CN118445755B - Intelligent fire-fighting open access method based on AI large model recognition algorithm - Google Patents
Intelligent fire-fighting open access method based on AI large model recognition algorithm Download PDFInfo
- Publication number
- CN118445755B CN118445755B CN202410606436.9A CN202410606436A CN118445755B CN 118445755 B CN118445755 B CN 118445755B CN 202410606436 A CN202410606436 A CN 202410606436A CN 118445755 B CN118445755 B CN 118445755B
- Authority
- CN
- China
- Prior art keywords
- fire
- data
- fighting
- model
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000012544 monitoring process Methods 0.000 claims abstract description 56
- 238000012549 training Methods 0.000 claims description 35
- 230000006399 behavior Effects 0.000 claims description 25
- 238000000605 extraction Methods 0.000 claims description 24
- 230000007246 mechanism Effects 0.000 claims description 7
- 230000009471 action Effects 0.000 claims description 5
- 230000001960 triggered effect Effects 0.000 claims description 5
- 230000008878 coupling Effects 0.000 claims description 4
- 238000010168 coupling process Methods 0.000 claims description 4
- 238000005859 coupling reaction Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 15
- 230000009286 beneficial effect Effects 0.000 description 12
- 238000013475 authorization Methods 0.000 description 11
- 238000012512 characterization method Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 239000000779 smoke Substances 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000004393 prognosis Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Alarm Systems (AREA)
Abstract
The invention provides an intelligent fire-fighting open access method based on A I large model recognition algorithm, which belongs to the technical field of artificial intelligence, and comprises the following steps: step 1: acquiring historical monitoring data of fire-fighting equipment; step 2: acquiring video identification data of fire-fighting equipment and fire-fighting related equipment data; step 3: based on A I large model learning history monitoring data, a target recognition model is obtained, and video recognition data and fire-fighting related equipment data are analyzed and recognized through the target recognition model, so that open access of fire-fighting equipment is performed. According to the intelligent fire-fighting open access method based on the AI large model recognition algorithm, AI large model learning history monitoring data is introduced, and a target recognition model is obtained. The large model can be used for identifying the states and parameters of various fire-fighting related equipment to realize open access, so that the follow-up system can comprehensively monitor the equipment states and conduct risk pre-judgment, and effective linkage is conducted when a fire disaster occurs, so that the system is more intelligent and flexible.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an intelligent fire-fighting open access method based on an AI large model identification algorithm.
Background
AI large model recognition algorithms generally refer to neural network models with a large number of parameters in the field of deep learning, which are capable of processing and learning complex data patterns. These algorithms exhibit powerful performance in many areas of natural language processing, image recognition, speech recognition, etc. The intelligent fire-fighting open access refers to allowing fire-fighting equipment, systems or applications of different manufacturers to be connected and cooperatively work through an open interface and a standardized protocol so as to realize intelligent and networked fire-fighting safety management.
The application number is: the invention patent of CN201510435245.1 discloses a design method and a system of open authorized access based on intelligent service, wherein the method comprises the following steps: step 1: describing the relation among the modules of the service authorization layer and the service authorization request flow; step 2: according to the relation between the service authorization layer modules in the step 1 and the service authorization request process; step 3: according to the steps 1 and 2, designing a service authentication authorization status response required by the service open authorization access, and describing a service authentication authorization status response process by using a pseudo code; step 4: designing a class of service authentication and authorization modules according to the step 2 and the step 3; step 5: according to the class of the service authentication authorization module in the step 4, designing a required database table structure comprising a table name, a field of the table and a field type; step 6: according to the steps, a service open authorization interface call timing diagram is designed, and the interface call implementation process of the service open authorization is described in the form of pseudo codes.
However, in the prior art, corresponding pseudo codes are required to be designed for the accessed main body for subsequent call, and the design of the pseudo codes in the prior art usually requires manual participation and is not intelligent enough; in addition, each accessed subject requires adaptive design pseudocode, which is not flexible enough.
In view of the foregoing, there is a need for an intelligent fire-fighting open access method based on an AI large model identification algorithm to solve at least the above-mentioned drawbacks.
Disclosure of Invention
The invention aims to provide an intelligent fire-fighting open access method based on an AI large model recognition algorithm, which introduces A I large model learning history monitoring data to obtain a target recognition model. The large model can be used for identifying the states and parameters of various fire-fighting related equipment to realize open access, so that the follow-up system can comprehensively monitor the equipment states and conduct risk pre-judgment, and effective linkage is conducted when a fire disaster occurs, so that the system is more intelligent and flexible.
The embodiment of the invention provides an intelligent fire-fighting open access method based on an AI large model identification algorithm, which comprises the following steps:
Step 1: acquiring historical monitoring data of fire-fighting equipment;
Step 2: acquiring video identification data of fire-fighting equipment and fire-fighting related equipment data;
step 3: based on A I large model learning history monitoring data, a target recognition model is obtained, and video recognition data and fire-fighting related equipment data are analyzed and recognized through the target recognition model, so that open access of fire-fighting equipment is performed.
