An intelligent firefighting operation and maintenance platform based on artificial intelligence

Abstract

The embodiments of the present specification disclose an intelligent fire protection operation and maintenance platform based on artificial intelligence, which includes a fire detection system, which is used to: obtain fire monitoring information of a target area; adapt the fire monitoring information to at least one fire deduction logic to obtain an adaptation result; wherein the fire deduction logic includes a mapping rule from fire monitoring information to a fire risk state, and the adaptation result includes a deduction formula for mapping the fire monitoring information to the fire risk state; input the fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first fire monitoring information feature; use a distance calculation model to determine a distance metric; input the distance metric and the first fire monitoring information feature into a feature enhancement network to perform a feature enhancement operation to obtain a second fire monitoring information feature; input the second fire monitoring information feature into a trained fire detection model for prediction to obtain the fire risk state of the target area.

Description

Technical Field

[0001] The present application relates to the field of smart firefighting technology, and more specifically, to an artificial intelligence-based smart firefighting operation and maintenance platform. Background Art

[0002] With the development of artificial intelligence technology, fire detection functions have begun to be widely used in the field of smart fire protection. For example, the fire detection function can be integrated into the smart fire protection operation and maintenance platform to intelligently assess the fire risk status based on the fire monitoring information of the target area.

[0003] At present, there are two main common fire detection schemes. The first is a fire reasoning scheme based on the experience of fire experts. This scheme mainly organizes relevant reasoning theories based on authoritative fire knowledge or historical fire experience, and performs fire detection by building an experience decision-making system based on these reasoning theories; the second is fire detection based on machine learning or deep learning models.

[0004] The above two fire detection schemes each have their own advantages and disadvantages. Among them, the advantage of the first scheme is that it is relatively simple and direct, and has a certain degree of evidence and comprehensibility. The disadvantage is that it requires a lot of manual work to organize the reasoning theory, and these reasoning theories cannot be automatically updated with the development of technology. The second scheme can achieve a higher accuracy based on the pre-trained model and a large amount of fire monitoring data, but at the same time lacks a certain degree of comprehensibility and cannot provide an intuitive derivation basis. In addition, for fire situations with a low probability of occurrence, the prediction effect may be poor due to insufficient training samples.

[0005] Therefore, it is hoped to study an intelligent fire protection operation and maintenance platform that can integrate the above two fire detection schemes, so as to improve the accuracy, reliability and comprehensibility of fire detection results. Summary of the invention

[0006] The embodiment of this specification provides an artificial intelligence-based smart fire operation and maintenance platform, the platform includes a fire detection system, and the fire detection system is used to:

[0007] Acquire fire monitoring information of a target area, wherein the fire monitoring information includes at least one of real-time monitoring data, historical fire information, and fire hazard information;

[0008] Adapting the fire monitoring information to at least one fire deduction logic to obtain an adaptation result of the fire monitoring information and the at least one fire deduction logic; wherein the fire deduction logic includes a mapping rule from the fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state;

[0009] Inputting the fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first fire monitoring information feature of the fire monitoring information;

[0010] Determine a distance metric based on the at least one deduction formula and a characteristic dimension corresponding to the at least one fire risk state using a distance calculation model;

[0011] Inputting the distance metric and the first fire monitoring information feature into a feature enhancement network, and performing a feature enhancement operation on a feature dimension corresponding to the at least one fire risk state of the first fire monitoring information feature by using the distance metric to obtain a second fire monitoring information feature;

[0012] The second fire monitoring information feature is input into a trained fire detection model for prediction to obtain the fire risk status of the target area.

[0013] In some embodiments, the fire simulation logic is obtained based on the following method:

[0014] Obtaining a fire knowledge description text, wherein the fire knowledge description text is used to describe the relationship between the fire monitoring information and the fire risk status;

[0015] Performing fire monitoring information identification and fire risk status identification on the fire knowledge description text to obtain key indicators in the fire monitoring information and risk status in the fire risk status corresponding to the key indicators;

[0016] The key indicators and their corresponding risk states are combined with logical statements to obtain the at least one fire deduction logic.

[0017] In some embodiments, the fire detection system is further configured to:

[0018] fusing the at least one fire deduction logic to obtain a fused deduction logic, wherein the fused deduction logic includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state;

[0019] The using the distance calculation model to determine the distance metric based on the at least one deduction formula and the characteristic dimension corresponding to the at least one fire risk state includes:

[0020] The distance metric is obtained by performing fuzzy operation on at least one deduction formula for mapping the fire monitoring information included in the fusion deduction logic to at least one fire risk state and a characteristic dimension corresponding to the at least one fire risk state through the distance calculation model.

[0021] In some embodiments, fusing the at least one fire deduction logic to obtain a fused deduction logic includes:

[0022] Based on the adaptation result, determine N fire deduction logics adapted to the fire monitoring information;

[0023] Performing fuzzy reasoning on the fire deduction logics corresponding to the same fire risk state in the N fire deduction logics to obtain a deduction formula corresponding to the same fire risk state in the fused deduction logic;

[0024] According to the fire deduction logic corresponding to other fire risk states except the same fire risk state, a deduction formula corresponding to the other fire risk states in the fused deduction logic is obtained.

[0025] In some embodiments, the fire detection model is trained based on the following method:

[0026] Acquire a fire monitoring sample, wherein the fire monitoring sample includes sample fire monitoring information of a sample target area and a sample fire risk status label;

[0027] Inputting the sample fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first sample fire monitoring information feature of the sample fire monitoring information;

[0028] Obtaining a distance metric determined based on the distance calculation model, and inputting the distance metric and the first sample fire monitoring information feature into a feature enhancement network, and then using the distance metric to perform a feature enhancement operation on a feature dimension corresponding to at least one fire risk state of the first sample fire monitoring information feature to obtain a second sample fire monitoring information feature;

[0029] Inputting the second sample fire monitoring information feature into the initial fire detection model for prediction to obtain a predicted fire risk state corresponding to the target area;

[0030] The model parameters are iteratively optimized according to the distance between the predicted fire risk state and the sample fire risk state label to obtain the trained fire detection model.

[0031] In some embodiments, the iterative optimization of model parameters based on the distance between the predicted fire risk state and the sample fire risk state label includes: collaboratively optimizing parameters of the feature extraction network, the feature enhancement network, and the initial fire detection model.

[0032] In some embodiments, the platform further comprises an escape and rescue planning system, wherein the escape and rescue planning system is used to:

[0033] When it is detected that the fire risk status of the target area meets the preset conditions, obtaining the building structure information and real-time monitoring data of the target area;

[0034] determining at least one candidate escape route based on the building structure information;

[0035] Calculating an estimated escape time corresponding to each candidate escape route, and predicting a risk index corresponding to each candidate escape route based on the building structure information and the real-time monitoring data;

[0036] The score corresponding to each candidate escape route is calculated according to the estimated escape time and the risk index, and a candidate escape route with the highest score is selected as the target escape route.

