A method and system for dynamic assessment of fire risk based on Internet of Things monitoring

Abstract

The present application provides a method and system for dynamic assessment of fire risk based on Internet of Things monitoring, the method obtains the current temperature and current environmental monitoring data of the target area in the target building, and generates information on whether the current fire source exists based on the current temperature and the current environmental monitoring data; obtains the current environmental information of the target area in the target building, and determines the current environmental information processing result corresponding to the current environmental information through the first relationship pre-established by the first neural network; determines the current fire risk level of the target area in the target building corresponding to the current environmental information processing result through the second relationship pre-established by the second neural network. Real-time acquisition of the current environmental information of the building through the Internet of Things provides a data basis for the subsequent dynamic assessment of fire risk, and real-time dynamic assessment of the fire risk level of the building through the neural network can effectively supervise the building.

Description

Technical Field

[0001] The present application relates to the technical field of fire risk assessment, and in particular to a method and system for dynamic assessment of fire risk based on Internet of Things monitoring. Background Art

[0002] With the increase in urban scale and the development of science and technology, the causes of building fires have gradually become more diverse, and the difficulty of building fire management has also increased. The accuracy of fire risk assessment can effectively help fire management personnel control the fire risk of buildings and propose targeted fire protection plans based on the existing situation.

[0003] In the era when software was not well developed, the assessment of fire risk in China was largely based on daily checklists for buildings, which lacked real-time data and could not effectively supervise and manage buildings scientifically, which hindered effective fire supervision of buildings. With the advancement of computer technology and software engineering, by analyzing the relationship between various entities, an entity relationship diagram of the building fire risk assessment system can be established, which can effectively establish a database to conduct effective fire risk assessment of buildings. However, the use of databases only reduces management costs and labor costs, and accurate assessment of fire risks is still a great challenge.

[0004] Foreign scholars have paid attention to the field of fire risk assessment early. In the 1970s, European scholars began to use statistical methods to study the relationship between fire frequency and building properties.

[0005] In recent years, with the improvement of mathematical theory foundation, machine learning algorithms have also made great breakthroughs, and more and more machine learning algorithms are applied to fire risk assessment of buildings. For example: applying accident tree and event tree analysis to fire risk assessment, assessing fire risk by analyzing the factors causing fire, fire fighting facilities and various fire scene characteristics; combining Bayesian network with dynamic fire assessment, combining network model with directed acyclic graph, adding prior knowledge to the network according to Bayesian thinking, which is conducive to using it to deal with uncertainty problems. However, these methods all have the following problems:

[0006] 1. The parameters of the network are difficult to adjust, which makes it difficult to achieve the optimal model;

[0007] 2. Using a lot of historical data places high demands on the performance of the model, making dynamic evaluation difficult. Summary of the invention

[0008] In view of the above problems, the present application is proposed to provide a method and system for dynamic fire risk assessment based on Internet of Things monitoring to overcome the above problems or at least partially solve the above problems, including:

[0009] A method for dynamic fire risk assessment based on Internet of Things monitoring, the method is applied to establish a corresponding relationship between relevant information of a building and a corresponding fire risk level through an artificial intelligence model; wherein the artificial intelligence model includes a first model;

[0010] The first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes a temperature of a target area in a target building, a smoke concentration, a dangerous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building;

[0011] The method comprises:

[0012] Acquire the current temperature and current environmental monitoring data of the target area in the target building, and generate information on whether a current fire source exists based on the current temperature and the current environmental monitoring data;

[0013] Acquire current environmental information of a target area in a target building, and determine a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke density, current dangerous gas density, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke density, a processing result of current dangerous gas density, and a processing result of current basic parameters of the target building;

[0014] The current fire risk level of the target area in the target building corresponding to the current environmental information processing result is determined through the second relationship pre-established by the second neural network.

[0015] Furthermore, the first model further includes a third neural network, and the corresponding relationship further includes a third relationship; the step of obtaining the current temperature and current environmental monitoring data of the target area in the target building, and generating information on whether the current fire source exists based on the current temperature and the current environmental monitoring data, includes:

[0016] Acquire the current temperature and current environmental monitoring data of the target area in the target building, and determine the current thermal map corresponding to the current temperature through the third relationship pre-established by the third neural network;

[0017] Generate information on whether a current fire source exists based on the current thermal map and the current environmental monitoring data.

[0018] Furthermore, the method further comprises:

[0019] The first relationship between the environmental information of the target area in the target building and the environmental information processing result is established through the first neural network.

[0020] Furthermore, the first neural network includes a first sub-neural network, a second sub-neural network, a third sub-neural network and a fourth sub-neural network; the first relationship includes a first sub-relationship, a second sub-relationship, a third sub-relationship and a fourth sub-relationship; the step of establishing the first relationship between the environmental information of the target area in the target building and the environmental information processing result through the first neural network includes:

[0021] Establishing a first sub-relationship between the information on whether a fire source exists and the processing result of the information on whether a fire source exists by using the first sub-neural network;

[0022] Establishing a second sub-relationship between smoke concentration and smoke concentration processing result through the second sub-neural network;

[0023] Establishing a third sub-relationship between the concentration of hazardous gas and the processing result of the concentration of hazardous gas by means of the third sub-neural network;

[0024] A fourth sub-relationship between the basic parameters of the target building and the processing results of the basic parameters of the target building is established through the fourth sub-neural network.

[0025] Furthermore, the method further comprises:

[0026] The second relationship between the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results and basic parameter processing results of the target building and the fire risk level of the target area in the target building is established through the second neural network.

[0027] Furthermore, the second neural network includes a first prediction subnetwork and a second prediction subnetwork, and the second relationship includes a fifth sub-relationship and a sixth sub-relationship; the step of establishing the second relationship between the fire source existence information processing result, the smoke concentration processing result, the dangerous gas concentration processing result, and the basic parameter processing result of the target building and the fire risk level of the target area in the target building through the second neural network includes:

[0028] A fifth sub-relationship among the fire source existence information processing result, the smoke concentration processing result, the dangerous gas concentration processing result, the basic parameter processing result of the target building and the comprehensive processing result is established through the first prediction sub-network;

[0029] A sixth sub-relationship between the comprehensive processing result and the fire risk level is established through the second prediction sub-network.

[0030] Furthermore, the second neural network further includes a third prediction sub-network, and the second relationship further includes a seventh sub-relationship; when the second neural network outputs that the fire risk level is a preset fire risk level, the third prediction sub-network is called; and the method further includes:

[0031] When the fire risk level is a dangerous level or a destructive level, the seventh sub-relationship pre-established by the third prediction sub-network is used to determine the current dangerous area coordinates corresponding to the target area whose current fire risk level is a dangerous level or a destructive level.

