A method for constructing and analyzing emergency incident portraits based on multimodal data

Abstract

The present invention belongs to the field of emergency management technology. A method for constructing and analyzing emergency portraits based on multimodal data is provided. The method comprises: training a convolutional neural network model based on a training data set, using the convolutional neural network model to identify key features of a second multimodal data source and determine a first portrait label, generating a first emergency portrait corresponding to each emergency type; using the convolutional neural network model to identify key features of a third multimodal data source acquired in real time and determine a second portrait label, generating a second emergency portrait; calculating the similarity between the second emergency portrait and each first emergency portrait in the portrait database, obtaining a matching first emergency portrait, and outputting early warning information of the emergency type corresponding to the first emergency portrait. The present invention analyzes the risk of an emergency under the current situation through an emergency portrait, and improves the efficiency and effect of responding to emergencies.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency event management, and in particular to a method for constructing and analyzing emergency event portraits based on multimodal data. Background Art

[0002] Emergency management is a complex process involving multiple aspects and links, including prevention and emergency preparedness, monitoring and early warning, emergency handling and rescue, post-event recovery and reconstruction and other key links. Existing emergency management basically handles the emergency events that have occurred according to fixed plans, but basically cannot identify the emergency events at an early stage, resulting in generally poor response effects to the emergency events, which cannot meet actual needs. The present invention aims to improve the efficiency and effect of responding to emergencies. Summary of the invention

[0003] In this regard, the present invention provides a method, system, electronic device, computer storage medium and computer program product for constructing and analyzing emergency incident portraits based on multimodal data to solve the above-mentioned technical problems.

[0004] The present invention discloses a method for constructing and analyzing emergency event portraits based on multimodal data, the method comprising the following steps: determining a plurality of emergency event types and corresponding first multimodal data sources, and constructing a training data set based on the first multimodal data source; wherein the first multimodal data source includes a plurality of different types of event source data; using the training data set to train a convolutional neural network model, using the trained convolutional neural network model to perform key feature recognition on a second multimodal data source, obtaining a plurality of groups of first key features, determining a first portrait label for each of the first key features in each group, and generating a first emergency event corresponding to each of the emergency event types based on all of the first portrait labels in each group. Portrait, storing each of the first emergency portraits generated into a portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; obtaining a third multimodal data source in real time, using the convolutional neural network model to perform key feature recognition on the third multimodal data source, obtaining a group of second key features, determining a second portrait label for each of the second key features, and generating a corresponding second emergency portrait based on all of the second portrait labels; calculating the similarity between the second emergency portrait and each of the first emergency portraits in the portrait database, obtaining a matched first emergency portrait, and outputting the warning information of the emergency type corresponding to the first emergency portrait.

[0005] In some embodiments, each of the event source data is configured with an event occurrence type label, and the event occurrence type label includes a first type label indicating that an emergency event has occurred and a second type label that the emergency event has not occurred; then, forming a training data set based on the first multimodal data source includes: dividing the first multimodal data source into a first type and a second type based on the event occurrence type label; evaluating the degree of emergency for each of the emergency event types to obtain a corresponding degree of emergency, and determining a mixing ratio corresponding to the emergency event type based on the degree of emergency; and selecting a corresponding number of the first multimodal data sources of the first type and a corresponding number of the first multimodal data sources of the second type based on the mixing ratio to form a training data set.

[0006] In some embodiments, the suddenness level of each sudden event type is evaluated to obtain the corresponding suddenness level, including: obtaining event source data corresponding to the sudden event type, extracting event suddenness time according to the event source data, and determining the suddenness level based on the event suddenness time; wherein, the shorter the event suddenness time is, the higher the suddenness level is, and vice versa.

[0007] In some embodiments, before the real-time acquisition of the third multimodal data source, the method also includes: retrieving event occurrence records corresponding to each of the emergency types, and obtaining an event occurrence regularity curve related to date and time based on statistics of the event occurrence records; determining the event occurrence volume corresponding to the current date and current time from the event occurrence regularity curve, and determining the emergency inspection frequency based on the event occurrence volume; wherein, the greater the event occurrence volume, the higher the emergency inspection frequency, and vice versa.

[0008] In some embodiments, the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database is calculated to obtain a matching first emergency event portrait, including: using a large language model to perform semantic understanding on the second emergency event portrait and each of the first emergency event portraits in the portrait database, performing feature transformation on the semantic understanding results, and obtaining a corresponding second semantic feature matrix and several first semantic feature matrices; using a similarity calculation formula to calculate the similarity between the second semantic feature matrix and each of the first semantic feature matrices, and screening out the first emergency event portraits corresponding to the first semantic feature matrices whose similarity is higher than a similarity threshold.

[0009] In some embodiments, the outputting of the warning information of the emergency event type corresponding to the first emergency event portrait includes: outputting the warning information of the emergency event type corresponding to the first emergency event portrait on the main monitoring interface of the monitoring center, and sending the warning information of the emergency event type corresponding to the first emergency event portrait to the terminal devices in the area corresponding to the warning information.

[0010] The present invention also discloses a system for constructing and analyzing emergency event portraits based on multimodal data, the system comprising a preprocessing module, a portrait construction module, a first event analysis module, and a second event analysis module; the preprocessing module is used to determine a number of emergency event types and corresponding first multimodal data sources, and to form a training data set based on the first multimodal data source; wherein the first multimodal data source includes a plurality of different types of event source data; the portrait construction module is used to train a convolutional neural network model using the training data set, perform key feature identification on a second multimodal data source using the trained convolutional neural network model, obtain a number of groups of first key features, determine a first portrait label for each of the first key features in each group, and generate a first portrait label corresponding to each of the emergency events based on all of the first portrait labels in each group. The first emergency event portrait corresponding to the event type is generated, and each of the first emergency event portraits is stored in a portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; the first event analysis module is used to obtain a third multimodal data source in real time, use the convolutional neural network model to identify key features of the third multimodal data source, obtain a group of second key features, determine a second portrait label for each of the second key features, and generate a corresponding second emergency event portrait based on all the second portrait labels; the second event analysis module is used to calculate the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database, obtain a matching first emergency event portrait, and output early warning information of the emergency event type corresponding to the first emergency event portrait.

[0011] The present invention also discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method described in any of the preceding items.

[0012] The present invention also discloses a computer storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any of the above methods.

[0013] The present invention also discloses a computer program product. When the computer program product is run on a terminal, the terminal implements any of the above methods when executing the computer program product.

[0014] The beneficial effect of the present invention is that the present invention pre-profiling various types of emergencies for subsequent correlation analysis between real-time situations and corresponding types of emergencies, thereby determining the risk of corresponding types of emergencies occurring under the current situation, so that relevant personnel or equipment can handle and respond in advance, thereby improving the efficiency and effectiveness of responding to emergencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 It is a flow chart of a method for constructing and analyzing an emergency incident portrait based on multimodal data disclosed in an embodiment of the present invention.

[0017] Figure 2 It is a schematic diagram of a process of forming a training data set based on a first multimodal data source disclosed in an embodiment of the present invention.

[0018] Figure 3 It is a structural schematic diagram of a system for constructing and analyzing emergency incident portraits based on multimodal data disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following is a description of the implementation of the present application by specific specific embodiments. People familiar with the technology can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0020] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0021] like Figure 1As shown, an embodiment of the present invention discloses a method for constructing and analyzing emergency event portraits based on multimodal data, the method comprising the following steps: determining a plurality of emergency event types and corresponding first multimodal data sources, and forming a training data set based on the first multimodal data source; wherein the first multimodal data source includes a plurality of different types of event source data; using the training data set to train a convolutional neural network model, using the trained convolutional neural network model to perform key feature recognition on a second multimodal data source to obtain a plurality of groups of first key features, determining a first portrait label for each of the first key features in each group, and generating a first emergency portrait label corresponding to each of the emergency event types based on all of the first portrait labels in each group. Generate an emergency portrait, and store each of the first emergency portraits generated into a portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; obtain a third multimodal data source in real time, use the convolutional neural network model to perform key feature recognition on the third multimodal data source, obtain a group of second key features, determine a second portrait label for each of the second key features, and generate a corresponding second emergency portrait based on all of the second portrait labels; calculate the similarity between the second emergency portrait and each of the first emergency portraits in the portrait database, obtain a matched first emergency portrait, and output warning information of the emergency type corresponding to the first emergency portrait.

[0022] The present invention first collects multimodal data sources corresponding to several types of emergencies that need to be monitored. The multimodal data sources contain data of different modes such as text, pictures, videos, audio, geography, and hydrology. These data are collected and collated through fixed monitoring systems (such as traffic monitoring systems, natural resource monitoring systems), mobile monitoring systems (such as floating vehicle sensors, personal mobile terminals, and drones). Emergencies may include traffic incidents, gatherings of people, natural disasters, and other types.

[0023] The multimodal data source is divided into two groups, namely, a first multimodal data source and a second multimodal data source, and the present invention does not limit the division ratio. A training data set is extracted based on the first multimodal data, and the constructed convolutional neural network model is trained using the training data set. The training goal is to enable the convolutional neural network model to identify key features corresponding to various types of emergencies in the multimodal data source, such as the key feature in a traffic accident is the running status of vehicles around the accident vehicle before the accident.

[0024] After completing the training of the convolutional neural network model, the convolutional neural network model is used to identify key features of the second multimodal data source to obtain multiple groups of first key features, each group of first key features representing a pre-incident feature group corresponding to the type of emergency. A first portrait label is determined for each first key feature, so that multiple first emergency portraits of the emergency type can be generated based on all the first portrait labels in each group, and the generated first emergency portraits are stored in the portrait database, thus completing the portraits of various types of emergencies.

[0025] Next, a convolutional neural network model is used to extract a set of second key features from the third multimodal data source acquired in real time, and the corresponding second emergency portrait is generated in the same manner as above. The similarity between the second emergency portrait and each first emergency portrait in the portrait database is calculated. If there is a first emergency portrait with a similarity higher than the similarity threshold, the current situation is likely to cause an emergency of the type of emergency corresponding to the matched first emergency portrait. At this time, the corresponding warning information should be output so that relevant personnel or equipment can deal with it in time, such as dispatching relevant personnel to the corresponding area for guidance, responding and guiding in advance through the road control system, vehicle control system, checkpoint system, etc., to reduce the probability of emergency, or reduce the adverse effects caused by the emergency.

[0026] Therefore, the present invention pre-profiling various types of emergencies for subsequent correlation analysis between real-time situations and corresponding types of emergencies, thereby determining the risk of corresponding types of emergencies occurring under the current situation, allowing relevant personnel or equipment to handle and respond in advance, thereby improving the efficiency and effectiveness of responding to emergencies.

[0027] In some embodiments, each of the event source data is configured with an event occurrence type label, and the event occurrence type label includes a first type label indicating that an emergency event has occurred and a second type label that the emergency event has not occurred; then, forming a training data set based on the first multimodal data source includes: dividing the first multimodal data source into a first type and a second type based on the event occurrence type label; evaluating the degree of emergency for each of the emergency event types to obtain a corresponding degree of emergency, and determining a mixing ratio corresponding to the emergency event type based on the degree of emergency; and selecting a corresponding number of the first multimodal data sources of the first type and a corresponding number of the first multimodal data sources of the second type based on the mixing ratio to form a training data set.

[0028] In the embodiment of the present invention, Figure 2As shown, the present invention also divides the first multimodal data source into a first type and a second type. The first type refers to the corresponding event source data corresponding to the occurrence of an emergency, that is, the corresponding emergency occurred in the subsequent period corresponding to the event source data; in contrast, the second type refers to the corresponding event source data corresponding to the non-occurrence of an emergency, that is, the corresponding emergency did not occur in the subsequent period corresponding to the event source data. The present invention constructs the event source data related to the occurrence and non-occurrence of emergencies as training data sets at the same time, so that the trained convolutional neural network model can grasp which key feature groups are strongly associated with the occurrence of emergencies and which key feature groups are weakly associated with the occurrence of emergencies, thereby ensuring that more accurate key features are extracted later.

[0029] The present invention also determines the respective ratios of the first type and the second type of first multimodal data in the training data set. Specifically, the suddenness degree of various emergency event types is evaluated to obtain the corresponding suddenness degree, and the mixing ratio corresponding to each emergency event type is determined according to the suddenness degree. The mixing ratios of the first type and the second type of first multimodal data of emergency event types with different suddenness degrees are different.

[0030] It should be noted that the first type of label and the second type of label are preferably added by automatic labeling. For example, the emergency monitoring system periodically obtains event source data and monitors whether a corresponding type of emergency occurs subsequently. If an emergency occurs, the event source data is labeled with the first type of label, otherwise, the second type of label is labeled.

[0031] In some embodiments, the suddenness level of each sudden event type is evaluated to obtain the corresponding suddenness level, including: obtaining event source data corresponding to the sudden event type, extracting event suddenness time according to the event source data, and determining the suddenness level based on the event suddenness time; wherein, the shorter the event suddenness time is, the higher the suddenness level is, and vice versa.

[0032] In an embodiment of the present invention, the event burst time of different types of emergencies is different, and the event burst time refers to the time span between the appearance of relevant signs and the confirmation of the occurrence of the corresponding type of emergencies. For example, based on monitoring information, a small number of people in uniforms gather in a certain area A, which means that signs related to the corresponding type of emergencies appear. After 10 minutes, the number of people in uniforms gather in area A reaches the gathering standard, which means that the event occurs, and the event burst time is 10 minutes. By performing a large amount of statistical analysis on the event source data corresponding to the type of emergencies, the event burst time that can reflect the universal law can be obtained. The degree of the emergencies of this type of emergencies is determined based on the event burst time and the corresponding negative correlation. However, the present invention does not limit the specific function form of the negative correlation between the degree of emergencies and the event burst time. Of course, it can also be in the form of a comparison table, which will not be described in detail.

[0033] In some embodiments, before the real-time acquisition of the third multimodal data source, the method also includes: retrieving event occurrence records corresponding to each of the emergency types, and obtaining an event occurrence regularity curve related to date and time based on statistics of the event occurrence records; determining the event occurrence volume corresponding to the current date and current time from the event occurrence regularity curve, and determining the emergency inspection frequency based on the event occurrence volume; wherein, the greater the event occurrence volume, the higher the emergency inspection frequency, and vice versa.

[0034] In an embodiment of the present invention, when inspecting emergencies, it is not possible to inspect all types of emergencies at full load, which will greatly increase the inspection load. The inspection system is prone to failure under high-load operation, and it is not economical. In this regard, the present invention sets a dynamic adjustment of the inspection frequency of various types of emergencies. Specifically, the event occurrence record corresponding to each type of emergency is retrieved. The event occurrence record records the date and time of each type of emergency that actually occurred in the monitoring area. Statistical analysis can be performed to obtain the event occurrence regularity curve of the monitoring area related to the date and time. The vertical axis of the event occurrence regularity curve is the event occurrence amount, and the horizontal axis is the date and time (for example, the 12:00-13:00 period on February 14). In this way, the event burst amount corresponding to the current date and event can be determined from the event occurrence regularity curve. The greater the number of sudden events, the greater the probability of this type of emergency at the current stage. At this time, the frequency of emergency inspections should be increased to avoid missing the identification and early warning of this type of emergency. Conversely, the smaller the probability of this type of emergency at the current stage, the frequency of emergency inspections should be appropriately reduced to avoid unnecessary high-frequency inspections, thereby reducing the processing load of the system.

[0035] In some embodiments, the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database is calculated to obtain a matching first emergency event portrait, including: using a large language model to perform semantic understanding on the second emergency event portrait and each of the first emergency event portraits in the portrait database, performing feature transformation on the semantic understanding results, and obtaining a corresponding second semantic feature matrix and several first semantic feature matrices; using a similarity calculation formula to calculate the similarity between the second semantic feature matrix and each of the first semantic feature matrices, and screening out the first emergency event portraits corresponding to the first semantic feature matrices whose similarity is higher than a similarity threshold.

[0036] In an embodiment of the present invention, the emergency portrait determined by the present invention is represented in the form of a portrait label, not a matrix, and it is difficult to directly calculate the similarity through the existing similarity calculation formula. In this regard, the present invention first uses a large language model (such as ChatGPT) to perform semantic understanding on the second emergency portrait generated in real time and each first emergency portrait in the portrait database, that is, to identify the meaning of each portrait label, and then fuse the meaning and perform feature transformation to obtain the corresponding second semantic feature matrix and several first semantic feature matrices. At this point, any existing similarity calculation formula can be used to calculate the similarity between the second semantic feature matrix and each first semantic feature matrix, and the first emergency portrait corresponding to the first semantic feature matrix with a similarity higher than the similarity threshold can be screened out.

[0037] The similarity calculation formula may be the commonly used Euclidean distance, cosine similarity, Manhattan distance, etc., and the present invention does not make any specific limitation thereto.

[0038] In some embodiments, the outputting of the warning information of the emergency event type corresponding to the first emergency event portrait includes: outputting the warning information of the emergency event type corresponding to the first emergency event portrait on the main monitoring interface of the monitoring center, and sending the warning information of the emergency event type corresponding to the first emergency event portrait to the terminal devices in the area corresponding to the warning information.

[0039] In an embodiment of the present invention, after obtaining the warning information, the warning information of the emergency type corresponding to the first emergency portrait can be directly output on the main monitoring interface of the monitoring center, so that the staff of the monitoring center or the emergency handling system can directly handle it. At the same time, the warning information of the emergency type corresponding to the first emergency portrait can also be sent to the terminal device in the area corresponding to the warning information. The terminal device can be a terminal device carried by urban management personnel (such as traffic police), or a terminal device used to manage urban equipment, such as a traffic signal controller, so that the monitoring center can be skipped to directly handle and respond to the emergency, thereby improving the handling efficiency.

[0040] like Figure 3 As shown, an embodiment of the present invention also discloses a system for constructing and analyzing emergency event portraits based on multimodal data, the system comprising a preprocessing module, a portrait construction module, a first event analysis module, and a second event analysis module; the preprocessing module is used to determine several emergency event types and corresponding first multimodal data sources, and to form a training data set based on the first multimodal data source; wherein the first multimodal data source includes multiple different types of event source data; the portrait construction module is used to train a convolutional neural network model using the training data set, and to use the trained convolutional neural network model to perform key feature identification on the second multimodal data source to obtain several groups of first key features, determine a first portrait label for each of the first key features in each group, and generate a first portrait label based on all the first portrait labels in each group. The first emergency event portrait corresponding to the emergency event type, and storing the generated first emergency event portraits into the portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; the first event analysis module is used to obtain the third multimodal data source in real time, use the convolutional neural network model to identify the key features of the third multimodal data source, obtain a group of second key features, determine the second portrait label for each of the second key features, and generate the corresponding second emergency event portrait based on all the second portrait labels; the second event analysis module is used to calculate the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database, obtain the matching first emergency event portrait, and output the warning information of the emergency event type corresponding to the first emergency event portrait.

[0041] An embodiment of the present invention further discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, characterized in that: the processor executes the computer program to implement the method described in the above embodiment.

[0042] An embodiment of the present invention further discloses a computer storage medium, wherein the computer storage medium stores a computer program, and is characterized in that: the computer program is executed by a processor to implement the method described in the above embodiment.

[0043] The embodiment of the present invention further discloses a computer program product. When the computer program product is run on a terminal, the terminal implements the method described in the above embodiment.

[0044] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0045] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0046] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention.

Claims

1. A method for constructing and analyzing emergency incident portraits based on multimodal data, characterized in that: The method comprises the following steps: determining a number of emergency event types and corresponding first multimodal data sources, and forming a training data set based on the first multimodal data source; wherein the first multimodal data source includes a plurality of different types of event source data; using the training data set to train a convolutional neural network model, and using the trained convolutional neural network model to perform key feature recognition on a second multimodal data source to obtain a number of groups of first key features, determining a first portrait label for each of the first key features in each group, and generating a first emergency event portrait corresponding to each of the emergency event types based on all of the first portrait labels in each group, and storing each of the generated first emergency event portraits in a portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; acquiring a third multimodal data source in real time, and using the convolutional neural network model to perform key feature recognition on the third multimodal data source to obtain a group of second key features, and determining a second portrait label for each of the second key features, and generating a corresponding second emergency event portrait based on all of the second portrait labels; calculating the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database, obtaining a matched first emergency event portrait, and outputting warning information of the emergency event type corresponding to the first emergency event portrait; Each of the event source data is configured with an event occurrence type label, and the event occurrence type label includes a first type label indicating that an emergency event has occurred and a second type label indicating that an emergency event has not occurred; then a training data set is formed based on the first multimodal data source, including: dividing the first multimodal data source into a first type and a second type based on the event occurrence type label; evaluating the emergency degree of each of the emergency event types to obtain a corresponding emergency degree, and determining a mixing ratio corresponding to the emergency event type based on the emergency degree; selecting a corresponding number of the first multimodal data sources of the first type and a corresponding number of the first multimodal data sources of the second type based on the mixing ratio to form a training data set; Perform a suddenness evaluation on each of the sudden event types to obtain a corresponding suddenness level, including: acquiring event source data corresponding to the sudden event type, extracting event suddenness time according to the event source data, and determining the suddenness level based on the event suddenness time; wherein, event suddenness time refers to the time span from the appearance of relevant signs to the confirmation of the occurrence of a sudden event of the corresponding type; and, the shorter the event suddenness time, the higher the suddenness level, and vice versa.

2. The method for constructing and analyzing emergency incident portraits based on multimodal data according to claim 1, characterized in that: Before the real-time acquisition of the third multimodal data source, the method also includes: retrieving the event occurrence records corresponding to each of the emergency types, and obtaining an event occurrence regularity curve related to date and time based on the event occurrence records; determining the event occurrence volume corresponding to the current date and current time from the event occurrence regularity curve, and determining the emergency inspection frequency based on the event occurrence volume; wherein, the greater the event occurrence volume, the higher the emergency inspection frequency, and vice versa.

3. The method for constructing and analyzing emergency incident portraits based on multimodal data according to claim 2, characterized in that: Calculate the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database to obtain a matching first emergency event portrait, including: using a large language model to perform semantic understanding on the second emergency event portrait and each of the first emergency event portraits in the portrait database, performing feature transformation on the semantic understanding results, and obtaining a corresponding second semantic feature matrix and several first semantic feature matrices; use a similarity calculation formula to calculate the similarity between the second semantic feature matrix and each of the first semantic feature matrices, and screen out the first emergency event portraits corresponding to the first semantic feature matrices whose similarity is higher than a similarity threshold.

4. The method for constructing and analyzing emergency incident portraits based on multimodal data according to claim 1, characterized in that: The output of the warning information of the emergency event type corresponding to the first emergency event portrait includes: outputting the warning information of the emergency event type corresponding to the first emergency event portrait on the main monitoring interface of the monitoring center, and sending the warning information of the emergency event type corresponding to the first emergency event portrait to the terminal equipment in the area corresponding to the warning information.

5. A system for constructing and analyzing emergency incident portraits based on multimodal data, the system being based on the method according to any one of claims 1 to 4, the system comprising a preprocessing module, a portrait construction module, a first incident analysis module, and a second incident analysis module; characterized in that: The preprocessing module is used to determine several emergency event types and corresponding first multimodal data sources, and to form a training data set based on the first multimodal data source; wherein the first multimodal data source includes multiple different types of event source data; the portrait construction module is used to use the training data set to train the convolutional neural network model, use the trained convolutional neural network model to perform key feature identification on the second multimodal data source, obtain several groups of first key features, determine a first portrait label for each of the first key features in each group, generate a first emergency event portrait corresponding to each of the emergency event types based on all of the first portrait labels in each group, and store the generated first emergency event portraits in a portrait database; wherein the second multimodal data source and the first multimodal data source are different parts of the same group of data sources; the first event analysis module is used to obtain a third multimodal data source in real time, use the convolutional neural network model to perform key feature identification on the third multimodal data source, obtain a group of second key features, determine a second portrait label for each of the second key features, and generate a corresponding second emergency event portrait based on all of the second portrait labels; the second event analysis module is used to calculate the similarity between the second emergency event portrait and each of the first emergency event portraits in the portrait database, obtain a matching first emergency event portrait, and output warning information of the emergency event type corresponding to the first emergency event portrait.

6. An electronic device comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 4.

7. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method according to any one of claims 1 to 4.

8. A computer program product comprising a computer program stored on a non-transitory computer readable medium, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images