[0037] In some embodiments, the predicting the risk index corresponding to each candidate escape route based on the building structure information and the real-time monitoring data includes:

[0038] Determining the level of escape risk factors that affect escape time and / or escape difficulty according to the building structure information and the real-time monitoring data, wherein the escape risk factors include personnel flow, passage capacity, and proximity to a fire source;

[0039] Obtain the escape risk weight corresponding to each escape risk factor;

[0040] Based on the level of the escape risk factor and the escape risk weight corresponding to the escape risk factor, a risk index corresponding to each candidate escape route is obtained.

[0041] In some embodiments, the escape and rescue planning system is further used to:

[0042] Get the user's real-time location information;

[0043] Based on the real-time location information and the target escape route, generating corresponding escape guidance information, the escape guidance information including at least one of an escape route map, voice guidance, and text guidance;

[0044] The escape guidance information is published to the user's mobile terminal device and / or the on-site indicating device.

[0045] In some embodiments, the escape and rescue planning system is further used to:

[0046] When receiving an emergency request for help, obtain the real-time location information of the user seeking help;

[0047] Generate an emergency rescue plan for the user based on the real-time location information of the user seeking help and the target escape route;

[0048] The emergency rescue plan is sent to rescue personnel and / or rescue equipment.

[0049] The beneficial effects that may be brought about by the intelligent fire protection operation and maintenance platform based on artificial intelligence provided in the embodiments of this specification include at least:

[0050] (1) By integrating the fire reasoning scheme based on fire expert experience and fire knowledge with the fire detection scheme based on the pre-trained model, the accuracy, reliability and understandability of fire detection can be improved;

[0051] (2) By fusing multiple fire deduction logics adapted to fire monitoring information, a fused deduction logic is generated, which can not only cope with complex and changeable fire scenes, but also integrate the advantages of different deduction logics, so that the fire detection system can better adapt to various fire types and environmental conditions, thereby improving the sensitivity and accuracy of fire detection;

[0052] (3) By introducing the feature enhancement network, the first fire monitoring information feature obtained by the feature extraction network through feature extraction of fire monitoring information can be enhanced in combination with the fire deduction logic based on fire knowledge or historical fire experience, thereby improving the accuracy and comprehensibility of the fire risk status detection results obtained based on the enhanced second fire monitoring information feature in the subsequent process;

[0053] (4) By configuring the fire detection system and the escape and rescue planning system on the smart fire operation and maintenance platform, the fire risk prediction and detection can be achieved. At the same time, it can respond quickly when a fire occurs, and combine the building structure information and real-time monitoring data to automatically trigger the calculation process of the escape and rescue planning system to generate corresponding escape guidance information. Through this smart fire operation and maintenance platform, the target escape route and emergency rescue plan can be quickly generated when a fire occurs, providing strong support for personnel evacuation and rescue operations.

[0054] Additional features will be described in part in the following description. For those skilled in the art, it will become apparent by reviewing the following and accompanying drawings, or it may be understood by the production or operation of the examples. The features of this specification may be implemented and obtained by practicing or using various aspects of the methods, tools, and combinations described in the following detailed examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0056] Figure 1 is an exemplary workflow diagram of a fire detection system of an intelligent fire protection operation and maintenance platform based on artificial intelligence according to some embodiments of this specification;

[0057] Figure 2 It is a module diagram of an exemplary working principle of a fire detection system of an intelligent fire protection operation and maintenance platform based on artificial intelligence according to some embodiments of this specification;

[0058] Figure 3 is a flow chart of an exemplary training method of a fire detection model according to some embodiments of this specification;

[0059] Figure 4 It is an exemplary workflow diagram of the escape and rescue planning system of the intelligent fire operation and maintenance platform based on artificial intelligence as shown in some embodiments of this specification. DETAILED DESCRIPTION

[0060] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, without paying creative work, this specification can also be applied to other similar scenarios based on these drawings. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0061] The embodiment of this specification provides an intelligent fire protection operation and maintenance platform based on artificial intelligence, which includes a fire detection system, which can integrate a fire reasoning solution based on fire protection expert experience and fire protection knowledge with a fire detection solution based on a pre-trained model, thereby improving the accuracy, reliability and understandability of fire detection. The intelligent fire protection operation and maintenance platform based on artificial intelligence provided by the embodiment of this specification is described in detail below in conjunction with the accompanying drawings.

[0062] Figure 1 The following is an exemplary workflow diagram of a fire detection system in an intelligent fire protection operation and maintenance platform based on artificial intelligence according to some embodiments of this specification. The steps in the flowchart 100 can be executed by any electronic device with data processing capabilities. For example, the electronic device can be implemented as a server or a terminal device, which is not limited in this specification. Figure 1 In the embodiment of this specification, the fire detection system can be used to perform the following steps S110 to S160.

[0063] Step S110, obtaining fire monitoring information of the target area, wherein the fire monitoring information includes at least one of real-time monitoring data, historical fire information, and fire hazard information.

[0064] In the embodiments of the present application, the target area may refer to any place where fire monitoring is required, such as buildings, factories, warehouses, etc. Fire monitoring information can be obtained through various sensors, such as smoke sensors, temperature sensors, flame sensors, etc. These sensors can be installed in multiple different locations within the target area to ensure comprehensive coverage and timely detection of signs of fire. In the embodiments of the present specification, the fire monitoring information may include at least one of real-time monitoring data, historical fire information, and fire hazard information, and the fire monitoring information may be used to determine the fire risk status of the target area.

[0065] Specifically, in the embodiments of the specification, the aforementioned real-time monitoring data may include smoke concentration, temperature change, flame intensity, etc. Among them, the smoke concentration is the current smoke level obtained by the smoke sensor (for example, the concentration of smoke particles in each cubic meter of air is detected by the smoke sensor); the temperature change is the rate of change calculated from the ambient temperature data obtained by the temperature sensor; the flame intensity is the flame brightness or heat radiation intensity obtained by the flame sensor. Historical fire information may include information such as the time, location, scale and cause of the fire, and fire hazard information may include the storage of flammable materials (for example, detailed information on the chemical storage area, including the type and quantity of stored chemicals, etc.), the safety status of electrical equipment, and the status of fire safety facilities. It should be noted that the above fire monitoring information is only an exemplary description. In some embodiments, the fire monitoring information may include more content. For example, in some embodiments, the real-time monitoring data may also include sound detection data, video monitoring data, wind speed detection data, etc.