[0032] The present invention also provides a system for dynamic assessment of fire risk based on Internet of Things monitoring, wherein the system is used to establish a corresponding relationship between relevant information of a building and a corresponding fire risk level through an artificial intelligence model; wherein the artificial intelligence model includes a first model;

[0033] The first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes a temperature of a target area in a target building, a smoke concentration, a dangerous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building;

[0034] The system comprises:

[0035] A current fire source existence information generation module is used to obtain the current temperature and current environmental monitoring data of the target area in the target building, and generate the current fire source existence information according to the current temperature and the current environmental monitoring data;

[0036] A current processing result determination module is used to obtain current environmental information of a target area in a target building, and determine a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke density, current dangerous gas density, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke density, a processing result of current dangerous gas density, and a processing result of current basic parameters of the target building;

[0037] The current fire risk level determination module is used to determine the current fire risk level of the target area in the target building corresponding to the current environmental information processing result through the second relationship pre-established by the second neural network.

[0038] The present invention also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the steps of the method for dynamic assessment of fire risk based on Internet of Things monitoring as described above are implemented.

[0039] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for dynamic assessment of fire risk based on Internet of Things monitoring as described above are implemented.

[0040] This application has the following advantages:

[0041] In an embodiment of the present application, a corresponding relationship between relevant information of a building and a corresponding fire risk level is established through an artificial intelligence model; wherein the artificial intelligence model includes a first model; the first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes a temperature, smoke concentration, hazardous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building in a target area of ​​the target building; the current temperature and current environmental monitoring data of the target area in the target building are obtained, and information on whether a current fire source exists is generated based on the current temperature and the current environmental monitoring data; the current environmental information of the target area in the target building is obtained, and a current environmental information processing result corresponding to the current environmental information is determined through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke concentration, current hazardous gas concentration, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke concentration, a processing result of current hazardous gas concentration, and a processing result of current basic parameters of the target building; the current fire risk level of the target area in the target building corresponding to the current environmental information processing result is determined through a second relationship pre-established by the second neural network. The Internet of Things can be used to obtain the current environmental information of the building in real time, providing a data basis for the subsequent dynamic assessment of fire risks. The fire risk level of the building can be dynamically assessed in real time through neural networks, which can effectively supervise the building. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0043] Figure 1This is a flowchart of a method for dynamic fire risk assessment based on Internet of Things monitoring provided by an embodiment of the present application;

[0044] Figure 2a It is a structural diagram of a lower computer of a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0045] Figure 2b It is a connection topology diagram between a lower computer and an upper computer of a method for dynamic fire risk assessment based on Internet of Things monitoring provided by an embodiment of the present application;

[0046] Figure 3 It is a structural schematic diagram of a multimodal attention neural network of a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0047] Figure 4 It is a structural schematic diagram of a fire source existence information generation module of a method for dynamic fire risk assessment based on Internet of Things monitoring provided by an embodiment of the present application;

[0048] Figure 5 It is a structural schematic diagram of an enhanced convolutional neural network of a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0049] Figure 6 It is a structural schematic diagram of a comprehensive processing module of a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0050] Figure 7a It is a visualization diagram of a combination of a thermal map of a safe area and environmental monitoring data in a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0051] Figure 7b This is a schematic diagram of predicting the fire risk level of a safe area according to a method for dynamic fire risk assessment based on Internet of Things monitoring provided by an embodiment of the present application;

[0052] Figure 8a It is a visualization diagram of a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application, which combines a heat map of a dangerous area with environmental monitoring data;

[0053] Figure 8b This is a schematic diagram of predicting the fire risk level of a dangerous area according to a method for dynamic fire risk assessment based on Internet of Things monitoring provided by an embodiment of the present application;

[0054] Fig. 9This is a structural block diagram of a system for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0055] Fig.10 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the objects, features and advantages of the present application more obvious and understandable, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0057] It should be noted that, with reference to FIG2 , in any embodiment of the present invention, a host computer and a slave computer are included. The function of the slave computer is to collect relevant data of the building environment in real time, and the function of the host computer is to analyze the real-time data of the building environment, and to evaluate the fire risk level of different areas of the building and determine the dangerous areas of the building. This application can use an industrial computer with an operating system of Ubuntu as the hardware part of the host computer, and the software part of the host computer is composed of an artificial intelligence model, and a multi-modal attention neural network (Multi-Modality and Attention Neural Network, MMANN) is selected, and its network structure refers to Figure 3 , which is composed of a temperature processing module, an image processing module, a smoke concentration processing module, a gas concentration processing module, a building attribute analysis module and a comprehensive processing module, and is used to jointly analyze data; the lower computer is composed of a single-chip microcomputer, a WIFI module and 4 sensors, wherein the 4 sensors are: temperature and humidity sensor, smoke sensor, gas detection sensor and camera module.

[0058] Figure 2a This is a schematic diagram of the hardware connection structure of the lower computer. The four sensors are connected to one end of the microcontroller respectively, and the other end of the microcontroller is connected to the WIFI transmission module; Figure 2b This is the connection topology diagram between the lower computer and the upper computer. First, the four sensors collect environmental information in real time and send the data to the single-chip microcomputer. After the single-chip microcomputer sorts out various types of data, it sends it to the upper computer through the WIFI module at regular intervals. Due to the different encoding methods of the collected environmental information, the upper computer needs to perform different preprocessing on various types of information, and then input them into the multimodal attention neural network to predict the fire risk level and the coordinates of the dangerous area.

[0059] Reference Figure 1 , showing a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application;

[0060] The method comprises:

[0061] S110, obtaining the current temperature and current environmental monitoring data of the target area in the target building, and generating information on whether a current fire source exists according to the current temperature and the current environmental monitoring data;

[0062] S120, obtaining current environmental information of a target area in a target building, and determining a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke density, current dangerous gas density, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke density, a processing result of current dangerous gas density, and a processing result of current basic parameters of the target building;

[0063] S130. Determine the current fire risk level of the target area in the target building corresponding to the current environmental information processing result through the second relationship pre-established by the second neural network.

[0064] In an embodiment of the present application, a corresponding relationship between relevant information of a building and a corresponding fire risk level is established through an artificial intelligence model; wherein the artificial intelligence model includes a first model; the first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes a temperature, smoke concentration, hazardous gas concentration, environmental monitoring data of a target area in a target building, and basic parameters of the corresponding target building; the method includes: obtaining a current temperature and current environmental monitoring data of a target area in a target building, and generating information on whether a current fire source exists based on the current temperature and the current environmental monitoring data; obtaining current environmental information of a target area in a target building, and determining a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke concentration, current hazardous gas concentration, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke concentration, a processing result of current hazardous gas concentration, and a processing result of current basic parameters of the target building; determining a current fire risk level of a target area in a target building corresponding to the current environmental information processing result through a second relationship pre-established by a second neural network. The Internet of Things can be used to obtain the current environmental information of the building in real time, providing a data basis for the subsequent dynamic assessment of fire risks. The fire risk level of the building can be dynamically assessed in real time through neural networks, which can effectively supervise the building.

[0065] Next, a method for dynamic fire risk assessment based on Internet of Things monitoring in this exemplary embodiment will be further described.

[0066] As described in step S110, the current temperature and current environmental monitoring data of the target area in the target building are obtained, and information on whether a current fire source exists is generated based on the current temperature and the current environmental monitoring data.