Preferably, step 3: based on AI large model learning history monitoring data, obtain target recognition model, carry out analysis and recognition to video recognition data and fire control related equipment data through target recognition model, carry out the open access of fire control equipment, include:
analyzing the historical monitoring data to obtain fire related data;
Based on the AI large model, identifying fire related features according to fire related data;
Learning fire related features, and obtaining a learning result, wherein the learning result is as follows: the association relationship between the fire disaster association characteristics and the fire disaster hidden danger;
Establishing a target identification model according to the learning result;
Analyzing video identification data and fire-fighting related equipment data through a target identification model to obtain an identification result;
judging whether fire hazards exist according to the identification result, and if the fire hazards exist, carrying out early warning based on a preset early warning mechanism.
Preferably, step 2: acquiring video identification data and fire-fighting related equipment data of fire-fighting equipment, including:
Acquiring an information tag of fire-fighting equipment;
Determining key parameters of the fire-fighting equipment according to the information label;
Judging whether the fire-fighting equipment is in normal operation or not according to the key parameters;
If the fire-fighting equipment operates normally, a monitoring network of the corresponding fire-fighting equipment is established;
And acquiring video identification data and fire-fighting related equipment data according to the monitoring network.
Preferably, establishing a monitoring network of the corresponding fire-fighting equipment comprises:
and establishing a monitoring network based on a loose coupling type seamless access mode.
The embodiment of the invention provides an intelligent fire-fighting open access method based on an AI large model identification algorithm, which further comprises the following steps:
step 4: acquiring access information of the linkage platform, performing problem prejudgment according to the access information, and triggering linkage measures according to prejudgment results of the problem prejudgment.
Preferably, based on A I big models, identifying fire-related features from the fire-related data includes:
extracting pre-training data in fire related data;
pre-training the AI large model through pre-training data to obtain a pre-training model;
and identifying fire related features in the fire related data according to the pre-training model.
Preferably, extracting pre-training data in fire-related data includes:
acquiring a monitoring data source of fire related data;
Dividing fire related data belonging to the same monitoring data source into the same data set and taking the same data set as a homologous data set;
Determining a comprehensive capability value of feature extraction according to the data type of fire related data in the homologous data set, and correlating with the corresponding homologous data set;
determining a feature extraction depth capability value according to the data quality of fire related data in the homologous data set, and correlating with the corresponding homologous data set;
And accumulating the feature extraction comprehensive capacity value and the feature extraction depth capacity value associated with the homologous data set to obtain a calculated value, and taking the homologous data set with the calculated value being greater than or equal to a preset calculated value threshold as pre-training data.
Preferably, the problem pre-judging according to the access information includes:
Acquiring recognition result characteristics of a recognition result of the target recognition model;
determining the combination characteristics of the linkage platform according to the identification result characteristics and the access information;
performing problem prejudgment according to the combined characteristics and the problem prejudgment model;
Wherein, according to recognition result characteristic and access information, confirm the combination characteristic of linkage platform, include:
Determining an influence information item according to the identification result characteristics and the access information, wherein the influence information item is associated with the identification result characteristics;
characterizing the influence information item information to obtain influence information characteristics;
the recognition result feature and the influence information feature are used together as a combined feature.
Preferably, triggering the linkage measure according to the prejudgment result of the prejudgment of the problem comprises:
determining the characteristics of the pre-judging result according to the pre-judging result;
acquiring historical execution measures of a linkage platform;
determining linkage measures according to the characteristics of the pre-judging result based on the triggering conditions of the history execution measures;
the triggering conditions of the history execution measure include:
the first target execution event of the history execution measure is similar to the second target execution event triggered by the linkage platform according to the characteristic of the pre-judging result;
the historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements, and the condition matching requirements are as follows: the method comprises the steps that matching items exist in historical execution conditions of key condition types in current execution conditions, and the sum of condition type weights of target condition types of matching items of non-key condition types in the current execution conditions is larger than or equal to a preset weight and a threshold.
Preferably, the step of obtaining the historical execution measure of the linkage platform includes:
Capturing the history execution measure based on the capturing condition of the history execution measure;
wherein the capturing conditions of the history execution measure include:
the first time interval of two adjacent capturing behaviors falls into a preset time interval;
The second time compartment of the historical time anchor point of each capturing action from the current time point is smaller than or equal to a preset time interval threshold value;
The vector matching result of the first behavior vector of the capturing behavior and the second behavior vector of the capturing strategy is matching;
The capture strategy is obtained by the following steps: expanding the history log on a time axis, determining the axis position of a history time anchor point on the time axis, traversing a preset number of log description items from the axis position to the front and back of the time axis, and acquiring a capture strategy according to the log description items.