[0066] In some embodiments, part of the content in the fire monitoring information can be processed to extract key indicators. In this specification, the key indicator can be understood as an important parameter for assessing fire risk and the possibility of fire. For example, by analyzing the trend of smoke concentration changes in real-time monitoring data, it can be determined whether there is abnormal smoke generation, which may indicate the occurrence of an initial fire. Similarly, an abnormal increase in the rate of temperature change may indicate the presence of a fire source. In addition, a sudden increase in flame intensity is also an important indicator of fire. For example, in some embodiments, the average and peak values ​​of smoke concentration, the trend and abnormal points of temperature changes, the rate of change of flame intensity, etc. can be calculated from real-time monitoring data. These key indicators can help more accurately assess fire risks and provide a basis for subsequent fire detection analysis. For example, in some embodiments, keywords and patterns can be extracted from the recorded text of historical fire information and fire hazard information, such as specific fire occurrence frequencies, types and quantities of flammable materials, common failure points of electrical equipment, etc.

[0067] In some embodiments, the fire monitoring information may be transmitted to the central processing unit of the fire detection system via a network (which may include a wired network and / or a wireless network). In some embodiments, in order to improve the reliability of the system, multiple redundant communication links may be set to ensure that when a communication link fails, other links can quickly take over the data transmission task.

[0068] Step S120, adapting the fire monitoring information to at least one fire deduction logic to obtain an adaptation result of the fire monitoring information and the at least one fire deduction logic; wherein the fire deduction logic includes a mapping rule from the fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state.

[0069] In the implementation of this specification, the fire deduction logic can be understood as a series of preset rules or algorithms for analyzing and interpreting fire monitoring information to predict the possibility and potential risk of fire. In the embodiments of this specification, the fire deduction logic can be constructed based on historical data, expert experience, statistical analysis and other methods to ensure its accuracy and reliability.

[0070] In some embodiments, the fire deduction logic may include a mapping rule from fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state, and the deduction formula can be understood as a mathematical formula or model that converts monitoring data into a fire risk assessment result. For example, a comprehensive assessment formula based on smoke concentration, temperature change and flame intensity can be set, which combines these key indicators according to certain weights to calculate the fire assessment index of the target area. The higher the index, the greater the possibility of a fire. In the present illustrative embodiment, the mapping rule from the aforementioned fire monitoring information to at least one fire risk state can be determined based on a variety of methods such as threshold rules, trend analysis, and combination conditions.

[0071] In some embodiments, the aforementioned fire simulation logic can be obtained based on the following method:

[0072] First, a large number of fire knowledge description texts can be obtained, and the fire knowledge description texts can be used to describe the relationship between fire monitoring information and fire risk status. Then, the fire knowledge description texts can be used to identify fire monitoring information and fire risk status, and obtain key indicators in the fire monitoring information and risk status corresponding to the key indicators in the fire risk status. Finally, the key indicators and their corresponding risk statuses are combined with logical statements to obtain at least one fire deduction logic. In some embodiments, the aforementioned fire risk status may include a fire type and a corresponding risk level.

[0073] Specifically, in the embodiments of this specification, the fire knowledge description text can be obtained based on authoritative fire knowledge or expert experience. Fire knowledge can include knowledge and experience formed in the field of fire technology over a long period of time. Exemplarily, the fire knowledge description text can be in the form of natural language description, such as "the typical precursors or current phenomena of fire type A include feature 1, feature 2, feature 3, etc.", "Area B is a fire-prone area", "Area C stores flammable items D in quantity E", etc.

[0074] In the embodiments of the present specification, the fire knowledge description text can be used to identify fire monitoring information and fire risk status, and at least one fire deduction logic can be obtained. Exemplarily, in some embodiments, the fire monitoring information and fire risk status identified from a fire knowledge description text can be combined with logical statements to obtain the fire deduction logic from the fire monitoring information to the fire risk status. As an example only, in some embodiments, the fire deduction logic can be expressed in the form of an IF statement. For example, for the fire knowledge description text "fire is usually accompanied by smoke, high temperature and abnormal sound", the precursor or current phenomenon information (an example of fire monitoring information) can be identified as smoke, high temperature, abnormal sound, and the fire risk status is identified as high. The corresponding fire deduction logic is "IF smoke detection value>threshold 1AND temperature detection value>threshold 2AND sound detection value>threshold 3, THEN fire risk status = high". Similarly, for the fire knowledge description text "fire may occur when the temperature of the electrical circuit is higher than the threshold value 4", it can be identified that the fire monitoring information is the temperature of the electrical circuit, and the fire risk status is identified as medium, and the corresponding fire deduction logic is obtained as "IF the electrical circuit temperature detection value > threshold value 4, THEN fire risk status = medium". In other embodiments of the present specification, the fire knowledge description text may also include a description of the chain reaction that may be triggered after the fire occurs. For example, the text may be described as "when a fire occurs in the storage area of ​​flammable and explosive items, there may be an explosion risk". Based on such a description, it can be identified that the fire monitoring information is a fire alarm signal in the storage area of ​​flammable and explosive items, and the fire risk status is identified as extremely high, and the corresponding fire deduction logic is obtained: "IF the fire alarm signal in the storage area of ​​flammable and explosive items is triggered, THEN fire risk status = extremely high".

[0075] Each fire deduction logic may correspond to a weight, and the weight may be used to characterize the confidence of the corresponding fire deduction logic. In the embodiments of this specification, the value of the weight may be between 0 and 1. Generally speaking, the more detailed the description of fire judgment in a fire knowledge description text, the higher the corresponding confidence and the greater the weight. In some embodiments, professional firefighters may label the fire knowledge description text with a weight, and use the weight as the weight of the corresponding fire deduction logic.

[0076] Reference Figure 2 In the embodiment of the present specification, after the fire knowledge description text is processed in the above manner to obtain at least one fire deduction logic, the at least one fire deduction logic can be stored in the deduction logic database. Accordingly, in the embodiment of the present specification, the weight corresponding to each fire deduction logic can also be stored in the deduction logic database.

[0077] In the above step S120, the fire monitoring information can be adapted to at least one fire deduction logic in the deduction logic database to obtain the adaptation result of the fire monitoring information and each fire deduction logic. The adaptation result may include the degree of adaptation of the fire monitoring information and each fire deduction logic. In the embodiment of the present specification, the degree of adaptation can be determined according to the consistency of the indicator items of the condition part between the fire monitoring information and the fire deduction logic.

[0078] For example, when part or all of the information in the fire monitoring information is completely consistent with the indicator items of the conditional part in a fire deduction logic, the degree of adaptation of the fire monitoring information to the fire deduction logic is 100%; when part or all of the information in the fire monitoring information is 50% consistent with the indicator items of the conditional part in a fire deduction logic, the degree of adaptation of the fire monitoring information to the fire deduction logic is 50%; when part or all of the information in the fire monitoring information is completely inconsistent with the indicator items of the conditional part in a fire deduction logic, the degree of adaptation of the fire monitoring information to the fire deduction logic is 0.

[0079] Step S130: input the fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first fire monitoring information feature of the fire monitoring information.

[0080] In some embodiments, each dimension of the first fire monitoring information feature may correspond to a fire type. For example, when the number of fire types that the feature extraction network can predict is W, the first fire monitoring information feature may be a W-dimensional feature vector, each dimension of which may correspond to a fire type, and accordingly, the feature of the dimension corresponds to the feature component of the fire monitoring information in the corresponding fire type. In some embodiments of the present specification, the number of fire types that the feature extraction network can predict is consistent with the number of fire types contained in the fire risk state corresponding to the fire deduction logic in the database.

[0081] It should be noted that in the embodiments of this specification, the feature extraction network may include CNN (Convolutional Neural Network) or RNN (Recurrent Neural Network), or other neural networks with feature extraction or encoding functions, such as Transformer encoders, etc., which are not limited in this specification.

[0082] Step S140: using a distance calculation model to determine a distance metric based on the at least one deduction formula and a characteristic dimension corresponding to the at least one fire risk state.

[0083] Reference Figure 2In an embodiment of the present specification, by adapting the fire monitoring information with the fire deduction logic in the deduction logic database, an adaptation result of the fire monitoring information and at least one fire deduction logic can be obtained, wherein the fire deduction logic includes a mapping rule from the fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state.

[0084] Furthermore, a distance calculation model may be used to determine a distance metric based on the characteristic dimensions corresponding to the at least one deduction formula and the at least one fire risk state. In an embodiment of the present specification, the distance metric may be used to characterize the similarity between the deduction formula and the corresponding fire risk state in the number of characteristic dimensions. When the characteristic dimensions between the deduction formula and the fire risk state are highly consistent, the distance metric is small; conversely, when the characteristic dimensions between the deduction formula and the fire risk state are significantly different, the distance metric is large. It can be understood that the higher the similarity between the deduction formula and the corresponding fire risk state in the number of characteristic dimensions, the higher the accuracy of the deduction formula in predicting the fire risk state. Therefore, by calculating the distance metric, the degree of match between the deduction formula and the fire risk state (the degree of match can be understood as the degree of activation of the deduction formula) can be effectively evaluated, thereby providing a basis for subsequent processing.

[0085] Specifically, in some embodiments, the distance metric can be obtained by performing fuzzy operations on at least one deduction formula that maps fire monitoring information to at least one fire risk state and the characteristic dimension corresponding to the at least one fire risk state. It can be understood that fuzzy operation is a mathematical method for dealing with uncertainty and ambiguity, which allows partial matching and incomplete consistency to be considered in the calculation process. In the embodiments of this specification, fuzzy operations may include operations on fuzzy sets, such as fuzzy intersections, fuzzy unions, and fuzzy complements. Through these operations, a fuzzy distance metric can be obtained, which can reflect the degree of similarity between the deduction formula and the fire risk state in the characteristic dimension.

[0086] Specifically, in some embodiments, the calculation process of the distance metric can be expressed as:

[0087]

[0088] Among them, D(F i ,R i) represents the distance measure between the deduction formula Fi and the fire risk status Ri, Fi∩Ri represents the intersection of the deduction formula Fi and the fire risk status Ri in the feature dimension, and Fi∪Ri represents the union of the deduction formula Fi and the fire risk status Ri in the feature dimension. The range of this distance measure is [0,1], where 0 represents that the deduction formula Fi and the fire risk status Ri are completely consistent in the feature dimension, and 1 represents that the deduction formula Fi and the fire risk status Ri are completely inconsistent in the feature dimension.

[0089] In some embodiments, the aforementioned fire detection system may be used to fuse the at least one fire deduction logic to obtain a fused deduction logic, wherein the fused deduction logic includes at least one deduction formula that maps the fire monitoring information to the at least one fire risk state. As an example only, in some embodiments, the fused deduction logic may be obtained based on the following method:

[0090] First, based on the above-mentioned adaptation results, N fire deduction logics adapted to the fire monitoring information are determined; then, fuzzy reasoning is performed on the fire deduction logics corresponding to the same fire risk state in the N fire deduction logics to obtain the deduction formula corresponding to the same fire risk state in the fused deduction logic; based on the fire deduction logics corresponding to other fire risk states except the same fire risk state, the deduction formula corresponding to the other fire risk states in the fused deduction logic is obtained.

[0091] It should be noted that in some embodiments of the present specification, N fire deduction logics with a degree of adaptation greater than a preset threshold (e.g., 0.3) can be selected from the deduction logic database based on the adaptation result, and then the N fire deduction logics can be fused. Fusion processing can adopt a variety of methods, such as weighted average, fuzzy logic fusion, or rule-based fusion strategy. In the weighted average method, the weight of each deduction logic can be assigned based on its degree of adaptation to ensure that the deduction logic with a higher degree of adaptation occupies a larger proportion in the fusion result. Fuzzy logic fusion can use fuzzy set operations to integrate the characteristics of different deduction logics to generate a comprehensive deduction formula. The rule-based fusion strategy can decide how to combine the characteristics of different deduction logics according to a set of predefined rules to generate the final fused deduction logic.

[0092] Specifically, in some embodiments of the present specification, fuzzy reasoning can be performed on the fire deduction logics corresponding to the same fire risk state in the N fire deduction logics to obtain the deduction formula corresponding to the same fire risk state in the fused deduction logic; then, according to the fire deduction logics corresponding to other fire risk states except the same fire risk state, the deduction formulas corresponding to the other fire risk states in the fused deduction logic are obtained. For example, assume that there are three fire deduction logics, marked as L1, L2 and L3 respectively. First, it is determined through the adaptation results that the three deduction logics are all adapted to the fire monitoring information, and they all correspond to the fire risk state R1. Next, fuzzy reasoning can be performed on the deduction formulas corresponding to the same fire risk state R1 in the three deduction logics to obtain the deduction formula corresponding to R1 in the fused deduction logic. In the fused deduction logic, in order to ensure the accuracy and reliability of the final deduction formula, fuzzy rules in fuzzy reasoning can be used to deal with conflicts and uncertainties between different deduction logics. For example, if there are certain differences between L1 and L2 in the deduction formula of the fire risk state R1, a compromise deduction formula can be determined through fuzzy rules, which can comprehensively consider the advantages of L1 and L2 while reducing the differences between them.

