A method and system for deducing urban fire emergency plans based on AI digitization

Abstract

The present invention discloses a method and system for deducing an urban fire emergency plan based on AI digitization, which relates to the field of intelligent emergency management technology, including collecting fire scene data and preprocessing it; using a multimodal fusion algorithm to fuse the preprocessed fire scene data to form a multimodal feature vector; based on historical fire data, training a fire situation prediction model, inputting the multimodal feature vector into the fire situation prediction model, and predicting the comprehensive index of fire development; according to the comprehensive index of fire development, analyzing the fire development trend, and generating targeted response strategies; simulating and exercising the generated response strategies, evaluating the execution effect of the response strategies, and optimizing the response strategies based on the execution effect to obtain emergency plans. The present invention predicts the fire development index through multimodal data fusion and the fire situation prediction model, generates and optimizes the response strategies, significantly enhances the efficiency of urban fire emergency management, and provides technical support for urban fire safety.

Description

Technical Field

[0001] The present invention relates to the field of intelligent emergency management technology, and in particular to an urban fire emergency plan deduction method and system based on AI digitization. Background Art

[0002] With the acceleration of urbanization, the frequency and harmfulness of urban fires are increasing. Therefore, it is urgent to develop efficient and intelligent urban fire emergency plans. In recent years, with the rapid development of artificial intelligence technology, especially breakthroughs in computer vision, natural language processing and machine learning, new solutions have been provided for fire emergency management and response. Traditional fire emergency management systems mainly rely on manual observation and decision-making, which are inefficient and prone to misjudgment. Modern AI digital systems can achieve comprehensive monitoring and intelligent analysis of fire scenes by integrating multi-source data and using advanced data processing technology and algorithms. In addition, with the application of Internet of Things technology, various sensors and monitoring equipment can collect on-site data in real time and transmit it to the data center through wireless networks, providing a solid foundation for the prediction of fire situations and the formulation of response strategies.

[0003] Although the current AI digital fire emergency management system has improved the speed and accuracy of emergency response to a certain extent, there are still some shortcomings. First, in terms of multimodal data fusion, existing systems often find it difficult to effectively integrate large amounts of data from different sources and types, resulting in low information utilization. Secondly, the accuracy of fire situation prediction still needs to be improved, especially in complex and changeable fire environments. How to accurately predict the development trend of the fire remains a challenge. In addition, the formulation of response strategies often lacks flexibility and adaptability, and cannot cope well with the variability of fires. Finally, when evaluating the effectiveness of response strategies, existing systems often rely on post-event manual evaluations and lack a real-time feedback mechanism, which makes the strategy optimization process relatively slow. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an urban fire emergency plan deduction method and system based on AI digitization to solve the problems of insufficient multimodal data fusion, low fire situation prediction accuracy, insufficient flexibility of response strategies and lack of strategy evaluation feedback mechanism in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides an urban fire emergency plan deduction method based on AI digitization, which includes collecting fire scene data and preprocessing it; using a multimodal fusion algorithm to fuse the preprocessed fire scene data to form a multimodal feature vector; based on historical fire data, training a fire situation prediction model, inputting the multimodal feature vector into the fire situation prediction model, and predicting a comprehensive fire development index; according to the comprehensive fire development index, analyzing the fire development trend and generating targeted response strategies; simulating and exercising the generated response strategies, evaluating the execution effects of the response strategies, optimizing the response strategies based on the execution effects, and obtaining emergency plans.

[0008] As a preferred solution of the AI-digitalized urban fire emergency plan deduction method of the present invention, the specific steps of collecting fire scene data and preprocessing are as follows:

[0009] Collect fire scene data through drones, thermal imagers, cameras and media platforms; fire scene data includes image data, video data, audio data and text data of the fire scene;

[0010] Pre-process the fire scene data by cleaning and standardizing;

[0011] Use convolutional neural networks to extract features from preprocessed image data;

[0012] Use recurrent neural network to extract preprocessed audio data features and video data features;

[0013] Use BERT to convert the preprocessed text data into numerical vectors.