The embodiment of the invention provides an intelligent fire-fighting open access system based on an AI large model identification algorithm, which comprises the following steps:
the historical monitoring data acquisition subsystem is used for acquiring historical monitoring data of the fire-fighting equipment;
the data acquisition subsystem is used for acquiring video identification data of the fire fighting equipment and fire fighting related equipment data;
And the open access subsystem is used for acquiring a target recognition model based on the AI large model learning history monitoring data, analyzing and recognizing the video recognition data and the fire-fighting related equipment data through the target recognition model, and performing open access of the fire-fighting equipment.
The beneficial effects of the invention are as follows:
The invention introduces AI large model learning history monitoring data to obtain a target recognition model. The large model can be used for identifying the states and parameters of various fire-fighting related equipment to realize open access, so that the follow-up system can comprehensively monitor the equipment states and conduct risk pre-judgment, and effective linkage is conducted when a fire disaster occurs, so that the system is more intelligent and flexible.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the application. The objects and other advantages of the application may be realized and obtained by means of the instrumentalities particularly pointed out in the specification.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a schematic diagram of an intelligent fire-fighting open access method based on an AI large model recognition algorithm in an embodiment of the invention;
fig. 2 is a schematic diagram of an intelligent fire-fighting open access system based on an AI large model recognition algorithm in an embodiment of the invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
The embodiment of the invention provides an intelligent fire-fighting open access method based on an AI large model identification algorithm, as shown in fig. 1, comprising the following steps:
Step 1: acquiring historical monitoring data of fire-fighting equipment; wherein, fire-fighting equipment is: the fire-fighting facilities to be managed and controlled, which are input in advance by staff, such as: a water tank, a smoke detector, a temperature sensor, a combustible gas detector and the like; the historical monitoring data are: data historically collected and recorded by fire equipment;
step 2: acquiring video identification data of fire-fighting equipment and fire-fighting related equipment data; the video identification data is: the system is connected with a monitoring camera to collect on-site monitoring video in real time; the fire-fighting related equipment data are: recorded data of fire-fighting related devices collected in real time based on the internet of things, such as: temperature sensor data, smoke sensor data, etc.;
Step 3: based on A I large model learning history monitoring data, a target recognition model is obtained, and video recognition data and fire-fighting related equipment data are analyzed and recognized through the target recognition model, so that open access of fire-fighting equipment is performed. Wherein, AI big model is: an artificial intelligence model capable of processing and analyzing a large amount of data and learning patterns and rules therefrom; the target recognition model is as follows: analyzing historical monitoring data to learn how to identify the artificial intelligent model which is accessed with data of different data types and predicts fire hazards, extracting useful features (such as features related to fire) in the historical monitoring data to train an AI large model during learning, and obtaining a target identification model after training is completed; the open access is: and integrating fire-fighting equipment of different manufacturers into the intelligent fire-fighting system through standardized communication protocols and interfaces.
The working principle and the beneficial effects of the technical scheme are as follows:
The application introduces AI large model learning history monitoring data to obtain a target recognition model. The large model can be used for identifying the states and parameters of various fire-fighting related equipment to realize open access, so that the follow-up system can comprehensively monitor the equipment states and conduct risk pre-judgment, and effective linkage is conducted when a fire disaster occurs, so that the system is more intelligent and flexible.
In one embodiment, step 3: based on AI large model learning history monitoring data, obtain target recognition model, carry out analysis and recognition to video recognition data and fire control related equipment data through target recognition model, carry out the open access of fire control equipment, include:
Analyzing the historical monitoring data to obtain fire related data; wherein, fire related data is: data potentially relevant to the occurrence of a fire, such as: temperature, smoke concentration, gas concentration, etc.;
based on the AI large model, identifying fire related features according to fire related data; wherein, fire related characteristics are: common features found in historical fire events, such as: the temperature reaches more than a few degrees;
learning fire related features, and obtaining a learning result, wherein the learning result is as follows: the association relationship between the fire disaster association characteristics and the fire disaster hidden danger; wherein, the association relation is: what fire associated features correspond to what fire hazards;
establishing a target identification model according to the learning result; wherein, the learning result is: model parameters and decision boundaries obtained after training an AI model;
Analyzing video identification data and fire-fighting related equipment data through a target identification model to obtain an identification result; wherein, the recognition result is: the target recognition model analyzes the video recognition data and the fire-fighting related equipment data to obtain a result, and the result can indicate what fire-fighting equipment the video recognition data and the fire-fighting related equipment data are and whether fire hazards are detected currently;
Judging whether fire hazards exist according to the identification result, and if the fire hazards exist, carrying out early warning based on a preset early warning mechanism. The preset early warning mechanism comprises the following steps: notifying the relevant personnel or fire department and taking corresponding action (e.g., activating the sprinkler system).
