Large model-based risk building identification method, system and equipment and medium

Abstract

The invention discloses a large model-based risk building identification method, system and equipment and a medium, belongs to the technical field of large model data training and risk building identification, and aims to solve the technical problem of how to automatically identify and evaluate the risk of a building and greatly improve the accuracy, efficiency and coverage of risk building identification. According to the technical scheme, the method comprises the steps of data collection and preprocessing: collecting data related to building safety, and preprocessing the data related to the building safety to obtain preprocessed data; feature engineering and data processing: carrying out feature extraction on the preprocessed data through a convolutional neural network, constructing a feature vector, and carrying out normalization and standardization processing on the feature vector to obtain normalized and standardized data; training the model; evaluating and optimizing the model; model deployment and reasoning; and subsequently monitoring and updating.

Description

Technical Field

[0001] The present invention relates to the technical field of large model data training and risky building identification, and in particular to a large model-based risky building identification method, system, equipment and medium. Background Art

[0002] Building risk identification is an important task in construction engineering and urban management. Over time, buildings will experience a variety of influences such as natural aging, earthquakes, weathering, and water intrusion, and may have problems such as structural cracks, material fatigue, foundation settlement, wall cracks, and water leakage. If these problems are not discovered and handled in a timely manner, they may lead to serious safety accidents such as building collapse, fire, or other safety hazards. Traditional building risk inspections usually rely on manual inspections, expert assessments, and infrastructure testing, which are inefficient and easily affected by subjective factors.

[0003] Therefore, how to automatically identify and assess the risks of buildings and significantly improve the accuracy, efficiency and coverage of risky building identification is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The technical task of the present invention is to provide a large-model-based risk building identification method, system, equipment and medium to solve the problem of how to automatically identify and evaluate the risks of buildings and greatly improve the accuracy, efficiency and coverage of risk building identification.

[0005] The technical task of the present invention is achieved in the following way: a risk building identification method based on a large model, the method is as follows:

[0006] Data collection and preprocessing: Collect data related to building safety, preprocess the data related to building safety, and obtain the preprocessed data;

[0007] Feature engineering and data processing: Use convolutional neural network (CNN) to extract features from preprocessed data, construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data;

[0008] Training model: Use the preprocessed data to train the convolutional neural network model, optimize the model through loss functions (such as cross entropy, mean square error, etc.), and adjust the hyperparameters of learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set;

[0009] Model evaluation and optimization: Use the validation set to evaluate and optimize the convolutional neural network model;

[0010] Model deployment and reasoning: The trained convolutional neural network model is optimized through model quantization to reduce the demand for computing resources in actual deployment, and deployed to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, reasoning is performed based on data related to building safety (sensor data and image data) to determine whether the building has potential dangers.

[0011] Subsequent monitoring and updating: The convolutional neural network model is continuously monitored to collect new data and update and fine-tune it based on the new data. The model is also regularly updated based on the aging of the building and changes in the external environment.

[0012] Preferably, the data related to building safety includes building structure data, monitoring data, historical data, image data and sensor data;

[0013] Among them, building structure data includes the age, material, number of floors and architectural design drawings of the building; monitoring data includes crack monitoring data, displacement data, vibration data, temperature and humidity of the building; historical data includes past building accidents, disaster data and natural disaster records; image data includes images and videos of the building's exterior, including high-definition images taken by drones and thermal imaging and crack detection images of the interior of the building; sensor data includes sensor data deployed inside and outside the building; sensor data includes pressure data collected by pressure sensors, temperature data collected by temperature sensors, and displacement data collected by displacement sensors.

[0014] Preferably, data collection and preprocessing are as follows:

[0015] Collect image data: Collect image datasets containing building information to ensure the angle, coordinates, lighting conditions, background diversity, longitude and latitude, and weather conditions of the buildings in the images to improve the generalization ability of the model;

[0016] Data cleaning and data labeling: Clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to the actual situation, so as to provide accurate labels for subsequent supervised learning;

[0017] Data enhancement: Use rotation, scaling, color transformation and flipping data enhancement techniques to increase the number of samples, expand the data set and increase the diversity of the data.