[0014] As a preferred solution of the AI-digitalized urban fire emergency plan deduction method of the present invention, wherein: the multimodal fusion algorithm is used to fuse the pre-processed fire scene data to form a multimodal feature vector, and the specific steps are as follows:

[0015] The extracted feature vectors are weighted and fused through the attention mechanism, and the expression is:

[0016]

[0017] Among them, F is the comprehensive feature vector obtained by weighted fusion, The i-th modality includes four modalities: image data features, audio data features, video data features, and text numerical vectors. represents the attention weight of the i-th modality, and i represents the index of different modalities.

[0018] As a preferred solution of the method for deducing urban fire emergency plans based on AI digitization of the present invention, wherein: based on historical fire data, a fire situation prediction model is trained, and the multimodal feature vector is input into the fire situation prediction model to predict the comprehensive index of fire development. The specific steps are as follows:

[0019] Using the LSTM model as the basic architecture, historical fire data is input into the LSTM model for training, and the LSTM model is trained into a fire situation prediction model that can predict fire situations;

[0020] The cross entropy loss function is used to measure the accuracy of the fire situation prediction model;

[0021] Reduce the loss function through the Adam optimizer;

[0022] The cross-validation method was used to optimize the parameters of the fire situation prediction model.

[0023] The multimodal feature vector , input into the fire situation prediction model to predict the comprehensive index of fire development, the expression is:

[0024]

[0025] in, Indicates at time When is the multimodal feature vector, is the comprehensive index of fire development. is the number of multimodal features, represents the attention weight of the i-th modality, represents the eigenvector of the i-th mode, T represents the maximum considered time range, t represents the time difference from the start of prediction to the current moment, c is the time attenuation coefficient, represents the normalized weight of the ith mode.

[0026] As a preferred solution of the AI-digitalized urban fire emergency plan deduction method of the present invention, the fire development trend is analyzed according to the fire development comprehensive index, and a targeted response strategy is generated. The specific steps are as follows:

[0027] when >g, it means that the current fire is rapidly deteriorating, where g is the fire controllability threshold;

[0028] when When ≤g, it means the current fire is within controllable range;

[0029] When the current fire is rapidly deteriorating, the response strategy includes implementing targeted firefighting operations, monitoring the fire development and maintaining the safety of the surrounding areas;

[0030] When the current fire is under control, the response strategy includes maintaining existing rescue plans and strengthening fire protection measures in surrounding buildings.

[0031] As a preferred solution of the AI-digitalized urban fire emergency plan deduction method of the present invention, the generated response strategy is simulated and exercised to evaluate the execution effect of the response strategy. The specific steps are as follows:

[0032] Based on the generated response strategies, simulated fire conditions are designed, including the location of the fire, wind speed and direction, to simulate different stages from small-scale initial fires to medium-scale spreading fires and large-scale uncontrolled fires;

[0033] Build a detailed virtual city model using ArcGIS, including accurate building layout, road network, and population density distribution. At the same time, integrate real-time meteorological data to implement response strategies in the virtual city model.

[0034] Calculate the effectiveness score of the coping strategy, the expression is:

[0035]

[0036] in, Indicates the efficiency of personnel evacuation. Indicates the extent of property damage. Indicates the speed of rescue operations. Represents the weight coefficient of personnel evacuation efficiency, The weight coefficient representing the degree of property loss, represents the weight coefficient of the rescue operation efficiency, represents the effectiveness score of the coping strategies;

[0037] when When ≥e, it means that the response strategy is effective and can effectively control the fire, where e is the strategy effectiveness threshold;

[0038] when When <e, it means that the response strategy is ineffective and it is difficult to control the fire.