[0067] In one embodiment of the present invention, the specific process of "obtaining the current temperature and current environmental monitoring data of the target area in the target building, and generating information on whether a current fire source exists based on the current temperature and the current environmental monitoring data" in step S110 can be further explained in combination with the following description.

[0068] In one embodiment of the present invention, the first model further includes a third neural network, and the corresponding relationship further includes a third relationship.

[0069] As described in the following steps, the current temperature and current environmental monitoring data of the target area in the target building are obtained, and the current thermal map corresponding to the current temperature is determined through the third relationship pre-established by the third neural network.

[0070] As an example, the temperature inside a building is an important indicator for judging the level of fire risk. An artificial neural network algorithm is used to analyze the thermal map patterns corresponding to the temperatures in different areas of the building. The mapping pattern between the temperature of the target area and the thermal map is found through the self-learning and adaptive characteristics of the artificial neural network.

[0071] The present invention proposes a temperature processing module, which is mainly composed of a third neural network. The third relationship between temperature and thermal map is established through the third neural network, so that the corresponding thermal map can be obtained after the temperature data is input into the module. The third neural network can use a fully connected neural network (Fully Connected Neural Network, FCNN), and its input size is 32×1, 32 represents the value of 32 temperature sensors. The fully connected layer is composed of 5 hidden layers, the number of neurons in the hidden layer is 2048, and the activation function of the hidden layer adopts ReLU. Finally, the neurons in the output layer are converted into a 224×224 two-dimensional matrix as a thermal map. The formula is as follows:

[0072] Map=Reshape(f(a(f(a(f(a(a(f(a(f(a(x)))))))))))) (1)

[0073] a(x)=Wx+b (2)

[0074] Among them, formula (1) indicates that the temperature is processed by a fully connected neural network to generate a heat map; formula (2), that is, a(x) represents the linear calculation of the hidden layer, W is the weight matrix, linear transformation, b is the bias; f(*) is the activation function ReLU; Reshape(*) function is to convert one-dimensional data into a matrix of size; Map represents the generated heat map.

[0075] In one embodiment of the present application, the input data of the third neural network is the temperature of the target area in the target building, and the temperature is input into the fully connected neural network model in a time sequence, so that the third neural network can theoretically obtain information from past temperature data and adjust the thermal map output according to the latest recorded temperature data. The output of the third neural network is the thermal map corresponding to the current temperature.

[0076] When training the third neural network, the corresponding thermal maps of the temperature conditions in different areas of a large number of buildings (including but not limited to one or more of the following: -10℃-10℃, 10℃-30℃, 30℃-40℃, 40℃-200℃, etc.) are collected and summarized, and the corresponding temperatures and thermal maps of several different areas are selected as sample data, and the neural network is trained in combination with the pre-stored expert experience information. For example: collect temperature data and corresponding thermal maps of buildings in different locations; and collect temperature data and corresponding thermal maps of the same building in different time periods; and collect temperature data and corresponding thermal maps of the same building under different usage conditions. Therefore, collecting data through multiple channels is conducive to increasing the amount of data, improving the learning ability of the artificial neural network, and then improving the accuracy and reliability of the determined corresponding relationship.

[0077] By using the artificial neural network algorithm and adjusting the network structure and the weights between network nodes, the neural network is made to fit the relationship between the temperature and thermal map corresponding to different areas. Ultimately, the neural network can accurately fit the corresponding relationship between the temperature and thermal map corresponding to different areas of different buildings.

[0078] The temperature processing module uses the third neural network to identify the current temperature data of the target area of ​​the building and output the thermal map. It should be noted that the temperature data is collected by the temperature and humidity sensor in the target area of ​​the building, and then uploaded to the host computer through the WIFI module as the input of the first sub-neural network, wherein the WIFI module is preferably an ESP8266 module.

[0079] In a specific implementation, after the third neural network is trained, it can be used to generate the corresponding current thermal map for the current temperature data of different areas in the building. 232 temperature and humidity sensors are installed to collect the current temperature in this area, so as to obtain a more realistic current temperature distribution in the target area. The DHT11 module is used as the hardware of the temperature and humidity sensor. After each temperature and humidity sensor uploads the collected current temperature data to the microcontroller, the microcontroller sends the data to the host computer through the WIFI module every 5 minutes. The time interval difference between the microcontrollers in sending data is seconds, n is the number of lower computers in the building. By regularly sending dynamic data of the building environment, it provides a data basis for the subsequent upper computer to dynamically evaluate the fire risk.

[0080] As described in the following steps, information on whether a current fire source exists is generated based on the current thermal map and the current environmental monitoring data.

[0081] It should be noted that the current environmental monitoring data is collected by a camera module in the building, and then uploaded to the host computer through the WIFI module.

[0082] As an example, see Figure 4 The present invention proposes a fire source existence information generation module, which crops the current environmental monitoring data collected by the camera module, and then splices it with the thermal map in the channel dimension to obtain the current fire source existence information. Through the current fire source existence information, detailed current temperature information and current open flame usage in the image area can be obtained. The fusion of data promotes the improvement of the accuracy of the network model.

[0083] In one specific implementation, every 50m in the building 2 A camera is installed to collect the current environmental monitoring data in this area, so as to better detect the use of open flames in the building. The hardware model of the camera module is OV7670 module. After each camera uploads the collected current environmental monitoring data to the microcontroller, the microcontroller sends the data to the host computer through the WIFI module every 5 minutes. The time interval difference between the microcontrollers sending data is seconds, n is the number of lower computers in the building. By regularly sending dynamic data of the building environment, it provides a data basis for the subsequent upper computer to dynamically evaluate the fire risk.

[0084] As described in step S120, the current environmental information of the target area in the target building is obtained, and the current environmental information processing result corresponding to the current environmental information is determined through the first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke concentration, current hazardous gas concentration and current basic parameters of the target building; the current environmental information processing result includes information processing result on whether a current fire source exists, current smoke concentration processing result, current hazardous gas concentration processing result and current basic parameter processing result of the target building.

[0085] In one embodiment of the present invention, the specific process of "obtaining current environmental information of a target area in a target building, and determining a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke concentration, current hazardous gas concentration, and current basic parameters of the target building; and the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke concentration, a processing result of current hazardous gas concentration, and a processing result of current basic parameters of the target building" in step S120 can be combined with the following description.

[0086] In one embodiment of the present invention, the first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes the temperature of a target area in a target building, smoke concentration, hazardous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building.

[0087] As described in the following steps, the first relationship between the environmental information of the target area in the target building and the environmental information processing result is established through the first neural network.

[0088] In one embodiment of the present invention, the specific process of the step of "establishing the first relationship between the environmental information of the target area in the target building and the environmental information processing result through the first neural network" can be further explained in combination with the following description.

[0089] In one embodiment of the present invention, the first neural network includes a first sub-neural network, a second sub-neural network, a third sub-neural network and a fourth sub-neural network; the first relationship includes a first sub-relationship, a second sub-relationship, a third sub-relationship and a fourth sub-relationship.