The working principle and the beneficial effects of the technical scheme are as follows:
According to the application, the historical monitoring data is analyzed to obtain fire related data potentially related to the occurrence of a fire, and the AI large model learning fire related data is introduced to obtain common fire related characteristics found in the historical fire event. And (3) learning fire related features to determine what fire related features correspond to what fire hidden danger, and establishing a target recognition model according to a learning result. According to the target recognition model, the video recognition data and the fire-fighting related equipment data, the fire-fighting equipment is opened and accessed, when the fire-fighting equipment is opened and accessed, the target recognition model recognizes the video recognition data and the fire-fighting related equipment data, determines what fire-fighting equipment is what data, and then carries out subsequent fire hidden danger recognition, and fire-fighting equipment of each manufacturer can be integrated into the existing system, so that the fire-fighting equipment is more flexible. And an early warning mechanism is introduced to timely early warn when fire hazards exist, so that the safety is improved.
In one embodiment, step 2: acquiring video identification data and fire-fighting related equipment data of fire-fighting equipment, including:
Acquiring an information tag of fire-fighting equipment; wherein, according to the information label, can confirm the public information of the fire-fighting equipment, such as: which vendor and which device;
Determining key parameters of the fire-fighting equipment according to the information label; wherein, key parameters are: parameters characterizing whether fire equipment is running;
Judging whether the fire-fighting equipment is in normal operation or not according to the key parameters; when judging whether the fire-fighting equipment normally operates, comparing and judging the key parameters with normal operation state parameters preset by the key parameters;
If the fire-fighting equipment operates normally, a monitoring network of the corresponding fire-fighting equipment is established; wherein, the monitoring network is: a network consisting of nodes of the Internet of things of the fire fighting equipment which operates normally;
And acquiring video identification data and fire-fighting related equipment data according to the monitoring network.
The working principle and the beneficial effects of the technical scheme are as follows:
the application acquires the information label of the fire-fighting equipment and judges whether the fire-fighting equipment is in normal operation or not according to the key parameters provided by the information label. If the fire fighting equipment is in normal operation, a monitoring network of corresponding fire fighting equipment is established, video identification data and fire fighting related equipment data are acquired according to the monitoring network, and reliability of acquiring the video identification data and the fire fighting related equipment data is improved.
In one embodiment, establishing a monitoring network for a respective fire apparatus includes:
and establishing a monitoring network based on a loose coupling type seamless access mode.
The working principle and the beneficial effects of the technical scheme are as follows:
the access mode of the Internet of things node of the intelligent fire protection system and the fire protection equipment which normally operates in the monitoring network is loose coupling type seamless access, so that the flexibility of acquiring video identification data and fire protection related equipment data is improved.
The embodiment of the invention provides an intelligent fire-fighting open access method based on an AI large model identification algorithm, which further comprises the following steps:
Step 4: acquiring access information of the linkage platform, performing problem prejudgment according to the access information, and triggering linkage measures according to prejudgment results of the problem prejudgment. Wherein, the linkage platform is: water systems, electrical systems, communication systems, and video systems; the access information is: the linkage platform is connected with the intelligent fire-fighting system to allow the intelligent fire-fighting system to read information; when the problem is prejudged according to the access information, carrying out early warning judgment according to an early warning judgment mechanism of the linkage platform; when the linkage measures are triggered, the control unit, triggering based on the technology of the Internet of things.
The working principle and the beneficial effects of the technical scheme are as follows:
The intelligent fire control system is introduced with the linkage platform, and access information of the linkage platform is read after the intelligent fire control system is linked with the linkage platform. And the problem judgment is carried out according to the early warning judgment mechanism of the linkage platform and the access information, and the linkage measures are triggered according to the prejudgment result of the problem judgment, so that the efficiency of the whole fire protection system is improved, and various emergency situations are effectively coped with.
In one embodiment, identifying fire-related features from fire-related data based on the AI large model includes:
Extracting pre-training data in fire related data; wherein, the pre-training data is: training data of AI model for extraction of fire related features:
Pre-training the AI large model through pre-training data to obtain a pre-training model; wherein, the pre-training model is: the AI model is used for extracting fire related features subsequently;
and identifying fire related features in the fire related data according to the pre-training model.
The working principle and the beneficial effects of the technical scheme are as follows:
The application acquires pre-training data for A I model training for fire associated feature extraction in fire associated data, pre-trains the AI large model through the pre-training data to acquire the pre-training model, performs subsequent fire hidden danger identification after the pre-training model identifies fire associated features in the fire associated data, performs model pre-training by using partial samples, reduces the dependence of the model on a large amount of labeling data, performs subsequent feature identification after the model has certain priori knowledge, and accelerates identification efficiency.