[0018] Preferably, the location and category information of the buildings in the data are accurately marked according to the actual situation. The specific situation is as follows:

[0019] For image data, mark areas with cracks and settlement problems;

[0020] For sensor data, normal and abnormal states need to be marked.

[0021] Preferably, feature engineering and data processing are as follows:

[0022] Feature extraction: For image data, convolutional neural network (CNN) is used to extract features; for sensor data, feature vectors are constructed using statistical features such as mean, standard deviation, maximum and minimum values.

[0023] Data normalization and standardization: Normalize and standardize the image data after feature extraction and the sensor data for constructing feature vectors;

[0024] Time series processing: Use the long short-term memory network (LSRM) to capture temporal changes in the normalized and standardized sensor data to obtain time series data.

[0025] Preferably, the model evaluation and optimization are as follows:

[0026] Overfitting and underfitting: During training, use data augmentation and L2 regularization to reduce overfitting;

[0027] Model interpretability: SHAP is used to explain the decision-making process of the model for building safety issues and obtain the most critical features for judging dangerous buildings.

[0028] A risk building identification system based on a large model, the system comprising:

[0029] A data collection and preprocessing module is used to collect data related to building safety, preprocess the data related to building safety, and obtain the preprocessed data;

[0030] The feature engineering and data processing module is used to extract features from the preprocessed data through a convolutional neural network (CNN), construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data;

[0031] The model training module is used to train the convolutional neural network model using preprocessed data, optimize the model through loss functions (such as cross entropy, mean square error, etc.), and adjust the hyperparameters of learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set;

[0032] Model evaluation and optimization module, used to evaluate and optimize the convolutional neural network model using the validation set;

[0033] The model deployment and reasoning module is used to optimize the trained convolutional neural network model through model quantization to reduce the demand for computing resources during actual deployment, and deploy it to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, it can make inferences based on data related to building safety (sensor data and image data) to determine whether the building has potential dangers.

[0034] The subsequent monitoring and updating module is used to continuously monitor the convolutional neural network model, collect new data, update and fine-tune it based on the new data, and regularly update the model according to the aging of the building and changes in the external environment.

[0035] Preferably, the data collection and preprocessing module includes:

[0036] The image data collection submodule is used to collect image datasets containing building information, ensuring the angles, coordinates, lighting conditions, background diversity, longitude and latitude, and weather conditions of the buildings in the images to improve the generalization ability of the model;

[0037] The data cleaning and data labeling submodule is used to clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to the actual situation, so as to provide accurate labels for subsequent supervised learning; for image data, the areas with cracks and settlement problems are marked; for sensor data, the normal and abnormal states need to be marked;

[0038] The data enhancement submodule is used to increase the number of samples using rotation, scaling, color transformation and flipping data enhancement techniques to expand the data set and increase the diversity of the data;

[0039] Feature engineering and data processing modules include:

[0040] The feature extraction submodule is used to extract features from image data using a convolutional neural network (CNN), obtain statistical features of sensor data using mean, standard deviation, maximum and minimum values, and construct feature vectors;

[0041] The data normalization and standardization submodule is used to normalize and standardize the image data after feature extraction and the sensor data for constructing feature vectors;

[0042] The time series processing submodule uses the long short-term memory network (LSRM) to capture the temporal changes of the normalized and standardized sensor data and obtain time series data.

[0043] An electronic device comprising: a memory and at least one processor;

[0044] Wherein, the memory stores computer-executable instructions;

[0045] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the risk building identification method based on a large model as described above.

[0046] A computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the above-mentioned large-model-based risk building identification method is implemented.

[0047] The large-model-based risk building identification method, system, device and medium of the present invention have the following advantages:

[0048] (1) The process of using the AI ​​big model to judge dangerous buildings in the present invention includes data collection and processing, feature engineering, model training and evaluation, deployment and reasoning, and continuous monitoring and updating, effectively using AI technology to predict and evaluate the safety of buildings, helping to discover potential dangers in advance and reduce the risk of accidents;

[0049] (ii) The present invention improves building safety: By intelligently identifying and analyzing buildings through AI big models, structural problems, aging or damage risk points in buildings can be discovered in a timely manner; for example, AI models can identify potential safety hazards such as wall cracks, floor subsidence, foundation settlement, etc., so that repair measures can be taken in advance to avoid the occurrence of building accidents;