[0093] Similarly, for other fire risk states R2, R3, etc., similar methods can be used for processing. Specifically, fuzzy reasoning can be performed on the deduction logic corresponding to R2 and R3 in L1, L2 and L3 respectively to obtain the deduction formulas corresponding to R2 and R3 in the fused deduction logic. When processing these different fire risk states, a rule-based fusion strategy can be adopted to ensure that the fused deduction logic can fully cover various possible fire risk states. In addition, for those fire risk states that are only corresponding to a single deduction logic in the adaptation results, such as the fire risk state R2 corresponding to L3, the deduction formula of the deduction logic can be directly incorporated into the fused deduction logic. In this way, the fused deduction logic can not only reflect the mapping relationship between fire monitoring information and fire risk states, but also integrate the advantages of multiple deduction logics to improve the accuracy and reliability of fire risk prediction.

[0094] It should be pointed out that in the embodiments of this specification, by generating the fused deduction logic, not only can complex and changeable fire scenes be coped with, but also the advantages of different deduction logics can be integrated, so that the fire detection system can better adapt to various fire types and environmental conditions, thereby improving the sensitivity and accuracy of fire detection.

[0095] It should also be noted that in the embodiments of the present specification, after the at least one fire deduction logic is fused to obtain the fused deduction logic, the distance calculation model can be used to perform fuzzy operations on at least one deduction formula that maps the fire monitoring information included in the fused deduction logic to at least one fire risk state and the characteristic dimension corresponding to the at least one fire risk state to obtain the distance metric.

[0096] Step S150, inputting the distance metric and the first fire monitoring information feature into a feature enhancement network, and performing a feature enhancement operation on the feature dimension corresponding to the at least one fire risk state of the first fire monitoring information feature by using the distance metric to obtain a second fire monitoring information feature.

[0097] Reference Figure 2 In an embodiment of the present specification, after obtaining the first fire monitoring information feature and the distance metric through the above steps, the first fire monitoring information feature and the distance metric can be input into a feature enhancement network for feature enhancement operation to obtain a second fire monitoring information feature.

[0098] In the embodiments of this specification, the feature enhancement network may be a deep learning network for enhancing the features of fire monitoring information to improve the accuracy and reliability of fire detection. Specifically, in some embodiments, the feature enhancement network may include multiple convolutional layers, pooling layers, and fully connected layers. Through the combination of these layers, the network can learn the key features in the fire monitoring information and enhance these features.

[0099] In the feature enhancement network, the first fire monitoring information feature and the distance metric can be used as input, and after layer-by-layer processing of the network, the second fire monitoring information feature can be finally obtained. The second fire monitoring information feature can not only contain the key information in the original monitoring information, but also enhance the feature dimensions related to the fire risk state through the guidance of the distance metric.

[0100] For example, assume that the first fire monitoring information feature contains three dimensions: temperature, smoke concentration, and flame intensity. In the feature enhancement network, through the combination of convolutional layers and pooling layers, the network can extract local features of temperature and smoke concentration, and combine these features with distance metrics through fully connected layers. The distance metric can indicate the correlation between the current monitoring information and the fire risk status, thereby guiding the network to enhance the feature dimensions that are closely related to the fire risk status.

[0101] For example, if the distance metric shows that the current monitoring information has a high correlation with the fire risk state R1, the feature enhancement network will focus on enhancing the feature dimensions related to R1, such as temperature and smoke concentration. In the embodiments of this specification, the above-mentioned enhancement operation may include increasing the weights of these feature dimensions, or further extracting and amplifying the detailed information of these features through nonlinear transformation to obtain the second fire monitoring information feature.

[0102] In the embodiments of the present specification, the degree of enhancement during the above-mentioned feature enhancement operation can be determined according to the aforementioned distance metric. Specifically, in some embodiments, the smaller the aforementioned distance metric is, the higher the degree of enhancement can be (a smaller distance metric means that the correlation between the current fire monitoring information and the fire risk state is higher, so the relevant feature dimensions need to be more significantly enhanced). Conversely, the larger the distance metric is, the lower the degree of enhancement (if the distance metric is larger, it indicates that the correlation between the current fire monitoring information and the fire risk state is lower, and the feature enhancement network will reduce the degree of enhancement of these feature dimensions). In some embodiments, the degree of enhancement can be expressed by the formula: degree of enhancement = α*(1-distance metric), where α is a pre-set constant used to control the maximum value of the degree of enhancement. As an example only, in some embodiments, α can be set to 0.15, that is, the weights corresponding to these feature dimensions can be increased by 15% at the maximum degree of enhancement.

[0103] In some embodiments of the present specification, the above-mentioned enhancement operation may also include further feature extraction and fusion of the first fire monitoring information feature and the deduction formula corresponding to the distance metric, so as to enhance the characterization capability of the first fire monitoring information feature.

[0104] It should be pointed out that in the embodiments of this specification, through the introduction of the feature enhancement network, the first fire monitoring information features obtained by the feature extraction network for feature extraction of fire monitoring information can be enhanced in combination with the fire deduction logic based on fire protection knowledge or historical fire protection experience, thereby improving the accuracy, reliability and comprehensibility of the fire risk status detection results obtained in the subsequent process.

[0105] Step S160: input the second fire monitoring information feature into the trained fire detection model for prediction to obtain the fire risk status of the target area.

[0106] Continue to refer to Figure 2After obtaining the second fire monitoring information feature through the above steps, the second fire monitoring information feature can be input into the trained fire detection model for prediction, so as to obtain the fire risk status of the target area. In the embodiment of the present specification, the fire risk status may include the possible types of fires and the corresponding risk levels, and the risk levels can be used to characterize the probability of fire occurrence and / or the potential degree of damage. Exemplarily, the types of fires that may occur may include but are not limited to electrical fires, oil fires, chemical fires, etc. The corresponding risk levels can be divided into three levels: low, medium, and high. A low risk level means that the probability of fire is low and the potential degree of damage is small; a medium risk level means that the probability of fire and the potential degree of damage are both at a medium level; and a high risk level indicates that the probability of fire is high and the potential degree of damage is large, and corresponding preventive and response measures need to be taken immediately.

[0107] Figure 3 A flowchart 200 of an exemplary training method of a fire detection model provided in some embodiments of the present specification. Figure 3 In the embodiment of this specification, the training process of the fire detection model may include:

[0108] Step S210, obtaining a fire monitoring sample, wherein the fire monitoring sample includes sample fire monitoring information of a sample target area and a sample fire risk status label.

[0109] In an embodiment of the present specification, the fire monitoring sample can be obtained by collecting historical data and actual cases. In an embodiment of the present specification, the fire monitoring sample may include sample fire monitoring information and sample fire risk status labels of the sample target area. Specifically, the sample fire monitoring information may include at least one of real-time monitoring data, historical fire information, and fire hazard information. For example, these sample fire monitoring information may not only cover different types of fire scenes (such as open areas, closed areas), but also include various environmental factors, such as temperature, humidity, wind speed, etc., as well as other factors that may affect the occurrence and development of fires, such as combustible types, building structures, area types, etc. The sample fire risk status label may include one or more fire types and their corresponding risk levels. In an embodiment of the present specification, the fire detection model is trained through these rich fire monitoring sample data, so that the fire detection model can learn the various characteristics and laws of fire occurrence.

[0110] Step S220: input the sample fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first sample fire monitoring information feature of the sample fire monitoring information.

[0111] In the embodiments of this specification, the feature extraction network may include CNN (Convolutional Neural Network) or RNN (Recurrent Neural Network), or other neural networks with feature extraction or encoding functions, such as Transformer encoders, etc., which are not limited in this specification.

[0112] The first sample fire monitoring information feature is similar to the aforementioned first fire monitoring information feature. The process of using the feature extraction network to perform feature extraction operations to obtain the first sample fire monitoring information feature can be referred to above and will not be repeated here.

[0113] Step S230, obtain a distance metric determined based on a distance calculation model, and input the distance metric and the first sample fire monitoring information feature into a feature enhancement network, and then use the distance metric to perform a feature enhancement operation on a feature dimension corresponding to at least one fire risk state of the first sample fire monitoring information feature to obtain a second sample fire monitoring information feature.

[0114] Similarly, in the embodiments of the present specification, the sample fire monitoring information can be adapted to at least one fire deduction logic in the deduction logic database to obtain an adaptation result of the sample fire monitoring information and at least one fire deduction logic. The fire deduction logic includes a mapping rule from the sample fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the sample fire monitoring information to the at least one fire risk state.

[0115] Furthermore, a distance metric may be determined using a distance calculation model, and then the distance metric and the first sample fire monitoring information feature are input into a feature enhancement network, and the distance metric is used to perform a feature enhancement operation on a feature dimension corresponding to at least one fire risk state of the first sample fire monitoring information feature to obtain a second sample fire monitoring information feature. The specific contents of the distance metric and feature enhancement operation can be found above and will not be elaborated here.

[0116] Step S240: input the second sample fire monitoring information feature into the initial fire detection model for prediction to obtain the predicted fire risk state corresponding to the target area.

[0117] Step S250, performing iterative optimization of model parameters according to the distance between the predicted fire risk state and the sample fire risk state label to obtain a trained fire detection model.

[0118] After obtaining the second sample fire monitoring information features through the above steps, the second sample fire monitoring information features can be input into the initial fire detection model for prediction, thereby obtaining the predicted fire risk state corresponding to the target area. The initial fire detection model can be a model built based on a machine learning algorithm, such as a support vector machine (SVM), a random forest, or a gradient boosted decision tree (GBDT). In this embodiment, the purpose of the initial fire detection model is to predict the fire risk state of the target area based on the input fire monitoring information features.

[0119] Next, the model parameters can be iteratively optimized according to the distance (i.e., difference) between the predicted fire risk state and the actual sample fire risk state label. In the embodiment of this specification, the optimization process can be implemented using a gradient descent algorithm or other optimization algorithms. By continuously adjusting the model parameters, the error between the predicted result and the actual label is minimized, thereby improving the accuracy and reliability of the model.

[0120] In the process of iterative optimization of model parameters, the cross-validation method can be used to evaluate the generalization ability of the fire detection model. Cross-validation ensures the performance of the model on unknown data by dividing the data set into multiple subsets and training and validating on different subsets. After multiple iterations of optimization, when the performance of the fire detection model reaches the predetermined threshold or no longer improves significantly, the model can be considered to have been trained. At this point, the trained fire detection model can be deployed to the fire monitoring system of the smart fire operation and maintenance platform for real-time monitoring and prediction of the fire risk status of the target area.

[0121] In practical applications, the fire detection model can also be integrated into the intelligent building management system and combined with hardware devices such as smoke detectors and temperature sensors to achieve early warning and rapid response to fires. In addition, the model can also be connected to the emergency response system of the fire department. Once a high-risk fire is detected, the system will automatically notify the firefighters to deal with it, thereby effectively reducing the losses caused by the fire.

[0122] It should be noted that in some embodiments of the present specification, the process of iteratively optimizing model parameters based on the distance between the predicted fire risk state and the actual sample fire risk state label may include optimizing the parameters of the aforementioned feature extraction network and feature enhancement network.

[0123] In some embodiments, the parameters of the feature extraction network, the feature enhancement network, and the initial fire detection model can be collaboratively optimized. Specifically, in the collaborative optimization process, the feature extraction network, the feature enhancement network, and the initial fire detection model can be regarded as an integral system, whose internal parameters are interrelated and jointly affect the final prediction results. It can be understood that the collaborative optimization method helps to improve the overall performance of the system. The specific implementation method of collaborative optimization can be regarded as a prior art and will not be discussed in detail in this specification.

[0124] In some embodiments of this specification, considering that for crowded areas such as shopping malls, factories, office buildings, and large public places, it is also very important to quickly guide people to evacuate the fire scene when a fire occurs. Based on this, in some embodiments, an escape and rescue planning system can also be configured in the aforementioned artificial intelligence-based smart fire operation and maintenance platform to quickly generate escape route and other escape guidance information. In some embodiments of this specification, the escape guidance information can be sent to an on-site indication device (such as a large screen, etc.) through the network for display.

[0125] Figure 4 This is an exemplary workflow 300 of an escape and rescue planning system of an intelligent firefighting operation and maintenance platform based on artificial intelligence as shown in some embodiments of this specification. Figure 4 In the embodiments of this specification, the escape and rescue planning system can be used to perform the following steps:

[0126] Step S310, when it is detected that the fire risk status of the target area meets the preset conditions, the building structure information and real-time monitoring data of the target area are obtained.

[0127] The building structure information may include key information such as floor layout, evacuation passage location, number and location of emergency exits, fire zoning, etc. Real-time monitoring data may include key parameters such as smoke concentration, temperature, fire source location, personnel distribution, etc. In the embodiments of this specification, real-time monitoring data collection can be performed on different locations of the target area through sensors (or other detection equipment) arranged at different locations.

[0128] Step S320: determining at least one candidate escape route based on the building structure information.

[0129] In the embodiments of this specification, at least one candidate escape route can be planned using an algorithm model based on the building structure information of the target area. Exemplarily, in some embodiments, a classic path planning algorithm such as the Dijkstra algorithm, the A* algorithm, or the genetic algorithm can be used to determine the candidate escape routes. These algorithms can plan one or more paths from the current position to the safe exit based on information such as the floor layout and the location of the evacuation passage. Among them, the current position can be the specific location where the fire occurred, or it can be the current location of the personnel (or the on-site indication device). During the planning process, the system will take into account various possible obstacles and dangerous areas to ensure the safety of the candidate escape routes.

[0130] Step S330, calculating the estimated escape time corresponding to each candidate escape route, and predicting the risk index corresponding to each candidate escape route based on the building structure information and the real-time monitoring data.

[0131] In the embodiment of the present specification, the time estimation and risk assessment of each candidate escape route can also be performed through the algorithm model. The calculation of the estimated escape time can take into account factors such as the length of the route, possible congestion, and the speed of movement of personnel. The prediction of the risk index needs to determine the level of escape risk factors that affect the escape time and / or escape difficulty based on the building structure information and real-time monitoring data to assess the degree of danger that may be encountered during the escape process. Exemplary escape risk factors may include personnel flow, channel capacity, and proximity to the fire source. For example, high smoke concentration and high temperature may mean a higher risk index, and the proximity to the fire source determines the probability of fire spread and smoke spread, and therefore directly affects the safety of the escape route. In the embodiment of the present specification, the aforementioned escape risk factors can be graded, for example, the escape risk factors are divided into three levels: low, medium, and high, and each level is assigned a corresponding value (for example, 1 for low, 2 for medium, and 3 for high).

[0132] Then, the escape risk weight corresponding to each escape risk factor is obtained, and the risk index of the route is obtained by calculating the weighted sum of the values ​​corresponding to each escape risk factor in each candidate escape route and normalizing the weighted sum. In the embodiment of this specification, the escape risk weight corresponding to each escape risk factor can be preset according to expert experience. As an example only, in some embodiments, the weight of the proximity to the fire source can be set to 0.4, the weight of the smoke concentration can be set to 0.3, the weight of the channel capacity can be set to 0.2, and the weight of the personnel flow can be set to 0.1.

[0133] Step S340, calculating the score corresponding to each candidate escape route according to the estimated escape time and the risk index, and selecting a candidate escape route with the highest score as the target escape route.

[0134] Through the above steps, the estimated escape time and risk index corresponding to each candidate escape route can be calculated. Further, the score corresponding to each candidate escape route can be calculated according to the estimated escape time and risk index, and the candidate escape route with the highest score can be selected as the target escape route.

[0135] For example, in some embodiments, the weight corresponding to the estimated escape time may be set to 0.4, the weight corresponding to the risk index may be set to 0.6, and then the score corresponding to each candidate escape route may be calculated by weighted summation.

[0136] After calculating the scores corresponding to all candidate escape routes, a candidate escape route with the highest score can be selected as the target escape route. It should be noted that in the embodiment of this specification, the candidate escape route with the highest score refers to the candidate escape route with the smallest score calculated by the above method.

[0137] In some embodiments of this specification, a user can connect to the escape and rescue planning system through a mobile terminal such as a mobile phone. In this case, the escape and rescue planning system can also be used to obtain the user's real-time location information, and then generate corresponding escape guidance information based on the real-time location information and the target escape route. In an embodiment of this specification, the escape guidance information may include at least one of an escape route map, voice guidance, and text guidance. Finally, the escape guidance information can be published to the user's mobile terminal device.

[0138] Similarly, in some embodiments, the escape and rescue planning system can also be used to obtain the location information of the on-site indication device, and then generate corresponding escape guidance information based on the location information of the on-site indication device and the target escape route, and publish the escape guidance information to the on-site indication device.

[0139] In some embodiments, the user can also connect to the escape and rescue planning system through a mobile terminal such as a mobile phone, and then send an emergency request for help to the escape and rescue planning system. When the escape and rescue planning system receives the emergency request for help, it can obtain the real-time location information of the user seeking help; then, based on the real-time location information of the user seeking help and the target escape route, it generates an emergency rescue plan corresponding to the user; further, the emergency rescue plan can be sent to rescue personnel and / or rescue equipment.

[0140] Specifically, in the embodiments of this specification, the emergency rescue plan may include key information such as rescue routes, rescue equipment requirements, rescue personnel allocation, etc. After receiving the emergency rescue plan, the rescue personnel can quickly carry out rescue operations according to the instructions in the plan to ensure the efficiency and accuracy of the rescue work.

[0141] It should be pointed out that in the embodiments of this specification, by configuring the above-mentioned fire detection system and escape and rescue planning system on the smart fire operation and maintenance platform, it is possible to predict and detect fire risks. At the same time, it is possible to respond quickly when a fire occurs, and automatically trigger the calculation process of the escape and rescue planning system in combination with building structure information and real-time monitoring data to generate corresponding escape guidance information. Through this smart fire operation and maintenance platform, target escape routes and emergency rescue plans can be quickly generated when a fire occurs, providing strong support for personnel evacuation and rescue operations.

[0142] In summary, the beneficial effects that may be brought about by the embodiments of this specification include but are not limited to: (1) In the intelligent fire operation and maintenance platform based on artificial intelligence provided in some embodiments of this specification, by integrating the fire reasoning scheme based on the experience of fire experts and fire knowledge with the fire detection scheme based on the pre-trained model, the accuracy, reliability and comprehensibility of fire detection can be improved; (2) In the intelligent fire operation and maintenance platform based on artificial intelligence provided in some embodiments of this specification, by integrating multiple fire deduction logics adapted to fire monitoring information to generate a fused deduction logic, it can not only cope with complex and changeable fire scenes, but also integrate the advantages of different deduction logics, so that the fire detection system can better adapt to various fire types and environmental conditions, thereby improving the sensitivity and accuracy of fire detection; (3) In some embodiments of this specification, the intelligent fire operation and maintenance platform based on artificial intelligence provided in some embodiments of this specification can be used to improve the fire detection accuracy. In the intelligent fire operation and maintenance platform based on artificial intelligence, by introducing the feature enhancement network, the first fire monitoring information feature obtained by the feature extraction network for feature extraction of fire monitoring information can be enhanced in combination with the fire deduction logic based on fire knowledge or historical fire experience, thereby improving the accuracy and understandability of the fire risk state detection results obtained based on the enhanced second fire monitoring information feature in the subsequent process; (4) In the intelligent fire operation and maintenance platform based on artificial intelligence provided in some embodiments of this specification, by configuring the fire detection system and the escape and rescue planning system on the intelligent fire operation and maintenance platform at the same time, the fire risk prediction and detection can be realized. At the same time, it can respond quickly when a fire occurs, and automatically trigger the calculation process of the escape and rescue planning system in combination with the building structure information and real-time monitoring data to generate corresponding escape guidance information. Through this intelligent fire operation and maintenance platform, the target escape route and emergency rescue plan can be quickly generated when a fire occurs, providing strong support for personnel evacuation and rescue operations.