In one embodiment, extracting pre-training data in fire-related data includes:
Acquiring a monitoring data source of fire related data; wherein, the monitoring data source is: raw data sources that can provide fire-related data, such as: fire control monitoring data of the number of months of a certain year;
Dividing fire related data belonging to the same monitoring data source into the same data set and taking the same data set as a homologous data set;
Determining a comprehensive capability value of feature extraction according to the data type of fire related data in the homologous data set, and correlating with the corresponding homologous data set; wherein, the data type is: data categories of fire-related data, such as: temperature, smoke concentration, etc.; the feature extraction comprehensive capability value is as follows: evaluating a quantized value of the overall degree of the data types of the fire related data in the homologous data set, dividing the type number of the data types by the total type number, and multiplying the divided data types by 100 to obtain a feature extraction overall capacity value;
Determining a feature extraction depth capability value according to the data quality of fire related data in the homologous data set, and correlating with the corresponding homologous data set; wherein, the data quality is: the data accuracy, integrity, consistency and credibility are comprehensively determined; the feature extraction depth capability value is: evaluating quantized values of depth or detail degree of features which can be extracted from a single homologous data set, wherein the higher the data quality is, the higher the feature extraction depth capability value is, and the conversion coefficient of the distance between the two values is determined manually;
And accumulating the feature extraction comprehensive capacity value and the feature extraction depth capacity value associated with the homologous data set to obtain a calculated value, and taking the homologous data set with the calculated value being greater than or equal to a preset calculated value threshold as pre-training data. Wherein the preset calculated value threshold is preset manually.
The working principle and the beneficial effects of the technical scheme are as follows:
the application acquires the monitoring data source capable of providing fire related data and determines the homologous data set belonging to the same monitoring data source. Analyzing the homologous data set, screening pre-training data, wherein the specific process is as follows:
acquiring the data type of fire related data in the homologous data set, evaluating the comprehensive degree of the data type to acquire a characteristic extraction comprehensive capacity value and correlating the characteristic extraction comprehensive capacity value with a corresponding homologous data set; and acquiring the data quality of fire association data in the homologous data sets, evaluating the depth acquisition feature extraction depth capability value of the feature which can be extracted from the single homologous data set, and associating with the corresponding homologous data set. And summing the feature extraction comprehensive capacity value and the feature extraction depth capacity value associated with the homologous data set to obtain a calculated value, taking the homologous data set with the calculated value larger than a calculated value threshold as pre-training data, and pre-training by using high-quality data under the condition that labeling data are scarce, wherein priori knowledge of model learning is more reliable, and the method is more beneficial to subsequent feature extraction.
In one embodiment, problem prejudging is performed according to access information, including:
Acquiring recognition result characteristics of a recognition result of the target recognition model; wherein, the recognition result is characterized in that: characterization of the recognition result, such as: where there is what fire hazard;
determining the combination characteristics of the linkage platform according to the identification result characteristics and the access information; wherein, the combination characteristic is: the recognition result features and feature combinations of the characterization representation of the impact of the recognition result features on the linked platform, such as: the recognition result characteristic is that the machine room fires, and the characterization of the influence of the recognition result characteristic on the linkage platform (such as an electric system) is expressed as follows: powering off the machine room;
performing problem prejudgment according to the combined characteristics and the problem prejudgment model; the problem prejudging model comprises the following steps: an AI model for judging linkage problems caused by fire hazards according to the combined characteristics;
Wherein, according to recognition result characteristic and access information, confirm the combination characteristic of linkage platform, include:
Determining an influence information item according to the identification result characteristics and the access information, wherein the influence information item is associated with the identification result characteristics; wherein the influence information items are: identifying access information of an information type influenced by the result characteristics;
characterizing the influence information item information to obtain influence information characteristics; wherein, influence information characteristic is: a characterizing representation of an item of influence information, such as: burning out the power supply;
the recognition result feature and the influence information feature are used together as a combined feature.
The working principle and the beneficial effects of the technical scheme are as follows:
The method comprises the steps of characterizing a recognition result of a target recognition model to obtain recognition result features, determining an influence information item influenced by the recognition result features, characterizing the influence information item information to obtain influence information features, and taking the recognition result features and the influence information features as combined features. And a problem prejudging model is introduced, the combination characteristic is input to prejudge the problem, the influence information item is determined first, and then the subsequent judgment is carried out, so that the prejudging efficiency is improved.
