing temperature calibration and noise reduction processing on the infrared thermal imaging data; Performing filtering processing on the deformation vibration signal; Performing feature scaling on the preprocessed visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal.

[0010] Preferably, the curtain wall deformation recognition model includes an input layer, an intermediate layer, and an output layer; Based on the input layer, splicing the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal to generate the multi-modal curtain wall data; Extracting features from the multi-modal curtain wall data through the dual-channel convolution module of the intermediate layer to generate intermediate curtain wall features; Based on the adaptive threshold decision module of the output layer, performing temporal weighted voting on the intermediate curtain wall features to obtain the curtain wall deformation detection result.

[0011] Preferably, the multi-modal curtain wall data is subjected to feature extraction by the dual-channel convolution module of the intermediate layer to generate intermediate curtain wall features, specifically including: The visible light curtain wall image and the infrared thermal imaging data are subjected to convolution operations through the convolution kernels of the first channel in the dual-channel convolution module, and the curtain wall spatial features are extracted; The time series corresponding to the deformation vibration signal is subjected to a convolution operation through the one-dimensional convolution layer of the second channel in the dual-channel convolution module, and the curtain wall time features are extracted; The curtain wall spatial features and the curtain wall time features are spliced to obtain the intermediate curtain wall features.

[0012] Preferably, based on the three-level alarm response mechanism, corresponding-level alarm responses are triggered according to the curtain wall deformation detection results, specifically including: Based on the material, structure, and service life of the target building curtain wall, the triggering conditions for each level of alarm in the three-level alarm response mechanism are determined; When the curtain wall deformation detection result meets the triggering condition of the first-level alarm, the edge node is started, and a lightweight algorithm is used to perform first-level processing on the curtain wall deformation detection result, generate first-level alarm response information, and send it to the management end; When the curtain wall deformation detection result meets the triggering condition of the second-level alarm, the regional server cluster is started, and a distributed computing framework is used to perform second-level processing on the curtain wall deformation detection result, generate second-level alarm response information, and send it to the management end; When the curtain wall deformation detection result meets the triggering condition of the third-level alarm, the cloud supercomputer is called, and a numerical simulation method is used to perform third-level processing on the curtain wall deformation detection result, generate third-level alarm response information, and send it to the management end.

[0013] Preferably, based on the reinforcement learning algorithm, the resource allocation ratios of the edge node, the regional server cluster, and the cloud supercomputer are dynamically adjusted to minimize the energy consumption of the three-level alarm response mechanism and generate an optimal mechanism resource allocation strategy.

[0014] A building curtain wall deformation degree detection system, the detection system includes: A data acquisition module for acquiring multi-modal curtain wall data of a target building curtain wall based on a multi-modal data acquisition network; A deformation detection module for constructing a curtain wall deformation detection model, inputting the multi-modal curtain wall data into the curtain wall deformation detection model, and generating a curtain wall deformation detection result; An alarm response module for triggering corresponding-level alarm responses based on a three-level alarm response mechanism according to the curtain wall deformation detection result; A closed-loop optimization module for correspondingly adjusting the parameters of the curtain wall deformation detection model based on a model closed-loop optimization mechanism.

[0015] Compared with related technologies, a method and system for detecting the deformation degree of a building curtain wall provided by the present invention have the following beneficial effects: The present invention can obtain multi-modal curtain wall data of a target building curtain wall based on a multi-modal data acquisition network; construct a curtain wall deformation detection model, input the multi-modal curtain wall data into the curtain wall deformation detection model to generate a curtain wall deformation detection result; based on a three-level alarm response mechanism, trigger an alarm response at the corresponding level according to the curtain wall deformation detection result; and correspondingly adjust the parameters of the curtain wall deformation detection model based on a model closed-loop optimization mechanism, so as to comprehensively, accurately and real-time detect and warn the deformation degree of the building curtain wall, and improve the safety and reliability of the building curtain wall.

[0016] By adopting a multi-modal data acquisition network, the present invention can comprehensively monitor the curtain wall, greatly improving the early detection rate of curtain wall deformation. By constructing a curtain wall deformation detection model, the present invention can automatically and accurately detect the deformation degree of the curtain wall, greatly reducing the need for manual intervention and improving the curtain wall detection efficiency. The present invention adopts a hierarchical response alarm mechanism, which can significantly reduce the energy consumption of the detection system, improve the system response speed, optimize the system performance and resource allocation. By means of a closed-loop optimization mechanism, the present invention can realize the continuous evolution of the model performance, enhance the adaptability and reliability of the model, ensure the continuous operation ability and instant response characteristics of the detection system, and strengthen the safety prevention level. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flowchart of a method for detecting the deformation degree of a building curtain wall according to the present invention; Figure 2 is a flowchart of obtaining multi-modal curtain wall data according to the present invention; Figure 3 is a system block diagram of a system for detecting the deformation degree of a building curtain wall according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The present invention will be further described below with reference to the drawings and embodiments.

[0019] Embodiment 1

[0020] As Figure 1 shown, a method for detecting the deformation degree of a building curtain wall, the detection method includes: S1. Obtain multi-modal curtain wall data of a target building curtain wall based on a multi-modal data acquisition network; Among them, the multi-modal data acquisition network refers to a comprehensive network that integrates various data acquisition technologies and devices. The target building curtain wall refers to the building curtain wall that needs to be detected for the degree of deformation.

[0021] It can be understood that through the multi-modal data acquisition network, a variety of data acquisition devices can be integrated, namely, a multi-view camera array, an infrared thermal imaging acquisition device, and an acceleration sensor, to collect information on the building curtain wall from different dimensions, and then obtain multi-modal curtain wall data, namely visible light curtain wall images, infrared thermal imaging data, and deformation vibration signals.

[0022] S2. Construct a curtain wall deformation detection model, input the multi-modal curtain wall data into the curtain wall deformation detection model, and generate a curtain wall deformation detection result; It should be noted that during the construction process of the curtain wall deformation detection model, the model can be trained using curtain wall data under different working conditions, enabling the model to learn the characteristic patterns of the curtain wall in the normal state and various deformation states.

[0023] Then, the obtained multi-modal curtain wall data can be input into this model. At this time, the model can analyze the input data and extract key features, and finally generate a curtain wall deformation detection result, thereby accurately judging whether the curtain wall has deformed, as well as information such as the location, degree, and development trend of the deformation.

[0024] S3. Based on a three-level alarm response mechanism, trigger the corresponding level of alarm response according to the curtain wall deformation detection result; In practical applications, the three-level alarm response mechanism is preset according to the safety standards and risk assessment results of the building curtain wall. This mechanism can divide the degree of curtain wall deformation into three levels: the first level represents minor deformation, which may only be slight displacement or cracks, with little impact on the overall structural safety of the curtain wall; the second level is significant deformation, at this time the deformation of the curtain wall is already relatively obvious, which may affect the stability of some structures; the third level is critical deformation, which means that the curtain wall is in a serious dangerous state and may cause safety accidents at any time.

[0025] S4. Based on the model closed-loop optimization mechanism, make corresponding adjustments to the parameters of the curtain wall deformation detection model.

[0026] Among them, based on the model closed-loop optimization mechanism, new multi-modal curtain wall data and actual detection feedback information can be continuously collected. By comparing the model prediction results with the actual situation, the performance of the curtain wall deformation detection model can be evaluated. If it is found that the model is inaccurate or has omissions in some cases, the parameters of the model can be appropriately adjusted.

[0027] For example, the weight values in the neural network of the model can be adjusted, the size and quantity of the convolutional kernels can be optimized, etc., so that the model can continuously adapt to new situations and changes, realize the continuous evolution of the model performance, and further improve the accuracy and reliability of deformation detection, enabling the entire detection system to always maintain the best working state.

[0028] In the specific implementation process, such as Figure 2 As shown, the multi-modal data acquisition network includes a multi-view camera array, an infrared thermal imaging acquisition device, and an acceleration sensor; The multi-view camera array, the infrared thermal imaging acquisition device, and the acceleration sensor are respectively used to obtain the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal of the target building curtain wall; Among them, the multi-modal curtain wall data includes the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal.

[0029] In practical applications, the multi-modal curtain wall data can reflect the curtain wall state from different dimensions, providing comprehensive and rich information for accurately detecting the deformation of the curtain wall.

[0030] Specifically, the multi-view camera array refers to a combination of multiple cameras installed around the building curtain wall according to a specific layout. These cameras can take pictures of the curtain wall from different angles and obtain all-round and dead-angle-free curtain wall information. By presetting multiple camera viewpoints, each camera can collect images of specific areas of the curtain wall, and then, these images can be integrated to form a complete visible light curtain wall image, providing an intuitive visual basis for subsequent analysis.

[0031] The infrared thermal imaging acquisition device detects the infrared radiation emitted by the curtain wall surface and converts it into a thermal image, thereby obtaining the infrared thermal imaging data. This data can reflect the temperature distribution on the curtain wall surface, helping to promptly discover temperature abnormal areas caused by internal defects or abnormal heat transfer, such as potential problems like hollowing and leakage inside the curtain wall.

[0032] In the building curtain wall detection scenario, the acceleration sensor is installed at key structural parts of the curtain wall, which can monitor the acceleration changes of the curtain wall under external actions in real time, and then obtain the deformation vibration signal. These signals can reflect the vibration state of the curtain wall, helping to judge the stability of the curtain wall under external forces such as wind and earthquake.

[0033] Based on the multi-view camera array to obtain the visible light curtain wall image of the target building curtain wall, specifically including: Collect the point cloud image data of the surface of the target building curtain wall by using a laser scanning device according to a preset scanning spacing; Adopt a surface fitting algorithm to construct a three-dimensional digital twin model corresponding to the target building curtain wall based on the point cloud image data; Taking the lowest overlap rate of the field of view angles of adjacent cameras and the largest curtain wall coverage as the optimization objectives, based on the particle swarm optimization algorithm, calculate the fitness function of the particle swarm in the three-dimensional digital twin model, and iteratively update the particle velocity and position until the optimization objectives are met, to obtain the camera positions on the surface of the three-dimensional digital twin model; For the camera positions that need to perform global detection, install fisheye lenses to obtain the global curtain wall images of the target building curtain wall. For the camera positions that need to perform local detection, install telephoto lenses to obtain the local curtain wall images of the target building curtain wall; Summarize the global curtain wall images and the local curtain wall images to generate the visible light curtain wall image.

[0034] It can be understood that the preset scanning spacing refers to the distance set in advance according to the size, complexity, and detection accuracy requirements of the curtain wall, usually in the range of 5 - 20 cm, to ensure that a sufficient and appropriate amount of point cloud data is obtained. Point cloud image data refers to the spatial coordinate information of a large number of discrete points on the curtain wall surface.

[0035] First, emit laser beams to the curtain wall surface through a laser scanning device and receive the reflected light, and the point cloud image data on the curtain wall surface can be recorded. Then, a surface fitting algorithm can be used to process this point cloud image data and construct a three-dimensional digital twin model. It should be noted that the surface fitting algorithm is a mathematical algorithm that can generate a smooth surface by fitting the discrete point cloud data, thereby constructing a virtual three-dimensional model corresponding to the actual height of the curtain wall, that is, the three-dimensional digital twin model, which accurately replicates the shape, size, and structure of the actual curtain wall in the virtual space.

[0036] Furthermore, the particle swarm optimization algorithm can be used to determine the camera positions with the lowest overlap rate of the field of view angles of adjacent cameras and the largest curtain wall coverage as the optimization objectives. The overlap rate of the field of view angles refers to the overlapping ratio of the captured images of adjacent cameras, and the particle swarm optimization algorithm is an optimization algorithm that simulates the foraging behavior of bird flocks. In the three-dimensional digital twin model, each particle represents the potential installation position of the camera. By calculating the fitness function of the particle swarm, the particle velocity and position are continuously iteratively updated until the optimization objectives are achieved, to obtain the optimal camera positions.

[0037] In addition, for the camera positions that need to perform global detection, install fisheye lenses, whose large field of view can obtain global curtain wall images; for the camera positions that need to perform local detection, install telephoto lenses, which can obtain local curtain wall images. Finally, these global and local curtain wall images can be integrated to generate a complete visible light curtain wall image.

[0038] Before inputting the multi-modal curtain wall data into the curtain wall deformation recognition model, preprocessing operations are performed on the multi-modal curtain wall data, specifically including: Perform grayscale and normalization processing on the visible light curtain wall image; Perform temperature calibration and noise reduction processing on the infrared thermal imaging data; Perform filtering processing on the deformation vibration signal; Perform feature scaling on the preprocessed visible light curtain wall image, infrared thermal imaging data, and deformation vibration signal.

[0039] Among them, by grayscaling and normalizing the visible light curtain wall image, the image information can be simplified, the brightness difference can be eliminated, the image data can be made more regular, facilitating the model to recognize image features and improving the detection accuracy. By performing temperature calibration and noise reduction processing on the infrared thermal imaging data, the temperature data can be ensured to be accurate and reliable, interference can be reduced, and the model can analyze thermal anomalies more accurately. By performing filtering processing on the deformation vibration signal, noise can be removed and key information can be retained, providing more effective vibration data for the curtain wall deformation recognition model.

[0040] In addition, by performing feature scaling and unifying the scales of different types of data, the curtain wall deformation recognition model can more efficiently fuse and analyze multi-modal data, significantly improving the detection performance of the model.

[0041] The curtain wall deformation recognition model includes an input layer, an intermediate layer, and an output layer; Based on the input layer, the visible light curtain wall image, infrared thermal imaging data, and deformation vibration signal are spliced to generate the multi-modal curtain wall data; The multi-modal curtain wall data is subjected to feature extraction through the dual-channel convolution module of the intermediate layer to generate intermediate curtain wall features; Based on the adaptive threshold decision module of the output layer, temporal weighted voting is performed on the intermediate curtain wall features to obtain the curtain wall deformation detection result.

[0042] The multi-modal curtain wall data is subjected to feature extraction through the dual-channel convolution module of the intermediate layer to generate intermediate curtain wall features, specifically including: Perform convolution operations on the visible light curtain wall image and infrared thermal imaging data through the convolution kernel of the first channel in the dual-channel convolution module, and extract the curtain wall spatial features; Perform convolution operations on the time series corresponding to the deformation vibration signal through the one-dimensional convolution layer of the second channel in the dual-channel convolution module, and extract the curtain wall time features; The curtain wall spatial features and curtain wall time features are spliced to obtain the intermediate curtain wall features.

[0043] It should be noted that through the input layer of the curtain wall deformation recognition model, visible light curtain wall images, infrared thermal imaging data, and deformation vibration signals can be spliced into multi-modal curtain wall data to achieve the integration of multi-source information and comprehensively reflect the curtain wall state from different dimensions.

[0044] Furthermore, in the dual-channel convolution module of the middle layer of the model, through the first channel, convolution operations can be performed on visible light and infrared data to extract curtain wall spatial features, such as the shape, position, and size of cracks on the curtain wall surface, etc.; through the one-dimensional convolution layer of the second channel, the time series of deformation vibration signals can be processed to extract time features, such as deformation rate, vibration frequency change, etc., thereby significantly improving the integrity and accuracy of feature extraction.

[0045] Finally, through the adaptive threshold decision module of the model output layer, temporal weighted voting can be performed on the middle curtain wall features to obtain the detection result, which can fully consider the change trend of data in the time dimension, effectively avoid the problem of misjudgment of data at a single moment, and enhance the reliability and stability of the detection result.

[0046] Through the curtain wall deformation recognition model, the accuracy and reliability of building curtain wall deformation detection can be significantly improved, and potential safety hazards of the curtain wall can be discovered in a timely and accurate manner.

[0047] Based on the three-level alarm response mechanism, corresponding-level alarm responses are triggered according to the curtain wall deformation detection results, specifically including: Based on the material, structure, and service life of the target building curtain wall, the triggering conditions for each level of alarm in the three-level alarm response mechanism are determined; When the curtain wall deformation detection result meets the triggering condition of the first-level alarm, the edge node is started, and a lightweight algorithm is used to perform the first-level processing on the curtain wall deformation detection result, generate the first-level alarm response information and send it to the management end; When the curtain wall deformation detection result meets the triggering condition of the second-level alarm, the regional server cluster is started, and a distributed computing framework is used to perform the second-level processing on the curtain wall deformation detection result, generate the second-level alarm response information and send it to the management end; When the curtain wall deformation detection result meets the triggering condition of the third-level alarm, the cloud supercomputer is called, and a numerical simulation method is used to perform the third-level processing on the curtain wall deformation detection result, generate the third-level alarm response information and send it to the management end.

[0048] Based on the reinforcement learning algorithm, the resource allocation ratio of the edge node, the regional server cluster, and the cloud supercomputer is dynamically adjusted to minimize the energy consumption of the three-level alarm response mechanism and generate the optimal mechanism resource allocation strategy.

[0049] In practical applications, the triggering conditions of each level of alarm can be determined based on the material, structure, and service life of the building curtain wall, so that the individual differences of the curtain wall can be fully considered. Different levels of alarm response mechanisms are triggered according to the severity of the curtain wall deformation, and the corresponding alarm response information is sent to the management terminal, enabling the management terminal to quickly understand the safety status of the curtain wall and make decisions in a timely manner. Here, the management terminal refers to the management personnel who monitor the deformation detection process of the curtain wall.

[0050] Furthermore, different computing resources can be enabled for different levels of alarms. Specifically, for a level-one alarm, an edge node can be started, and a lightweight algorithm can be used to process the detection results, thus avoiding excessive resource consumption; for a level-two alarm, a regional server cluster can be started for distributed computing, thereby balancing the processing efficiency and resource occupancy; for a level-three alarm, a cloud supercomputer can be called, and a numerical simulation method can be used to deeply analyze the detection results, so as to accurately evaluate severe deformations.

[0051] In addition, the resource allocation ratios of the edge node, regional server cluster, and cloud supercomputer can be dynamically adjusted based on the reinforcement learning algorithm, effectively reducing the energy consumption of the level-three alarm response mechanism, generating an optimal resource allocation strategy, achieving the maximization of resource utilization and the minimization of energy consumption, greatly improving the efficiency, reliability, and economy of the detection system, and ensuring the safe operation of the building curtain wall.

[0052] Embodiment 2

[0053] As Figure 3 shown, a building curtain wall deformation degree detection system, the detection system includes: A data acquisition module, configured to acquire multi-modal curtain wall data of a target building curtain wall based on a multi-modal data acquisition network; A deformation detection module, configured to build a curtain wall deformation detection model, input the multi-modal curtain wall data into the curtain wall deformation detection model, and generate a curtain wall deformation detection result; An alarm response module, configured to trigger a corresponding level of alarm response based on a three-level alarm response mechanism according to the curtain wall deformation detection result; A closed-loop optimization module, configured to correspondingly adjust the parameters of the curtain wall deformation detection model based on a model closed-loop optimization mechanism.

[0054] Through the introduction of the above embodiments, the present invention can, through the building curtain wall deformation degree detection method and system, obtain multi-modal curtain wall data of the target building curtain wall based on a multi-modal data acquisition network; construct a curtain wall deformation detection model, input the multi-modal curtain wall data into the curtain wall deformation detection model to generate a curtain wall deformation detection result; based on a three-level alarm response mechanism, trigger an alarm response at the corresponding level according to the curtain wall deformation detection result; and based on a model closed-loop optimization mechanism, make corresponding adjustments to the parameters of the curtain wall deformation detection model, so as to comprehensively, accurately and real-time detect and warn of the deformation degree of the building curtain wall, and improve the safety and reliability of the building curtain wall.

[0055] By adopting a multi-modal data acquisition network, the present invention can comprehensively monitor the curtain wall, greatly improving the early detection rate of curtain wall deformation. By constructing a curtain wall deformation detection model, the present invention can automatically and accurately detect the degree of curtain wall deformation, greatly reducing the need for manual intervention and improving the curtain wall detection efficiency. The present invention adopts a hierarchical response alarm mechanism, which can significantly reduce the energy consumption of the detection system, improve the system response speed, and optimize the system performance and resource allocation. By means of a closed-loop optimization mechanism, the present invention can achieve continuous evolution of the model performance, enhance the adaptability and reliability of the model, ensure the continuous operation ability and instant response characteristics of the detection system, and strengthen the safety prevention level.

[0056] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0057] Those of ordinary skill in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disc memories, tape memories, or any other medium that can be used to carry or store data and is computer-readable.

[0058] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the element.

Claims

1. A method for detecting the deformation degree of a building curtain wall, characterized in that, The detection method includes: Based on a multi-modal data acquisition network, obtaining multi-modal curtain wall data of the target building curtain wall; Constructing a curtain wall deformation detection model, inputting the multi-modal curtain wall data into the curtain wall deformation detection model to generate a curtain wall deformation detection result; Based on a three-level alarm response mechanism, triggering an alarm response at the corresponding level according to the curtain wall deformation detection result; Based on a model closed-loop optimization mechanism, correspondingly adjusting the parameters of the curtain wall deformation detection model.

2. The method for detecting the deformation degree of a building curtain wall according to claim 1, characterized in that, The multi-modal data acquisition network includes a multi-view camera array, an infrared thermal imaging acquisition device, and an acceleration sensor; The multi-view camera array, the infrared thermal imaging acquisition device, and the acceleration sensor are respectively used to obtain a visible light curtain wall image, infrared thermal imaging data, and a deformation vibration signal of the target building curtain wall; Among them, the multi-modal curtain wall data includes the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal.

3. A method for detecting the deformation degree of a building curtain wall according to claim 2, characterized in that, Based on the multi-view camera array to obtain a visible light curtain wall image of the target building curtain wall, specifically including: According to a preset scanning spacing, using a laser scanning device to collect point cloud image data on the surface of the target building curtain wall; Adopting a surface fitting algorithm, constructing a three-dimensional digital twin model corresponding to the target building curtain wall based on the point cloud image data; Taking the lowest overlap rate of the field of view angles of adjacent cameras and the largest curtain wall coverage as the optimization goal, based on the particle swarm optimization algorithm, calculating the fitness function of the particle swarm in the three-dimensional digital twin model, and iteratively updating the particle velocity and position until the optimization goal is met to obtain the camera positions on the surface of the three-dimensional digital twin model; For the camera positions that need to perform global detection, install a fish-eye lens to obtain a global curtain wall image of the target building curtain wall. For the camera positions that need to perform local detection, install a telephoto lens to obtain a local curtain wall image of the target building curtain wall; Summarize the global curtain wall image and the local curtain wall image to generate the visible light curtain wall image.

4. A method for detecting the deformation degree of a building curtain wall according to claim 2, characterized in that, Before inputting the multi-modal curtain wall data into the curtain wall deformation recognition model, perform preprocessing operations on the multi-modal curtain wall data, specifically including: Performing grayscale and normalization processing on the visible light curtain wall image; Performing temperature calibration and noise reduction processing on the infrared thermal imaging data; Performing filtering processing on the deformation vibration signal; Performing feature scaling on the preprocessed visible light curtain wall image, infrared thermal imaging data, and deformation vibration signal.

5. A method for detecting the deformation degree of a building curtain wall according to claim 1, characterized in that, The curtain wall deformation recognition model includes an input layer, an intermediate layer, and an output layer; Based on the input layer, splicing the visible light curtain wall image, the infrared thermal imaging data, and the deformation vibration signal to generate the multi-modal curtain wall data; Performing feature extraction on the multi-modal curtain wall data through a dual-channel convolution module in the intermediate layer to generate intermediate curtain wall features; Based on the adaptive threshold decision module in the output layer, performing temporal weighted voting on the intermediate curtain wall features to obtain the curtain wall deformation detection result.

6. The method for detecting the deformation degree of a building curtain wall according to claim 5, characterized in that, Performing feature extraction on the multi-modal curtain wall data through the dual-channel convolution module of the intermediate layer to generate intermediate curtain wall features, specifically including: Performing a convolution operation on the visible light curtain wall image and the infrared thermal imaging data through the convolution kernel of the first channel in the dual-channel convolution module, and extracting the curtain wall spatial features; Performing a convolution operation on the time series corresponding to the deformation vibration signal through the one-dimensional convolutional layer of the second channel in the dual-channel convolution module, and extracting the curtain wall time features; Splicing the curtain wall spatial features and the curtain wall time features to obtain the intermediate curtain wall features.

7. A method for detecting the deformation degree of a building curtain wall according to claim 1, characterized in that, Based on the three-level alarm response mechanism, triggering the corresponding level of alarm response according to the curtain wall deformation detection result, specifically including: Determining the triggering conditions of each level of alarm in the three-level alarm response mechanism based on the material, structure, and service life of the target building curtain wall; When the curtain wall deformation detection result meets the triggering condition of the first-level alarm, starting the edge node, adopting a lightweight algorithm, performing first-level processing on the curtain wall deformation detection result, generating first-level alarm response information, and sending it to the management end; When the curtain wall deformation detection result meets the triggering condition of the second-level alarm, starting the regional server cluster, adopting a distributed computing framework, performing second-level processing on the curtain wall deformation detection result, generating second-level alarm response information, and sending it to the management end; When the curtain wall deformation detection result meets the triggering condition of the third-level alarm, calling the cloud supercomputer, adopting a numerical simulation method, performing third-level processing on the curtain wall deformation detection result, generating third-level alarm response information, and sending it to the management end.

8. A method for detecting the deformation degree of a building curtain wall according to claim 7, characterized in that, Based on the reinforcement learning algorithm, dynamically adjusting the resource allocation ratio of the edge node, the regional server cluster, and the cloud supercomputer, minimizing the energy consumption of the three-level alarm response mechanism, and generating an optimal mechanism resource allocation strategy.

9. A building curtain wall deformation degree detection system, applied to a building curtain wall deformation degree detection method according to any one of claims 1-8, characterized in that, The detection system includes: A data acquisition module for acquiring multi-modal curtain wall data of a target building curtain wall based on a multi-modal data acquisition network; A deformation detection module for constructing a curtain wall deformation detection model, inputting the multi-modal curtain wall data into the curtain wall deformation detection model, and generating a curtain wall deformation detection result; An alarm response module for triggering the corresponding level of alarm response based on the three-level alarm response mechanism according to the curtain wall deformation detection result; A closed-loop optimization module for correspondingly adjusting the parameters of the curtain wall deformation detection model based on the model closed-loop optimization mechanism.

Images