An AI city emergency management service system and method based on multi-source data fusion

Abstract

The present invention relates to the field of digital data processing technology, and specifically to an AI city emergency management service system and method based on multi-source data fusion, including: continuously monitoring streets through a public camera network to obtain monitoring image information; using a convolutional neural network to identify fire conditions based on the monitoring image information, and generating disaster information when a disaster is discovered; when a user discovers a fire, uploading disaster photo information through a wireless terminal; based on the disaster photo information, calling a camera in a corresponding area to obtain corresponding disaster information, and notifying a fire fighting system; distinguishing disaster levels based on the flame area in the disaster information, and when the disaster level is lower than a preset value, sending the location information of fire extinguishing resources and disaster information to user terminals within a preset range near the disaster location. Thus, the existing video surveillance network and public mobile terminals are directly used to manage fires, improve work efficiency, and reduce the difficulty of system construction.

Description

Technical Field

[0001] The present invention relates to the technical field of digital data processing, and in particular to an AI city emergency management service system and method based on multi-source data fusion. Background Art

[0002] Urban emergency management services are a key mechanism to ensure the safety of citizens and social stability. They cover the entire process of management from prevention, preparation to response and recovery, and are designed to effectively respond to various emergencies, including natural disasters (such as earthquakes, floods, and typhoons) and man-made disasters (such as fires, traffic accidents, and terrorist attacks). By establishing a sound early warning system, formulating emergency plans, organizing training and drills, and coordinating resources from all parties, this service can be quickly activated when an emergency occurs, providing timely rescue and support to minimize casualties and property losses, and help cities quickly resume normal operations.

[0003] There is an existing fire monitoring method that uses smoke data, temperature data, drone image data, and manual inspection data to comprehensively warn of fire situations. However, the above data is difficult to collect on a large scale through the Internet, and the solution is difficult to implement, thereby reducing work efficiency. Summary of the invention

[0004] The purpose of the present invention is to provide an AI city emergency management service system and method based on multi-source data fusion, aiming to directly utilize the existing video surveillance network and public mobile terminals to manage fires, improve work efficiency and reduce the difficulty of system installation.

[0005] To achieve the above-mentioned object, in a first aspect, the present invention provides an AI city emergency management service method based on multi-source data fusion, comprising continuously monitoring streets through a public camera network to obtain monitoring image information;

[0006] A convolutional neural network is used to identify fire conditions based on monitoring image information, and disaster information is generated when a disaster is discovered, wherein the disaster information includes the flame area and the disaster location;

[0007] When a user discovers a fire, he uploads disaster photo information through a wireless terminal. The photo information includes photo data, location information and sending time;

[0008] Based on the disaster photo information, the camera in the corresponding area is called to obtain the corresponding disaster information and notify the fire fighting system;

[0009] The disaster level is distinguished based on the flame area in the disaster information, and when the disaster level is lower than a preset value, the positioning information of the fire-fighting resources and the disaster information are sent to user terminals within a preset range near the disaster location.

[0010] The specific steps of continuously monitoring the street through the public camera network to obtain monitoring image information include:

[0011] Configure the camera to capture video data at set intervals;

[0012] Extracting surveillance images from video data at preset intervals;

[0013] Compress the monitoring image and upload it to the host computer.

[0014] The specific steps of using a convolutional neural network to identify fire conditions based on monitoring image information and generating disaster information when a disaster is discovered include:

[0015] Construct a convolutional neural network model based on fire image information to obtain a judgment model;

[0016] Get the uploaded surveillance images;

[0017] The monitoring image is input into the trained judgment model to identify the fire situation. When a disaster is found, disaster information is generated and a video clip acquisition request is sent to the corresponding target camera.

[0018] The target camera captures video clips within the preset time range of the surveillance image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flame area and disaster location.

[0019] The specific steps of constructing a convolutional neural network model based on fire image information to obtain a judgment model include:

[0020] Collect images containing normal conditions and different types of fire scenes;

[0021] Add labels to each image to get the dataset;

[0022] The convolutional neural network model is trained using the labeled data set, and the judgment model is obtained through multiple iterations.

[0023] The target camera captures video clips within a preset time range of the monitoring image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flame area and disaster location. The specific steps include:

[0024] According to the timestamp of the disaster, determine the time range of the video clip to be captured, and obtain the corresponding video clip;

[0025] Apply video compression technology to compress video clips before uploading;

[0026] Extract sample images from the video clips and identify the fire area and damage location.

[0027] Among them, when the user discovers a fire, the user uploads disaster photo information through a wireless terminal, and the photo information includes photo data, location information and sending time. The specific steps include:

[0028] Users use wireless terminals to take photos of fires;

[0029] Record the location coordinates when the fire photo was taken to obtain positioning information;

[0030] Upload fire photos and location information to the host computer and record the sending time.

[0031] The specific steps of calling the camera in the corresponding area based on the disaster information to obtain the corresponding disaster image group and notifying the fire fighting system include:

[0032] Centered on the location information, define the sphere of influence on the map;

[0033] Extract all available cameras within the impact range, and obtain the target camera closest to the fire location that captured the fire situation;

[0034] For the target camera, the disaster image group within the period before and after the sending time is extracted and transmitted to the fire protection system.

[0035] In a second aspect, the present invention also provides an AI city emergency management service system based on multi-source data fusion, including a camera monitoring module, a fire identification module, a fire information uploading module, a disaster information notification module and a user notification module;

[0036] The camera monitoring module is used to continuously monitor the street through a public camera network to obtain monitoring image information;

[0037] The fire identification module is used to identify the fire situation using a convolutional neural network based on the monitoring image information, and generate disaster information when a disaster is found, wherein the disaster information includes the flame area and the disaster location;

[0038] The fire information uploading module is used for users to upload disaster photo information through a wireless terminal when they discover a fire. The photo information includes photo data, location information and sending time;

[0039] The disaster information notification module is used to call the camera in the corresponding area to obtain the corresponding disaster information based on the disaster photo information, and notify the fire protection system;

[0040] The user notification module is used to distinguish the disaster level based on the flame area in the disaster information, and when the disaster level is lower than a preset value, send the location information of the fire extinguishing resources and the disaster information to the user terminals within a preset range near the disaster location.

[0041] The present invention discloses an AI city emergency management service system and method based on multi-source data fusion, which utilizes a public camera network throughout the city to realize 24-hour uninterrupted monitoring of street conditions. The image information captured by these cameras will be used as one of the basic data sources to provide materials for subsequent analysis and processing. For the monitoring of emergencies such as fires, the system introduces the convolutional neural network (CNN) technology in deep learning, which can automatically identify fire characteristics from a large number of monitoring images, such as the presence or absence of flames, their coverage area and specific location. Once a fire is detected, the system will immediately generate detailed disaster information, including but not limited to the scope of the flame and the precise geographical location. At the same time, considering that not all areas can be fully covered by cameras, or in some cases the cameras may fail to detect the disaster in time, the system also encourages citizens to participate in emergency management. Users can upload disaster photos containing photo data, geographic positioning information and timestamps through personal wireless terminal devices, such as smart phones, when witnessing a fire. After receiving the photos uploaded by the user, the system will quickly call nearby cameras to obtain more detailed on-site conditions based on the location information in the photos, and compare and confirm them with the information previously analyzed by CNN. If a fire does occur, the system will immediately activate the emergency response mechanism and send an alarm to the relevant fire department to ensure that rescue forces can arrive at the scene as soon as possible. In order to allocate fire-fighting resources more scientifically, the system will also judge the severity of the disaster based on the flame area in the disaster information, and divide the disaster into different levels accordingly. For small-scale fires whose assessment results are lower than the preset safety threshold, the system will take more flexible response measures. It sends the location information of nearby fire-fighting resources (such as portable fire extinguishers, small fire trucks, etc.) together with the details of the disaster to the mobile terminals of residents or staff within a specific range around the disaster site. This allows nearby personnel to initially control the fire before the arrival of professional fire brigades, thereby minimizing losses and improving management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 It is a flow chart of an AI city emergency management service method based on multi-source data fusion of the present invention.

[0044] Figure 2 It is a flow chart of the present invention for continuously monitoring streets through a public camera network to obtain monitoring image information.

[0045] Figure 3 The present invention adopts a convolutional neural network to identify fire conditions based on monitoring image information and generates a flow chart of disaster information when a disaster is discovered.

[0046] Figure 4 The present invention constructs a convolutional neural network model based on fire image information to obtain a flow chart of a judgment model.

[0047] Figure 5 The target camera of the present invention intercepts video clips within a preset time range of the monitoring image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flow chart of the flame area and the disaster location.

[0048] Figure 6 It is a flowchart of the user of the present invention uploading disaster photo information through a wireless terminal when discovering a fire, wherein the photo information includes photo data, location information and sending time.

[0049] Figure 7 It is a flowchart of the present invention for calling the camera in the corresponding area based on the disaster information to obtain the corresponding disaster image group and notifying the fire fighting system. DETAILED DESCRIPTION

[0050] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0051] First embodiment

[0052] See also Figure 1 to Figure 7 The present invention provides an AI city emergency management service method based on multi-source data fusion, comprising:

[0053] S101 continuously monitors the streets through a public camera network to obtain monitoring image information;

[0054] The specific steps include:

[0055] S201 configures the camera to capture video data at a set time interval;

[0056] Configure each public camera according to actual needs and environmental conditions. This includes selecting a suitable installation location, adjusting the angle and focal length of the camera to ensure coverage of all key areas that need to be monitored. At the same time, technicians will also set the camera's operating parameters, such as frame rate (the number of frames captured per second), resolution, and time interval. This time interval determines when the camera starts capturing video data, which is crucial to ensuring monitoring continuity and reducing unnecessary data volume.

[0057] S202 extracts surveillance images from the video data at preset intervals;

[0058] Process the video stream captured by the camera. The key here is "preset time" - that is, extracting single-frame images from the continuous video stream at a preset time interval as monitoring images. This process not only helps to reduce the burden of subsequent data processing, but also ensures that the obtained images are representative enough to accurately reflect the actual situation during that time period.

[0059] S203 compresses the monitoring image and uploads it to the host computer.

[0060] Compression technology significantly reduces file size while maintaining image quality, thereby reducing bandwidth usage and increasing transmission speed. The compressed image is then uploaded to a host computer (usually located in a data center or cloud server) for further analysis.

[0061] S102 uses a convolutional neural network to identify the fire situation based on the monitoring image information, and generates disaster information when a disaster is found, wherein the disaster information includes the flame area and the disaster location;

[0062] The specific steps include:

[0063] S301 constructs a convolutional neural network model based on fire image information to obtain a judgment model;

[0064] The specific steps include:

[0065] S401 collects images including normal conditions and different types of fire scenes;

[0066] Build a comprehensive and representative image library. This process involves collecting image data from a wide range of environments, including not only normal, non-fire conditions, but also multiple different types of fire scenarios. This includes indoor and outdoor scenes, images of different times and weather conditions, and fires caused by different burning materials.

[0067] S402 adds a label to each image to obtain a data set;

[0068] In order for machine learning algorithms to understand and learn from them, each image must be properly labeled. The label should accurately describe the content of the image, such as whether it contains a fire, and if so, the location and size of the flames. This process requires manual annotation or the use of semi-automatic tools to improve efficiency and accuracy.

[0069] S403 uses the labeled data set to train the convolutional neural network model, and obtains the judgment model through multiple iterations.

[0070] Use the prepared labeled dataset to train the convolutional neural network model. In this process, the model will learn how to distinguish between normal conditions and fire scenes, and gradually optimize its internal parameters to improve prediction accuracy. Training is usually an iterative process, during which the model structure and hyperparameters are constantly adjusted until satisfactory performance indicators are achieved. After sufficient training, the convolutional neural network will become a reliable judgment model that responds quickly when receiving new surveillance images, accurately identifies whether there is a fire, and assesses the flame area and determines the location of the disaster.

[0071] S302 obtains the uploaded surveillance image;

[0072] First, it is necessary to obtain the uploaded surveillance images from each distributed camera node. These images are uploaded to the central server or cloud platform through the S101 process described earlier, after time interval setting, image extraction and compression. In order to ensure the security and reliability of data transmission, encrypted communication protocols are usually used, and load balancing strategies are involved to optimize data distribution between multiple servers.

[0073] After receiving the image, some preliminary preprocessing operations are performed, such as decompression, format conversion, etc., to ensure that the image meets the input standards required for subsequent analysis.

[0074] S303 inputs the monitoring image into the trained judgment model to identify the fire situation, generates disaster information when a disaster is found, and sends a video clip acquisition request to the corresponding target camera;

[0075] The pre-processed surveillance images will be fed into the trained convolutional neural network (CNN) model for inference. This process uses the powerful pattern recognition capabilities of deep learning to quickly and accurately determine whether there are signs of fire in the image. The model can not only identify the flame itself, but also analyze factors such as smoke morphology and color changes, thereby improving the accuracy of detection.

[0076] Once the model detects a fire, it will immediately generate detailed disaster information. This includes but is not limited to the specific location of the fire, the area covered, and the severity assessment of the fire. This information is crucial for emergency response teams as it helps them plan rescue operations more effectively, prioritize resource allocation, and take the most appropriate firefighting measures.

[0077] In addition to generating disaster information, the system automatically sends instructions to relevant target cameras, asking them to provide complete video clips before and after the fire. This measure helps with post-incident investigation and evidence preservation, and also provides richer contextual information for real-time monitoring. The selection of video clips is usually based on the precise time and location of the disaster to ensure that the most relevant scenes are captured.

[0078] The S304 target camera captures video clips within a preset time range of the monitoring image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flame area and disaster location.

[0079] The specific steps include:

[0080] S501 determines the time range of the video clip to be captured according to the timestamp of the disaster occurrence, and obtains the corresponding video clip;

[0081] The specific time range of the video clips to be captured is determined based on the timestamp generated when the fire is detected. This time range usually includes a period of time before the fire starts (e.g. 1-2 minutes) and during the duration of the fire, so as to fully understand the occurrence of the incident. Selecting an appropriate time window is crucial for subsequent analysis, because it should not be too short to miss important information, nor too long to cause excessive data volume and affect processing efficiency.

[0082] Once the time range is determined, the system sends instructions to the relevant target cameras, asking them to capture and provide video clips within the specified time period. To ensure data integrity and accuracy, some redundancy measures are taken, such as requesting additional seconds of video from adjacent time periods as a buffer.

[0083] S502 compresses the video clip using video compression technology and then uploads it;

[0084] The received raw video clips will be processed by applying efficient compression algorithms. This step is designed to reduce the file size, thereby speeding up uploads and saving storage space. Commonly used compression technologies include encoding formats such as H.264 and H.265, which significantly reduce the bit rate while maintaining visual quality. The compression process also needs to consider retaining enough details to support subsequent image recognition tasks.

[0085] The compressed video clips are then uploaded to a central server or cloud platform via a secure communication protocol for further analysis by the trained convolutional neural network (CNN) model. To ensure the security of data transmission, encryption technology and authentication mechanisms are used to protect sensitive information from unauthorized access.

[0086] S503 extracts sample images from the video clips and identifies the fire area and the disaster location.

[0087] The system will extract sample images from the uploaded video clips at a certain frequency (such as one or more frames per second). These sample images will become the basis for the next step of analysis, helping to more accurately locate the flames and assess the fire. The selection of sampling frequency requires a balance between analysis accuracy and computing resource consumption.

[0088] The trained CNN model is used to perform in-depth analysis on the sample images to identify the specific location and coverage area of ​​the flames. The model distinguishes the flame area from other background elements based on the feature patterns learned in previous training, such as color distribution and texture changes. For each identified flame area, the system calculates its geometric dimensions and then derives the flame area. At the same time, the location information of the camera and other auxiliary data are combined to accurately locate the geographical location of the disaster.

[0089] After completing the above analysis, the system will make a final confirmation of the fire situation and update the relevant disaster information. If new situations or changes are found, such as the fire spreading to a new area, the system will immediately adjust the response strategy and send an update notification to the relevant departments.

[0090] S103 When a user discovers a fire, he uploads disaster photo information through a wireless terminal, and the photo information includes photo data, location information and sending time;

[0091] The specific steps include:

[0092] S601 User uses a wireless terminal to take photos of the fire;

[0093] When users perceive a fire, they immediately use their smart devices (such as smartphones or tablets) as wireless terminals to record the fire scene. These devices are usually equipped with high-quality cameras that can capture clear photos and provide valuable information for subsequent analysis.

[0094] Users should take as many photos of the fire from different angles as possible to fully reflect the scale, location and scope of the fire. To ensure the quality of the photos, users should pay attention to the lighting conditions to avoid overexposure or underexposure that affects the readability of the image.

[0095] S602 records the location coordinates when the fire photo was taken to obtain positioning information;

[0096] Almost all modern wireless terminals have built-in global positioning system (GPS) modules, which can obtain precise location coordinates while taking photos. This step is crucial to determine the exact location of the fire. The system automatically records the latitude and longitude information at the time of shooting and associates it with the photo to form a complete disaster report.

[0097] In some cases, such as indoors or in areas with poor signal, GPS positioning is not accurate enough. Users also manually enter the approximate address or reference description for subsequent confirmation.

[0098] S603 uploads the fire photo and location information to the host computer and records the sending time.

[0099] Once the photo is taken and the location information is obtained, the user packages the data into a file package through the application of the wireless terminal. The application is responsible for organizing all the necessary information, including photo data, location coordinates, and the current timestamp, and then uploads it to the host computer (i.e., central server or cloud platform) through an Internet connection (such as 4G / 5G, Wi-Fi, etc.). In order to ensure the security and integrity of the transmission, encryption technology and error checking mechanisms are usually used. During the upload process, the system will record the exact sending time on the server side, which is very important for tracking the timeline of events and evaluating the response efficiency. Each uploaded data packet will be assigned a unique identifier for future query and management. In addition, the server will check whether the received data is complete and feedback a confirmation message to the user's wireless terminal to inform whether the upload is successful or not.

[0100] S104 calls the camera in the corresponding area to obtain corresponding disaster information based on the disaster photo information, and notifies the fire fighting system;

[0101] The specific steps include:

[0102] S701 uses the location information as the center and defines the impact area on the map;

[0103] The exact location of the fire is determined based on the location information in the photos uploaded by the user. This location information provides the geographic coordinates (such as longitude and latitude) of the fire scene as the basis for subsequent analysis. Next, the system will delineate a reasonable disaster impact range on the digital map with the location information as the center. The size of this range depends on many factors, including the type of fire, environmental conditions, and the radius of impact. For example, for a small indoor fire, the impact range will be smaller; while for a forest fire or a chemical plant fire, a larger area needs to be delineated. In addition, considering the impact of factors such as wind direction and terrain on the spread of the fire, the system also dynamically adjusts the boundaries of the impact range. In order to better understand the situation in the affected area, the system will mark and conduct risk assessments on important buildings, transportation routes, densely populated areas, etc. in the area. This information helps to provide decision support for rescue operations and ensure the optimal allocation of resources.

[0104] S702 extracts all available cameras within the impact range, and obtains a target camera that is closest to the fire location and has captured the fire situation;

[0105] Query the database to extract all available public cameras within the defined impact area. This involves integrating camera resources from different departments (such as traffic management, urban security, etc.) to form a comprehensive monitoring network. The location, coverage and current status of each camera will be recorded in detail. From these cameras, the system will select those devices that can clearly capture the location of the fire, especially the cameras that are closest to the fire and have the best viewing angle as "target cameras." The selection criteria not only consider distance, but also include factors such as camera resolution, angle, and whether there are no obstacles. If there are multiple qualified cameras, the system will give priority to those with night vision or high-definition quality to ensure the best quality image data.

[0106] S703 extracts the disaster image group within the period before and after the sending time from the target camera and transmits it to the fire protection system.

[0107] According to the timestamp of the disaster photos uploaded by the user, a reasonable time window is set, which usually includes a few minutes to more than ten minutes before and after the fire photos are sent. The selection of this time period is intended to capture the key stages of the fire development, neither missing important information nor increasing unnecessary data due to excessive time. The system will extract all video frames or regularly extracted image samples within this time period from the selected target camera to form a complete "disaster image group". In order to facilitate transmission and storage, these images will be appropriately compressed, but at the same time ensure sufficient clarity for subsequent analysis.

[0108] Finally, the system will quickly transmit the extracted disaster image group to the fire system through a secure channel for reference by the command center and front-line rescue personnel. At the same time, the system will automatically generate a detailed disaster report containing information such as the fire location, impact range, image links, etc., and immediately notify the relevant fire department.

[0109] S105 distinguishes the disaster level based on the flame area in the disaster information, and when the disaster level is lower than the preset value, sends the location information of the fire-fighting resources and the disaster information to the user terminals within a preset range near the disaster location. Specifically, it includes calculating the number of pixels covered by the flame and converting it into the actual area in combination with the camera parameters; when the disaster level is lower than the preset safety threshold, the fire resource scheduling program is immediately started. This process includes finding the available fire trucks, personnel and other necessary equipment closest to the fire location; determining the preset geographical range where the user terminals that need to receive the alarm are located, such as within a certain radius centered on the fire point; filtering out all user terminals within the geographical range from the registered services; constructing the message content containing the location information of the fire-fighting resources, the estimated arrival time and a brief description of the disaster, and then pushing it to the selected user terminal via SMS, application notification or other appropriate means.

[0110] The system uses image processing algorithms to analyze the disaster image set obtained from the camera and identify the fire area. This usually involves analyzing color, texture and other features to accurately distinguish between flames and other background elements.

[0111] After identifying the flame areas, the system calculates the number of pixels covered by these areas. In order to convert the number of pixels into actual area, it is necessary to combine the parameters of the camera, such as focal length, installation height, tilt angle, etc., as well as the shooting distance. Through these parameters, a mapping relationship is established to accurately calculate the actual coverage area of ​​the flame. This process also needs to consider factors such as image distortion correction to improve measurement accuracy.

[0112] Based on the pre-set safety threshold (such as the size of the flame area), the system automatically assesses the severity of the current disaster and divides it into different levels. These levels are a series of levels from minor to severe, and each level corresponds to specific response measures. When the disaster level is lower than the preset safety threshold, it means that the fire has not yet reached the level that requires large-scale rescue, but it still needs to be taken seriously. At this time, the system will immediately start the fire resource dispatching program so that quick action can be taken to prevent the fire from further expanding.

[0113] The system monitors the location and status of all registered fire trucks, personnel and their equipment in real time to ensure that the resources closest to the fire site and on standby can be quickly found. This includes but is not limited to fire trucks, fire extinguishers, firefighters, etc. In order to achieve the fastest response time, the system also calculates the best path from each resource point to the fire scene, taking into account factors such as traffic conditions and road restrictions to plan the optimal route.

[0114] With the fire point as the center, the system will define a geographical range with a preset radius as the warning zone. This range is flexibly adjusted according to factors such as fire type, environmental conditions and population density to ensure that all affected areas are covered. Next, the system will filter all user terminals within this geographical range from the registered services. These terminals are smartphones, tablets or other devices with notification functions. In this way, the individuals or organizations that need to be notified can be accurately targeted.

[0115] The system constructs a message body containing the following information:

[0116] Location information of fire-fighting resources: including the specific locations of fire-fighting vehicles and personnel that are about to arrive.

[0117] Estimated time of arrival: The estimated time of arrival calculated based on the dispatch path.

[0118] Brief description of the disaster: Provide basic information about the fire, such as the area and location of the flames, to help the recipient understand the severity of the situation.

[0119] Finally, the system will select the appropriate method to push the above message to the selected user terminal. This is achieved through multiple channels, such as SMS, in-app notification, email or voice call, etc. The choice of each method depends on the user's preferences and urgency.

[0120] Second embodiment

[0121] The present invention also provides an AI city emergency management service system based on multi-source data fusion, including a camera monitoring module, a fire identification module, a fire information uploading module, a disaster information notification module and a user notification module; the camera monitoring module is used to continuously monitor the streets through a public camera network to obtain monitoring image information; the fire identification module is used to use a convolutional neural network to identify the fire situation based on the monitoring image information, and generate disaster information when a disaster is discovered, and the disaster information includes the flame area and the disaster location; the fire information uploading module is used for the user to upload disaster photo information through a wireless terminal when a fire is discovered, and the photo information includes photo data, positioning information and sending time; the disaster information notification module is used to call the camera in the corresponding area based on the disaster photo information to obtain corresponding disaster information, and notify the fire protection system; the user notification module is used to distinguish the disaster level based on the flame area in the disaster information, and when the disaster level is lower than the preset value, the positioning information of the fire extinguishing resources and the disaster information are sent to the user terminal within a preset range near the disaster location.

[0122] In this embodiment, the camera monitoring module continuously monitors streets and other key areas through a widely distributed public camera network. It can capture video streams in real time and extract monitoring images at preset time intervals. These images are compressed and uploaded to a central server or cloud platform, providing a rich raw data foundation for subsequent analysis.

[0123] The fire identification module uses convolutional neural network (CNN) technology to conduct in-depth analysis of surveillance images to identify potential fire situations. Once signs of fire are found, the module immediately generates detailed disaster information, including the flame area and specific location, to provide decision-making basis for subsequent emergency response. When ordinary citizens or on-site personnel discover a fire, they upload disaster photo information through wireless terminals (such as smartphones). This information not only contains the photo data of the fire, but also includes the location information when shooting and the sending timestamp, providing valuable first-hand information for the system. The disaster information notification module integrates the disaster photo information uploaded by users with existing monitoring resources, calls the cameras in the corresponding area to obtain more comprehensive disaster images, and notifies the fire protection system in time. This process ensures the integrity and accuracy of the information and improves the speed of emergency response. The user notification module distinguishes the disaster level based on the flame area in the disaster information, and sends the location information of the fire-fighting resources and the disaster information to the user terminals within a preset range near the disaster location when the disaster level is lower than the preset safety threshold. This helps to improve the community's self-protection awareness and emergency response capabilities.

[0124] Through close collaboration between modules, the whole process from fire monitoring to disaster confirmation to public notification is automated. It not only improves the efficiency and accuracy of urban emergency management, but also enhances the community's ability to respond to emergencies.

[0125] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art understands that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. An AI city emergency management service method based on multi-source data fusion, characterized in that: Includes: Continuously monitor streets through a public camera network to obtain monitoring image information; A convolutional neural network is used to identify fire conditions based on monitoring image information, and disaster information is generated when a disaster is discovered, wherein the disaster information includes the flame area and the disaster location; When a user discovers a fire, he uploads disaster photo information through a wireless terminal. The photo information includes photo data, location information and sending time; Based on the disaster photo information, the camera in the corresponding area is called to obtain the corresponding disaster information and notify the fire fighting system; The disaster level is distinguished based on the flame area in the disaster information, and when the disaster level is lower than a preset value, the positioning information of the fire-fighting resources and the disaster information are sent to user terminals within a preset range near the disaster location.

2. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 1, characterized in that: The specific steps of continuously monitoring the streets through the public camera network and obtaining monitoring image information include: Configure the camera to capture video data at set intervals; Extracting surveillance images from video data at preset intervals; Compress the monitoring image and upload it to the host computer.

3. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 2, characterized in that: The specific steps of using a convolutional neural network to identify fire conditions based on monitoring image information and generating disaster information when a disaster is discovered include: Construct a convolutional neural network model based on fire image information to obtain a judgment model; Get the uploaded surveillance images; The surveillance images are input into the trained judgment model to identify the fire situation. When a disaster is found, disaster information is generated and a video clip acquisition request is sent to the corresponding target camera. The target camera captures video clips within the preset time range of the surveillance image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flame area and disaster location.

4. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 3, characterized in that: The specific steps of constructing a convolutional neural network model based on fire image information to obtain a judgment model include: Collect images containing normal conditions and different types of fire scenes; Add labels to each image to get the dataset; The convolutional neural network model is trained using the labeled data set, and the judgment model is obtained through multiple iterations.

5. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 4, characterized in that: The target camera intercepts the video clips within the preset time range of the monitoring image and compresses and uploads them to the judgment model to confirm the fire situation, and then extracts the flame area and disaster location. The specific steps include: According to the timestamp of the disaster, determine the time range of the video clip to be captured, and obtain the corresponding video clip; Apply video compression technology to compress video clips before uploading; Extract sample images from the video clips and identify the fire area and damage location.

6. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 5, characterized in that: When the user discovers a fire, the user uploads disaster photo information through a wireless terminal, and the photo information includes photo data, location information, and sending time. The specific steps include: Users use wireless terminals to take photos of fires; Record the location coordinates when the fire photo was taken to obtain positioning information; Upload fire photos and location information to the host computer and record the sending time.

7. The AI ​​city emergency management service method based on multi-source data fusion as claimed in claim 6, characterized in that: The specific steps of calling the camera in the corresponding area based on the disaster information to obtain the corresponding disaster image group and notifying the fire fighting system include: Centered on the location information, define the sphere of influence on the map; Extract all available cameras within the impact range, and obtain the target camera closest to the fire location that captured the fire situation; For the target camera, the disaster image group within the period before and after the sending time is extracted and transmitted to the fire protection system.

8. An AI city emergency management service system based on multi-source data fusion, applied to an AI city emergency management service method based on multi-source data fusion as claimed in any one of claims 1 to 7, characterized in that: It includes camera monitoring module, fire identification module, fire information uploading module, disaster information notification module and user notification module; The camera monitoring module is used to continuously monitor the street through a public camera network to obtain monitoring image information; The fire identification module is used to identify the fire situation using a convolutional neural network based on the monitoring image information, and generate disaster information when a disaster is found, wherein the disaster information includes the flame area and the disaster location; The fire information uploading module is used for users to upload disaster photo information through a wireless terminal when they discover a fire. The photo information includes photo data, location information and sending time; The disaster information notification module is used to call the camera in the corresponding area to obtain the corresponding disaster information based on the disaster photo information, and notify the fire protection system; The user notification module is used to distinguish the disaster level based on the flame area in the disaster information, and when the disaster level is lower than a preset value, send the location information of the fire extinguishing resources and the disaster information to the user terminals within a preset range near the disaster location.

Images