In one embodiment, triggering the linkage measure according to the prejudgment result of the prejudgment of the problem comprises:
determining the characteristics of the pre-judging result according to the pre-judging result; wherein, the prejudgement result is characterized in that: characterization of the prognosis results, such as: the machine room fires, and the prejudgement result is characterized in that: power down time, place and area;
acquiring historical execution measures of a linkage platform; wherein, history executive measures are: the linkage platform records the management measures or actions taken in the past;
determining linkage measures according to the characteristics of the pre-judging result based on the triggering conditions of the history execution measures;
the triggering conditions of the history execution measure include:
The first target execution event of the history execution measure is similar to the second target execution event triggered by the linkage platform according to the characteristic of the pre-judging result; the first target execution event is: history of events for performing measure management; the second target execution event is: the event caused by the link platform is prejudged result characteristics;
the historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements, and the condition matching requirements are as follows: the method comprises the steps that matching items exist in historical execution conditions of key condition types in current execution conditions, and the sum of condition type weights of target condition types of matching items of non-key condition types in the current execution conditions is larger than or equal to a preset weight and a threshold. Wherein, the history execution condition is: specific conditions when the history execution conditions are executed; the current execution conditions are: under the current situation, the conditions faced by the linkage platform are as follows; the key condition types are preset by manpower, are the condition types which are necessary to be provided, and are non-key condition types except the condition types of the key condition types; the matching items are: the condition matching result of the current execution condition and the historical execution condition is the matched current execution condition; the target condition types are: when the current execution condition is matched with the historical execution condition of the non-key condition type, the corresponding non-key condition type; the condition type weight represents the importance degree of the matching condition of the non-key condition type, and the larger the condition type weight is, the higher the importance degree is; the preset weight and threshold are preset manually.
The working principle and the beneficial effects of the technical scheme are as follows:
the application characterizes the pre-judgment result and obtains the features of the pre-judgment result. And introducing historical execution measures adopted by the linkage platform in the past, introducing trigger conditions, and determining linkage measures according to the characteristics of the pre-judgment result. The triggering conditions include: the event managed by the history execution measure is similar to the event caused by the link platform of the prejudgement result characteristic; the historical execution conditions of the historical execution measures are matched with the conditions of the key condition types of the current execution conditions of the linkage platform, and the conditions of the non-key condition types are adaptively matched, so that the linkage platform is more flexible.
In one embodiment, obtaining historical execution measures of the linkage platform includes:
capturing the history execution measure based on the capturing condition of the history execution measure; wherein, the capturing conditions are: acquiring a standard of historical execution measures;
wherein the capturing conditions of the history execution measure include:
the first time interval of two adjacent capturing behaviors falls into a preset time interval; wherein the first time compartment is: time difference of behavior occurrence time of two time-adjacent capturing behaviors; the preset time interval is preset manually, for example: 12 hours to 24 hours;
The second time compartment of the historical time anchor point of each capturing action from the current time point is smaller than or equal to a preset time interval threshold value; wherein, the historical time anchor point is: capturing the execution time of the historical execution measures of behavior capture; the current time point is: at the current moment; the second time compartment is: a time difference between the historical time anchor and the current time point; the preset time interval threshold is preset manually;
The vector matching result of the first behavior vector of the capturing behavior and the second behavior vector of the capturing strategy is matching; wherein, the first behavior vector is: a vector describing the capture behavior feature; the second behavior vector is: describing a vector of capturing behavior features corresponding to the capturing strategy; the capture strategy is: extracting behavior schemes of historical execution measures;
The capture strategy is obtained by the following steps: expanding the history log on a time axis, determining the axis position of a history time anchor point on the time axis, traversing a preset number of log description items from the axis position to the front and back of the time axis, and acquiring a capture strategy according to the log description items. Wherein, the history log is: logging of the linkage platform; the preset number is as follows: 10 can also be set by the person himself; the log description items are: log description semantics; when the capture strategy is determined, a scheme suitable for extracting corresponding historical execution measures is determined according to log description semantics and is used as the capture strategy.
The working principle and the beneficial effects of the technical scheme are as follows:
The application introduces the capturing condition of the history execution measure, and obtains the history execution measure according to the capturing condition, wherein the specific constraint of the capturing condition is as follows: the first time interval of two adjacent capturing behaviors falls into a preset time interval, and capturing computing power resources are reasonably distributed; the execution time of the historical execution measures of each capturing action is smaller than or equal to a time interval threshold value from a second time interval of the current time point, and the historical execution measures are more time-efficient; the first behavior feature description vector of the capturing behavior is matched with the vector matching result of the second behavior feature description vector of the capturing strategy, the capturing strategy is determined according to log semantics near the execution time of the historical execution measures, and the capturing behavior is more adaptive and further more reasonable.