[0050] (III) The present invention realizes automated risk assessment: Traditional building risk assessment often relies on manual inspection, which is not only time-consuming but also has the risk of missed inspections. The present invention can automatically analyze data from various aspects of building design, construction, maintenance, etc., and conduct a comprehensive assessment, thereby achieving more efficient and comprehensive risk identification and prediction;

[0051] (IV) The present invention realizes post-disaster recovery and emergency response: After a natural disaster or emergency occurs, the building risk identification technology based on the AI ​​big model can help quickly assess the risk level of damaged buildings and provide accurate guidance for post-disaster recovery work, which not only helps to reduce casualties but also accelerates post-disaster reconstruction work;

[0052] (V) This invention optimizes building design and construction: During the building design and construction phase, AI can analyze the impact of different materials, construction methods, and environmental conditions on building safety, identify potential design flaws or construction problems in advance, and reduce the possibility of future risks by optimizing design solutions;

[0053] (VI) The present invention realizes a long-term monitoring and early warning system: by combining sensors, drones and other technologies, the AI ​​big model can monitor buildings over a long period of time, analyze the building status in real time, promptly identify potential structural problems, and issue prompts through the early warning system to help managers take preventive measures;

[0054] (VII) The present invention saves resources and costs: AI big models can efficiently analyze a large amount of building data, thereby reducing unnecessary manual inspections and excessive maintenance, reducing the cost and resource waste of building management; at the same time, it can optimize maintenance cycles and capital use, and improve the economy and benefits of building management;

[0055] (VIII) The present invention promotes sustainable urban development: With the acceleration of urbanization, the aging, maintenance and safety management of buildings have become challenges for urban development; the application of AI big models can help urban managers better understand the distribution of building risks and identify dangerous building areas in advance, thereby improving the safety of urban infrastructure and promoting sustainable urban development;

[0056] (IX) The present invention can accurately identify structural problems and potential risks of buildings by training deep learning models; through comprehensive analysis of building images, design drawings and historical data, the system can detect early risk signs such as subtle cracks and tilts, greatly improving the accuracy of risk identification;

[0057] (10) The large-model-based automated identification method of the present invention can process a large amount of building image data in a short time, automatically identify potential risk points, and significantly improve the efficiency of building inspections; it has obvious advantages, especially in the monitoring of large-scale urban construction and old buildings; it reduces the need for manual inspections and reduces labor costs; at the same time, early detection of building hidden dangers and structural problems can avoid larger-scale losses and maintenance costs, reducing long-term operating costs, and can not only identify current risks, but also predict potential risk areas and provide early warnings by analyzing historical building data and environmental factors; the construction party can take preventive measures before problems occur, thereby avoiding the occurrence of safety accidents;

[0058] (XI) The present invention can provide more accurate and timely data support for building management departments by automatically identifying risky buildings, helping decision makers to quickly understand the risk status of various types of buildings, so as to better allocate resources and determine maintenance priorities. It can be integrated with intelligent building management systems to provide real-time monitoring and data analysis, helping urban builders, property companies and other relevant departments to more efficiently manage and maintain buildings, and improve the overall safety and sustainability of buildings through data-driven intelligent decision-making;

[0059] (XII) The present invention can conduct a comprehensive assessment of old buildings, help urban planners identify building groups with potential risks, and formulate reasonable repair or demolition plans to ensure the safety and stability of urban development. It has strong adaptability and can identify various types of building risks, whether modern high-rise buildings or old buildings left over from history, and can effectively identify potential problems therein. The model can be optimized as data continues to accumulate, improving the ability to identify different building structures and materials; at the same time, through automated and accurate risk identification, the public's confidence in building safety will be greatly enhanced; especially in public places and areas with high-density crowds, timely discovery and repair of building hazards will help protect the lives and property of citizens. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The present invention is further described below in conjunction with the accompanying drawings.

[0061] Attached Figure 1 This is a flowchart of the risk building identification method based on a large model. DETAILED DESCRIPTION

[0062] The large-model-based risk building identification method, system, device and medium of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments of the specification.

[0063] Embodiment 1:

[0064] As attached Figure 1 As shown, this embodiment provides a risk building identification method based on a large model, and the method is specifically as follows:

[0065] S1. Data collection and preprocessing: collecting data related to building safety, preprocessing the data related to building safety, and obtaining the preprocessed data;

[0066] S2. Feature engineering and data processing: extract features from the preprocessed data through a convolutional neural network (CNN), construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data;

[0067] S3. Training model: Use the preprocessed data to train the convolutional neural network model, optimize the model through loss functions (such as cross entropy, mean square error, etc.), and adjust the hyperparameters of learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set;

[0068] S4. Model evaluation and optimization: Use the validation set to evaluate and optimize the convolutional neural network model;

[0069] S5. Model deployment and reasoning: The trained convolutional neural network model is optimized through model quantization to reduce the demand for computing resources in actual deployment, and deployed to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, reasoning is performed based on data related to building safety (sensor data and image data) to determine whether the building has potential dangers.

[0070] S6. Subsequent monitoring and updating: Continuously monitor the convolutional neural network model, collect new data, update and fine-tune it based on the new data, and regularly update the model based on the aging of the building and changes in the external environment.

[0071] The data related to building safety in step S1 of this embodiment includes building structure data, monitoring data, historical data, image data and sensor data;

[0072] Among them, building structure data includes the age, material, number of floors and architectural design drawings of the building; monitoring data includes crack monitoring data, displacement data, vibration data, temperature and humidity of the building; historical data includes past building accidents, disaster data and natural disaster records; image data includes images and videos of the building's exterior, including high-definition images taken by drones and thermal imaging and crack detection images of the interior of the building; sensor data includes sensor data deployed inside and outside the building; sensor data includes pressure data collected by pressure sensors, temperature data collected by temperature sensors, and displacement data collected by displacement sensors.

[0073] The data collection and preprocessing in step S1 of this embodiment are specifically as follows:

[0074] S101. Collect image data: Collect image datasets containing building information to ensure the angle, coordinates, lighting conditions, background diversity, longitude and latitude, and weather conditions of the buildings in the images to improve the generalization ability of the model;

[0075] S102, Data cleaning and data labeling: Clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to the actual situation, so as to provide accurate labels for subsequent supervised learning;

[0076] S103, Data enhancement: Use rotation, scaling, color transformation and flipping data enhancement techniques to increase the number of samples, so as to expand the data set and increase the diversity of the data.

[0077] In step S102 of this embodiment, the location and category information of the buildings in the data are accurately marked according to the actual situation. The specific situation is as follows:

[0078] For image data, mark areas with cracks and settlement problems;

[0079] For sensor data, normal and abnormal states need to be marked.

[0080] The feature engineering and data processing in step S2 of this embodiment are specifically as follows:

[0081] S201, feature extraction: for image data, use convolutional neural network (CNN) to extract features; for sensor data, use statistical features of mean, standard deviation, maximum and minimum values ​​to construct feature vectors;

[0082] S202, data normalization and standardization: normalizing and standardizing the image data after feature extraction and the sensor data for constructing feature vectors;

[0083] S203, time series processing: Use the long short-term memory network LSRM to capture the temporal changes of the sensor data after normalization and standardization to obtain time series data.

[0084] The model evaluation and optimization in step S4 of this embodiment are specifically as follows:

[0085] S401, Overfitting and Underfitting: During the training process, use data enhancement and L2 regularization to reduce overfitting;

[0086] S402, Model Interpretability: For building safety issues, SHAP is used to explain the decision-making process of the model and obtain the most critical features for judging dangerous buildings.

[0087] Embodiment 2:

[0088] This embodiment provides a risk building identification system based on a large model, the system comprising:

[0089] A data collection and preprocessing module is used to collect data related to building safety, preprocess the data related to building safety, and obtain the preprocessed data;

[0090] The feature engineering and data processing module is used to extract features from the preprocessed data through a convolutional neural network (CNN), construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data;

[0091] The model training module is used to train the convolutional neural network model using preprocessed data, optimize the model through loss functions (such as cross entropy, mean square error, etc.), and adjust the hyperparameters of learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set;

[0092] Model evaluation and optimization module, used to evaluate and optimize the convolutional neural network model using the validation set;

[0093] The model deployment and reasoning module is used to optimize the trained convolutional neural network model through model quantization to reduce the demand for computing resources during actual deployment, and deploy it to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, it can make inferences based on data related to building safety (sensor data and image data) to determine whether the building has potential dangers.

[0094] The subsequent monitoring and updating module is used to continuously monitor the convolutional neural network model, collect new data, update and fine-tune it based on the new data, and regularly update the model according to the aging of the building and changes in the external environment.

[0095] The data collection and preprocessing module in this embodiment includes:

[0096] The image data collection submodule is used to collect image datasets containing building information, ensuring the angles, coordinates, lighting conditions, background diversity, longitude and latitude, and weather conditions of the buildings in the images to improve the generalization ability of the model;

[0097] The data cleaning and data labeling submodule is used to clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to the actual situation, so as to provide accurate labels for subsequent supervised learning; for image data, the areas with cracks and settlement problems are marked; for sensor data, the normal and abnormal states need to be marked;

[0098] The data enhancement submodule is used to increase the number of samples by using rotation, scaling, color transformation and flipping data enhancement techniques to expand the data set and increase the diversity of the data.

[0099] The feature engineering and data processing modules in this embodiment include:

[0100] The feature extraction submodule is used to extract features from image data using a convolutional neural network (CNN), obtain statistical features of sensor data using mean, standard deviation, maximum and minimum values, and construct feature vectors;

[0101] The data normalization and standardization submodule is used to normalize and standardize the image data after feature extraction and the sensor data for constructing feature vectors;

[0102] The time series processing submodule uses the long short-term memory network (LSRM) to capture the temporal changes of the normalized and standardized sensor data and obtain time series data.

[0103] Embodiment 3:

[0104] This embodiment also provides an electronic device, including: a memory and at least one processor;

[0105] Wherein, the memory stores computer-executable instructions;

[0106] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the large model-based risk building identification method according to any one of the present inventions.

[0107] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor may be a microprocessor or any conventional processor, etc.

[0108] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the terminal, etc. In addition, the memory can also include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, at least one disk storage period, a flash memory device, or other volatile solid-state storage devices.

[0109] Embodiment 4:

[0110] This embodiment also provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions are loaded by a processor, so that the processor executes the large model-based risk building identification method in any embodiment of the present invention. Specifically, a system or device equipped with a storage medium can be provided, on which a software program code that implements the functions of any of the above embodiments is stored, and a computer (or CPU or MPU) of the system or device reads and executes the program code stored in the storage medium.

[0111] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.

[0112] The storage medium embodiments for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer by a communication network.

[0113] In addition, it should be clear that the functions of any of the above embodiments can be implemented not only by executing the program code read by the computer, but also by enabling an operating system operating on the computer to complete part or all of the actual operations based on instructions from the program code.

[0114] In addition, it can be understood that the program code read from the storage medium is written to a memory provided in an expansion board inserted into the computer or written to a memory provided in an expansion unit connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or the expansion unit is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above-mentioned embodiments.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A risk building identification method based on a large model, characterized in that: The method is as follows: Data collection and preprocessing: Collect data related to building safety, preprocess the data related to building safety, and obtain the preprocessed data; Feature engineering and data processing: Use convolutional neural networks to extract features from preprocessed data, construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data; Training model: Use the preprocessed data to train the convolutional neural network model, optimize the model through the loss function, and adjust the hyperparameters of learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set; Model evaluation and optimization: Use the validation set to evaluate and optimize the convolutional neural network model; Model deployment and reasoning: The trained convolutional neural network model is optimized through model quantization and deployed to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, reasoning is performed based on data related to building safety to determine whether the building has potential dangers. Subsequent monitoring and updating: The convolutional neural network model is continuously monitored to collect new data and update and fine-tune it based on the new data. The model is also regularly updated based on the aging of the building and changes in the external environment.

2. The risk building identification method based on large model according to claim 1 is characterized in that: Data related to building safety include building structure data, monitoring data, historical data, and image data and sensor data; Among them, building structure data includes the age, material, number of floors and architectural design drawings of the building; monitoring data includes crack monitoring data, displacement data, vibration data, temperature and humidity of the building; historical data includes past building accidents, disaster data and natural disaster records; image data includes images and videos of the building's exterior, including high-definition images taken by drones and thermal imaging and crack detection images of the interior of the building; sensor data includes sensor data deployed inside and outside the building; sensor data includes pressure data collected by pressure sensors, temperature data collected by temperature sensors, and displacement data collected by displacement sensors.

3. The large model-based risk building identification method according to claim 1 or 2, characterized in that: The data collection and preprocessing are as follows: Collect image data: Collect image datasets containing building information, ensuring the angles, coordinates, lighting conditions, background diversity, longitude and latitude, and weather conditions of the buildings in the images; Data cleaning and data labeling: Clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to actual conditions; Data enhancement: Use rotation, scaling, color transformation and flipping data enhancement techniques to increase the number of samples, expand the data set and increase the diversity of the data.

4. The risk building identification method based on large model according to claim 3 is characterized in that: According to the actual situation, the location and category information of the buildings in the data are accurately marked. The specific situation is as follows: For image data, mark areas with cracks and settlement problems; For sensor data, normal and abnormal states need to be marked.

5. The risk building identification method based on large model according to claim 4 is characterized in that: Feature engineering and data processing are as follows: Feature extraction: For image data, convolutional neural network (CNN) is used to extract features; for sensor data, feature vectors are constructed using statistical features such as mean, standard deviation, maximum and minimum values. Data normalization and standardization: Normalize and standardize the image data after feature extraction and the sensor data for constructing feature vectors; Time series processing: Use the long short-term memory network (LSRM) to capture temporal changes in the normalized and standardized sensor data to obtain time series data.

6. The risk building identification method based on large model according to claim 5 is characterized in that: Model evaluation and optimization are as follows: Overfitting and underfitting: During training, use data augmentation and L2 regularization to reduce overfitting; Model interpretability: SHAP is used to explain the decision-making process of the model for building safety issues and obtain the most critical features for judging dangerous buildings.

7. A risk building identification system based on a large model, characterized in that: The system includes: A data collection and preprocessing module is used to collect data related to building safety, preprocess the data related to building safety, and obtain the preprocessed data; The feature engineering and data processing module is used to extract features from the preprocessed data through a convolutional neural network, construct feature vectors, and normalize and standardize the feature vectors to obtain normalized and standardized data; The model training module is used to train the convolutional neural network model using preprocessed data, optimize the model through the loss function, and adjust the hyperparameters of the learning rate, batch size, and number of network layers to ensure that the model can achieve the best performance on the validation set; Model evaluation and optimization module, used to evaluate and optimize the convolutional neural network model using the validation set; The model deployment and reasoning module is used to optimize the trained convolutional neural network model through model quantization and deploy it to the cloud platform to monitor the building in real time and predict possible dangers. At the same time, it can also make inferences based on data related to building safety to determine whether the building has potential dangers. The subsequent monitoring and updating module is used to continuously monitor the convolutional neural network model, collect new data, update and fine-tune it based on the new data, and regularly update the model according to the aging of the building and changes in the external environment.

8. The risk building identification system based on large model according to claim 7 is characterized in that: The data collection and preprocessing modules include: The image data collection submodule is used to collect image datasets containing building information, ensuring the angles, coordinates, lighting conditions and background diversity, longitude and latitude, and weather conditions of the buildings in the images; The data cleaning and data labeling submodule is used to clean the collected data, remove noise data, and accurately label the location and category information of the buildings in the data according to the actual situation; for image data, mark the areas with cracks and settlement problems; for sensor data, it is necessary to mark the normal and abnormal states; The data enhancement submodule is used to increase the number of samples using rotation, scaling, color transformation and flipping data enhancement techniques to expand the data set and increase the diversity of the data; Feature engineering and data processing modules include: The feature extraction submodule is used to extract features from image data using a convolutional neural network (CNN), obtain statistical features of sensor data using mean, standard deviation, maximum and minimum values, and construct feature vectors; The data normalization and standardization submodule is used to normalize and standardize the image data after feature extraction and the sensor data for constructing feature vectors; The time series processing submodule uses the long short-term memory network (LSRM) to capture the temporal changes of the normalized and standardized sensor data and obtain time series data.

9. An electronic device, characterized in that: include: memory and at least one processor; Wherein, the memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the large model-based risk building identification method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executable instructions, the large-model-based risk building identification method as described in any one of claims 1 to 6 is implemented.

Images