[0143] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or may be any other beneficial effects that may be obtained.

[0144] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0145] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0146] In addition, it will be understood by those skilled in the art that various aspects of this specification may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of this specification may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of this specification may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0147] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0148] The computer program code required for the operation of each part of this specification can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program code can be run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0149] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing processing device or mobile device.

[0150] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0151] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0152] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0153] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. An intelligent firefighting operation and maintenance platform based on artificial intelligence, characterized in that: The platform includes a fire detection system, which is used to: Acquire fire monitoring information of a target area, wherein the fire monitoring information includes at least one of real-time monitoring data, historical fire information, and fire hazard information; Adapting the fire monitoring information to at least one fire deduction logic to obtain an adaptation result of the fire monitoring information and the at least one fire deduction logic; wherein the fire deduction logic includes a mapping rule from the fire monitoring information to at least one fire risk state, and the adaptation result includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state; Inputting the fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first fire monitoring information feature of the fire monitoring information; Determine a distance metric based on the at least one deduction formula and a characteristic dimension corresponding to the at least one fire risk state using a distance calculation model; Inputting the distance metric and the first fire monitoring information feature into a feature enhancement network, and performing a feature enhancement operation on a feature dimension corresponding to the at least one fire risk state of the first fire monitoring information feature by using the distance metric to obtain a second fire monitoring information feature; Inputting the second fire monitoring information feature into a trained fire detection model for prediction to obtain a fire risk status of the target area; The fire simulation logic is obtained based on the following method: Obtaining a fire knowledge description text, wherein the fire knowledge description text is used to describe the relationship between the fire monitoring information and the fire risk status; Performing fire monitoring information identification and fire risk status identification on the fire knowledge description text to obtain key indicators in the fire monitoring information and risk status in the fire risk status corresponding to the key indicators; The key indicators and their corresponding risk states are combined with logical statements to obtain the at least one fire deduction logic.

2. The intelligent firefighting operation and maintenance platform based on artificial intelligence as claimed in claim 1, characterized in that: The fire detection system is also used for: fusing the at least one fire deduction logic to obtain a fused deduction logic, wherein the fused deduction logic includes at least one deduction formula for mapping the fire monitoring information to the at least one fire risk state; The using the distance calculation model to determine the distance metric based on the at least one deduction formula and the characteristic dimension corresponding to the at least one fire risk state includes: The distance metric is obtained by performing fuzzy operation on at least one deduction formula for mapping the fire monitoring information included in the fusion deduction logic to at least one fire risk state and a characteristic dimension corresponding to the at least one fire risk state through the distance calculation model.

3. The intelligent firefighting operation and maintenance platform based on artificial intelligence as claimed in claim 2 is characterized in that: The fusing the at least one fire deduction logic to obtain a fused deduction logic includes: Based on the adaptation result, determine N fire deduction logics adapted to the fire monitoring information; Performing fuzzy reasoning on the fire deduction logics corresponding to the same fire risk state in the N fire deduction logics to obtain a deduction formula corresponding to the same fire risk state in the fused deduction logic; According to the fire deduction logic corresponding to other fire risk states except the same fire risk state, a deduction formula corresponding to the other fire risk states in the fused deduction logic is obtained.

4. The intelligent firefighting operation and maintenance platform based on artificial intelligence as claimed in any one of claims 1 to 3, characterized in that: The fire detection model is trained based on the following method: Acquire a fire monitoring sample, wherein the fire monitoring sample includes sample fire monitoring information of a sample target area and a sample fire risk status label; Inputting the sample fire monitoring information into a feature extraction network to perform a feature extraction operation to obtain a first sample fire monitoring information feature of the sample fire monitoring information; Obtaining a distance metric determined based on the distance calculation model, and inputting the distance metric and the first sample fire monitoring information feature into a feature enhancement network, and then using the distance metric to perform a feature enhancement operation on a feature dimension corresponding to at least one fire risk state of the first sample fire monitoring information feature to obtain a second sample fire monitoring information feature; Inputting the second sample fire monitoring information feature into the initial fire detection model for prediction to obtain a predicted fire risk state corresponding to the target area; The model parameters are iteratively optimized according to the distance between the predicted fire risk state and the sample fire risk state label to obtain the trained fire detection model.

5. The artificial intelligence-based smart firefighting operation and maintenance platform according to claim 4, characterized in that: The iterative optimization of model parameters according to the distance between the predicted fire risk state and the sample fire risk state label includes: collaboratively optimizing parameters of the feature extraction network, the feature enhancement network and the initial fire detection model.

6. The artificial intelligence-based smart firefighting operation and maintenance platform according to claim 1, characterized in that: The platform also includes an escape and rescue planning system, which is used to: When it is detected that the fire risk status of the target area meets the preset conditions, obtaining the building structure information and real-time monitoring data of the target area; determining at least one candidate escape route based on the building structure information; Calculating an estimated escape time corresponding to each candidate escape route, and predicting a risk index corresponding to each candidate escape route based on the building structure information and the real-time monitoring data; The score corresponding to each candidate escape route is calculated according to the estimated escape time and the risk index, and the candidate escape route with the highest score is selected as the target escape route.

7. The artificial intelligence-based smart firefighting operation and maintenance platform according to claim 6, characterized in that: The predicting of the risk index corresponding to each candidate escape route based on the building structure information and the real-time monitoring data includes: Determining the level of escape risk factors that affect escape time and / or escape difficulty according to the building structure information and the real-time monitoring data, wherein the escape risk factors include personnel flow, passage capacity, and proximity to a fire source; Obtain the escape risk weight corresponding to each escape risk factor; Based on the level of the escape risk factor and the escape risk weight corresponding to the escape risk factor, a risk index corresponding to each candidate escape route is obtained.

8. The artificial intelligence-based smart firefighting operation and maintenance platform according to claim 7, characterized in that: The escape and rescue planning system is also used for: Get the user's real-time location information; Based on the real-time location information and the target escape route, generating corresponding escape guidance information, the escape guidance information including at least one of an escape route map, voice guidance, and text guidance; The escape guidance information is published to the user's mobile terminal device and / or the on-site indicating device.

9. The artificial intelligence-based smart firefighting operation and maintenance platform according to claim 8, characterized in that: The escape and rescue planning system is also used for: When receiving an emergency request for help, obtain the real-time location information of the user seeking help; Generate an emergency rescue plan for the user based on the real-time location information of the user seeking help and the target escape route; The emergency rescue plan is sent to rescue personnel and / or rescue equipment.

Images