The embodiment of the invention provides an intelligent fire-fighting open access system based on an AI large model identification algorithm, as shown in fig. 2, comprising:
The historical monitoring data acquisition subsystem 1 is used for acquiring historical monitoring data of the fire-fighting equipment;
The data acquisition subsystem 2 is used for acquiring video identification data of the fire-fighting equipment and fire-fighting related equipment data;
And the open access subsystem 3 is used for acquiring a target recognition model based on the AI large model learning history monitoring data, analyzing and recognizing the video recognition data and the fire-fighting related equipment data through the target recognition model, and performing open access of the fire-fighting equipment.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (3)
1. An intelligent fire-fighting open access method based on an AI large model recognition algorithm is characterized by comprising the following steps:
Step 1: acquiring historical monitoring data of fire-fighting equipment;
Step 2: acquiring video identification data of fire-fighting equipment and fire-fighting related equipment data;
Step 3: acquiring a target recognition model based on the AI large model learning history monitoring data, analyzing and recognizing the video recognition data and the fire-fighting related equipment data through the target recognition model, and performing open access of the fire-fighting equipment;
Step 3: based on AI large model learning history monitoring data, obtain target recognition model, carry out analysis and recognition to video recognition data and fire control related equipment data through target recognition model, carry out the open access of fire control equipment, include:
analyzing the historical monitoring data to obtain fire related data;
Based on the AI large model, identifying fire related features according to fire related data;
Learning fire related features, and obtaining a learning result, wherein the learning result is as follows: the association relationship between the fire disaster association characteristics and the fire disaster hidden danger;
Establishing a target identification model according to the learning result;
Analyzing video identification data and fire-fighting related equipment data through a target identification model to obtain an identification result;
judging whether fire hazards exist according to the identification result, and if the fire hazards exist, carrying out early warning based on a preset early warning mechanism;
based on the AI large model, identifying fire associated features from the fire associated data, including:
extracting pre-training data in fire related data;
pre-training the AI large model through pre-training data to obtain a pre-training model;
identifying fire related features in the fire related data according to the pre-training model;
extracting pre-training data in fire-related data, comprising:
acquiring a monitoring data source of fire related data;
Dividing fire related data belonging to the same monitoring data source into the same data set and taking the same data set as a homologous data set;
Determining a comprehensive capability value of feature extraction according to the data type of fire related data in the homologous data set, and correlating with the corresponding homologous data set;
determining a feature extraction depth capability value according to the data quality of fire related data in the homologous data set, and correlating with the corresponding homologous data set;
Accumulating the feature extraction comprehensive capacity value and the feature extraction depth capacity value associated with the homologous data set to obtain a calculated value, and taking the homologous data set with the calculated value being greater than or equal to a preset calculated value threshold as pre-training data;
step 4: acquiring access information of a linkage platform, performing problem prejudgment according to the access information, and triggering linkage measures according to prejudgment results of the problem prejudgment;
Problem prejudging is carried out according to the access information, and the method comprises the following steps:
Acquiring recognition result characteristics of a recognition result of the target recognition model;
determining the combination characteristics of the linkage platform according to the identification result characteristics and the access information;
performing problem prejudgment according to the combined characteristics and the problem prejudgment model;
Wherein, according to recognition result characteristic and access information, confirm the combination characteristic of linkage platform, include:
Determining an influence information item according to the identification result characteristics and the access information, wherein the influence information item is associated with the identification result characteristics;
characterizing the influence information item information to obtain influence information characteristics;
the identification result features and the influence information features are used as combined features together;
triggering linkage measures according to a prejudgment result of the prejudgment of the problem, wherein the linkage measures comprise:
determining the characteristics of the pre-judging result according to the pre-judging result;
acquiring historical execution measures of a linkage platform;
determining linkage measures according to the characteristics of the pre-judging result based on the triggering conditions of the history execution measures;
the triggering conditions of the history execution measure include:
the first target execution event of the history execution measure is similar to the second target execution event triggered by the linkage platform according to the characteristic of the pre-judging result;
The historical execution conditions of the historical execution measures and the current execution conditions of the linkage platform meet the condition matching requirements, and the condition matching requirements are as follows: the method comprises the steps that matching items exist in historical execution conditions of key condition types in current execution conditions, and the sum of condition type weights of target condition types of the matching items of non-key condition types in the current execution conditions is larger than or equal to a preset weight and a threshold;
acquiring historical execution measures of the linkage platform, including:
Capturing the history execution measure based on the capturing condition of the history execution measure;
wherein the capturing conditions of the history execution measure include:
the first time interval of two adjacent capturing behaviors falls into a preset time interval;
The second time compartment of the historical time anchor point of each capturing action from the current time point is smaller than or equal to a preset time interval threshold value;
The vector matching result of the first behavior vector of the capturing behavior and the second behavior vector of the capturing strategy is matching;
The capture strategy is obtained by the following steps: expanding the history log on a time axis, determining the axis position of a history time anchor point on the time axis, traversing a preset number of log description items from the axis position to the front and back of the time axis, and acquiring a capture strategy according to the log description items.
2. The intelligent fire-fighting open access method based on the AI large model recognition algorithm as set forth in claim 1, wherein step 2: acquiring video identification data and fire-fighting related equipment data of fire-fighting equipment, including:
Acquiring an information tag of fire-fighting equipment;
Determining key parameters of the fire-fighting equipment according to the information label;
Judging whether the fire-fighting equipment is in normal operation or not according to the key parameters;
If the fire-fighting equipment operates normally, a monitoring network of the corresponding fire-fighting equipment is established;
And acquiring video identification data and fire-fighting related equipment data according to the monitoring network.
