Method and device for identifying fire protection hidden danger through AI visual analysis technology

Abstract

The embodiment of the invention provides a method and device for identifying fire protection hidden dangers through an AI visual analysis technology, and the method and device achieve the integration of image video streaming, three-dimensional space, thermal imaging, environment perception and other multi-modal data through the innovative construction of a multi-source data collection and fusion mechanism. And designing a feature extraction model based on transfer learning, and realizing high-precision hidden danger feature recognition through integrated learning in combination with a hierarchical classification structure and an attention mechanism. Time sequence analysis and space positioning technologies are introduced, a hidden danger feature association network and an evolution model are constructed, and hidden danger development trend prediction and common hidden danger discovery are achieved. According to the method, the defects of the traditional technology in the aspects of multi-modal data processing, feature recognition, trend prediction and the like are effectively overcome, and the intelligent level and the early warning capability of fire protection hidden danger recognition are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and particularly to a method and device for identifying fire and fire prevention hazards through AI vision analysis technology. Background Art

[0002] Existing methods for identifying fire and fire prevention hazards have obvious deficiencies. Traditional systems have limitations in multi-source data collection and fusion, making it difficult to effectively integrate multi-modal data such as images, videos, depth information, thermal imaging, and acoustic signals, which affects the comprehensiveness of hazard identification.

[0003] In addition, existing technologies have bottlenecks in feature extraction and model training. Most systems fail to make full use of transfer learning and attention mechanisms, lacking in-depth exploration of the hierarchical relationship of features, resulting in low recognition accuracy.

[0004] Existing systems have technical shortcomings in hazard trend analysis. Lack of the ability to predict hazards based on temporal evolution, making it difficult to accurately predict the development trend of hazards, which affects the timeliness of preventive measures. Solving these problems is of great significance for improving the effect of fire and fire prevention hazard identification. Summary of the Invention

[0005] In view of the problems in the prior art, this application provides a method and device for identifying fire and fire prevention hazards through AI vision analysis technology, which can effectively solve the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improve the intelligent level and early warning ability of fire and fire prevention hazard identification.

[0006] To solve at least one of the above problems, this application provides the following technical solutions: In a first aspect, this application provides a method for identifying fire and fire prevention hazards through AI vision analysis technology, including: Collecting image and video stream data from security monitoring systems, mobile patrol terminals, and fire patrol robots, obtaining scene depth information through binocular cameras to construct a three-dimensional space model, collecting infrared thermal imaging data to obtain temperature distribution characteristics, collecting environmental perception signals to extract environmental characteristics, preprocessing the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental characteristics to obtain normalized training data, extracting multi-modal feature vectors from the normalized training data, annotating the multi-modal feature vectors to construct training samples, selecting a basic feature extraction layer from a pre-trained model library to construct a transfer learning model, inputting the training samples into the transfer learning model, training a fire hazard feature neural network model group based on a hierarchical classification structure and an attention mechanism, and performing ensemble learning on the recognition results of the fire hazard feature neural network model group to obtain a fusion recognition model; Receive multi-source data collected in real time, perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain hidden danger feature data, and analyze the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data; Perform spatial positioning on the hidden danger feature data to construct hidden danger marking data, construct a hidden danger feature association network based on the hidden danger marking data and the trend prediction data, establish a hidden danger evolution model using the hidden danger feature association network, predict the development trend of the hidden danger through the hidden danger evolution model, perform clustering analysis on the hidden danger feature data to discover common hidden dangers, and integrate the hidden danger marking data, the hidden danger development trend and the common hidden dangers into structured data and output it to the monitoring terminal.

[0007] Further, it also includes: performing convolutional decomposition and pooling operations on the normalized training data to extract image texture features, extracting three-dimensional spatial features through a depth image analysis module, obtaining temperature gradient features using an infrared feature extraction module, extracting acoustic features based on a spectrogram analysis module, fusing the image texture features, the three-dimensional spatial features, the temperature gradient features and the acoustic features to obtain a multi-modal feature vector, normalizing the feature components of the multi-modal feature vector and adding time tags, and performing class annotation and attribute annotation on the multi-modal feature vector based on manual annotation rules to construct training samples; Perform feature layer analysis on the models in the pre-trained model library, select representative basic feature extraction layers according to the application scenario, copy the weight parameters and network structure of the basic feature extraction layers to the target network to construct a transfer learning model, adjust the structure of the input layer and output layer of the transfer learning model to adapt to the data format of the training samples, and input the training samples into the adjusted transfer learning model for feature mapping.

[0008] Further, it also includes: constructing a hierarchical classification structure based on a preset fire hazard classification system, dividing the hazard types into multiple sub-category nodes according to physical characteristics, assigning corresponding feature recognition models to each sub-category node, embedding a self-attention calculation unit and a cross-modal attention calculation unit in the feature recognition models, using the self-attention calculation unit to perform importance weighting on the feature vectors, calculating the association weights between different modal features through the cross-modal attention calculation unit, and inputting the training samples into each feature recognition model for backpropagation training to obtain a group of fire hazard feature neural network models; Calculate the recognition accuracy and confidence scores of each model in the fire hazard feature neural network model group, assign fusion weight coefficients to each model based on the recognition accuracy and confidence scores, perform weighted combination on the recognition results of the model group and the corresponding fusion weight coefficients, perform decision fusion on the weighted combination results through a voting mechanism to obtain a fusion recognition model, and iteratively optimize the fusion weight coefficients of the fusion recognition model.

[0009] Further, it also includes: receiving data streams collected by multiple sensors, dividing the data streams into data segments through time window division, aligning and synchronizing the data segments according to timestamps, performing wavelet transform on the synchronized data to obtain time-frequency components, extracting the amplitude features and phase features of the time-frequency components, constructing a dynamic feature sequence by sequentially splicing the amplitude features and phase features based on the sliding window method, and performing normalization processing and noise reduction filtering on the dynamic feature sequence; Parse the dynamic feature sequence into feature tensors according to a preset time step, separate the channels of the feature tensors to obtain multiple feature components, input the feature components into the corresponding input ends of the fusion recognition model, calculate the recognition results and confidence levels of each feature component through the fusion recognition model, screen and combine the recognition results according to the confidence levels to obtain hazard feature data, and perform time series correlation analysis on the hazard feature data.

[0010] Further, it also includes: constructing a time series sample matrix for the dynamic feature sequence and the hazard feature data according to the time dimension, performing singular value decomposition on the time series sample matrix to extract principal component features, calculating the time increment and change rate of the principal component features, performing trend decomposition on the principal component features based on the change rate to obtain periodic components and trend components, reconstructing the periodic components and trend components into a time series curve, and performing piecewise fitting on the time series curve to calculate fitting parameters; Construct a state prediction model using the fitting parameters, input historical time series data into the state prediction model for parameter training, perform extrapolation calculation on the development trend of the hazard features based on the trained state prediction model, perform probability assessment on the hazard state at the predicted time point according to the extrapolation calculation results to obtain trend prediction data, and quantitatively evaluate the credibility of the trend prediction data.

[0011] Further, it also includes: performing target detection and localization on the hazard feature data to obtain bounding box coordinates, mapping the bounding box coordinates into a three-dimensional space model to obtain spatial position parameters, calculating the relative distance and azimuth angle between the spatial position parameters and a preset reference point, integrating the type label, the spatial position parameters, the relative distance, and the azimuth angle of the hazard feature data to construct hazard marking data, and performing spatial clustering on the hazard marking data according to the relative distance; Construct a node feature matrix based on the hidden danger marked data, construct a node transition probability matrix using the trend prediction data, combine the node feature matrix and the node transition probability matrix into a hidden danger feature association network, calculate the association strength and propagation path between nodes in the hidden danger feature association network, construct a state transition equation according to the association strength and the propagation path, and establish a hidden danger evolution model based on the state transition equation.

[0012] Further, it also includes: numerically solve the state transition equation of the hidden danger evolution model to obtain an evolution trajectory, calculate the development direction and change rate of the evolution trajectory, extrapolate the evolution trajectory according to a preset time interval to obtain a hidden danger state prediction value, compare the hidden danger state prediction value with the threshold value at the corresponding time point to obtain the hidden danger development trend, perform density clustering analysis on the hidden danger feature data to obtain feature clusters, calculate the centroid position and distribution range of the feature clusters, and identify common hidden dangers according to the number and scale of the feature clusters; Map the hidden danger marked data according to a preset data structure format, perform time serialization processing on the hidden danger development trend data, convert the common hidden danger data into an attribute relationship table, integrate the field mapping data, the time serialization data, and the attribute relationship table into structured data based on a unified data template, and transmit the structured data to a specified storage area of the monitoring terminal through a data interface.

[0013] In a second aspect, the present application provides a device for identifying fire and fire protection hazards through AI vision analysis technology, including: A model fusion module, which is used to collect image and video stream data from a security monitoring system, a mobile patrol terminal, and a fire patrol robot, obtain the depth information of the scene through a binocular camera to construct a three-dimensional space model, collect infrared thermal imaging data to obtain temperature distribution characteristics, collect environmental perception signals to extract environmental characteristics, preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental characteristics to obtain normalized training data, extract multi-modal feature vectors from the normalized training data, label the multi-modal feature vectors to construct training samples, select a basic feature extraction layer from a pre-trained model library to construct a transfer learning model, input the training samples into the transfer learning model, train a fire hazard feature neural network model group based on a hierarchical classification structure and an attention mechanism, and perform ensemble learning on the recognition results of the fire hazard feature neural network model group to obtain a fusion recognition model; A trend prediction module, configured to receive multi-source data collected in real time, perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain potential hazard feature data, and analyze the time evolution trend of the potential hazard feature data based on the dynamic feature sequence to obtain trend prediction data; A potential hazard recognition module, configured to perform spatial positioning on the potential hazard feature data to construct potential hazard marking data, construct a potential hazard feature association network based on the potential hazard marking data and the trend prediction data, establish a potential hazard evolution model by using the potential hazard feature association network, predict the development trend of potential hazards through the potential hazard evolution model, perform clustering analysis on the potential hazard feature data to discover common potential hazards, and integrate the potential hazard marking data, the potential hazard development trend, and the common potential hazards into structured data and output it to the monitoring terminal.

[0014] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the method for identifying fire and fire protection hazards through AI vision analysis technology are implemented.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for identifying fire and fire protection hazards through AI vision analysis technology are implemented.

[0016] In a fifth aspect, the present application provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the method for identifying fire and fire protection hazards through AI vision analysis technology are implemented.

[0017] As can be seen from the above technical solutions, the present application provides a method and device for identifying fire and fire protection hazards through AI vision analysis technology. By innovatively constructing a multi-source data acquisition and fusion mechanism, multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals are integrated. A feature extraction model based on transfer learning is designed, combined with a hierarchical classification structure and an attention mechanism, and high-precision potential hazard feature recognition is achieved through ensemble learning. Time series analysis and spatial positioning technologies are introduced to construct a potential hazard feature association network and an evolution model, realizing the prediction of the development trend of potential hazards and the discovery of common potential hazards. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire protection hazard identification. Description of the Drawings

[0018] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a schematic flowchart of the method for identifying fire and fire safety hazards through AI vision analysis technology in the embodiments of the present application; Figure 2 It is a structural diagram of the device for identifying fire and fire safety hazards through AI vision analysis technology in the embodiments of the present application; Figure 3 It is a schematic structural diagram of the electronic device in the embodiments of the present application.

[0020] Reference numerals: Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed implementation manners

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0022] In the technical solutions of the present application, the acquisition, storage, use, processing, etc. of data all comply with the relevant regulations of national laws and regulations.

[0023] In view of the problems existing in the prior art, the present application provides a method and device for identifying fire and fire prevention hazards through AI vision analysis technology. By innovatively constructing a multi-source data acquisition and fusion mechanism, multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals are integrated. A feature extraction model based on transfer learning is designed, combined with a hierarchical classification structure and an attention mechanism, and high-precision hazard feature recognition is achieved through ensemble learning. Time series analysis and spatial positioning technologies are introduced to construct a hazard feature correlation network and an evolution model to realize the prediction of the development trend of hazards and the discovery of common hazards. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire prevention hazard identification.

[0024] In order to effectively solve the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improve the intelligent level and early warning ability of fire and fire prevention hazard identification, an embodiment of a method for identifying fire and fire prevention hazards through AI vision analysis technology is provided in the present application. See Figure 1 , the method for identifying fire and fire prevention hazards through AI vision analysis technology specifically includes the following contents: Step S101: Collect image and video stream data from the security monitoring system, mobile patrol terminals, and fire patrol robots. Obtain the scene depth information through a binocular camera to construct a three-dimensional space model. Collect infrared thermal imaging data to obtain temperature distribution characteristics. Collect environmental perception signals to extract environmental features. Preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental features to obtain normalized training data. Extract multi-modal feature vectors from the normalized training data. Label the multi-modal feature vectors to construct training samples. Select a basic feature extraction layer from a pre-trained model library to construct a transfer learning model. Input the training samples into the transfer learning model. Train a group of fire hazard feature neural network models based on a hierarchical classification structure and an attention mechanism. Perform ensemble learning on the recognition results of the group of fire hazard feature neural network models to obtain a fusion recognition model; Optionally, in view of the problems such as the large limitation of a single data source, incomplete feature extraction, and unsatisfactory model training effect in the process of fire hazard recognition in this embodiment, a set of hazard recognition model training scheme based on multi-modal data fusion is innovatively designed. In this embodiment, a multi-source data acquisition network is first constructed, integrating multiple acquisition terminals such as fixed security monitoring cameras, patrol personnel's handheld terminals, and intelligent patrol robots to achieve full-round monitoring of key areas in the building. Collect image and video stream data through high-definition cameras to capture the real-time status of key parts such as fire corridors, safety exits, and fire-fighting facilities.

[0025] This embodiment optimizes the method for obtaining scene depth information. By using a binocular camera to collect a stereo image pair, a three-dimensional spatial structure is reconstructed based on the principle of parallax calculation: Depth = (f B) / d where f is the camera focal length, B is the binocular baseline distance, and d is the parallax of the corresponding point. This method can accurately obtain the spatial position and size information of an object, providing basic data for subsequent potential hazard spatial positioning.

[0026] This embodiment innovatively realizes the acquisition of temperature distribution characteristics. An infrared thermal imaging device is used to scan key areas to obtain temperature field distribution data. In the detection of electrical equipment, this method can effectively detect potential faults such as local overheating and poor contact. At the same time, temperature anomalies are also important early warning indicators for early fires.

[0027] This embodiment designs an environmental perception feature extraction method. For example, environmental acoustic feature extraction, where an environmental perception signal is collected through a high-sensitivity microphone array, and environmental features are extracted using spectrogram analysis. These features can reflect information such as the operating state of equipment and abnormal vibrations, providing important evidence for early detection of potential hazards.

[0028] This embodiment optimizes the data preprocessing process. Image data is denoised, enhanced, and normalized; 3D model data is subjected to point cloud filtering and mesh optimization; temperature data is calibrated and smoothed; acoustic data is denoised and spectroscopically analyzed. Through preprocessing, the quality and consistency of various types of data are ensured.

[0029] This embodiment demonstrates significant advantages in multi-modal feature extraction. Visual features such as texture, shape, and color of an image are extracted through a deep learning network; spatial structure features are extracted using a 3D convolutional network; temperature gradient features are extracted using heat map analysis; acoustic features are extracted through spectrogram analysis. These features describe the manifestation characteristics of fire hazards from different dimensions.

[0030] This embodiment innovatively designs a feature annotation strategy. A annotation system including multiple categories such as equipment failures, channel blockages, and illegal electricity use is established. For each type of potential hazard, its location, type, severity, and other attribute information are detailedly annotated. These annotation information provide accurate supervision signals for model training.

[0031] This embodiment optimizes the construction method of the transfer learning model. A model that performs excellently in tasks such as object detection and scene understanding is selected from a pre-trained model library, and the weights and structure of its basic feature layer are extracted. This method makes full use of the advantages of the pre-trained model in feature extraction, significantly reducing the training data requirements.

[0032] This embodiment innovatively implements a model training strategy. By adopting a hierarchical classification structure, the hidden danger identification task is decomposed into multiple specialized subtasks. An attention mechanism is introduced in each subtask to enhance the model's perception ability of key features. This divide-and-conquer method improves the recognition accuracy of the model.

[0033] This embodiment significantly improves the recognition effect through model ensemble learning. The prediction results of different sub-models are weighted and fused, comprehensively considering the advantages of each model. For example, when identifying electrical fire hazards, temperature features and image features have different importance, and the recognition accuracy is improved through reasonable weight allocation.

[0034] This embodiment establishes a complete model training system. From data collection to feature extraction, and then to model training and integration, a systematic training process is formed. This training system can adapt to the hidden danger identification needs of different scenarios and has good scalability.

[0035] This embodiment provides intelligent technical support for fire safety management. Through multi-modal data analysis and deep learning technology, automatic and intelligent identification of fire hazards is achieved. This method not only improves the accuracy of hazard detection but also reduces the workload of manual inspections.

[0036] This embodiment realizes the intelligent upgrade of hidden danger identification. Through multi-source data fusion and deep learning technology, hidden danger identification becomes more comprehensive and accurate. This intelligent identification method provides strong technical support for fire safety management.

[0037] Step S102: Receive multi-source data collected in real time, perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain hidden danger feature data, and analyze the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data; Optionally, in view of the problems that it is difficult to capture the dynamic evolution characteristics of fire hazards and the real-time recognition effect is not good, this embodiment innovatively designs a dynamic feature recognition scheme based on time series analysis. This embodiment first establishes a real-time reception mechanism for multi-source data, and synchronously collects video streams, depth information, thermal imaging data, and acoustic signals through a data bus. To ensure the real-time and integrity of the data, a buffer queue management mechanism is adopted to perform time alignment and synchronization processing on data streams with different sampling rates.

[0038] This embodiment optimizes the time series decomposition algorithm. The multi-source data is divided into time windows, and the window length is dynamically adjusted according to the characteristic time scales of different types of hidden dangers. For example, for electrical fire hidden dangers, the evolution period of abnormal temperature is relatively long, and a larger time window is required; while for sudden hidden dangers such as passage blockages, a smaller time window is needed to capture rapid changes.

[0039] This embodiment innovatively realizes multi-scale feature extraction. The time-frequency analysis of the signal is carried out through wavelet transform: WT(a,τ) = ∫ x(t)ψ((t-τ) / a)dt where x(t) is the input signal, ψ is the wavelet function, a is the scale parameter, and τ is the translation parameter. This method can effectively extract the feature changes on different time scales and provide rich time-frequency features for hidden danger identification.

[0040] This embodiment designs a method for constructing a dynamic feature sequence. The sliding window technique is adopted to sequentially splice the features extracted within consecutive time windows. The window sliding step size is set according to the actual monitoring requirements, which not only ensures the continuity of the features but also avoids the accumulation of redundant data. Through this method, the discrete feature observations are transformed into a continuous time series feature stream.

[0041] This embodiment optimizes the feature recognition strategy. The dynamic feature sequence is input into the trained fusion recognition model, and the attention mechanism inside the model can automatically capture the importance changes of different modal features. For example, in the night scene, the weight of the thermal imaging feature will be automatically increased; in the monitoring of the equipment operating state, the importance of the acoustic feature will increase accordingly.

[0042] This embodiment shows significant advantages in real-time hidden danger identification. Through the dynamic fusion of multi-modal features, the early signs of hidden dangers can be accurately captured. For example, in the monitoring of electrical equipment, potential equipment failures can be detected in advance through the comprehensive analysis of multi-dimensional features such as abnormal temperature, abnormal sound, and changes in the appearance of the equipment.

[0043] This embodiment innovatively designs a trend analysis method. The time series analysis is carried out on the identified hidden danger feature data to extract its evolution trend features. An autoregressive moving average model is used to model the feature sequence: X(t) = φ1X(t-1) +... + φpX(t-p) + ε(t) where X(t) is the time series feature value, φi is the autoregressive coefficient, and ε(t) is the random error term. The development trend of the feature can be predicted through this model.

[0044] This embodiment optimizes the generation strategy of prediction data. Based on the trend model trained with historical data, it predicts the development trend of potential hazard characteristics. The model takes into account factors such as seasonal changes and periodic fluctuations, improving the accuracy of prediction. For example, in the prediction of passageway blockage in crowded places, the model will consider the daily variation pattern of the pedestrian flow.

[0045] This embodiment innovatively realizes the credibility evaluation of prediction results. By calculating the prediction interval and confidence level, it quantitatively evaluates the reliability of prediction results. This evaluation mechanism provides a decision-making reference for managers and helps to reasonably arrange the priorities of inspection and rectification work.

[0046] This embodiment significantly improves the foresight of potential hazard identification through dynamic feature analysis and trend prediction. By analyzing and predicting the evolution law of potential hazards, it transforms the fire safety management from passive response to active prevention.

[0047] This embodiment establishes a complete dynamic monitoring system. From real-time data collection to feature extraction and then to trend prediction, it forms a systematic monitoring process. This dynamic monitoring system can timely detect potential hazards and provide a basis for preventive maintenance.

[0048] This embodiment provides an intelligent early warning means for fire safety management. Through real-time feature recognition and trend prediction, it realizes the early warning of fire hazards. This early warning mechanism not only improves the timeliness of hazard discovery but also enhances the initiative of safety management.

[0049] This embodiment realizes the intelligent upgrade of potential hazard identification. Through dynamic feature analysis and trend prediction, it makes potential hazard identification more timely and accurate. This intelligent identification method provides strong technical support for fire safety management.

[0050] Step S103: Conduct spatial positioning on the potential hazard characteristic data to construct potential hazard marking data, construct a potential hazard feature association network based on the potential hazard marking data and the trend prediction data, establish a potential hazard evolution model using the potential hazard feature association network, predict the development trend of potential hazards through the potential hazard evolution model, conduct clustering analysis on the potential hazard characteristic data to discover common potential hazards, and integrate the potential hazard marking data, the potential hazard development trend, and the common potential hazards into structured data and output it to the monitoring terminal.

[0051] Optionally, in view of the problems such as the complex spatial distribution of fire hazards, the difficulty in grasping the correlation between hazards, and the inaccurate prediction of development trends, this embodiment innovatively designs a hazard evolution prediction scheme based on spatial positioning and correlation analysis. First, this embodiment performs spatial positioning on the identified hazard feature data, and uses a deep learning object detection algorithm to calculate the bounding box coordinates of the hazard target. Combining with the three-dimensional space model constructed by the binocular camera, the two-dimensional coordinates are mapped into the three-dimensional space to obtain the spatial parameters of the hazard position.

[0052] This embodiment optimizes the spatial positioning method. A relative coordinate system is established based on a preset reference point, and the distance and azimuth angle between the hazard position and the reference point are calculated: Position = [r cos(θ) cos(φ), r cos(θ) sin(φ), r sin(θ)] where r is the relative distance, θ is the pitch angle, and φ is the azimuth angle. This positioning method is convenient for describing the spatial distribution characteristics of hazards.

[0053] This embodiment innovatively realizes the construction of hazard marked data. Information such as the type label, spatial position, and relative distance of the hazard is integrated into structured marked data. For example, for electrical fire hazards, the marked data includes multi-dimensional information such as equipment type, fault characteristics, temperature anomaly degree, and spatial position. These marked data provide a basis for subsequent correlation analysis.

[0054] This embodiment designs a method for constructing a feature correlation network. A node feature matrix is constructed based on the hazard marked data, and each node contains the attribute features of the hazard. The transition probability matrix between nodes is constructed using trend prediction data to describe the evolution law of the hazard state. Through the combination of the feature matrix and the transition matrix, a complete hazard feature correlation network is formed.

[0055] This embodiment optimizes the establishment strategy of the hazard evolution model. Calculate the correlation strength and propagation path between nodes in the correlation network, and construct a state transition equation: S(t+1) = T S(t) + ε(t) where S(t) is the state vector at time t, T is the transition matrix, and ε(t) is the random perturbation term. This model can describe the dynamic evolution process of the hazard state.

[0056] This embodiment has demonstrated significant advantages in potential hazard prediction. By numerically solving the evolution model, the evolution trajectory of the potential hazard state is obtained. Based on the development direction and change rate of the evolution trajectory, the development trend of the potential hazard can be predicted. For example, in the monitoring of electrical equipment, the model can predict the diffusion range and development speed of abnormal temperatures.

[0057] This embodiment innovatively designs a common potential hazard discovery method. Conduct density clustering analysis on potential hazard feature data, and discover feature clusters based on distance metrics in the feature space: Cluster_density = Σ(exp(-||xi - xj||² / 2σ²)) where xi and xj are feature vectors, and σ is the density parameter. By analyzing the distribution characteristics of the feature clusters, common potential hazard types are identified.

[0058] This embodiment optimizes the data integration strategy. Map the fields of potential hazard labeled data according to a unified data structure, perform time serialization processing on potential hazard development trend data, and convert common potential hazard data into an attribute relationship table. Integrate this information through a data template to form standardized structured data.

[0059] This embodiment innovatively implements a data output mechanism. Configure data interface parameters and establish a secure connection channel with the monitoring terminal. Transmit the structured data to the specified storage area of the monitoring terminal through the data interface to ensure the reliability and security of data transmission.

[0060] This embodiment significantly improves the scientific nature of potential hazard management through spatial analysis and evolution prediction. The analysis of potential hazard distribution based on spatial positioning helps optimize the inspection route and resource allocation. The prediction results of the evolution model provide a decision-making basis for potential hazard rectification.

[0061] This embodiment establishes a complete potential hazard analysis system. From spatial positioning to correlation analysis, and then to trend prediction, a systematic analysis process is formed. This analysis system can comprehensively grasp the distribution characteristics and development laws of potential hazards.

[0062] This embodiment provides scientific decision-making support for fire safety management. Through spatial analysis and trend prediction, precise management of fire hazards is achieved. This management method not only improves the pertinence of potential hazard treatment but also enhances the effectiveness of preventive measures.

[0063] This embodiment realizes the intelligent upgrade of potential hazard management. Through spatial positioning and evolution prediction, potential hazard management becomes more precise and efficient. This intelligent management method provides strong technical support for fire safety work.

[0064] As can be seen from the above description, the method for identifying fire and fire prevention hazards through AI vision analysis technology provided by the embodiments of the present application can integrally collect and fuse multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals by innovatively constructing a multi-source data collection and fusion mechanism. Design a feature extraction model based on transfer learning, combine a hierarchical classification structure and an attention mechanism, and achieve high-precision hazard feature recognition through ensemble learning. Introduce time series analysis and spatial positioning technologies, construct a hazard feature association network and an evolution model, and realize the prediction of hazard development trends and the discovery of common hazards. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire prevention hazard recognition.

[0065] In an embodiment of the method for identifying fire and fire prevention hazards through AI vision analysis technology of the present application, the following content may also be specifically included: Step S201: Perform convolution decomposition and pooling operations on the normalized training data to extract image texture features, extract three-dimensional space features through a depth image analysis module, obtain temperature gradient features using an infrared feature extraction module, extract acoustic features based on a spectrogram analysis module, fuse the image texture features, the three-dimensional space features, the temperature gradient features, and the acoustic features to obtain a multi-modal feature vector, normalize the feature components of the multi-modal feature vector and add time tags, and perform class labeling and attribute labeling on the multi-modal feature vector based on manual labeling rules to construct training samples; Step S202: Perform feature layer analysis on the models in the pre-trained model library, select representative basic feature extraction layers according to the application scenario, copy the weight parameters and network structure of the basic feature extraction layers to the target network to construct a transfer learning model, adjust the structure of the input layer and output layer of the transfer learning model to adapt to the data format of the training samples, and input the training samples into the adjusted transfer learning model for feature mapping.

[0066] Optionally, in view of the problems such as incomplete extraction of fire prevention hazard features, difficulty in multi-modal data fusion, and low model training efficiency in this embodiment, a deep learning training scheme based on multi-modal feature fusion is innovatively designed. In this embodiment, deep feature extraction is first performed on the normalized training data, and a multi-layer convolutional neural network is used to process the image data. The design of the convolutional kernel takes into account texture features of different scales. Small-sized convolutional kernels are used to extract local details, and large-sized convolutional kernels are used to capture global structures.

[0067] This embodiment optimizes the method for extracting three-dimensional spatial features. The depth information obtained by the binocular camera is processed by the depth image analysis module, and 3D reconstruction is achieved using the point cloud registration algorithm: Point_cloud = f(depth_map, camera_params), where depth_map is the depth map and camera_params are the camera parameters. This method can accurately reconstruct the spatial structure features of potential hazard targets.

[0068] This embodiment innovatively realizes the extraction of temperature gradient features. The thermal imaging data is analyzed by the infrared feature extraction module to calculate the spatial gradient of the temperature field:

[0069] where T is the temperature value, and x, y are the spatial coordinates. Through gradient analysis, temperature anomaly regions and heat transfer directions can be effectively identified.

[0070] This embodiment designs an acoustic feature extraction method. Based on the spectrogram analysis module to process acoustic signals, the short-time Fourier transform is used to obtain time-frequency features: STFT(t,f) = ∫ x(τ)w(τ - t)e^(-j2πfτ)dτ where x(τ) is the acoustic signal, w is the window function, t is the time, and f is the frequency. This analysis method can capture the abnormal features of the sound during device operation.

[0071] This embodiment optimizes the feature fusion strategy. The features of different modalities are adaptively weighted and fused: Feature_fusion = Σ(wi Feature_i) where wi is the feature weight, and Feature_i are the features of each modality. The weight coefficients are automatically optimized through backpropagation to make the fused features better represent the potential hazard features.

[0072] This embodiment shows significant advantages in the feature extraction process. Through the collaborative analysis of multi-modal features, the manifestation features of potential hazards can be comprehensively captured. For example, in the monitoring of electrical equipment, by combining multi-dimensional features such as temperature anomalies, abnormal operation sounds, and appearance changes, equipment failures can be accurately identified.

[0073] This embodiment innovatively designs an annotation rule system. An annotation rule including multiple dimensions such as equipment type, fault type, and danger level is established. For each feature vector, its category attributes and time information are detailedly annotated to provide complete supervision information for model training.

[0074] This embodiment optimizes the transfer learning strategy. A model that performs excellently in tasks such as object detection and scene understanding is selected from the pre-trained model library as the base network. Through feature layer parsing, a base feature extraction layer with good generalization ability is extracted.

[0075] This embodiment innovatively realizes the adjustment of the model structure. The input layer and output layer of the transfer learning model are reconstructed to adapt to the data format of multi-modal features. Multiple feature channels are designed in the input layer to receive feature data of different modalities respectively; the output layer is adjusted according to the requirements of the hidden danger classification task.

[0076] This embodiment significantly improves the model training effect through feature mapping optimization. Utilizing the feature extraction ability of the pre-trained model reduces the computational overhead of training from scratch. At the same time, fine-tuning is performed to ensure that the model can adapt to the specific requirements of fire hazard identification.

[0077] This embodiment establishes a complete feature learning system. From multi-modal feature extraction to feature fusion and then to transfer learning, a systematic feature learning process is formed. This learning system can effectively extract and utilize the feature information of hidden dangers.

[0078] This embodiment provides an intelligent feature analysis method for fire safety management. Through deep learning technology, the automatic extraction and analysis of fire hazard features are realized. This analysis method not only improves the comprehensiveness of feature extraction but also enhances the efficiency of model training.

[0079] This embodiment realizes the intelligent upgrade of feature analysis. Through multi-modal feature fusion and transfer learning, feature extraction becomes more accurate and efficient. This intelligent analysis method provides strong technical support for fire safety management.

[0080] In an embodiment of the method for identifying fire hazards through AI vision analysis technology in this application, the following content may also be specifically included: Step S301: Construct a hierarchical classification structure based on a preset fire hazard classification system, divide the hazard types into multiple sub-category nodes according to physical characteristics, assign corresponding feature recognition models to each sub-category node, embed a self-attention calculation unit and a cross-modal attention calculation unit in the feature recognition models, use the self-attention calculation unit to perform importance weighting on the feature vectors, calculate the correlation weights between different modal features through the cross-modal attention calculation unit, and input the training samples into each feature recognition model for backpropagation training to obtain a group of fire hazard feature neural network models; Step S302: Calculate the recognition accuracy and confidence scores of each model in the fire hazard feature neural network model group, assign fusion weight coefficients to each model based on the recognition accuracy and confidence scores, perform weighted combination on the recognition results of the model group and the corresponding fusion weight coefficients, and perform decision fusion on the weighted combination results through a voting mechanism to obtain a fusion recognition model, and iteratively optimize the fusion weight coefficients of the fusion recognition model.

[0081] Optionally, in view of the problems such as complex fire hazard types, low feature recognition accuracy, and poor model fusion effect in this embodiment, an innovative model training scheme based on hierarchical classification and attention mechanism is designed. In this embodiment, a hierarchical hazard classification system is first constructed, and hazards are scientifically classified according to physical characteristics. For example, electrical fire hazards can be further divided into subcategories such as wire aging, poor contact, and overloading; flammable and explosive hazards can be further divided into subcategories such as gas leakage, illegal storage, and non-compliant environment.

[0082] This embodiment optimizes the allocation strategy of the feature recognition model. For each subcategory node, a dedicated feature recognition model is designed. The network structure of each model is optimized according to the complexity of the hazard features to ensure that the model has appropriate feature extraction capabilities. For example, for temperature anomaly detection, the model focuses on thermal imaging features; for channel blockage detection, it focuses on spatial feature analysis.

[0083] This embodiment innovatively implements a self-attention mechanism. Embed a self-attention calculation unit in the feature recognition model: Attention(Q,K,V) = softmax(QK^T / √d)V where Q, K, and V are the query matrix, key matrix, and value matrix respectively, and d is the feature dimension. Through the attention mechanism, the model can adaptively focus on important features and improve the recognition accuracy.

[0084] This embodiment designs a cross-modal attention calculation method. Process the association between different modal features through a cross-modal attention unit: Cross_attention = Σ(wi,j Feature_i Feature_j) where wi,j is the inter-modal association weight, and Feature_i and Feature_j are features of different modalities. This mechanism can capture the cooperative relationship between multi-modal features.

[0085] This embodiment optimizes the model training strategy. An end-to-end backpropagation training method is adopted to optimize multiple links such as feature extraction, attention calculation, and classification prediction simultaneously. The learning rate is dynamically adjusted during the training process to ensure that the model converges to the optimal solution.

[0086] This embodiment demonstrates significant advantages in terms of model organization. Through a hierarchical model structure, each sub-model can focus on the identification of specific types of hidden dangers, improving the professionalism of the model. At the same time, the introduction of the attention mechanism enables the model to intelligently process complex feature relationships.

[0087] This embodiment innovatively designs a model fusion strategy. Calculate the recognition accuracy of each model: Accuracy = (TP + TN) / (TP + TN + FP + FN) where TP is true positive, TN is true negative, FP is false positive, and FN is false negative. Based on the accuracy, initial weights are assigned to the models.

[0088] This embodiment optimizes the confidence evaluation method. The confidence score is calculated by analyzing the probability distribution of the model output. The higher the confidence, the more certain the model is about the prediction result, and the corresponding weight coefficient is also larger. This mechanism ensures the dominant role of high-quality prediction results in the fusion process.

[0089] This embodiment innovatively implements a decision fusion mechanism. The prediction results of each model are integrated by means of weighted voting: Final_result = argmax(Σ(wi Prediction_i)) where wi is the model weight and Prediction_i is the prediction result of each model. In this way, the advantages of each model are fully utilized.

[0090] This embodiment improves the fusion effect through iterative optimization. Regularly evaluate the performance of the fusion model and dynamically adjust the weight coefficients according to the evaluation results. This adaptive optimization mechanism ensures that the fusion model can continuously maintain good recognition effects.

[0091] This embodiment establishes a complete model training and fusion system. From hierarchical classification to attention calculation, and then to model fusion, a systematic recognition framework is formed. This framework can effectively handle complex fire hazard identification tasks.

[0092] This embodiment provides an intelligent identification tool for fire safety management. Through multi-model collaboration and intelligent fusion, accurate identification of fire hazards is achieved. This identification method not only improves the accuracy of identification but also enhances the adaptability of the system.

[0093] This embodiment realizes the intelligent upgrade of hidden danger identification. Through hierarchical classification and model fusion, the hidden danger identification becomes more accurate and reliable. This intelligent identification method provides strong technical support for fire safety management.

[0094] In an embodiment of the method for identifying fire safety hidden dangers through AI vision analysis technology in this application, the following specific content may also be included: Step S401: Receive the data streams collected by multiple sensors, divide the data streams into data segments through time window division, align and synchronize the data segments according to timestamps, perform wavelet transform on the synchronized data to obtain time-frequency components, extract the amplitude features and phase features of the time-frequency components, construct a dynamic feature sequence by sequentially splicing the amplitude features and phase features based on the sliding window method, and perform normalization processing and noise reduction filtering on the dynamic feature sequence; Step S402: Parse the dynamic feature sequence into a feature tensor according to a preset time step, perform channel separation on the feature tensor to obtain multiple feature components, input the feature components into the corresponding input ends of the fusion recognition model, calculate the recognition results and confidence levels of each feature component through the fusion recognition model, screen and combine the recognition results according to the confidence levels to obtain hidden danger feature data, and perform time series correlation analysis on the hidden danger feature data.

[0095] Optionally, in response to problems such as difficulties in real-time processing of multi-source data, discontinuous feature extraction, and unstable recognition results, this embodiment innovatively designs a dynamic feature processing solution based on time-frequency analysis. This embodiment first establishes a multi-channel data reception mechanism to simultaneously receive multi-channel sensor data such as video streams, thermal imaging, depth information, and acoustic signals. A high-speed cache queue is used to manage the data streams to ensure the real-time and continuous reception of data.

[0096] This embodiment optimizes the time window division strategy. According to the sampling characteristics of different types of data, an adaptive window division method is designed: Window_size = max(fs Tc, min_size) where fs is the sampling frequency, Tc is the feature correlation time, and min_size is the minimum window size. This method ensures that the window size can not only completely capture feature changes but also maintain computational efficiency.

[0097] This embodiment innovatively achieves data synchronization alignment. Based on timestamp information, interpolation and resampling techniques are used to achieve time alignment of different data streams. For data with different sampling rates, linear interpolation is used to ensure the consistency of data points on the time axis. For example, the low-frame-rate thermal imaging data is matched to the high-frame-rate video data timing through interpolation.

[0098] This embodiment designs an efficient time-frequency analysis method. Perform continuous wavelet transform on the synchronized data: CWT(a,b) = (1 / √a)∫x(t)ψ ((t-b) / a)dt where a is the scale parameter, b is the translation parameter, and ψ is the wavelet basis function. Through multi-scale analysis, the time-domain and frequency-domain characteristics of the signal can be obtained simultaneously.

[0099] This embodiment optimizes the feature extraction strategy. Extract amplitude features and phase features from the time-frequency components. The amplitude features reflect the signal intensity changes, and the phase features contain the signal structure information. The combination of these two types of features can comprehensively describe the dynamic characteristics of the signal. For example, in equipment anomaly detection, abnormal amplitudes of vibration signals often indicate the occurrence of faults, while phase changes reflect the fault types.

[0100] This embodiment shows significant advantages in dynamic feature construction. The sliding window method is used to sequentially splice the features: Feature_sequence = [F(t-n), ..., F(t-1), F(t)] where F(t) is the feature vector at time t, and n is the sequence length. Through continuous update of the feature sequence, real-time tracking of dynamic changes is achieved.

[0101] This embodiment innovatively designs a feature preprocessing method. Normalize the dynamic feature sequence to make features of different dimensions comparable. The wavelet threshold denoising method is used to eliminate random perturbations and improve the signal-to-noise ratio of the features.

[0102] This embodiment optimizes the construction strategy of the feature tensor. The dynamic feature sequence is reorganized into a three-dimensional feature tensor according to a fixed time step, maintaining the structured organization of the time dimension, feature dimension, and batch dimension. This organizational method facilitates subsequent deep learning processing.

[0103] This embodiment innovatively implements a feature separation mechanism. Channel separation is performed on the feature tensor, and features of different modalities are assigned to corresponding processing channels. For example, visual features, thermal features, depth features, etc. are respectively input into the dedicated input terminals of the fusion recognition model.

[0104] In this embodiment, the recognition reliability is improved through confidence evaluation. The recognition result and confidence of each feature component are calculated: Confidence = Softmax(output_logits) where output_logits are the original scores output by the model. Reliable recognition results are screened based on the confidence threshold.

[0105] This embodiment establishes a complete time series analysis framework. From data reception to feature extraction and then to result recognition, a continuous processing flow is formed. This framework can effectively process real-time data streams and ensure the continuity of feature extraction and recognition.

[0106] This embodiment provides real-time analysis capabilities for fire safety monitoring. Through dynamic feature processing and real-time recognition, continuous monitoring of fire hazards is achieved. This monitoring method not only improves the timeliness of recognition but also enhances the reliability of the system.

[0107] This embodiment realizes the intelligent upgrade of feature processing. Through time-frequency analysis and dynamic feature construction, feature extraction becomes more accurate and stable. This intelligent processing method provides strong technical support for fire safety monitoring.

[0108] In an embodiment of the method for identifying fire hazards through AI vision analysis technology in this application, the following content may also be specifically included: Step S501: Construct a time series sample matrix for the dynamic feature sequence and the hazard feature data according to the time dimension, perform singular value decomposition on the time series sample matrix to extract the principal component features, calculate the time increment and change rate of the principal component features, perform trend decomposition on the principal component features based on the change rate to obtain the periodic component and the trend component, reconstruct the periodic component and the trend component into a time series curve, and perform piecewise fitting on the time series curve to calculate the fitting parameters; Step S502: Use the fitting parameters to construct a state prediction model, input historical time series data into the state prediction model for parameter training, perform extrapolation calculation on the development trend of the hazard features based on the trained state prediction model, perform probability evaluation on the hazard state at the predicted time point according to the extrapolation calculation result to obtain trend prediction data, and perform quantitative evaluation on the credibility of the trend prediction data.

[0109] Optionally, in view of problems such as inaccurate prediction of the development trend of fire hazards, difficult recognition of periodic changes, and low reliability of prediction results, this embodiment innovatively designs a set of trend prediction schemes based on time series analysis. This embodiment first aligns and integrates the dynamic feature sequence and the hazard feature data in the time dimension to construct a time series sample matrix containing multi-dimensional features. This matrix structure retains the complete information of the features evolving over time.

[0110] This embodiment optimizes the principal component feature extraction method. Perform singular value decomposition on the time series sample matrix: X = UΣV^T where X is the time series sample matrix, U is the left singular matrix, Σ is the singular value matrix, and V is the right singular matrix. By selecting the eigenvectors corresponding to the important singular values, the main change patterns in the data are extracted.

[0111] This embodiment innovatively realizes feature change analysis. Calculate the increment of the principal component feature at adjacent time points: ΔF(t) = F(t) - F(t-1) where F(t) is the eigenvalue at time t. Calculate the change rate based on the increment to reflect the dynamic change trend of the feature. This method can effectively capture the speed characteristics of potential hazard development.

[0112] This embodiment designs a trend decomposition strategy. Use the seasonal decomposition method to process the principal component feature and decompose it into a periodic component and a trend component. The periodic component reflects the regular changes in potential hazard characteristics, such as equipment usage cycles, environmental temperature changes, etc.; the trend component reflects the long-term development trend.

[0113] This embodiment optimizes the time series reconstruction method. Recombine the decomposed periodic component and trend component to construct a time series curve with physical meaning. This reconstruction method retains the main features of the data while reducing the influence of random noise.

[0114] This embodiment shows significant advantages in curve fitting. Perform piecewise fitting on the reconstructed time series: y(t) = Σ(ai φi(t)) where φi(t) is the basis function and ai is the fitting parameter. By selecting appropriate piecewise points and basis functions, an accurate description of the trends in different stages is achieved.

[0115] This embodiment innovatively designs a state prediction model. Based on the fitting parameters, construct a nonlinear state space model, and the model structure takes into account the historical dependence and nonlinear evolution characteristics of the features. This model design makes the prediction results more in line with the actual laws of potential hazard development.

[0116] This embodiment optimizes the model training strategy. Input the historical time series data into the prediction model in chronological order and use the sliding window method for parameter training. Dynamically adjust the learning rate during the training process to ensure that the model converges to the optimal solution.

[0117] This embodiment innovatively implements the trend extrapolation algorithm. Based on the trained state prediction model, the hidden danger state at future time points is predicted: State(t + k) = f(State(t), Parameters) where k is the prediction step, and f is the state transition function. The prediction of the long-term trend is achieved through multi-step iteration.

[0118] This embodiment improves the prediction reliability through probability evaluation. Probability modeling is performed on the prediction results to calculate the occurrence probabilities of different states. This probabilistic prediction method provides more comprehensive information for risk assessment.

[0119] This embodiment establishes a complete prediction evaluation system. The prediction credibility evaluation index is constructed: Confidence = g(prediction_error, sample_size, trend_stability) where g is the evaluation function, comprehensively considering factors such as prediction error, sample size, and trend stability.

[0120] This embodiment provides predictive decision-making support for fire safety management. Through trend prediction and credibility evaluation, a scientific prediction of the development trend of fire hazards is achieved. This prediction method not only improves the accuracy of early warning but also provides a basis for preventive maintenance.

[0121] This embodiment realizes the intelligent upgrade of predictive analysis. Through time series analysis and trend prediction, the hidden danger early warning becomes more accurate and reliable. This intelligent prediction method provides strong technical support for fire safety management.

[0122] In an embodiment of the method for identifying fire hazards through AI vision analysis technology in this application, the following content may also be specifically included: Step S601: Perform target detection and positioning on the hidden danger feature data to obtain the bounding box coordinates, map the bounding box coordinates into a three-dimensional space model to obtain spatial position parameters, calculate the relative distance and azimuth angle between the spatial position parameters and a preset reference point, integrate the type label, the spatial position parameters, the relative distance, and the azimuth angle of the hidden danger feature data to construct hidden danger marking data, and perform spatial clustering on the hidden danger marking data according to the relative distance; Step S602: Construct a node feature matrix based on the hidden danger marking data, construct a node transition probability matrix using the trend prediction data, combine the node feature matrix and the node transition probability matrix into a hidden danger feature association network, calculate the association strength and propagation path between nodes in the hidden danger feature association network, construct a state transition equation based on the association strength and the propagation path, and establish a hidden danger evolution model based on the state transition equation.

[0123] Optionally, in this embodiment, in view of problems such as inaccurate spatial positioning of fire hazards, difficult to grasp the correlation between hazards, and difficult to predict the evolution law, an innovative hidden danger modeling scheme based on spatial analysis and network evolution is designed. In this embodiment, a deep learning object detection network is first used to locate the hidden danger feature data to obtain the bounding box coordinates of the target. The detection network adopts a multi-scale feature fusion strategy, which improves the detection ability for targets of different sizes.

[0124] This embodiment optimizes the spatial mapping method. Map the two-dimensional bounding box coordinates to the three-dimensional space through the camera projection model: P_3D = R K^(-1) P_2D + T where P_2D is the two-dimensional coordinate, K is the camera internal parameter matrix, R is the rotation matrix, and T is the translation vector. This mapping method realizes the accurate conversion from the image space to the physical space.

[0125] This embodiment innovatively realizes the calculation of spatial parameters. Establish a spatial coordinate system based on a preset reference point, and calculate the relative distance and azimuth angle of the hidden danger position: Distance = √((x - x0)² + (y - y0)² + (z - z0)²) Angle = arctan2((y - y0), (x - x0)) where (x0, y0, z0) are the coordinates of the reference point. This representation method is convenient for describing the spatial distribution characteristics of hidden dangers.

[0126] This embodiment designs a construction method for hidden danger marking data. Integrate the hidden danger type label, spatial position parameters, relative distance, and azimuth angle into structured marking data. For example, for electrical fire hazards, the marking data includes multi-dimensional information such as equipment type, fault characteristics, and spatial position.

[0127] This embodiment optimizes the spatial clustering strategy. Use the density clustering algorithm to group the hidden danger marking data: Cluster = DBSCAN(Points, eps, min_samples) Among them, eps is the neighborhood radius and min_samples is the minimum number of samples. Hidden danger groups that are spatially close are discovered through clustering analysis.

[0128] This embodiment shows significant advantages in the construction of the feature network. A node feature matrix is constructed based on the hidden danger marking data, and each node contains the attribute features and spatial information of the hidden danger. The node transition probability matrix is constructed using trend prediction data to describe the evolution law of the hidden danger state.

[0129] This embodiment innovatively designs the associated network structure. The node feature matrix and the transition probability matrix are combined into a hidden danger feature association network. The edge weights in the network reflect the intensity of the mutual influence between nodes, and the direction represents the propagation path of the influence.

[0130] This embodiment optimizes the calculation method of the association intensity. The association intensity is calculated based on the similarity of node features and the spatial distance: Correlation = f(Feature_similarity, Spatial_distance) where f is the association function, comprehensively considering the feature similarity and spatial proximity. This calculation method can accurately describe the interaction between hidden dangers.

[0131] This embodiment innovatively realizes the propagation path analysis. The shortest path and critical path in the network are calculated using graph theory algorithms to identify the main channels of hidden danger propagation. This path information has important guiding significance for formulating preventive measures.

[0132] This embodiment describes the system evolution through the state transition equation. The state transition equation is constructed based on the association intensity and the propagation path: dS / dt = A S + B U + ε where S is the state vector, A is the state transition matrix, U is the control input, and ε is the random perturbation term.

[0133] This embodiment establishes a complete evolution model system. From spatial positioning to network construction and then to state evolution, a systematic modeling process is formed. This model system can accurately describe the spatial distribution and evolution law of fire hidden dangers.

[0134] This embodiment provides a scientific analysis tool for fire safety management. Through spatial analysis and evolution modeling, the precise positioning and development prediction of fire hidden dangers are realized. This analysis method not only improves the accuracy of hidden danger identification but also enhances the pertinence of preventive measures.

[0135] This embodiment realizes the intelligent upgrade of hidden danger analysis. Through spatial modeling and network analysis, the management of hidden dangers becomes more accurate and efficient. This intelligent analysis method provides strong technical support for fire safety management.

[0136] In an embodiment of the method for identifying fire safety hidden dangers through AI vision analysis technology in the present application, the following content may also be specifically included: Step S701: Numerically solve the state transition equation of the hidden danger evolution model to obtain the evolution trajectory, calculate the development direction and change rate of the evolution trajectory, extrapolate the evolution trajectory according to a preset time interval to obtain the predicted value of the hidden danger state, compare the predicted value of the hidden danger state with the threshold value at the corresponding time point to obtain the development trend of the hidden danger, perform density clustering analysis on the hidden danger feature data to obtain feature clusters, calculate the centroid position and distribution range of the feature clusters, and identify common hidden dangers according to the number and scale of the feature clusters; Step S702: Map the fields of the hidden danger marking data according to a preset data structure format, perform time serialization processing on the hidden danger development trend data, convert the common hidden danger data into an attribute relationship table, integrate the field mapping data, the time serialization data, and the attribute relationship table into structured data based on a unified data template, and transmit the structured data to the specified storage area of the monitoring terminal through a data interface.

[0137] Optionally, in view of the problems such as the difficult accurate prediction of the evolution trajectory of fire hidden dangers, the difficult identification of common hidden dangers, and the non-standard data integration, this embodiment innovatively designs a set of hidden danger assessment schemes based on evolution analysis and data structuring. This embodiment first solves the state transition equation of the hidden danger evolution model by using a numerical integration method to obtain the evolution trajectory of the hidden danger state over time. The Runge-Kutta method is used for high-precision numerical calculation to ensure the stability and accuracy of the solution.

[0138] This embodiment optimizes the evolution trajectory analysis method. Calculate the tangent vector and curvature of the trajectory at each time point: Direction = dS / dt / ||dS / dt|| Rate = ||d²S / dt²|| / ||dS / dt|| where S is the state vector and t is the time variable. These parameters reflect the direction and speed characteristics of the development of hidden dangers.

[0139] This embodiment innovatively realizes state prediction. Perform extrapolation calculation based on the known trajectory: S(t+Δt) = S(t) + v(t) Δt + 0.5 a(t) Δt² Where v(t) is the instantaneous change rate, a(t) is the acceleration term, and Δt is the prediction time interval. By this method, the hidden danger state at future time points can be predicted.

[0140] This embodiment designs a trend evaluation strategy. Compare the predicted value with the dynamic threshold: Risk_level = f(S_pred -S_threshold) Where S_pred is the predicted state value, S_threshold is the threshold, and f is the evaluation function. This evaluation method can detect potential risks in a timely manner.

[0141] This embodiment optimizes the density clustering algorithm. Use the improved DBSCAN method to cluster the hidden danger feature data: Density(p) = Σ exp(-||p-qi||² / 2σ²) Where p is the feature point, qi is other points in the neighborhood, and σ is the density parameter. Identify the aggregation area in the feature space through density analysis.

[0142] This embodiment shows significant advantages in the identification of common hidden dangers. Calculate the statistical features of the feature clusters, including the centroid position: Centroid = (1 / n)Σxi Where xi is the position vector of the points in the cluster and n is the number of points. Discover the common hidden danger patterns by analyzing the spatial distribution characteristics of the feature clusters.

[0143] This embodiment innovatively designs a data structuring scheme. Map the hidden danger marked data according to a unified field structure to ensure the standardization of the data format. For example, map information such as spatial location, hidden danger type, and risk level to predefined data fields.

[0144] This embodiment optimizes the time series processing. Organize the hidden danger development trend data in time series: TimeSeries = {<t1,S1>,<t2,S2>,...,<tn,Sn>} Where ti is the time stamp and Si is the corresponding state value. This organization method facilitates subsequent trend analysis and query.

[0145] This embodiment innovatively realizes the conversion of the attribute relationship table. Convert the common hidden danger data into a standard relational structure and establish the association relationship between the hidden danger features. This structured representation facilitates data management and analysis.

[0146] In this embodiment, various types of information are integrated through a unified data template. The template design takes into account the integrity, consistency, and scalability of the data to ensure the effective integration of data from different sources.

[0147] In this embodiment, a complete data transmission mechanism is established. Configure a secure data interface and establish a reliable connection with the monitoring terminal: Connection = establish_connection(endpoint, protocol, auth_token) Transmit the structured data to the specified storage area through an encrypted channel.

[0148] This embodiment provides a systematic analysis tool for fire safety management. Through evolutionary analysis and data structuring, a comprehensive assessment of fire hazards is achieved. This analysis method not only improves the accuracy of hazard prediction but also enhances the standardization of data management.

[0149] This embodiment realizes the intelligent upgrade of hazard management. Through evolutionary prediction and data integration, hazard management becomes more scientific and efficient. This intelligent management method provides strong technical support for fire safety work.

[0150] In order to effectively solve the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improve the intelligent level and early warning ability of fire hazard identification, this application provides an embodiment of a device for implementing all or part of the method for identifying fire hazards through AI vision analysis technology. See Figure 2 , the device for identifying fire hazards through AI vision analysis technology specifically includes the following: The model fusion module 10 is used to collect image and video stream data from the security monitoring system, mobile patrol terminals, and fire patrol robots, obtain the scene depth information through a binocular camera to construct a three-dimensional space model, collect infrared thermal imaging data to obtain the temperature distribution characteristics, collect environmental perception signals to extract environmental features, preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental features to obtain normalized training data, extract multi-modal feature vectors from the normalized training data, label the multi-modal feature vectors to construct training samples, select a basic feature extraction layer from the pre-trained model library to construct a transfer learning model, input the training samples into the transfer learning model, train a group of fire hazard feature neural network models based on a hierarchical classification structure and an attention mechanism, and perform ensemble learning on the recognition results of the group of fire hazard feature neural network models to obtain a fusion recognition model; A trend prediction module 20, configured to receive multi-source data collected in real time, perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain potential hazard feature data, and analyze the time evolution trend of the potential hazard feature data based on the dynamic feature sequence to obtain trend prediction data; A potential hazard identification module 30, configured to perform spatial positioning on the potential hazard feature data to construct potential hazard marking data, construct a potential hazard feature association network based on the potential hazard marking data and the trend prediction data, establish a potential hazard evolution model by using the potential hazard feature association network, predict the development trend of potential hazards through the potential hazard evolution model, perform clustering analysis on the potential hazard feature data to discover common potential hazards, and integrate the potential hazard marking data, the potential hazard development trend, and the common potential hazards into structured data and output it to the monitoring terminal.

[0151] As can be seen from the above description, the device for identifying fire and fire protection hazards through AI vision analysis technology provided by the embodiments of the present application can innovatively construct a multi-source data collection and fusion mechanism to integrate multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals. Design a feature extraction model based on transfer learning, combine a hierarchical classification structure and an attention mechanism, and achieve high-precision potential hazard feature recognition through ensemble learning. Introduce time series analysis and spatial positioning technologies, construct a potential hazard feature association network and an evolution model, and realize the prediction of the development trend of potential hazards and the discovery of common potential hazards. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire protection hazard identification.

[0152] From the hardware level, in order to effectively solve the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improve the intelligent level and early warning ability of fire and fire protection hazard identification, the embodiments of the present application provide an electronic device for implementing all or part of the content in the method for identifying fire and fire protection hazards through AI vision analysis technology. The electronic device specifically includes the following content: A processor, a memory, a communications interface, and a bus; wherein, the processor, the memory, and the communications interface complete communication with each other through the bus; the communications interface is used to implement information transmission between a device for identifying fire and fire prevention hazards through AI vision analysis technology and related devices such as a core business system, a user terminal, and a related database, etc.; the logic controller may be a desktop computer, a tablet computer, a mobile terminal, etc., and this embodiment is not limited thereto. In this embodiment, the logic controller may be implemented with reference to the embodiments of the method for identifying fire and fire prevention hazards through AI vision analysis technology and the embodiments of the device for identifying fire and fire prevention hazards through AI vision analysis technology, the content of which is incorporated herein, and the repeated parts will not be elaborated again.

[0153] It can be understood that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0154] In practical applications, part of the method for identifying fire and fire prevention hazards through AI vision analysis technology may be executed on the electronic device side as described above, or all operations may be completed in the client device. Specifically, it can be selected according to the processing capacity of the client device and the limitations of the user usage scenario, etc. This application does not make any limitations in this regard. If all operations are completed in the client device, the client device may further include a processor.

[0155] The above-mentioned client device may have a communication module (i.e., a communication unit), and may be communicatively connected to a remote server to implement data transmission with the server. The server may include a server on the task scheduling center side, and may also include a server on an intermediate platform in other implementation scenarios, such as a server on a third-party server platform communicatively linked to the task scheduling center server. The server may include a single computer device, or may include a server cluster composed of multiple servers, or a server structure of a distributed device.

[0156] Figure 3 This is a schematic block diagram of the system composition of the electronic device 9600 according to an embodiment of the present application. As Figure 3 shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It should be noted that this Figure 3 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0157] In one embodiment, the method function for identifying fire and fire protection hazards through AI vision analysis technology can be integrated into the central processing unit 9100. Among them, the central processing unit 9100 can be configured to perform the following controls: Step S101: Collect image and video stream data from the security monitoring system, mobile patrol terminals, and fire patrol robots. Obtain the depth information of the scene through a binocular camera to construct a three-dimensional space model. Collect infrared thermal imaging data to obtain the temperature distribution characteristics. Collect environmental perception signals to extract environmental features. Preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental features to obtain normalized training data. Extract multi-modal feature vectors from the normalized training data. Label the multi-modal feature vectors to construct training samples. Select a basic feature extraction layer from the pre-trained model library to construct a transfer learning model. Input the training samples into the transfer learning model. Based on the hierarchical classification structure and attention mechanism, train a group of neural network models for fire hazard features. Perform ensemble learning on the recognition results of the group of neural network models for fire hazard features to obtain a fusion recognition model; Step S102: Receive multi-source data collected in real time. Perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence. Input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain hazard feature data. Analyze the time evolution trend of the hazard feature data based on the dynamic feature sequence to obtain trend prediction data; Step S103: Perform spatial positioning on the hazard feature data to construct hazard marking data. Based on the hazard marking data and the trend prediction data, construct a hazard feature association network. Use the hazard feature association network to establish a hazard evolution model. Predict the development trend of hazards through the hazard evolution model. Perform clustering analysis on the hazard feature data to discover common hazards. Integrate the hazard marking data, the hazard development trend, and the common hazards into structured data and output it to the monitoring terminal.

[0158] As can be seen from the above description, the electronic device provided in the embodiment of the present application innovatively constructs a multi-source data collection and fusion mechanism to integrate multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals. Designs a feature extraction model based on transfer learning, combines a hierarchical classification structure and an attention mechanism, and achieves high-precision hazard feature recognition through ensemble learning. Introduces time series analysis and spatial positioning technologies, constructs a hazard feature association network and an evolution model, and realizes the prediction of hazard development trends and the discovery of common hazards. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire protection hazard recognition.

[0159] In another embodiment, the device for identifying fire and fire protection hazards through AI vision analysis technology can be separately configured from the central processor 9100. For example, the device for identifying fire and fire protection hazards through AI vision analysis technology can be configured as a chip connected to the central processor 9100, and the method function of identifying fire and fire protection hazards through AI vision analysis technology can be realized through the control of the central processor.

[0160] As Figure 3 shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It should be noted that the electronic device 9600 does not necessarily have to include Figure 3 all the components shown in Figure 3 ; in addition, the electronic device 9600 may further include

[0161] As Figure 3 shown, the central processor 9100, sometimes also referred to as a controller or an operation control, may include a microprocessor or other processor devices and / or logic devices. The central processor 9100 receives inputs and controls the operations of the various components of the electronic device 9600.

[0162] Among them, the memory 9140, for example, may be one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. The above information related to failures can be stored, and in addition, programs for executing relevant information can also be stored. And the central processor 9100 can execute the program stored in the memory 9140 to implement information storage or processing, etc.

[0163] The input unit 9120 provides inputs to the central processor 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to supply power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0164] The memory 9140 can be a solid-state memory, for example, a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that stores information even when power is off, can be selectively erased and has more data. Examples of such a memory are sometimes referred to as EPROMs, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 can include an application / function storage section 9142, which is used to store application programs and function programs or the processes for operating the electronic device 9600 by the central processing unit 9100.

[0165] The memory 9140 can also include a data storage section 9143, which is used to store data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage section 9144 of the memory 9140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (such as a messaging application, an address book application, etc.).

[0166] The communication module 9110 is a transmitter / receiver that transmits and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.

[0167] Based on different communication technologies, multiple communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module 9110 (transmitter / receiver) is also coupled to the speaker 9131 and the microphone 9132 via the audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, so as to implement normal telecommunication functions. The audio processor 9130 can include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processing unit 9100, so that recording can be performed on the local machine via the microphone 9132, and the sound stored on the local machine can be played via the speaker 9131.

[0168] Embodiments of the present application also provide a computer-readable storage medium that can implement all steps of the method for identifying fire and fire prevention hazards through AI vision analysis technology where the execution subject in the above embodiments is a server or a client. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements all steps of the method for identifying fire and fire prevention hazards through AI vision analysis technology where the execution subject in the above embodiments is a server or a client. For example, when the processor executes the computer program, the following steps are implemented: Step S101: Collect image and video stream data from the security monitoring system, mobile patrol terminals, and fire patrol robots. Obtain the depth information of the scene through a binocular camera to construct a three-dimensional space model. Collect infrared thermal imaging data to obtain temperature distribution characteristics. Collect environmental perception signals to extract environmental features. Preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental features to obtain normalized training data. Extract multi-modal feature vectors from the normalized training data. Label the multi-modal feature vectors to construct training samples. Select a basic feature extraction layer from the pre-trained model library to construct a transfer learning model. Input the training samples into the transfer learning model. Train a fire hazard feature neural network model group based on a hierarchical classification structure and an attention mechanism. Perform ensemble learning on the recognition results of the fire hazard feature neural network model group to obtain a fusion recognition model; Step S102: Receive real-time collected multi-source data. Perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence. Input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain hazard feature data. Analyze the time evolution trend of the hazard feature data based on the dynamic feature sequence to obtain trend prediction data; Step S103: Perform spatial positioning on the hazard feature data to construct hazard marking data. Construct a hazard feature association network based on the hazard marking data and the trend prediction data. Use the hazard feature association network to establish a hazard evolution model. Predict the hazard development trend through the hazard evolution model. Perform clustering analysis on the hazard feature data to discover common hazards. Integrate the hazard marking data, the hazard development trend, and the common hazards into structured data and output it to the monitoring terminal.

[0169] As can be seen from the above description, the computer-readable storage medium provided by the embodiments of the present application integrates multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals by innovatively constructing a multi-source data acquisition and fusion mechanism. A feature extraction model based on transfer learning is designed, combined with a hierarchical classification structure and an attention mechanism, and high-precision hidden danger feature recognition is achieved through ensemble learning. Temporal analysis and spatial positioning technologies are introduced to construct a hidden danger feature association network and an evolution model to realize the prediction of the development trend of hidden dangers and the discovery of common hidden dangers. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and fire hazard recognition.

[0170] An embodiment of the present application also provides a computer program product that can implement all the steps of the method for identifying fire and fire hazards through AI vision analysis technology, where the execution subject in the above embodiment is a server or a client. When the computer program / instructions are executed by a processor, the steps of the method for identifying fire and fire hazards through AI vision analysis technology are implemented. For example, the computer program / instructions implement the following steps: Step S101: Collect image and video stream data from the security monitoring system, mobile patrol terminals, and fire patrol robots. Obtain the depth information of the scene through a binocular camera to construct a three-dimensional space model. Collect infrared thermal imaging data to obtain temperature distribution characteristics. Collect environmental perception signals to extract environmental characteristics. Preprocess the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental characteristics to obtain normalized training data. Extract multi-modal feature vectors from the normalized training data. Label the multi-modal feature vectors to construct training samples. Select a basic feature extraction layer from the pre-trained model library to construct a transfer learning model. Input the training samples into the transfer learning model. Train a group of fire hazard feature neural network models based on the hierarchical classification structure and the attention mechanism. Perform ensemble learning on the recognition results of the group of fire hazard feature neural network models to obtain a fusion recognition model; Step S102: Receive real-time collected multi-source data. Perform temporal decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence. Input the dynamic feature sequence into the fusion recognition model for multi-modal feature recognition to obtain hidden danger feature data. Analyze the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data; Step S103: Perform spatial positioning on the hidden danger feature data to construct hidden danger marking data, construct a hidden danger feature association network based on the hidden danger marking data and the trend prediction data, establish a hidden danger evolution model using the hidden danger feature association network, predict the development trend of the hidden danger through the hidden danger evolution model, perform clustering analysis on the hidden danger feature data to discover common hidden dangers, and integrate the hidden danger marking data, the hidden danger development trend, and the common hidden dangers into structured data and output it to the monitoring terminal.

[0171] As can be seen from the above description, the computer program product provided by the embodiments of the present application innovatively constructs a multi-source data acquisition and fusion mechanism to integrate multi-modal data such as image and video streams, three-dimensional space, thermal imaging, and acoustic signals. Designs a feature extraction model based on transfer learning, combines a hierarchical classification structure and an attention mechanism, and realizes high-precision hidden danger feature recognition through ensemble learning. Introduces time series analysis and spatial positioning technologies, constructs a hidden danger feature association network and an evolution model, and realizes the prediction of the development trend of hidden dangers and the discovery of common hidden dangers. This method effectively solves the deficiencies of traditional technologies in multi-modal data processing, feature recognition, and trend prediction, and significantly improves the intelligent level and early warning ability of fire and hidden danger recognition.

[0172] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also 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 one Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0174] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one process or Figure 1 one or more processes and / or blocks Figure 1 specified in a block or blocks.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process or Figure 1 one or more processes and / or blocks Figure 1 specified in a block or blocks.

[0176] Specific embodiments are used in the present invention to illustrate the principles and implementation manners of the present invention. The description of the above embodiments is only for helping to understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A method for identifying fire hazards through AI visual analysis technology, characterized in that: The method comprises: Collecting image and video stream data from security monitoring systems, mobile inspection terminals and fire inspection robots, acquiring scene depth information through binocular cameras to build a three-dimensional space model, collecting infrared thermal imaging data to obtain temperature distribution features, collecting environmental perception signals to extract environmental features, preprocessing the image and video stream data, the three-dimensional space model, the temperature distribution features and the environmental features to obtain normalized training data, extracting multimodal feature vectors from the normalized training data, annotating the multimodal feature vectors to construct training samples, selecting a basic feature extraction layer from a pre-trained model library to construct a transfer learning model, inputting the training samples into the transfer learning model, training a fire hazard feature neural network model group based on a hierarchical classification structure and an attention mechanism, and integrating the recognition results of the fire hazard feature neural network model group to obtain a fusion recognition model; Receiving multi-source data collected in real time, performing time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, inputting the dynamic feature sequence into the fusion recognition model to perform multimodal feature recognition to obtain hidden danger feature data, and analyzing the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data; The hidden danger feature data are spatially located to construct hidden danger marking data, a hidden danger feature association network is constructed based on the hidden danger marking data and the trend prediction data, a hidden danger evolution model is established using the hidden danger feature association network, hidden danger development trends are predicted through the hidden danger evolution model, cluster analysis is performed on the hidden danger feature data to discover common hidden dangers, and the hidden danger marking data, the hidden danger development trends and the common hidden dangers are integrated into structured data and output to a monitoring terminal.

2. The method for identifying fire hazards by using AI visual analysis technology according to claim 1, characterized in that: The step of extracting a multimodal feature vector from the normalized training data, annotating the multimodal feature vector to construct a training sample, selecting a basic feature extraction layer from a pre-trained model library to construct a transfer learning model, and inputting the training sample into the transfer learning model includes: Performing convolution decomposition and pooling operations on the normalized training data to extract image texture features, extracting three-dimensional spatial features through a deep image analysis module, obtaining temperature gradient features using an infrared feature extraction module, extracting acoustic features based on a spectrogram analysis module, fusing the image texture features, the three-dimensional spatial features, the temperature gradient features, and the acoustic features to obtain a multimodal feature vector, normalizing the feature components of the multimodal feature vector and adding time tags, and performing category and attribute annotation on the multimodal feature vector based on manual annotation rules to construct training samples; Perform feature layer analysis on the models in the pre-trained model library, select a representative basic feature extraction layer according to the application scenario, copy the weight parameters and network structure of the basic feature extraction layer to the target network to build a transfer learning model, adjust the input layer and output layer of the transfer learning model to adapt to the data format of the training samples, and input the training samples into the adjusted transfer learning model for feature mapping.

3. The method for identifying fire hazards through AI visual analysis technology according to claim 1, characterized in that: The fire hazard feature neural network model group is trained based on the hierarchical classification structure and the attention mechanism, and the recognition results of the fire hazard feature neural network model group are integrated and learned to obtain a fusion recognition model, including: A hierarchical classification structure is constructed based on a preset fire hazard classification system, and the hazard types are divided into multiple subcategory nodes according to physical characteristics. A corresponding feature recognition model is assigned to each subcategory node, and a self-attention calculation unit and a cross-modal attention calculation unit are embedded in the feature recognition model. The self-attention calculation unit is used to weight the importance of the feature vector, and the cross-modal attention calculation unit is used to calculate the association weights between different modal features. The training samples are input into each feature recognition model for back-propagation training to obtain a fire hazard feature neural network model group; Calculate the recognition accuracy and confidence score of each model in the fire hazard characteristic neural network model group, assign a fusion weight coefficient to each model based on the recognition accuracy and confidence score, perform weighted combination on the recognition results of the model group and the corresponding fusion weight coefficient, perform decision fusion on the weighted combination results through a voting mechanism to obtain a fusion recognition model, and iteratively optimize the fusion weight coefficient of the fusion recognition model.

4. The method for identifying fire hazards through AI visual analysis technology according to claim 1, characterized in that: The receiving of multi-source data collected in real time, performing time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, and inputting the dynamic feature sequence into the fusion recognition model to perform multi-modal feature recognition to obtain hidden danger feature data, includes: Receive data streams collected by multiple sensors, divide the data streams into time windows to obtain data segments, align and synchronize the data segments according to timestamps, perform wavelet transform on the synchronized data to obtain time-frequency components, extract amplitude features and phase features of the time-frequency components, sequentially splice the amplitude features and phase features based on a sliding window method to construct a dynamic feature sequence, and perform normalization and noise reduction filtering on the dynamic feature sequence; The dynamic feature sequence is parsed into a feature tensor according to a preset time step, the feature tensor is channel-separated to obtain multiple feature components, the feature components are input into the corresponding input ends of the fusion recognition model, the recognition result and confidence of each feature component are calculated by the fusion recognition model, the recognition results are screened and combined according to the confidence to obtain hidden danger feature data, and the hidden danger feature data is subjected to time series correlation analysis.

5. The method for identifying fire hazards through AI visual analysis technology according to claim 1, characterized in that: The step of analyzing the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data includes: The dynamic feature sequence and the hidden danger feature data are used to construct a time series sample matrix according to the time dimension, singular value decomposition is performed on the time series sample matrix to extract the principal component features, the time increment and the change rate of the principal component features are calculated, the principal component features are trend decomposed based on the change rate to obtain the periodic component and the trend component, the periodic component and the trend component are reconstructed into a time series curve, and the time series curve is piecewise fitted to calculate the fitting parameters; A state prediction model is constructed using the fitting parameters, and historical time series data is input into the state prediction model for parameter training. The development trend of hidden danger characteristics is extrapolated based on the trained state prediction model. According to the extrapolated calculation results, a probability assessment of the hidden danger state at the predicted time point is performed to obtain trend prediction data, and the credibility of the trend prediction data is quantitatively assessed.

6. The method for identifying fire hazards through AI visual analysis technology according to claim 1, characterized in that: The step of spatially locating the hidden danger feature data to construct hidden danger marking data, constructing a hidden danger feature association network based on the hidden danger marking data and the trend prediction data, and establishing a hidden danger evolution model using the hidden danger feature association network includes: Performing target detection and positioning on the hidden danger feature data to obtain bounding box coordinates, mapping the bounding box coordinates to a three-dimensional space model to obtain spatial position parameters, calculating the relative distance and azimuth between the spatial position parameters and a preset reference point, integrating the type label of the hidden danger feature data, the spatial position parameters, the relative distance and the azimuth to construct hidden danger marking data, and performing spatial clustering on the hidden danger marking data according to the relative distance; A node feature matrix is ​​constructed based on the hidden danger marking data, a node transfer probability matrix is ​​constructed using the trend prediction data, the node feature matrix and the node transfer probability matrix are combined into a hidden danger feature association network, the association strength and propagation path between nodes in the hidden danger feature association network are calculated, a state transfer equation is constructed according to the association strength and the propagation path, and a hidden danger evolution model is established based on the state transfer equation.

7. The method for identifying fire hazards through AI visual analysis technology according to claim 1, characterized in that: The method of predicting the hidden danger development trend by the hidden danger evolution model, performing cluster analysis on the hidden danger feature data to find common hidden dangers, and integrating the hidden danger marking data, the hidden danger development trend and the common hidden dangers into structured data and outputting them to the monitoring terminal includes: Numerically solving the state transfer equation of the hidden danger evolution model to obtain an evolution trajectory, calculating the development direction and change rate of the evolution trajectory, extrapolating the evolution trajectory according to a preset time interval to obtain a hidden danger state prediction value, comparing the hidden danger state prediction value with the threshold value of the corresponding time point to obtain the hidden danger development trend, performing density clustering analysis on the hidden danger feature data to obtain feature clusters, calculating the center of gravity position and distribution range of the feature clusters, and identifying common hidden dangers according to the number and scale of feature clusters; The hidden danger marking data is field mapped according to a preset data structure format, the hidden danger development trend data is time-serialized, the common hidden danger data is converted into an attribute relationship table, the field mapping data, the time-serialized data and the attribute relationship table are integrated into structured data based on a unified data template, and the structured data is transmitted to the storage area specified by the monitoring terminal through a data interface.

8. A device for identifying fire hazards through AI visual analysis technology, characterized in that: The device comprises: A model fusion module is used to collect image and video stream data from a security monitoring system, a mobile inspection terminal, and a fire inspection robot, obtain scene depth information through a binocular camera to build a three-dimensional space model, collect infrared thermal imaging data to obtain temperature distribution characteristics, collect environmental perception signals to extract environmental characteristics, pre-process the image and video stream data, the three-dimensional space model, the temperature distribution characteristics, and the environmental characteristics to obtain normalized training data, extract multimodal feature vectors from the normalized training data, annotate the multimodal feature vectors to construct training samples, select a basic feature extraction layer from a pre-trained model library to construct a transfer learning model, input the training samples into the transfer learning model, train a fire hazard feature neural network model group based on a hierarchical classification structure and an attention mechanism, and integrate the recognition results of the fire hazard feature neural network model group to obtain a fusion recognition model; A trend prediction module is used to receive multi-source data collected in real time, perform time series decomposition and feature extraction on the multi-source data to construct a dynamic feature sequence, input the dynamic feature sequence into the fusion recognition model to perform multimodal feature recognition to obtain hidden danger feature data, and analyze the time evolution trend of the hidden danger feature data based on the dynamic feature sequence to obtain trend prediction data; The hidden danger identification module is used to spatially locate the hidden danger feature data to construct hidden danger marking data, construct a hidden danger feature association network based on the hidden danger marking data and the trend prediction data, establish a hidden danger evolution model using the hidden danger feature association network, predict the hidden danger development trend through the hidden danger evolution model, perform cluster analysis on the hidden danger feature data to discover common hidden dangers, and integrate the hidden danger marking data, the hidden danger development trend and the common hidden dangers into structured data and output them to the monitoring terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for identifying fire hazards through AI visual analysis technology as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying fire hazards by using AI visual analysis technology as described in any one of claims 1 to 7 are implemented.

Images