[0039] As a preferred solution of the AI-digitalized urban fire emergency plan deduction method of the present invention, wherein: the response strategy is optimized based on the execution effect to obtain the emergency plan, and the specific steps are as follows:

[0040] When the response strategy is ineffective and it is difficult to control the fire, review the evacuation drill records, check the evacuation efficiency, property loss rate and rescue operation speed, and make corresponding adjustments based on the inspection results, including adjusting the evacuation route, adding evacuation exits and adjusting the evacuation sequence;

[0041] After adjustment, the implementation effect of the response strategy is evaluated again until the response strategy is effective and an urban fire emergency plan is generated;

[0042] When the response strategy is effective and can effectively control the fire, an urban fire emergency plan is generated based on the current response strategy.

[0043] In a second aspect, the present invention provides an urban fire emergency plan simulation system based on AI digitization, comprising:

[0044] Data acquisition and preprocessing module, used to collect fire scene data and perform preprocessing;

[0045] A feature extraction and fusion module is used to fuse the pre-processed fire scene data into a multi-modal feature vector using a multi-modal fusion algorithm;

[0046] The fire situation prediction module is used to train a fire situation prediction model based on historical fire data, input the multimodal feature vector into the fire situation prediction model, and predict the comprehensive index of fire development;

[0047] The response strategy generation module is used to analyze the fire development trend and generate targeted response strategies based on the comprehensive fire development index;

[0048] The strategy evaluation and optimization module is used to simulate and exercise the generated response strategies, evaluate the execution effect of the response strategies, optimize the response strategies based on the execution effect, and obtain emergency plans.

[0049] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the urban fire emergency plan simulation method based on AI digitization as described in the first aspect of the present invention is implemented.

[0050] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the urban fire emergency plan deduction method based on AI digitization as described in the first aspect of the present invention.

[0051] The beneficial effects of the present invention are: collecting and preprocessing multimodal data at the fire scene, using a multimodal fusion algorithm to extract features and construct a fire situation prediction model, and achieving accurate prediction of the comprehensive index of fire development. Furthermore, a response strategy is generated based on the fire development trend, and the effectiveness of the strategy is evaluated through simulation exercises, ultimately achieving the purpose of optimizing the response strategy to obtain an emergency plan. This method not only improves the accuracy of fire prediction, but also can flexibly adjust the response strategy according to changes in the fire, significantly enhancing the efficiency and effectiveness of urban fire emergency management. Ultimately, the invention can provide strong technical support for urban fire safety and protect people's lives and property. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0053] Figure 1 This is a flow chart of the urban fire emergency plan deduction method based on AI digitization in Example 1.

[0054] Figure 2 This is a flow chart of the AI-based digital urban fire emergency plan simulation system in Example 1. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0057] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0058] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for deducing an urban fire emergency plan based on AI digitization, comprising the following steps:

[0059] S1. Collect fire scene data and perform preprocessing.

[0060] Furthermore, fire scene data is collected through drones, thermal imagers, cameras and media platforms; the fire scene data includes image data, video data, audio data and text data of the fire scene;

[0061] Use drones equipped with high-definition cameras and thermal imagers to conduct aerial photography and collect high-definition image data and video data of the fire scene;

[0062] Collect audio data from the scene, such as the sound of burning fire and explosion, through microphones installed on drones or fixed locations;

[0063] Collect real-time text data related to fires from social media platforms (such as Weibo, WeChat, etc.) through API interfaces;

[0064] Furthermore, the fire scene data are preprocessed by cleaning and standardization.

[0065] The image data preprocessing is as follows: remove irrelevant background parts, retain the core area of ​​the fire scene, adjust the images to the same size for subsequent processing, and use the median filter algorithm to remove noise in the image.

[0066] The video data preprocessing is as follows: adjust the frame rate of the video to ensure that all videos have the same playback speed.

[0067] Resize the video resolution to the same size, adjust the brightness and contrast of the video, and improve the clarity.

[0068] Audio data preprocessing specifically involves using digital signal processing techniques (such as frequency domain filtering) to remove background noise from the audio and adjust the volume of the audio file to a consistent level.

[0069] The specific preprocessing of text data is as follows: Use natural language processing technology (such as spell checker) to correct grammatical errors in the text. Use text analysis tools (such as TF-IDF algorithm) to extract key information related to the fire.

[0070] Use convolutional neural networks to extract features from preprocessed image data, including key information such as fire size and direction of spread;

[0071] The specific process includes inputting the preprocessed image into the CNN model, and extracting local features in the image, such as edges and textures, through a series of convolution kernels. Then, the pooling layer is used to reduce the spatial dimension and the amount of calculation while maintaining the key features of the image. The features extracted by the convolution layer and the pooling layer are then mapped to the fully connected layer for feature fusion and classification. Finally, the output layer obtains key information such as the size of the fire and the direction of spread.

[0072] Use recurrent neural network to extract preprocessed audio data features and video data features;

[0073] Among them, audio data features include alarm sounds, cries for help, etc.; video data features include key visual features such as fire spread speed and changes in flame shape.

[0074] The specific process includes inputting the preprocessed audio data into the RNN model. The hidden layer of the RNN can capture the dependencies in the time series and extract the key features in the audio. Finally, the output layer obtains the key information in the audio, such as the time of the alarm sound, the intensity of the cry for help, etc. The processing of the preprocessed video data is similar and will not be repeated here.

[0075] Use BERT to convert preprocessed text data into numerical vectors for fusion with other types of data;

[0076] The specific process includes inputting the preprocessed text into the BERT model, which captures the semantic information in the text through a multi-layer Transformer encoder. Finally, the output layer obtains the embedding vector of each word, which contains rich semantic information and can be used for subsequent data fusion.

[0077] S2. Use the multimodal fusion algorithm to fuse the preprocessed fire scene data to form a multimodal feature vector.

[0078] The extracted feature vectors are weighted and fused through the attention mechanism, and the expression is:

[0079]

[0080] Among them, F is the comprehensive feature vector F obtained by weighted fusion, represents the i-th mode, which includes four modes: image data features, audio data features, video data features, and text numerical vectors. represents the attention weight of the i-th modality, and i represents the index of different modalities.

[0081] It should also be noted that for each mode’s eigenvector , calculate the attention weight , the expression is:

[0082]

[0083] in, is the dimension of the feature vector, is normalized, is the query vector used to calculate the importance of each modality feature.

[0084] S3. Based on historical fire data, a fire situation prediction model is trained, and the multimodal feature vector is input into the fire situation prediction model to predict the comprehensive index of fire development.

[0085] Furthermore, using the LSTM model as the infrastructure, historical fire data is input into the LSTM model for training, and the LSTM model is trained into a fire situation prediction model that can predict the fire situation;

[0086] The cross entropy loss function is used to measure the accuracy of the fire situation prediction model. The expression is:

[0087]

[0088] in, represents the average cross entropy loss over the entire training set, represents the total number of samples in the training set, represents the cross entropy loss at the tth time point, represents the predicted output of the model at time point t, Indicates the actual fire intensity at the true value of the comprehensive index at the tth time point;

[0089] The loss function is reduced by the Adam optimizer, and the expression is:

[0090]

[0091] in, represents the updated value of the model parameters after the tth iteration, represents the value of the model parameters after the t-1th iteration, represents the learning rate, represents the weighted moving average of the gradient, represents the weighted moving average of the squared gradient, represents a minimum positive value;

[0092] The cross-validation method is used to optimize the parameters of the fire situation model to ensure that the model performs stably on different data subsets. The specific process includes dividing the entire data set into multiple equal subsets, selecting one of the subsets as the validation set each time, and the remaining subsets as the training set for model training, and evaluating the model performance on the validation set. By performing multiple iterations in this way, the present invention can evaluate the performance of the model on different data subsets, and adjust the model parameters accordingly to achieve the goal of maintaining stable performance on different data subsets.

[0093] The multimodal feature vector , input into the fire situation prediction model to predict the comprehensive index of fire development, the expression is:

[0094]

[0095] in, Indicates at time When is the multimodal feature vector, is the comprehensive index of fire development. is the number of multimodal features, represents the attention weight of the i-th modality, represents the eigenvector of the i-th mode, T represents the maximum considered time range, t represents the time difference from the start of prediction to the current moment, c is the time attenuation coefficient, represents the normalized weight of the ith mode.

[0096] It should be noted that the time attenuation coefficient c is a positive real number, which determines the comprehensive index of fire development. The decay rate over time t. A larger c value means that the exponential decays faster over time, while a smaller c value means that the decay is slower.

[0097] During the model training process, the cross-validation method is used to find the optimal c value to ensure that the model performs stably on different data subsets.

[0098] S4. Analyze the fire development trend based on the comprehensive fire development index and generate targeted response strategies.

[0099] Furthermore, by analyzing past fire cases, a comprehensive fire development index was extracted. Based on the data, a critical value g was determined. It represents a critical point in the development of the fire.

[0100] when >g, it means the current fire is rapidly deteriorating;

[0101] when When ≤g, it means the current fire is within controllable range;

[0102] When the fire is rapidly deteriorating, the response strategy includes immediately initiating emergency evacuation procedures, ensuring the safe evacuation of personnel, and dispatching more firefighting teams, ambulances and rescue supplies, including fire-fighting equipment, first aid medicines, food, water, tents and other necessary survival supplies;

[0103] When the fire is currently under control, the response strategy includes implementing targeted firefighting operations, monitoring the fire development and maintaining the safety of the surrounding areas.

[0104] It should be noted that directional firefighting means to select firefighting strategies and firefighting points in a targeted manner according to the specific conditions at the fire scene, so as to extinguish the fire in the most effective way. Specifically, it includes accurately determining the location of the fire source, selecting appropriate firefighting equipment according to the nature of the fire source, concentrating firefighting resources in the areas where the fire is most serious, and taking measures to prevent the fire from spreading to surrounding areas.

[0105] S5. Conduct simulation drills on the generated response strategies, evaluate the implementation effects of the response strategies, optimize the response strategies based on the evaluation results, and obtain emergency plans.

[0106] Furthermore, based on the generated response strategies, simulated fire conditions are designed, including factors such as the location of the fire, wind speed, and wind direction, to simulate different stages from small-scale initial fires to medium-scale spreading fires and large-scale uncontrolled fires;

[0107] A detailed virtual city model was built using ArcGIS, including accurate building layout, road network, population density distribution and other elements. At the same time, real-time meteorological data was integrated to implement response strategies in the virtual city model to ensure that the simulation exercise was as close to the actual situation as possible;

[0108] When implementing the response strategy in the virtual environment, the evacuation efficiency, property loss degree and rescue operation speed are considered. A strategy evaluation function is designed to calculate the effectiveness score of the response strategy. The expression is:

[0109]

[0110] in, Indicates the efficiency of personnel evacuation. Indicates the extent of property damage. Indicates the speed of rescue operations. Represents the weight coefficient of personnel evacuation efficiency, The weight coefficient representing the degree of property loss, represents the weight coefficient of the rescue operation efficiency, represents the effectiveness score of the coping strategies;

[0111] It should be noted that different response strategies directly affect the efficiency of personnel evacuation, the extent of property damage, and the speed of rescue operations, thus changing the results of the effectiveness score;

[0112] Evacuation efficiency E: It is evaluated by calculating the proportion of people successfully evacuated within a given time. This involves the speed and efficiency of evacuation in simulation exercises.

[0113] Extent of property damage : Calculated by assessing the property damage caused by the fire. This includes estimating the value of property that was damaged or completely destroyed.

[0114] Rescue operation speed R: It is evaluated by measuring the time from the occurrence of fire to the arrival of rescue forces at the scene and the start of rescue operations. This reflects the response speed and efficiency of the rescue operation.

[0115] Define a strategy effectiveness threshold e, and combine it with the strategy effectiveness score to evaluate the effectiveness of the strategy, which involves multiple aspects such as personnel evacuation efficiency, property loss degree, and rescue operation speed;

[0116] when When ≥e, it means that the response strategy is effective and can effectively control the fire, where e is the strategy effectiveness threshold;

[0117] when When <e, it means that the response strategy is ineffective and it is difficult to control the fire.

[0118] It should also be noted that the definition of the strategy effectiveness threshold e is specifically determined based on statistical analysis of a large number of historical fire case data. , a reasonable interval can be found, indicating that the strategy is highly effective, which aims to ensure that the strategy evaluation can reflect the efficiency requirements in practical applications, thereby providing a scientific basis for decision-making.

[0119] When the response strategy is ineffective and it is difficult to control the fire, review the evacuation drill records, check the evacuation efficiency, property loss rate and rescue operation speed, and make corresponding adjustments based on the inspection results, including adjusting the evacuation route, adding evacuation exits and adjusting the evacuation sequence;

[0120] After adjustment, the implementation effect of the response strategy is evaluated again until the response strategy is effective and an urban fire emergency plan is generated;

[0121] When the response strategy is effective and can effectively control the fire, an urban fire emergency plan is generated based on the current response strategy.

[0122] It should also be noted that adjusting the evacuation route specifically means: if the evacuation efficiency is not high, the evacuation route can be replanned, the number of evacuation exits can be increased, or evacuation signs can be improved to increase the evacuation speed.

[0123] To reduce the extent of property loss, specifically: If the property loss is serious, the protection measures for important assets can be strengthened, such as using fire-proof materials to reinforce key facilities, or optimizing the configuration of property protection materials.

[0124] To improve the speed of rescue operations, specifically: if the response speed of the rescue operation is not fast enough, the deployment and dispatch of the rescue team can be optimized to ensure that rescue supplies and personnel can arrive at the scene quickly.

[0125] This embodiment also provides an urban fire emergency plan simulation system based on AI digitization, including:

[0126] Data acquisition and preprocessing module, used to collect fire scene data and perform preprocessing;

[0127] A feature extraction and fusion module is used to fuse the pre-processed fire scene data into a multi-modal feature vector using a multi-modal fusion algorithm;

[0128] The fire situation prediction module is used to train a fire situation prediction model based on historical fire data, input the multimodal feature vector into the fire situation prediction model, and predict the comprehensive index of fire development;

[0129] The response strategy generation module is used to analyze the fire development trend and generate targeted response strategies based on the comprehensive fire development index;

[0130] The strategy evaluation and optimization module is used to simulate and exercise the generated response strategies, evaluate the execution effect of the response strategies, optimize the response strategies based on the execution effect, and obtain emergency plans.

[0131] This embodiment also provides a computer device, which is suitable for the case of an urban fire emergency plan deduction method based on AI digitization, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the urban fire emergency plan deduction method based on AI digitization as proposed in the above embodiment.

[0132] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0133] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing the urban fire emergency plan deduction method based on AI digitization as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0134] In summary, the present invention achieves accurate prediction of the comprehensive index of fire development by: collecting and preprocessing multimodal data at the fire scene, using a multimodal fusion algorithm to extract features and construct a fire situation prediction model. Furthermore, a response strategy is generated based on the fire development trend, and the effectiveness of the strategy is evaluated through simulation exercises, ultimately achieving the purpose of optimizing the response strategy. This method not only improves the accuracy of fire prediction, but also can flexibly adjust the response strategy according to changes in the fire, significantly enhancing the efficiency and effectiveness of urban fire emergency management. Ultimately, this invention can provide strong technical support for urban fire safety and protect people’s lives and property.

[0135] Example 2, referring to Table 1, is the second example of the present invention. To further verify the effect of the present invention, experimental simulation data of the urban fire emergency plan deduction method based on AI digitization is provided.

[0136] In order to better demonstrate the effect of the present invention in practical application, this experiment conducted three simulated fire drills, namely a small-scale initial fire, a medium-scale spreading fire, and a large-scale uncontrolled fire. In each drill, the execution effect of the response strategy in the two cases of using the prior art and the present invention was compared.

[0137] First, images, videos, audio, and text data of the fire scene were collected through various means such as drones, thermal imagers, cameras, and social media platforms, and necessary preprocessing was performed, including noise removal, unification of size and format, to ensure data quality and consistency.

[0138] Secondly, advanced deep learning technologies such as convolutional neural networks (CNN), recurrent neural networks (RNN) and BERT are used to extract key features from images, videos and texts, and weighted fusion is performed through the attention mechanism to form a comprehensive multimodal feature vector for building a fire situation prediction model.

[0139] Next, in a virtual urban environment, fire scenarios were simulated according to different fire scales (small-scale initial fire, medium-scale spreading fire, and large-scale uncontrolled fire), and response strategies were generated based on the constructed prediction model.

[0140] Finally, the execution effects of these strategies were evaluated, including key indicators such as personnel evacuation efficiency, property loss degree, and rescue operation speed, and the effectiveness scores of the strategies were calculated through the strategy evaluation function, as shown in Table 1:

[0141] Table 1

[0142]

[0143] By analyzing the data in the above table, we can clearly see the effect of the method of the present invention. Compared with traditional methods, the present invention significantly improves the accuracy of fire prediction by integrating multiple data sources and using advanced deep learning technology. For example, in terms of personnel evacuation efficiency, the method of the present invention reduces the evacuation time by an average of 20%, and in terms of reducing property losses, the method of the present invention can provide early warning, reducing property losses by about 35%.

[0144] The AI-digitalized urban fire emergency plan deduction method of the present invention can not only significantly improve the accuracy of fire prediction, but also formulate the best emergency response strategy in the first place, especially for complex and changeable fire scenes, which can respond quickly. This reflects the significant advantages of the present invention in improving rescue efficiency, reducing property losses and ensuring personnel safety.

[0145] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for deducing urban fire emergency plans based on AI digitization, characterized by: include, Collect fire scene data and perform pre-processing; The pre-processed fire scene data are fused using a multi-modal fusion algorithm to form a multi-modal feature vector; Based on historical fire data, a fire situation prediction model is trained, and the multimodal feature vector is input into the fire situation prediction model to predict the comprehensive index of fire development; Analyze the fire development trend and generate targeted response strategies based on the comprehensive fire development index; Conduct simulation exercises on the generated response strategies, evaluate the execution effects of the response strategies, optimize the response strategies based on the execution effects, and obtain emergency plans; The specific steps of collecting fire scene data and preprocessing are as follows: Collect fire scene data through drones, thermal imagers, cameras and media platforms; fire scene data includes image data, video data, audio data and text data of the fire scene; Pre-process the fire scene data by cleaning and standardizing; Use convolutional neural networks to extract features from preprocessed image data; Use recurrent neural network to extract preprocessed audio data features and video data features; Use BERT to convert the preprocessed text data into numerical vectors; The multimodal fusion algorithm is used to fuse the pre-processed fire scene data to form a multimodal feature vector. The specific steps are as follows: The extracted feature vectors are weighted and fused through the attention mechanism, and the expression is: Among them, F is the comprehensive feature vector obtained by weighted fusion, represents the i-th mode, which includes four modes: image data features, audio data features, video data features, and text numerical vectors. represents the attention weight of the i-th modality, i represents the index of different modalities; Based on the historical fire data, a fire situation prediction model is trained, and the multimodal feature vector is input into the fire situation prediction model to predict the comprehensive index of fire development. The specific steps are as follows: Using the LSTM model as the basic architecture, historical fire data is input into the LSTM model for training, and the LSTM model is trained into a fire situation prediction model that can predict fire situations; The cross entropy loss function is used to measure the accuracy of the fire situation prediction model; Reduce the loss function through the Adam optimizer; Use cross-validation method to optimize the parameters of fire situation prediction model; The multimodal feature vector , input into the fire situation prediction model to predict the comprehensive index of fire development, the expression is: in, Indicates at time When is the multimodal feature vector, is the comprehensive index of fire development. is the number of multimodal features, represents the attention weight of the i-th modality, represents the eigenvector of the ith mode, T represents the maximum considered time range, t represents the time difference from the start of prediction to the current moment, c is the time attenuation coefficient, represents the normalized weight of the ith modality.

2. The method for deducing urban fire emergency plans based on AI digitization according to claim 1, characterized in that: According to the fire development comprehensive index, the fire development trend is analyzed and a targeted response strategy is generated. The specific steps are as follows: when When >g, it means that the current fire is rapidly deteriorating; where g is the fire controllability threshold; when When ≤g, it means the current fire is within controllable range; When the current fire is rapidly deteriorating, the response strategy includes immediately initiating emergency evacuation procedures, dispatching more firefighting teams, ambulances and rescue supplies; When the fire is currently under control, the response strategy includes implementing targeted firefighting operations, monitoring the fire development and maintaining the safety of the surrounding areas.

3. The method for deducing urban fire emergency plans based on AI digitization according to claim 2 is characterized in that: Conduct simulation exercises on the generated response strategies to evaluate their implementation effects. The specific steps are as follows: Based on the generated response strategies, simulated fire conditions are designed, including the location of the fire, wind speed and direction, to simulate different stages from small-scale initial fires to medium-scale spreading fires and large-scale uncontrolled fires; Build a detailed virtual city model using ArcGIS, including accurate building layout, road network, and population density distribution. At the same time, integrate real-time meteorological data to implement response strategies in the virtual city model. Calculate the effectiveness score of the coping strategy, the expression is: in, Indicates the efficiency of personnel evacuation. Indicates the extent of property damage, Indicates the speed of rescue operations. Represents the weight coefficient of personnel evacuation efficiency, The weight coefficient representing the degree of property loss, represents the weight coefficient of the rescue operation efficiency, represents the effectiveness score of the coping strategies; when When ≥e, it means that the response strategy is effective and can effectively control the fire, where e is the strategy effectiveness threshold; when When <e, it means that the response strategy is ineffective and it is difficult to control the fire.

4. The AI-digitalized urban fire emergency plan deduction method according to claim 3 is characterized by: The response strategy is optimized based on the execution effect to obtain an emergency plan. The specific steps are as follows: When the response strategy is ineffective and it is difficult to control the fire, review the evacuation drill records, check the evacuation efficiency, property loss rate and rescue operation speed, and make corresponding adjustments based on the inspection results, including adjusting the evacuation route, adding evacuation exits and adjusting the evacuation sequence; After adjustment, the implementation effect of the response strategy is evaluated again until the response strategy is effective and an urban fire emergency plan is generated; When the response strategy is effective and can effectively control the fire, an urban fire emergency plan is generated based on the current response strategy.

5. A system using the AI ​​digitalized urban fire emergency plan deduction method according to any one of claims 1 to 4, characterized in that: include: Data acquisition and preprocessing module, used to collect fire scene data and perform preprocessing; A feature extraction and fusion module is used to fuse the pre-processed fire scene data into a multi-modal feature vector using a multi-modal fusion algorithm; The fire situation prediction module is used to train a fire situation prediction model based on historical fire data, input the multimodal feature vector into the fire situation prediction model, and predict the comprehensive index of fire development; The response strategy generation module is used to analyze the fire development trend and generate targeted response strategies based on the comprehensive fire development index; The strategy evaluation and optimization module is used to simulate and exercise the generated response strategies, evaluate the execution effect of the response strategies, optimize the response strategies based on the execution effect, and obtain emergency plans.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the urban fire emergency plan deduction method based on AI digitization are implemented as described in any one of claims 1 to 4.

7. 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 urban fire emergency plan deduction method based on AI digitization are implemented as described in any one of claims 1 to 4.

Images