3. The intelligent fire-fighting open access method based on the AI large model identification algorithm of claim 2, wherein establishing a monitoring network of the corresponding fire-fighting equipment comprises: and establishing a monitoring network based on a loose coupling type seamless access mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410606436.9A CN118445755B (en) | 2024-05-16 | 2024-05-16 | Intelligent fire-fighting open access method based on AI large model recognition algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410606436.9A CN118445755B (en) | 2024-05-16 | 2024-05-16 | Intelligent fire-fighting open access method based on AI large model recognition algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118445755A CN118445755A (en) | 2024-08-06 |
| CN118445755B true CN118445755B (en) | 2024-11-01 |
Family
ID=92319575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410606436.9A Active CN118445755B (en) | 2024-05-16 | 2024-05-16 | Intelligent fire-fighting open access method based on AI large model recognition algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118445755B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108053627A (en) * | 2017-11-23 | 2018-05-18 | 深圳市泰和安科技有限公司 | A kind of fire alarm method, device and equipment |
| CN116822964A (en) * | 2023-08-25 | 2023-09-29 | 石家庄长川电气科技有限公司 | Fire-fighting equipment management system and method based on Internet of things |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180105510A (en) * | 2017-03-15 | 2018-09-28 | 신한대학교 산학협력단 | Method for estimating probability of stopping a fire railway vehicle in a tunnel for fire danger rate prediction of railway tunnels |
| PL3422222T3 (en) * | 2017-06-29 | 2024-07-15 | F. Hoffmann-La Roche Ag | Method and state machine system for detecting an operation status for a sensor |
| CN111489006A (en) * | 2019-01-29 | 2020-08-04 | 深圳富桂精密工业有限公司 | Fire development situation prediction method and device and computer-readable storage medium |
| CN111882810B (en) * | 2020-07-31 | 2022-07-01 | 广州市微智联科技有限公司 | Fire identification and early warning method and system |
| CN112863105A (en) * | 2021-01-20 | 2021-05-28 | 强悠悠 | Fire-fighting early warning system based on fire-fighting relevance |
| KR102626476B1 (en) * | 2021-12-27 | 2024-01-17 | 주식회사 현대케피코 | Disaster response method and system |
| CN117910811B (en) * | 2024-03-18 | 2024-05-28 | 深圳原世界科技有限公司 | Intelligent fire control management method and system based on multi-mode AI large model |
-
2024
- 2024-05-16 CN CN202410606436.9A patent/CN118445755B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108053627A (en) * | 2017-11-23 | 2018-05-18 | 深圳市泰和安科技有限公司 | A kind of fire alarm method, device and equipment |
| CN116822964A (en) * | 2023-08-25 | 2023-09-29 | 石家庄长川电气科技有限公司 | Fire-fighting equipment management system and method based on Internet of things |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118445755A (en) | 2024-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111445106B (en) | Safety control method and system for fault processing operation site of electricity acquisition equipment | |
| CN111609883B (en) | Communication machine room protection monitoring management system based on big data | |
| CN109145789A (en) | Power supply system safety work support method and system | |
| CN112990656B (en) | Health evaluation system and health evaluation method for IT equipment monitoring data | |
| CN104821060A (en) | Robbery early warning method and device | |
| CN117474357B (en) | Power distribution room operation and maintenance management method and system based on deep learning | |
| CN113936239A (en) | Intelligent fire fighting condition identification method and system based on neural network algorithm | |
| CN112578690B (en) | Intelligent building energy consumption management and control system and method | |
| CN108776452B (en) | Special equipment field maintenance monitoring method and system | |
| CN112800910A (en) | Communication machine room maintenance operation efficiency evaluation method and system | |
| CN112488898A (en) | Emergency handling system and method based on intelligent building management and control | |
| CN112528860A (en) | Safety tool management method and system based on image recognition | |
| CN113935645A (en) | High-risk operation management and control system and method based on identification analysis technology | |
| CN117439916A (en) | Network security test evaluation system and method | |
| CN115567690A (en) | Intelligent monitoring system capable of automatically identifying dangerous points of field operation | |
| CN116010886B (en) | Security monitoring method, device, electronic equipment and storage medium | |
| CN112863105A (en) | Fire-fighting early warning system based on fire-fighting relevance | |
| CN115424340A (en) | Method for identifying protective clothing based on YOLOV5 neural network technology | |
| CN108537920A (en) | Visitor's monitoring method based on recognition of face and system | |
| CN115394036A (en) | Monitoring and early warning method and system for building fire | |
| CN118445755B (en) | Intelligent fire-fighting open access method based on AI large model recognition algorithm | |
| CN117180687B (en) | Intelligent fire-fighting remote monitoring system and method | |
| CN116665419B (en) | Intelligent fault early warning system and method based on AI analysis in power production operation | |
| CN117899409A (en) | Remote management system for intelligent fire fighting | |
| CN117081917A (en) | Intelligent gateway based on AI (advanced technology attachment) internet of things sewage treatment operation and maintenance management |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |