A smart firefighting open access method based on AI large model recognition algorithm

Through the AI ​​big model identification algorithm, learn the historical monitoring data of fire-fighting equipment and obtain the target recognition model, solve the problem of insufficient intelligence and flexibility of pseudo-code design in the existing technology, and realize intelligent monitoring of fire-fighting equipment status and fire hazards. Identification improves the intelligence and linkage capabilities of the system.

CN118445755BActive Publication Date: 2024-11-01JIANGSU TIANFENG HAIZHIYUAN COMM POWER TECH CO LTD
View PDF 10 Cites 0 Cited by

Patent Information

Application Number
CN202410606436.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-16
Publication Date
2024-11-01
Estimated Expiration
2044-05-16

AI Technical Summary

Technical Problem

In existing smart fire protection systems, pseudo-code design requires manual participation, lacks intelligence and flexibility, and it is difficult to adapt to the access needs of different fire protection equipment.

Method used

Using AI large-scale model recognition algorithm, we learn the historical monitoring data of fire-fighting equipment, obtain target recognition models, realize open access, identify the status of fire-fighting equipment and fire hazards, and conduct risk prediction and linkage.

Benefits of technology

It realizes intelligent monitoring of fire equipment status and effective identification of fire hazards, improves the intelligence and flexibility of the system, and can effectively link when a fire occurs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118445755B_ABST
    Figure CN118445755B_ABST
Patent Text Reader

Abstract

The present invention provides a smart firefighting open access method based on an AI large model recognition algorithm, which belongs to the field of artificial intelligence technology. The method includes: step 1: obtaining historical monitoring data of firefighting equipment; step 2: obtaining video recognition data and firefighting-related equipment data of firefighting equipment; step 3: learning historical monitoring data based on the AI ​​large model, obtaining a target recognition model, analyzing and identifying the video recognition data and firefighting-related equipment data through the target recognition model, and performing open access to firefighting equipment. A smart firefighting open access method based on an AI large model recognition algorithm of the present invention introduces an AI large model to learn historical monitoring data and obtain a target recognition model. The large model can be used to identify the status and parameters of various firefighting-related equipment to achieve open access, which is convenient for the subsequent system to comprehensively monitor the equipment status and prejudge risks, and to effectively link up when a fire occurs, which is more intelligent and flexible.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a smart firefighting open access method based on an AI large model recognition algorithm. Background Art

[0002] AI large model recognition algorithms usually refer to neural network models with a large number of parameters in the field of deep learning, which can process and learn complex data patterns. These algorithms have demonstrated strong performance in many fields such as natural language processing, image recognition, and speech recognition. Smart firefighting open access refers to the use of open interfaces and standardized protocols to allow firefighting equipment, systems or applications from different manufacturers to connect and work together to achieve intelligent and networked fire safety management.

[0003] The invention patent with application number: CN201510435245.1 discloses a design method and system for open authorization access based on smart services, wherein the method comprises: step 1: describing the relationship between each module of the service authorization layer and the service authorization request process; step 2: according to the relationship between the modules of the service authorization layer and the service authorization request process of the above step 1; step 3: according to the above steps 1 and 2, design the service authentication and authorization status response required for the service open authorization access, and describe the service authentication and authorization status response process in pseudo code; step 4: according to the above steps 2 and 3, design the class of the service authentication and authorization module; step 5: according to the class of the service authentication and authorization module of the above step 4, design the required database table structure, including the table name, table fields and field types; step 6: according to the above steps, design the service open authorization interface call sequence diagram, and describe the interface call implementation process of the service open authorization in the form of pseudo code.

[0004] However, the above-mentioned prior art requires designing corresponding pseudocodes for the connected subject for subsequent calls. The design of pseudocodes in the prior art usually requires manual participation and is not intelligent enough. In addition, each connected subject needs to design pseudocodes adaptively, which is not flexible enough.

[0005] In view of this, there is an urgent need for an open access method for smart fire protection based on AI large model recognition algorithm to at least solve the above-mentioned shortcomings. Summary of the invention

[0006] One of the purposes of the present invention is to provide an open access method for intelligent fire protection based on AI large model recognition algorithm, introduce AI large model to learn historical monitoring data, and obtain target recognition model. The large model can be used to identify the status and parameters of various fire-related equipment to achieve open access, which is convenient for the subsequent system to comprehensively monitor the equipment status and prejudge risks, and effectively link up in the event of a fire, which is more intelligent and flexible.

[0007] An embodiment of the present invention provides a smart fire protection open access method based on an AI large model recognition algorithm, comprising:

[0008] Step 1: Obtain historical monitoring data of fire-fighting equipment;

[0009] Step 2: Obtain video recognition data of firefighting equipment and firefighting-related equipment data;

[0010] Step 3: Based on the AI ​​big model, learn the historical monitoring data to obtain the target recognition model, analyze and identify the video recognition data and fire-related equipment data through the target recognition model, and provide open access to fire-fighting equipment.

[0011] Preferably, step 3: learning historical monitoring data based on the AI ​​big model to obtain a target recognition model, analyzing and identifying the video recognition data and fire-related equipment data through the target recognition model, and providing open access to fire-fighting equipment, including:

[0012] Analyze historical monitoring data and obtain fire-related data;

[0013] Based on the AI ​​big model, fire-related features are identified according to fire-related data;

[0014] Learning fire-related features and obtaining learning results, where the learning results are: the correlation between the fire-related features and fire hazards;

[0015] According to the learning results, establish a target recognition model;

[0016] Analyze video recognition data and fire-related equipment data through the target recognition model to obtain recognition results;

[0017] Based on the identification results, determine whether there is a fire hazard. If there is a fire hazard, an early warning will be issued based on the preset early warning mechanism.

[0018] Preferably, step 2: obtaining video recognition data of firefighting equipment and firefighting-related equipment data includes:

[0019] Obtain information tags for firefighting equipment;

[0020] Determine the key parameters of firefighting equipment based on information labels;

[0021] Determine whether the fire-fighting equipment is operating normally based on key parameters;

[0022] If the fire-fighting equipment is operating normally, a monitoring network for the corresponding fire-fighting equipment shall be established;

[0023] Based on the monitoring network, video recognition data and fire-related equipment data are obtained.

[0024] Preferably, a monitoring network for corresponding fire-fighting equipment is established, including:

[0025] Establish a monitoring network based on loosely coupled seamless access.

[0026] An embodiment of the present invention provides a smart fire protection open access method based on an AI large model recognition algorithm, further comprising:

[0027] Step 4: Obtain access information of the linkage platform, and predict problems based on the access information, and trigger linkage measures based on the prediction results of the problem prediction.

[0028] Preferably, based on the AI ​​big model, fire-related features are identified according to fire-related data, including:

[0029] Extract pre-training data from fire-related data;

[0030] Pre-train the AI ​​large model through pre-training data to obtain the pre-training model;

[0031] Identify fire-related features in fire-related data based on the pre-trained model.

[0032] Preferably, extracting pre-training data from fire-related data includes:

[0033] Acquire monitoring data sources for fire-related data;

[0034] Divide the fire-related data belonging to the same monitoring data source into the same data set and use it as a homologous data set;

[0035] According to the data type of fire-related data in the homologous data set, the comprehensive feature extraction capability value is determined and associated with the corresponding homologous data set;

[0036] According to the data quality of the fire-related data in the homologous data set, the feature extraction depth capability value is determined and associated with the corresponding homologous data set;

[0037] The feature extraction comprehensive capability value and the feature extraction depth capability value associated with the homologous data set are accumulated to obtain a calculated value, and the homologous data set whose calculated value is greater than or equal to a preset calculated value threshold is used as pre-training data.

[0038] Preferably, predicting the problem based on the access information includes:

[0039] Obtaining recognition result features of a recognition result of a target recognition model;

[0040] Determine the combined features of the linkage platform based on the identification result features and access information;

[0041] Predict problems based on combined features and problem prediction models;

[0042] Among them, according to the identification result characteristics and access information, the combined characteristics of the linkage platform are determined, including:

[0043] Determine an impact information item according to the identification result feature and the access information, wherein the impact information item is associated with the identification result feature;

[0044] Characterizing the impact information item information to obtain impact information features;

[0045] The identification result features and the impact information features are taken together as combined features.

[0046] Preferably, according to the prediction result of the problem prediction, the linkage measures are triggered, including:

[0047] According to the prediction result, determine the characteristics of the prediction result;

[0048] Obtain historical execution measures of the linkage platform;

[0049] Based on the triggering conditions of historical execution measures and the characteristics of the predicted results, determine the linkage measures;

[0050] The triggering conditions for historical execution measures include:

[0051] The first target execution event of the historical execution measure is similar to the second target execution event triggered by the predicted result characteristics on the linkage platform;

[0052] The historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements. The condition matching requirements are: the historical execution conditions of the key condition types all have matching items in the current execution conditions, and the sum of the condition type weights of the target condition types of the matching items of the historical execution conditions of the non-key condition types in the current execution conditions is greater than or equal to the preset weights and thresholds.

[0053] Preferably, obtaining historical execution measures of the linkage platform includes:

[0054] Capturing historical execution measures based on the capturing conditions of historical execution measures;

[0055] Among them, the capture conditions of historical execution measures include:

[0056] The first time interval between two adjacent capture behaviors falls within a preset time interval;

[0057] The second time interval between the historical time anchor point of each captured behavior and the current time point is less than or equal to a preset time interval threshold;

[0058] A vector matching result of the first behavior vector of the capture behavior and the second behavior vector of the capture strategy is a match;

[0059] The capture strategy is obtained by expanding the historical logs on the timeline, determining the axis position of the historical time anchor point on the timeline, traversing a preset number of log description items from the axis position to the front and back of the timeline, and obtaining the capture strategy based on the log description items.

[0060] An embodiment of the present invention provides a smart fire protection open access system based on an AI large model recognition algorithm, including:

[0061] The historical monitoring data acquisition subsystem is used to obtain the historical monitoring data of fire-fighting equipment;

[0062] A data acquisition subsystem, used to acquire video recognition data of fire-fighting equipment and data of fire-fighting related equipment;

[0063] The open access subsystem is used to learn historical monitoring data based on the AI ​​large model, obtain the target recognition model, analyze and identify the video recognition data and fire-related equipment data through the target recognition model, and provide open access to fire-fighting equipment.

[0064] The beneficial effects of the present invention are:

[0065] The present invention introduces an AI big model to learn historical monitoring data and obtain a target recognition model. The big model can be used to identify the status and parameters of various fire-related equipment to achieve open access, which is convenient for subsequent systems to comprehensively monitor the status of equipment and make risk predictions, and to effectively link up when a fire occurs, which is more intelligent and flexible.

[0066] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the present application documents.

[0067] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0069] Figure 1 A schematic diagram of an open access method for intelligent fire protection based on an AI large model recognition algorithm in an embodiment of the present invention;

[0070] Figure 2This is a schematic diagram of a smart fire protection open access system based on an AI large model recognition algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0072] The embodiment of the present invention provides a smart fire protection open access method based on AI large model recognition algorithm, such as Figure 1 As shown, including:

[0073] Step 1: Obtain historical monitoring data of firefighting equipment; firefighting equipment refers to firefighting facilities that need to be controlled and pre-entered by staff, such as water tanks, smoke detectors, temperature sensors, and combustible gas detectors; historical monitoring data refers to data collected and recorded in the history of firefighting equipment;

[0074] Step 2: Obtain the video recognition data of fire-fighting equipment and fire-related equipment data; the video recognition data is: the on-site monitoring video collected in real time by the monitoring camera connected to the system; the fire-related equipment data is: the recorded data of fire-related equipment collected in real time based on the Internet of Things, such as: temperature sensor data and smoke sensor data, etc.;

[0075] Step 3: Based on the AI ​​big model, learn the historical monitoring data, obtain the target recognition model, analyze and identify the video recognition data and fire-related equipment data through the target recognition model, and provide open access to fire-fighting equipment. Among them, the AI ​​big model is: an artificial intelligence model that can process and analyze a large amount of data and learn patterns and rules from it; the target recognition model is: an artificial intelligence model that analyzes historical monitoring data to learn how to identify and access data of different types and predict fire hazards. When learning, extract useful features from historical monitoring data (such as: features related to fire) to train the AI ​​big model. After training, obtain the target recognition model; open access is: the process of integrating fire-fighting equipment from different manufacturers into the smart fire-fighting system through standardized communication protocols and interfaces.

[0076] The working principle and beneficial effects of the above technical solution are:

[0077] This application introduces an AI big model to learn historical monitoring data and obtain a target recognition model. The big model can identify the status and parameters of various fire-related equipment to achieve open access, which is convenient for subsequent systems to comprehensively monitor equipment status and predict risks, and effectively link up in the event of a fire, which is more intelligent and flexible.

[0078] In one embodiment, step 3: learning historical monitoring data based on the AI ​​big model to obtain a target recognition model, analyzing and identifying video recognition data and fire-related equipment data through the target recognition model, and providing open access to fire-fighting equipment, including:

[0079] Analyze historical monitoring data to obtain fire-related data; fire-related data refers to data that is potentially related to the occurrence of fire, such as temperature, smoke concentration, and gas concentration;

[0080] Based on the AI ​​big model, fire-related features are identified according to fire-related data; fire-related features are: common features found in historical fire events, such as: the temperature reaches more than a few degrees;

[0081] Learning fire-related features and obtaining learning results, the learning results are: the correlation between fire-related features and fire hazards; wherein the correlation is: which fire-related features correspond to which fire hazards;

[0082] According to the learning results, a target recognition model is established; wherein the learning results are: model parameters and decision boundaries obtained after training the AI ​​model;

[0083] The video recognition data and the fire-related equipment data are analyzed by the target recognition model to obtain a recognition result; wherein the recognition result is: the result obtained after the target recognition model analyzes the video recognition data and the fire-related equipment data, which can indicate what kind of fire-fighting equipment the video recognition data and the fire-related equipment data are and whether a fire hazard is currently detected;

[0084] Based on the identification results, it is determined whether there is a fire hazard. If there is a fire hazard, an early warning is issued based on the preset early warning mechanism. Among them, the preset early warning mechanism is: notifying relevant personnel or fire departments and taking corresponding measures (such as: starting the sprinkler system).

[0085] The working principle and beneficial effects of the above technical solution are:

[0086] This application parses historical monitoring data to obtain fire-related data that is potentially related to the occurrence of fires, and introduces an AI large model to learn fire-related data to obtain common fire-related features found in historical fire events. Fire-related features are learned to determine which fire-related features correspond to which fire hazards, and a target recognition model is established based on the learning results. Open access to firefighting equipment is performed based on the target recognition model, video recognition data, and firefighting-related equipment data. During open access, the target recognition model identifies video recognition data and firefighting-related equipment data to determine which data is which type of firefighting equipment, and then performs subsequent fire hazard identification. Firefighting equipment from various manufacturers can be integrated into the existing system, which is more flexible. The introduction of an early warning mechanism can provide timely warnings when there are fire hazards, thereby improving safety.

[0087] In one embodiment, step 2: obtaining video recognition data of firefighting equipment and firefighting-related equipment data includes:

[0088] Obtaining information tags of fire-fighting equipment; wherein, according to the information tags, public information of the fire-fighting equipment can be determined, such as: which manufacturer and what type of equipment;

[0089] According to the information tag, determine the key parameters of the fire-fighting equipment; wherein the key parameters are: parameters that characterize whether the fire-fighting equipment is in operation;

[0090] According to the key parameters, whether the fire-fighting equipment is operating normally is judged; when judging whether the fire-fighting equipment is operating normally, the key parameters are compared with the normal operating state parameters preset by the key parameters;

[0091] If the fire-fighting equipment operates normally, a monitoring network for the corresponding fire-fighting equipment is established; wherein the monitoring network is a network composed of IoT nodes of the fire-fighting equipment that operates normally;

[0092] Based on the monitoring network, video recognition data and fire-related equipment data are obtained.

[0093] The working principle and beneficial effects of the above technical solution are:

[0094] This application obtains the information tag of the fire-fighting equipment and determines whether the fire-fighting equipment is operating normally based on the key parameters provided by the information tag. If it is operating normally, a monitoring network for the corresponding fire-fighting equipment is established, and video recognition data and fire-related equipment data are obtained based on the monitoring network, thereby improving the reliability of obtaining video recognition data and fire-related equipment data.

[0095] In one embodiment, a monitoring network of corresponding fire-fighting equipment is established, including:

[0096] Establish a monitoring network based on loosely coupled seamless access.

[0097] The working principle and beneficial effects of the above technical solution are:

[0098] The access method of the intelligent fire protection system of the present application and the IoT node of the fire protection equipment operating normally in the monitoring network is loosely coupled and seamless, which improves the flexibility of obtaining video recognition data and fire protection related equipment data.

[0099] The embodiment of the present invention provides a smart fire protection open access method based on an AI large model recognition algorithm, further comprising:

[0100] Step 4: Get the access information of the linkage platform, and make a problem prediction based on the access information, and trigger the linkage measures based on the prediction results of the problem prediction. The linkage platform is: water system, electrical system, communication system and video system; the access information is: the information that the linkage platform and the smart fire protection system allow the smart fire protection system to read after the linkage platform is connected; when predicting the problem based on the access information, make an early warning judgment based on the early warning judgment mechanism of the linkage platform; when triggering the linkage measures, it is triggered based on the Internet of Things technology.

[0101] The working principle and beneficial effects of the above technical solution are:

[0102] This application introduces a linkage platform, obtains the smart fire protection system and reads the access information of the linkage platform after linkage. According to the early warning judgment mechanism of the linkage platform and the access information, the problem is judged, and the linkage measures are triggered according to the predicted results of the problem judgment, which improves the efficiency of the overall fire protection system and effectively responds to various emergencies.

[0103] In one embodiment, based on the AI ​​big model, fire-related features are identified according to fire-related data, including:

[0104] Extract pre-trained data from fire-related data; wherein the pre-trained data is: training data of an AI model for extracting fire-related features:

[0105] Pre-training the AI ​​large model through pre-training data to obtain a pre-training model; wherein the pre-training model is: an AI model used for subsequent extraction of fire-related features;

[0106] Identify fire-related features in fire-related data based on the pre-trained model.

[0107] The working principle and beneficial effects of the above technical solution are:

[0108] The present application obtains pre-training data for AI model training for extracting fire-related features from fire-related data, pre-trains the AI ​​large model through the pre-training data, obtains the pre-training model, and performs subsequent fire hazard identification after the pre-training model identifies the fire-related features in the fire-related data. The model is pre-trained using some samples to reduce the model's dependence on a large amount of labeled data. After the model has certain prior knowledge, subsequent feature recognition is performed, which speeds up the recognition efficiency.

[0109] In one embodiment, extracting pre-training data from fire-related data includes:

[0110] Obtaining a monitoring data source for fire-related data; wherein the monitoring data source is: an original data source that can provide fire-related data, such as: fire monitoring data of a certain date in a certain month in a certain year;

[0111] Divide the fire-related data belonging to the same monitoring data source into the same data set and use it as a homologous data set;

[0112] According to the data type of the fire-related data in the homologous data set, determine the comprehensive feature extraction capability value and associate it with the corresponding homologous data set; wherein the data type is: the data category of the fire-related data, such as: temperature, smoke concentration, etc.; the comprehensive feature extraction capability value is: a quantitative value for evaluating the comprehensiveness of the data type of the fire-related data in the homologous data set. When evaluating, divide the number of data types by the total number of types and multiply by 100 to obtain the comprehensive feature extraction capability value;

[0113] According to the data quality of the fire-related data in the homologous data set, the feature extraction depth capability value is determined and associated with the corresponding homologous data set; the data quality is an indicator determined by the accuracy, completeness, consistency and credibility of the data; the feature extraction depth capability value is a quantitative value for evaluating the depth or detail of the features that can be extracted from a single homologous data set. The higher the data quality, the higher the feature extraction depth capability value. The conversion coefficient of the distance between the two is determined manually.

[0114] The feature extraction comprehensive capability value and the feature extraction depth capability value associated with the homologous data set are accumulated to obtain a calculated value, and the homologous data set whose calculated value is greater than or equal to a preset calculated value threshold is used as pre-training data. The preset calculated value threshold is manually set in advance.

[0115] The working principle and beneficial effects of the above technical solution are:

[0116] This application obtains a monitoring data source that can provide fire-related data, and determines the homologous data set belonging to the same monitoring data source. The homologous data set is analyzed and pre-training data is screened. The specific process is as follows:

[0117] Obtain the data type of fire-related data in the homologous dataset, evaluate the comprehensiveness of the data type, obtain the comprehensive feature extraction capability value, and associate it with the corresponding homologous dataset; obtain the data quality of fire-related data in the homologous dataset, evaluate the depth of features that can be extracted from a single homologous dataset, obtain the feature extraction depth capability value, and associate it with the corresponding homologous dataset. Sum the feature extraction comprehensive capability value and feature extraction depth capability value associated with the homologous dataset to obtain the calculated value, and use the homologous dataset with a calculated value greater than the calculated value threshold as pre-training data. When labeled data is scarce, use high-quality data for pre-training, so that the prior knowledge learned by the model is more reliable and more conducive to subsequent feature extraction.

[0118] In one embodiment, predicting a problem based on access information includes:

[0119] Obtaining recognition result features of the recognition result of the target recognition model; wherein the recognition result features are: a characteristic representation of the recognition result, such as: where and what kind of fire hazard exists;

[0120] Determine the combined features of the linkage platform according to the identification result features and the access information; wherein the combined features are: a feature combination of the identification result features and a characteristic representation of the impact of the identification result features on the linkage platform, for example: the identification result feature is a fire in the computer room, and the characteristic representation of the impact of the identification result features on the linkage platform (for example, the electrical system) is: the computer room is powered off;

[0121] Predict problems based on combined features and a problem prediction model; the problem prediction model is an AI model that determines linkage problems caused by fire hazards based on combined features;

[0122] Among them, according to the identification result characteristics and access information, the combined characteristics of the linkage platform are determined, including:

[0123] Determine an impact information item according to the identification result feature and the access information, where the impact information item is associated with the identification result feature; wherein the impact information item is: access information of the information type affected by the identification result feature;

[0124] Characterizing the impact information item to obtain the impact information feature; wherein the impact information feature is: a characterization representation of the impact information item, such as: burning power supply;

[0125] The identification result features and the impact information features are taken together as combined features.

[0126] The working principle and beneficial effects of the above technical solution are:

[0127] This application characterizes the recognition result of the target recognition model to obtain the recognition result feature, determines the impact information item affected by the recognition result feature, characterizes the impact information item to obtain the impact information feature, and uses the recognition result feature and the impact information feature as a combined feature. The problem prediction model is introduced and the combined feature is input to predict the problem, and the impact information item is determined first and then the subsequent judgment is made, which improves the prediction efficiency.

[0128] In one embodiment, according to the prediction result of the problem prediction, the linkage measures are triggered, including:

[0129] According to the prediction result, the prediction result feature is determined; wherein the prediction result feature is: a characteristic representation of the prediction result, for example: a fire in a computer room, the prediction result feature is: power outage time, location and area;

[0130] Obtain the historical execution measures of the linkage platform; where the historical execution measures are: management measures or action records taken by the linkage platform in the past;

[0131] Based on the triggering conditions of historical execution measures and the characteristics of the predicted results, determine the linkage measures;

[0132] The triggering conditions for historical execution measures include:

[0133] The first target execution event of the historical execution measure is similar to the second target execution event triggered by the predicted result feature on the linkage platform; wherein the first target execution event is: an event managed by the historical execution measure; the second target execution event is: an event caused by the predicted result feature on the linkage platform;

[0134] The historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements. The condition matching requirements are: the historical execution conditions of the key condition type all have matching items in the current execution conditions, and the sum of the condition type weights of the target condition type of the matching items of the historical execution conditions of the non-key condition type in the current execution conditions is greater than or equal to the preset weight and threshold. Among them, the historical execution conditions are: the specific conditions when executing the historical execution conditions; the current execution conditions are: the conditions faced by the linkage platform under the current circumstances; the key condition types are manually pre-set and are the required condition types. Except for the condition types of the key condition types, all other condition types are non-key condition types; the matching items are: the current execution conditions whose condition matching results with the historical execution conditions are matched; the target condition type is: the corresponding non-key condition type when the current execution conditions and the historical execution conditions of the non-key condition types have the same condition matching; the condition type weight represents the importance of the matching conditions of the non-key condition type. The larger the condition type weight, the higher the importance; the preset weight and threshold are manually pre-set.

[0135] The working principle and beneficial effects of the above technical solution are:

[0136] This application characterizes the prediction results and obtains the prediction result characteristics. The historical execution measures taken by the linkage platform in the past are introduced, trigger conditions are introduced, and linkage measures are determined based on the prediction result characteristics. The trigger conditions include: the events managed by the historical execution measures are similar to the events caused by the prediction result characteristics on the linkage platform; the historical execution conditions of the historical execution measures match the key condition types of the current execution conditions of the linkage platform, and the non-key condition types are adaptively matched, which is more flexible.

[0137] In one embodiment, obtaining historical execution measures of the linkage platform includes:

[0138] Based on the capturing condition of the historical execution measures, the historical execution measures are captured; wherein the capturing condition is: a standard for acquiring the historical execution measures;

[0139] Among them, the capture conditions of historical execution measures include:

[0140] The first time interval between two adjacent capture behaviors falls within a preset time interval; wherein the first time interval is: the time difference between the occurrence times of the two adjacent capture behaviors; the preset time interval is manually preset, such as: 12 hours to 24 hours;

[0141] The second time interval between the historical time anchor point of each captured behavior and the current time point is less than or equal to a preset time interval threshold; wherein the historical time anchor point is: the execution time of the historical execution measure captured by the captured behavior; the current time point is: the current moment; the second time interval is: the time difference between the historical time anchor point and the current time point; the preset time interval threshold is preset manually;

[0142] The vector matching result of the first behavior vector of the capture behavior and the second behavior vector of the capture strategy is a match; wherein the first behavior vector is: a vector describing the characteristics of the capture behavior; the second behavior vector is: a vector describing the characteristics of the capture behavior corresponding to the capture strategy; the capture strategy is: an extraction behavior plan of historical execution measures;

[0143] The method of obtaining the capture strategy is: expand the historical log on the time axis, determine the axis position of the historical time anchor point on the time axis, traverse the preset number of log description items from the axis position to the front and back of the time axis, and obtain the capture strategy according to the log description items. Among them, the historical log is: the record log of the linkage platform; the preset number is: 10, which can also be set manually; the log description item is: log description semantics; when determining the capture strategy, according to the log description semantics, determine the appropriate solution for extracting the corresponding historical execution measures and use it as the capture strategy.

[0144] The working principle and beneficial effects of the above technical solution are:

[0145] The present application introduces the capture conditions of historical execution measures. According to the capture conditions, historical execution measures are obtained. The specific constraints of the capture conditions are: the first time interval between two adjacent capture behaviors falls within the preset time interval range, and the capture computing resources are reasonably allocated; the execution time of the historical execution measures of each capture behavior is less than or equal to the time interval threshold from the second time interval of the current time point, and the historical execution measures are more timely; the vector matching result of the first behavior feature description vector of the capture behavior and the second behavior feature description vector of the capture strategy is a match. The capture strategy is determined based on the log semantics near the execution time of the historical execution measures, which is more adaptable. Furthermore, the capture behavior is also more reasonable.

[0146] The embodiment of the present invention provides a smart fire protection open access system based on AI large model recognition algorithm, such as Figure 2 As shown, including:

[0147] The historical monitoring data acquisition subsystem 1 is used to acquire the historical monitoring data of fire-fighting equipment;

[0148] Data acquisition subsystem 2, used to acquire video recognition data of fire-fighting equipment and data of fire-fighting related equipment;

[0149] The open access subsystem 3 is used to learn historical monitoring data based on the AI ​​large model, obtain the target recognition model, analyze and identify the video recognition data and fire-related equipment data through the target recognition model, and provide open access to fire-fighting equipment.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A smart fire protection open access method based on AI large model recognition algorithm, characterized in that: include: Step 1: Obtain historical monitoring data of fire-fighting equipment; Step 2: Obtain video recognition data of firefighting equipment and firefighting-related equipment data; Step 3: Based on the AI ​​big model, learn the historical monitoring data to obtain the target recognition model, analyze and identify the video recognition data and fire-related equipment data through the target recognition model, and provide open access to fire-fighting equipment; Step 3: Based on the AI ​​big model, learn the historical monitoring data to obtain the target recognition model. Use the target recognition model to analyze and identify the video recognition data and fire-related equipment data, and provide open access to fire-fighting equipment, including: Analyze historical monitoring data and obtain fire-related data; Based on the AI ​​big model, fire-related features are identified according to fire-related data; Learning fire-related features and obtaining learning results, where the learning results are: the correlation between the fire-related features and fire hazards; According to the learning results, establish a target recognition model; Analyze video recognition data and fire-related equipment data through the target recognition model to obtain recognition results; Based on the identification results, determine whether there is a fire hazard. If there is a fire hazard, issue an early warning based on the preset early warning mechanism; Based on the AI ​​big model, fire-related features are identified according to fire-related data, including: Extract pre-training data from fire-related data; Pre-train the AI ​​large model through pre-training data to obtain the pre-training model; identifying fire-related features in fire-related data according to a pre-trained model; Extract pre-training data from fire-related data, including: Acquire monitoring data sources for fire-related data; Divide the fire-related data belonging to the same monitoring data source into the same data set and use it as a homologous data set; According to the data type of fire-related data in the homologous data set, the comprehensive feature extraction capability value is determined and associated with the corresponding homologous data set; According to the data quality of the fire-related data in the homologous data set, the feature extraction depth capability value is determined and associated with the corresponding homologous data set; The feature extraction comprehensive capability value and the feature extraction deep capability value associated with the homologous data set are accumulated to obtain a calculated value, and the homologous data set whose calculated value is greater than or equal to a preset calculated value threshold is used as pre-training data; Step 4: Obtain access information of the linkage platform, and predict problems based on the access information, and trigger linkage measures based on the prediction results of the problem prediction; Predict problems based on access information, including: Obtaining recognition result features of a recognition result of a target recognition model; Determine the combined features of the linkage platform based on the identification result features and access information; Predict problems based on combined features and problem prediction models; Among them, according to the identification result characteristics and access information, the combined characteristics of the linkage platform are determined, including: Determine an impact information item according to the identification result feature and the access information, wherein the impact information item is associated with the identification result feature; Characterizing the impact information item information to obtain impact information features; The identification result feature and the impact information feature are taken together as the combined feature; According to the prediction results of problem prediction, linkage measures are triggered, including: According to the prediction result, determine the characteristics of the prediction result; Obtain historical execution measures of the linkage platform; Based on the triggering conditions of historical execution measures and the characteristics of the predicted results, determine the linkage measures; The triggering conditions for historical execution measures include: The first target execution event of the historical execution measure is similar to the second target execution event triggered by the predicted result characteristics on the linkage platform; The historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements. The condition matching requirements are: the historical execution conditions of the key condition types all have matching items in the current execution conditions, and the sum of the condition type weights of the target condition types of the matching items of the historical execution conditions of the non-key condition types in the current execution conditions is greater than or equal to the preset weight and threshold; Obtain historical execution measures of the linkage platform, including: Capturing historical execution measures based on the capturing conditions of historical execution measures; Among them, the capture conditions of historical execution measures include: The first time interval between two adjacent capture behaviors falls within a preset time interval; The second time interval between the historical time anchor point of each captured behavior and the current time point is less than or equal to a preset time interval threshold; A vector matching result of the first behavior vector of the capture behavior and the second behavior vector of the capture strategy is a match; The capture strategy is obtained by expanding the historical logs on the timeline, determining the axis position of the historical time anchor point on the timeline, traversing a preset number of log description items from the axis position to the front and back of the timeline, and obtaining the capture strategy based on the log description items.

2. The method for intelligent fire protection open access based on AI large model recognition algorithm as claimed in claim 1, characterized in that: Step 2: Obtain video recognition data of firefighting equipment and firefighting-related equipment data, including: Obtain information tags for firefighting equipment; Determine the key parameters of firefighting equipment based on information labels; Determine whether the fire-fighting equipment is operating normally based on key parameters; If the fire-fighting equipment is operating normally, a monitoring network for the corresponding fire-fighting equipment shall be established; Based on the monitoring network, video recognition data and fire-related equipment data are obtained.

3. The method for intelligent fire protection open access based on AI large model recognition algorithm as claimed in claim 2, characterized in that: Establish a monitoring network for corresponding fire-fighting equipment, including: establishing a monitoring network based on loosely coupled seamless access.

Citation Information

Patent Citations

  • Smart service based open authorization access design method and system

    CN105117657A

  • Fire alarm method, apparatus, and facility

    CN108053627A

  • Fire identification and early warning method and system

    CN111882810A

  • Fire-fighting early warning system based on fire-fighting relevance

    CN112863105A

  • Fire-fighting equipment management system and method based on Internet of Things

    CN116822964A