[0090] As described in the following steps, a first sub-relationship between the information on whether a fire source exists and the processing result of the information on whether a fire source exists is established through the first sub-neural network.

[0091] As an example, the present invention proposes an image processing module, which is mainly composed of a first sub-neural network, and is used to pre-process information on whether a fire source exists, and use an artificial neural network algorithm to analyze the rules of fire source existence information processing results corresponding to the fire source existence information in different areas of a building.

[0092] In a specific implementation, referring to Figure 5 The first sub-neural network can use the enhanced convolutional neural network (Improved Convolutional Neural Network). The network architecture design of the enhanced convolutional neural network draws on the idea of ​​Encoder-Decoder (encoding-decoding), which can effectively reduce the amount of calculation without losing performance. The structure of the enhanced convolutional neural network is shown below Figure 5 As shown in the figure, it is divided into a main network and a branch network. The main network has 10 convolutional layers, 3 pooling layers, and 3 upsampling layers; in the branch network, each branch has only one convolutional layer. The input size of the enhanced convolutional neural network is 4×224×224, where 4 represents the sum of the number of channels of the input image data and the heat map, and 224×224 represents the resolution of the image. The convolution kernels of all convolutional layers are 3×3, the pooling layer uses average pooling, and the resolution is reduced to half of the original. The upsampling layer uses transposed convolution, and the resolution is doubled. The output results of the 2nd, 3rd, 4th, and 5th convolutional layers on the main network are respectively input to the branch convolutional layers for feature enhancement, and then fused with the features of the 6th, 7th, 8th, and 9th convolutional layers. Finally, the extracted features are pooled and flattened to obtain the final result, and the final output size is 588×1. The formula is as follows:

[0093] F1=g(I input ) (3)

[0094] F2=z(g(F1)) (4)

[0095] F′2=z(g(F2)) (5)

[0096] F3=z(g(P(F2))) (6)

[0097] F′3=z(g(F3)) (7)

[0098] F4=z(g(P(F3))) (8)

[0099] F′4=z(g(F4)) (9)

[0100] F5=z(g(P(F4))) (10)

[0101] F′5=z(g(F5)) (11)

[0102] F6=z(g(F5)) (12)

[0103] V output =R(g(z(g(Up(z(g(Up(z(g(Up(F6+F′5)+F′4))))+F′3)))+F′2)))) (13)

[0104] Where, formula (3) represents the input I input After the first convolution layer, feature F1 is calculated. Formula (4) indicates that feature F1 is calculated after the second convolution layer to obtain feature F2. Formula (5) indicates that feature F2 is calculated after the branch convolution layer to obtain F′2. Formula (6) indicates that feature F2 is calculated after the pooling layer and the third convolution layer to obtain feature F3. Formula (7) indicates that feature F3 is calculated after the branch convolution layer to obtain feature F′3. Formula (8) indicates that feature F3 is calculated after the pooling layer and the fourth convolution layer to obtain feature F4. Formula (9) indicates that feature F4 is calculated after the branch convolution layer. The feature F′4 is obtained. Formula (10) indicates that the feature F4 is calculated through the pooling layer and the 5th convolution layer to obtain the feature F5. Formula (11) indicates that the feature F5 is calculated through the branch convolution layer to obtain the feature F′5. Formula (12) indicates that the feature F5 is calculated through the 6th convolution layer to obtain the feature F6. Formula (13) indicates that the features F6, F′2, F′3, F′4 and F′5 are calculated through the upsampling layer and the 6th, 7th, 8th and 9th convolution layers and downsampling layers, and the two-dimensional features are flattened into a one-dimensional vector to obtain the output vector V. output ; g represents the convolution operation of the convolution layer, z represents the activation function ReLU, P represents the average pooling operation of the pooling layer, Up represents the transposed convolution operation of the upsampling layer, and R represents the average pooling operation and flattens the two-dimensional matrix into a one-dimensional vector operation.

[0105] As an example, when training the first sub-neural network, a large amount of fire source existence information in different areas of the target building is used as the input of the training set, and the fire source existence information processing result corresponding to each data of the fire source existence information in the training set is used as the training output corresponding to each corresponding data point, so as to construct the structure of the first sub-neural network and conduct supervised training on it. After the first sub-neural network is trained, it can be used to generate corresponding fire source existence information processing results for the fire source existence information in different areas of the building.

[0106] As described in the following steps, a second sub-relationship between smoke concentration and smoke concentration processing result is established through the second sub-neural network.

[0107] As an example, when training the second sub-neural network, the smoke concentrations in different areas of the building and the smoke concentration processing results are collected as sample data to train the second sub-neural network. For example: sample collection: smoke concentrations in different buildings at the same time and the corresponding smoke concentration processing results; and smoke concentration processing results corresponding to smoke concentrations in the same building at different times, etc. After the second sub-neural network is trained, it can be used to generate corresponding smoke concentration processing results for smoke concentrations in different areas of the building.

[0108] In one specific implementation, every 50m in the building 2 32 smoke density sensors are installed to collect smoke density in this area, so as to obtain a distribution of smoke density closer to the real one. The hardware selected is MQ-2 sensor module. After each smoke density sensor uploads the collected smoke density to the microcontroller, the microcontroller sends the data to the host computer through the WIFI module every 5 minutes. The time interval difference between the microcontrollers sending data is seconds, n is the number of lower computers in the building. By regularly sending dynamic data of the building environment, it provides a data basis for the subsequent upper computer to dynamically evaluate the fire risk.

[0109] As described in the following steps, a third sub-relationship between the hazardous gas concentration and the hazardous gas concentration processing result is established through the third sub-neural network;

[0110] As an example, when training the third sub-neural network, the concentration of dangerous gases in different areas of the building and the results of dangerous gas concentration processing are collected as sample data to train the third sub-neural network. For example: Sample collection: before a building fire occurs, the concentration of flammable and explosive gases (such as ethanol, methane and liquefied gas, etc.) is collected; after a building fire occurs, the concentration of special gases (such as carbon monoxide and carbon dioxide, etc.) is collected. After the third sub-neural network is trained, it can be used to generate corresponding dangerous gas concentration processing results for the concentration of dangerous gases in different areas of the building.

[0111] In a specific implementation, due to the diversity of flammable and explosive gases and dangerous gases in the natural world, a gas filter is installed every 50m in a building. 232 gas detection sensors of three different types are installed in each area to collect the concentrations of methane, ethanol, liquefied gas, carbon monoxide and carbon dioxide at different locations in the building, which can more accurately analyze the hidden dangers of different dangerous gases to the building. Among them, the hardware selection of the three gas sensors is MQ-3 module, MQ-5 module and MQ-7 module. After each gas sensor uploads the collected temperature data to the microcontroller, the microcontroller sends the data to the host computer through the WIFI module every 5 minutes. The time interval difference between the microcontrollers sending data is seconds, n is the number of lower computers in the building. By regularly sending dynamic data of the building environment, it provides a data basis for the subsequent upper computer to dynamically evaluate the fire risk.

[0112] As described in the following steps, a fourth sub-relationship between the basic parameters of the target building and the processing results of the basic parameters of the target building is established through the fourth sub-neural network.

[0113] As an example, when training the fourth sub-neural network, the basic parameters of the target building and the basic parameter processing results of the target building are collected as sample data to train the fourth sub-neural network. The basic parameters of the target building include but are not limited to the type of building, fire protection facilities, building area and building service life. After the fourth sub-neural network is trained, it can be used to generate corresponding dangerous gas concentration processing results for the dangerous gas concentrations in different areas of the building.

[0114] In a specific implementation, the present invention proposes a smoke concentration processing module, a hazardous gas concentration processing module and a building attribute analysis module, the above modules are respectively composed of a second sub-neural network, a third sub-neural network and a fourth sub-neural network, the above second sub-neural network, the third sub-neural network and the fourth sub-neural network can all use a fully connected neural network with the same network structure, which is used to analyze the smoke concentration data, the hazardous gas concentration data and the basic parameters of the target building respectively. The input size of the above fully connected neural network is 32×1, there are 5 hidden layers, the number of neurons in each hidden layer is 6144, and the activation function of the hidden layer neurons adopts the Sigmoid function; the number of neurons in the output layer is 588, and the formula is as follows:

[0115] Out=f(a(f(a(f(a(f(a(f(a(x))))))))))) (14)

[0116] a(x)=Wx+b (15)

[0117] Among them, formula (14) represents the smoke concentration data, hazardous gas concentration data or basic parameters of the target building analyzed by the fully connected neural network; formula (15) that is, a(x) represents the linear calculation of the hidden layer; W is the weight matrix; b is the bias; f(*) is the activation function Sigmoid function of the hidden layer neuron, and Out is the output of the fully connected neural network.

[0118] As described in step S130, the current fire risk level of the target area in the target building corresponding to the current environmental information processing result is determined through the second relationship pre-established by the second neural network.

[0119] In one embodiment of the present invention, the specific process of "determining the current fire risk level of the target area in the target building corresponding to the current environmental information processing result through the second relationship pre-established by the second neural network" in step S130 can be further explained in combination with the following description.

[0120] In one embodiment of the invention, the first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes the temperature of a target area in a target building, smoke concentration, hazardous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building.

[0121] As described in the following steps, the second relationship between the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results and basic parameter processing results of the target building and the fire risk level of the target area in the target building is established through the second neural network.

[0122] In one embodiment of the present invention, the specific process of the step of "establishing the second relationship between the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results and basic parameter processing results of the target building and the fire risk level of the target area in the target building through the second neural network" can be combined with the following description.

[0123] In one embodiment of the present invention, the second neural network includes a first prediction sub-network and a second prediction sub-network, and the second relationship includes a fifth sub-relationship and a sixth sub-relationship.

[0124] As described in the following steps, a fifth sub-relationship among the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results, and basic parameter processing results of the target building and the comprehensive processing results is established through the first prediction sub-network.

[0125] In a specific implementation, referring to Figure 6The present invention proposes a comprehensive processing module, which is mainly composed of a second neural network, which includes a first prediction sub-network. The first prediction sub-network can use an attention mechanism network, which is composed of a self-attention function, so as to obtain a comprehensive processing result after inputting the information processing result of whether there is a fire source, the smoke concentration processing result, the dangerous gas concentration processing result and the basic parameter processing result of the target building into the neural network. The formula is as follows:

[0126] Out attention =f self-attention (W q x,W k x,W v x) (16)

[0127]

[0128] Wherein, formula (16) represents the self-attention calculation of input x, f self-attention (*) represents the self-attention function, W q , W k , W v They represent three different matrices, which are input into the self-attention function after three different linear transformations; Formula (17) represents f self-attention (*) represents the calculation process, softmax(*) represents the softmax function, and d represents the normalization factor.

[0129] By taking the sum of the processing results of whether there is a fire source, smoke concentration, hazardous gas concentration and the basic parameter processing results of the target building as the input of the first prediction sub-network, a comprehensive processing result is generated. Through the attention mechanism network, more important neurons can be found, and then greater weight can be given to the input data, which can greatly improve the performance of the network model.

[0130] As an example, when training the first prediction sub-network, the information processing results of whether there is a fire source in different areas of the building, the processing results of smoke concentration, the processing results of dangerous gas concentration, and the basic parameter processing results of the target building are collected as sample data to train the first prediction sub-network. After the first prediction sub-network is trained, it can be used to generate corresponding comprehensive processing results for the information processing results of whether there is a fire source in different areas of the building, the processing results of smoke concentration, the processing results of dangerous gas concentration, and the basic parameter processing results of the target building.

[0131] As described in the following steps, a sixth sub-relationship between the comprehensive processing result and the fire risk level is established through the second prediction sub-network.

[0132] In a specific implementation, the second neural network also includes a second prediction subnetwork, which is a fire risk level prediction network, and a fully connected neural network can be selected. The input size of the network is 588×1, there are 5 hidden layers, the number of neurons in the hidden layer is 6,144, and the activation function of the hidden layer neurons is a Sigmoid function; the number of neurons in the output layer is 4. The formula is as follows:

[0133] Out 等级 =f(a(f(a(f(a(f(a(f(a(Out attention ))))))))))) (18)

[0134] a(x)=Wx+b (19)

[0135] Among them, Out attention represents the output of the attention mechanism network; Formula (18) represents the prediction of the fire risk level by the output of the attention mechanism network through the fully connected neural network; Out 等级 Represents the predicted fire risk level. Formula (19) means that a(x) represents the linear calculation of a hidden layer, W is the weight matrix, b is the bias; f(*) is the activation function Sigmoid function of the hidden layer neuron.

[0136] As an example, when training the second prediction sub-network, the comprehensive processing results of different areas in the building are collected as sample data to train the second prediction sub-network. After the second prediction sub-network is trained, it can be used to generate corresponding fire risk levels based on the comprehensive processing results of different areas in the building.

[0137] In a specific implementation, the fire risk levels predicted for different areas in a building can be divided into four levels, namely: Level 1, safe; Level 2, critical; Level 3, dangerous; Level 4, destructive; the corresponding preventive measures are: Level 1, normal; Level 2, need to ensure the implementation of measures; Level 3, take emergency measures; Level 4, rectification.

[0138] In one embodiment of the present invention, the specific process of "establishing the second relationship between the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results and basic parameter processing results of the target building and the fire risk level of the target area in the target building through the second neural network" can be combined with the following description.

[0139] In one embodiment of the present invention, the second neural network includes a third prediction sub-network, and the second relationship includes a seventh sub-relationship; when the second neural network outputs that the fire risk level is a preset fire risk level, the third prediction sub-network is called.

[0140] As described in the following steps, when the fire risk level is a dangerous level or a destructive level, the current dangerous area coordinates corresponding to the target area where the current fire risk level is a dangerous level or a destructive level are determined through the seventh sub-relationship pre-established by the third prediction subnetwork.

[0141] As an example, the second neural network also includes a third prediction subnetwork. The function of the third prediction subnetwork is to predict the coordinates of the current dangerous area when the second prediction subnetwork predicts that there is a dangerous or damaged area in the target building. The third prediction subnetwork is a dangerous area coordinate prediction network, which can also use a fully connected neural network with the same network structure as the second prediction subnetwork, and its formula is as follows:

[0142] Out 坐标 =f(a(f(a(f(a(f(a(f(a(Out attention ))))))))))) (20)

[0143] a(x)=Wx+b (21)

[0144] Among them, Out attention represents the output of the attention mechanism network; Formula (20) represents the prediction of the dangerous area by the output of the attention mechanism network after the fully connected neural network, Out 坐标 Represents the four coordinates of the predicted dangerous area; Formula (21), that is, a(x) represents the linear calculation of a hidden layer, W is the weight matrix, and b is the bias; f(x) is the activation function Sigmoid function of the hidden layer neuron.

[0145] As an example, when training the third prediction sub-network, the fire risk levels of different areas in the target building are collected as sample data to train the third prediction sub-network. After the third prediction sub-network is trained, it can be used to generate corresponding dangerous area coordinates for the fire risk levels of different areas in the target building, and mark the dangerous areas with rectangles.

[0146] In a specific embodiment, taking two areas of a supervised building as an example, Figure 7a and Figure 7b The same area of ​​the building under supervision, where Figure 7a It is a visualization diagram of the combination of the thermal map output by the temperature processing module in the host computer and the environmental monitoring data. Figure 7bRepresents the output of the multimodal neural network. At this time, the current area is safe and the fire risk level is 1.

[0147] Figure 8a and Figure 8b Another area of ​​the building under supervision, where Figure 8a This is a visualization diagram of the combination of the thermal map output by the temperature processing module in the lower computer and the environmental monitoring data. Figure 8b Represents the output of the multimodal attention neural network. At this time, the current area is dangerous and the fire risk level is 4. The method provided in this embodiment identifies the dangerous area in the figure and marks the coordinates predicted by the multimodal attention neural network with a rectangle.

[0148] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0149] Reference Fig. 9 , showing a system for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present application; the system is used to establish a corresponding relationship between relevant information of a building and a corresponding fire risk level through an artificial intelligence model; wherein the artificial intelligence model includes a first model;

[0150] The first model includes a first neural network and a second neural network, the corresponding relationship includes a first relationship and a second relationship, and the relevant information includes a temperature of a target area in a target building, a smoke concentration, a dangerous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building;

[0151] Specifically include:

[0152] A current fire source existence information generating module 910 is used to obtain the current temperature and current environmental monitoring data of the target area in the target building, and generate the current fire source existence information according to the current temperature and the current environmental monitoring data;

[0153] The current processing result determination module 920 is used to obtain the current environmental information of the target area in the target building, and determine the current environmental information processing result corresponding to the current environmental information through the first relationship pre-established by the first neural network; wherein the current environmental information includes the information whether there is a current fire source, the current smoke concentration, the current dangerous gas concentration and the current basic parameters of the target building; the current environmental information processing result includes the information processing result of whether there is a current fire source, the current smoke concentration processing result, the current dangerous gas concentration processing result and the current basic parameter processing result of the target building;

[0154] The current fire risk level determination module 930 is used to determine the current fire risk level of the target area in the target building corresponding to the current environmental information processing result through the second relationship pre-established by the second neural network.

[0155] In one embodiment of the present invention, the current fire source existence information generating module 910 includes:

[0156] A current temperature processing module, used for obtaining the current temperature of the target area in the target building, and determining a current thermal map corresponding to the current temperature through a third relationship pre-established by the third neural network;

[0157] The current fire source existence information generation module is used to generate the current fire source existence information based on the current thermal map and the current environmental monitoring data.

[0158] In one embodiment of the present invention, the current processing result determination module 920 includes:

[0159] The first relationship determination module is used to establish the first relationship between the environmental information of the target area in the target building and the environmental information processing result through the first neural network.

[0160] In one embodiment of the present invention, the first relationship determination module includes:

[0161] A first sub-relationship determination submodule, used for establishing a first sub-relationship between the information on whether a fire source exists and the processing result of the information on whether a fire source exists through the first sub-neural network;

[0162] A second sub-relationship determination submodule, used for establishing a second sub-relationship between smoke concentration and smoke concentration processing result through the second sub-neural network;

[0163] A third sub-relationship determination submodule, used for establishing a third sub-relationship between the hazardous gas concentration and the hazardous gas concentration processing result through the third sub-neural network;

[0164] The fourth sub-relationship determination submodule is used to establish a fourth sub-relationship between the basic parameters of the target building and the processing results of the basic parameters of the target building through the fourth sub-neural network.

[0165] In one embodiment of the present invention, the current fire risk level determination module 930 includes:

[0166] The second relationship determination module is used to establish the second relationship between the fire source existence information processing results, smoke concentration processing results, hazardous gas concentration processing results and basic parameter processing results of the target building and the fire risk level of the target area in the target building through the second neural network.

[0167] In one embodiment of the present invention, the second relationship determination module includes:

[0168] a fifth sub-relationship determination submodule, configured to establish, through the first prediction sub-network, a fifth sub-relationship between the fire source existence information processing result, the smoke concentration processing result, the dangerous gas concentration processing result, the basic parameter processing result of the target building, and the comprehensive processing result;

[0169] The sixth sub-relationship determination submodule is used to establish a sixth sub-relationship between the comprehensive processing result and the fire risk level through the second prediction sub-network.

[0170] In one embodiment of the present invention, the current fire risk level determination module 930 further includes:

[0171] The seventh sub-relationship determination module is used to determine the current dangerous area coordinates corresponding to the target area whose current fire risk level is dangerous or destructive through the seventh sub-relationship pre-established by the third prediction sub-network when the fire risk level is dangerous or destructive.

[0172] Reference Fig.10 , showing a computer device of a method for dynamic fire risk assessment based on Internet of Things monitoring of the present application, which may specifically include the following:

[0173] The computer device 12 is in the form of a general-purpose computing device, and the components of the computer device 12 may include but are not limited to: one or more processors or processing units 16, a memory 28, and a bus 18 connecting different system components (including the memory 28 and the processing unit 16).

[0174] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, an Audio Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0175] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0176] Memory 28 may include computer system readable media in the form of volatile memory, such as random access memory 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Fig.10 Not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, which are configured to perform the functions of the various embodiments of the present application.

[0177] A program / utility 40 having a set (at least one) of program modules 42 may be stored in, for example, a memory, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules 42, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0178] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, cameras, etc.), one or more devices that enable an operator to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed through the I / O interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., local area networks (LANs)), wide area networks (WANs), and / or public networks (e.g., the Internet) through a network adapter 20. Fig.10 As shown, the network adapter 20 communicates with other modules of the computer device 12 via the bus 18. It should be understood that although Fig.10 Not shown, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units 16, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 34, etc.

[0179] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, for example, implementing a method for dynamic assessment of fire risk based on Internet of Things monitoring provided by an embodiment of the present invention.

[0180] In an embodiment of the present invention, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for dynamic assessment of fire risk based on Internet of Things monitoring as provided in all embodiments of the present application.

[0181] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0182] Computer-readable signal media may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0183] The computer program code for performing the operation of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as an independent software package, partially on the operator's computer, partially on the remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the operator's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other.

[0184] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0185] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0186] The above is a detailed introduction to a method and system for dynamic assessment of fire risks based on Internet of Things monitoring provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for dynamic assessment of fire risk based on Internet of Things monitoring, characterized in that: The method is applied to establish a corresponding relationship between relevant information of a building and a corresponding fire risk level through an artificial intelligence model; wherein the artificial intelligence model includes a first model; the first model includes a first neural network, a second neural network, and a third neural network, the corresponding relationship includes a first relationship, a second relationship, and a third relationship, and the relevant information includes the temperature of a target area in a target building, smoke concentration, hazardous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building; wherein the first neural network includes a first sub-neural network, a second sub-neural network, a third sub-neural network, and a fourth sub-neural network; the first relationship includes a first sub-relationship, a second sub-relationship, a third sub-relationship, and a fourth sub-relationship; wherein the second neural network includes a first prediction sub-network and a second prediction sub-network, and the second relationship includes a fifth sub-relationship and a sixth sub-relationship; The method comprises: Acquire the current temperature and current environmental monitoring data of the target area in the target building, and determine the current thermal map corresponding to the current temperature through the third relationship pre-established by the third neural network; generate information on whether a current fire source exists based on the current thermal map and the current environmental monitoring data; Acquire current environmental information of a target area in a target building, and determine a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke density, current dangerous gas density, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke density, a processing result of current dangerous gas density, and a processing result of current basic parameters of the target building; A fifth sub-relationship between the fire source existence information processing result, smoke concentration processing result, dangerous gas concentration processing result, and basic parameter processing result of the target building and the comprehensive processing result is established through the first prediction sub-network; a sixth sub-relationship between the comprehensive processing result and the fire risk level is established through the second prediction sub-network; The method further comprises: A first sub-relationship between the information on whether a fire source exists and the processing result of the information on whether a fire source exists is established through the first sub-neural network; a second sub-relationship between the smoke concentration and the processing result of the smoke concentration is established through the second sub-neural network; a third sub-relationship between the dangerous gas concentration and the processing result of the dangerous gas concentration is established through the third sub-neural network; a fourth sub-relationship between the basic parameters of the target building and the processing result of the basic parameters of the target building is established through the fourth sub-neural network; Among them, the third neural network uses a fully connected neural network, and its input size is 32×1, 32 represents the values ​​of 32 temperature sensors; the fully connected layer consists of 5 hidden layers, the number of neurons in the hidden layers is 2048, and the activation function of the hidden layer uses ReLU. Finally, the neurons in the output layer are converted into a 224×224 two-dimensional matrix as a heat map. The formula is as follows: Map=Reshape(f(a(f(a(f(a(a(f(a(f(a(x))))))))))))(1) a(x)=Wx+b(2) Wherein, formula (1) indicates that the temperature is processed by a fully connected neural network to generate a heat map; formula (2) that is, a(x) represents the linear calculation of the hidden layer, W is the weight matrix, linear transformation, b is the bias; f(*) is the activation function ReLU; Reshape(*) function is to convert one-dimensional data into a matrix of size; Map represents the generated heat map; Among them, the first sub-neural network uses an enhanced convolutional neural network. The enhanced convolutional neural network structure is divided into a backbone network and a branch network. The backbone network has 10 convolutional layers, 3 pooling layers, and 3 upsampling layers. In the branch network, there is only one convolutional layer on each branch. The input size of the enhanced convolutional neural network is 4×224×224, 4 represents the sum of the number of channels of the input image data and the heat map, and 224×224 represents the resolution of the image. The convolution kernels of all convolutional layers are 3×3, the pooling layer uses average pooling, and the resolution is reduced to half of the original. The upsampling layer uses transposed convolution, and the resolution is doubled. The output results of the 2nd, 3rd, 4th and 5th convolutional layers on the backbone are respectively input to the branch convolutional layers for feature enhancement, and then fused with the features of the 6th, 7th, 8th and 9th convolutional layers. Finally, the extracted features are pooled and flattened to obtain the final result, and the final output size is 588×1. The formula is as follows: F1=g(I input )(3) F2=z(g(F1))(4) F2′=z(g(F2))(5) F3=z(g(P(F2)))(6) F3′=z(g(F3))(7) F4=z(g(P(F3)))(8) F4′=z(g(F4))(9) F5=z(g(P(F4)))(10) F5′=z(g(F5))(11) F6=z(g(F5))(12) V output =R(g(z(g(Up(z(g(Up(z(g(Up(F6+F5′)+F4′)))+F3′)))+F2′))))(13) In the formula, formula (3) represents the input I input After the first convolution layer, feature F1 is calculated. Formula (4) indicates that feature F1 is calculated through the second convolution layer to obtain feature F2. Formula (5) indicates that feature F2 is calculated through the branch convolution layer to obtain feature F2′. Formula (6) indicates that feature F2 is calculated through the pooling layer and the third convolution layer to obtain feature F3. Formula (7) indicates that feature F3 is calculated through the branch convolution layer to obtain feature F3′. Formula (8) indicates that feature F3 is calculated through the pooling layer and the fourth convolution layer to obtain feature F4. Formula (9) indicates that feature F4 is calculated through the branch convolution layer. The feature F4′ is obtained. Formula (10) indicates that the feature F4 is calculated through the pooling layer and the 5th convolution layer to obtain the feature F5. Formula (11) indicates that the feature F5 is calculated through the branch convolution layer to obtain the feature F5′. Formula (12) indicates that the feature F5 is calculated through the 6th convolution layer to obtain the feature F6. Formula (13) indicates that the features F6, F2′, F3′, F4′ and F5′ are calculated through the upsampling layer and the 6th, 7th, 8th and 9th convolution layers and downsampling layers, and the two-dimensional features are flattened into a one-dimensional vector to obtain the output vector V. output ; g represents the convolution operation of the convolution layer, z represents the activation function ReLU, P represents the average pooling operation of the pooling layer, Up represents the transposed convolution operation of the upsampling layer, and R represents the average pooling operation and flattens the two-dimensional matrix into a one-dimensional vector operation; Among them, the second sub-neural network, the third sub-neural network and the fourth sub-neural network all use fully connected neural networks with the same network structure, which are used to analyze the smoke concentration data, the dangerous gas concentration data and the basic parameters of the target building respectively. The input size of the above fully connected neural network is 32×1, with a total of 5 hidden layers, and the number of neurons in each hidden layer is 6144. The activation function of the hidden layer neurons uses the Sigmoid function; the number of neurons in the output layer is 588, and the formula is as follows: Out=f(a(f(a(f(a(f(a(f(a(x)))))))))))(14) a(x)=Wx+b(15) Wherein, formula (14) indicates that smoke concentration data, hazardous gas concentration data or basic parameters of target buildings are analyzed by fully connected neural network; formula (15) that is, a(x) represents the linear calculation of hidden layer; W is the weight matrix; b is the bias; f(*) is the activation function Sigmoid function of hidden layer neurons, and Out is the output of fully connected neural network; The input size of the second prediction subnetwork is 588×1, with 5 hidden layers, 6144 neurons in the hidden layers, and the activation function of the neurons in the hidden layers is the Sigmoid function; the number of neurons in the output layer is 4; the formula is as follows: Out 等级 =f(a(f(a(f(a(f(a(f(a(Out attention )))))))))) (18) a(x)=Wx+b (19) In the formula, Out attention represents the output of the attention mechanism network; Formula (18) represents the prediction of the fire risk level by the output of the attention mechanism network through the fully connected neural network; Out 等级 represents the predicted fire risk level; formula (19) that is, a(x) represents the linear calculation of a hidden layer, W is the weight matrix, b is the bias; f(*) is the activation function Sigmoid function of the hidden layer neuron.

2. The method according to claim 1, characterized in that The second neural network further includes a third prediction sub-network, and the second relationship further includes a seventh sub-relationship; when the second neural network outputs that the fire risk level is a preset fire risk level, the third prediction sub-network is called; and the method further includes: When the fire risk level is a dangerous level or a destructive level, the seventh sub-relationship pre-established by the third prediction sub-network is used to determine the current dangerous area coordinates corresponding to the target area whose current fire risk level is a dangerous level or a destructive level.

3. A system for dynamic assessment of fire risk based on Internet of Things monitoring, characterized in that: The system is used to establish a corresponding relationship between relevant information of a building and a corresponding fire risk level through an artificial intelligence model; wherein the artificial intelligence model includes a first model; the first model includes a first neural network, a second neural network, and a third neural network, the corresponding relationship includes a first relationship, a second relationship, and a third relationship, and the relevant information includes a temperature of a target area in a target building, a smoke concentration, a dangerous gas concentration, environmental monitoring data, and basic parameters of the corresponding target building; wherein the first neural network includes a first sub-neural network, a second sub-neural network, a third sub-neural network, and a fourth sub-neural network; the first relationship includes a first sub-relationship, a second sub-relationship, a third sub-relationship, and a fourth sub-relationship; wherein the second neural network includes a first prediction sub-network and a second prediction sub-network, and the second relationship includes a fifth sub-relationship and a sixth sub-relationship; The system comprises: A current fire source existence information generation module is used to obtain the current temperature and current environmental monitoring data of the target area in the target building, and determine the current thermal map corresponding to the current temperature through the third relationship pre-established by the third neural network; and generate the current fire source existence information according to the current thermal map and the current environmental monitoring data; A current processing result determination module is used to obtain current environmental information of a target area in a target building, and determine a current environmental information processing result corresponding to the current environmental information through a first relationship pre-established by the first neural network; wherein the current environmental information includes information on whether a current fire source exists, current smoke density, current dangerous gas density, and current basic parameters of the target building; the current environmental information processing result includes a processing result of information on whether a current fire source exists, a processing result of current smoke density, a processing result of current dangerous gas density, and a processing result of current basic parameters of the target building; The current fire risk level determination module is used to establish a fifth sub-relationship between the fire source existence information processing result, the smoke concentration processing result, the dangerous gas concentration processing result and the basic parameter processing result of the target building and the comprehensive processing result through the first prediction sub-network; and establish a sixth sub-relationship between the comprehensive processing result and the fire risk level through the second prediction sub-network; The system further comprises: A first sub-relationship between the information on whether a fire source exists and the processing result of the information on whether a fire source exists is established through the first sub-neural network; a second sub-relationship between the smoke concentration and the processing result of the smoke concentration is established through the second sub-neural network; a third sub-relationship between the dangerous gas concentration and the processing result of the dangerous gas concentration is established through the third sub-neural network; a fourth sub-relationship between the basic parameters of the target building and the processing result of the basic parameters of the target building is established through the fourth sub-neural network; Among them, the third neural network uses a fully connected neural network, and its input size is 32×1, 32 represents the values ​​of 32 temperature sensors; the fully connected layer consists of 5 hidden layers, the number of neurons in the hidden layers is 2048, and the activation function of the hidden layer uses ReLU. Finally, the neurons in the output layer are converted into a 224×224 two-dimensional matrix as a heat map. The formula is as follows: Map=Reshape(f(a(f(a(f(a(a(f(a(f(a(x))))))))))))(1) a(x)=Wx+b(2) Wherein, formula (1) indicates that the temperature is processed by a fully connected neural network to generate a heat map; formula (2) that is, a(x) represents the linear calculation of the hidden layer, W is the weight matrix, linear transformation, b is the bias; f(*) is the activation function ReLU; Reshape(*) function is to convert one-dimensional data into a matrix of size; Map represents the generated heat map; Among them, the first sub-neural network uses an enhanced convolutional neural network. The enhanced convolutional neural network structure is divided into a backbone network and a branch network. The backbone network has 10 convolutional layers, 3 pooling layers, and 3 upsampling layers. In the branch network, there is only one convolutional layer on each branch. The input size of the enhanced convolutional neural network is 4×224×224, 4 represents the sum of the number of channels of the input image data and the heat map, and 224×224 represents the resolution of the image. The convolution kernels of all convolutional layers are 3×3, the pooling layer uses average pooling, and the resolution is reduced to half of the original. The upsampling layer uses transposed convolution, and the resolution is doubled. The output results of the 2nd, 3rd, 4th and 5th convolutional layers on the backbone are respectively input to the branch convolutional layers for feature enhancement, and then fused with the features of the 6th, 7th, 8th and 9th convolutional layers. Finally, the extracted features are pooled and flattened to obtain the final result, and the final output size is 588×1. The formula is as follows: F1=g(I input )(3) F2=z(g(F1))(4) F2′=z(g(F2))(5) F3=z(g(P(F2)))(6) F3′=z(g(F3))(7) F4=z(g(P(F3)))(8) F4′=z(g(F4))(9) F5=z(g(P(F4)))(10) F5′=z(g(F5))(11) F6=z(g(F5))(12) V output =R(g(z(g(Up(z(g(Up(z(g(Up(F6+F5′)+F4′)))+F3′)))+F2′))))(13) In the formula, formula (3) represents the input I input After the first convolution layer, feature F1 is calculated. Formula (4) indicates that feature F1 is calculated through the second convolution layer to obtain feature F2. Formula (5) indicates that feature F2 is calculated through the branch convolution layer to obtain feature F2′. Formula (6) indicates that feature F2 is calculated through the pooling layer and the third convolution layer to obtain feature F3. Formula (7) indicates that feature F3 is calculated through the branch convolution layer to obtain feature F3′. Formula (8) indicates that feature F3 is calculated through the pooling layer and the fourth convolution layer to obtain feature F4. Formula (9) indicates that feature F4 is calculated through the branch convolution layer. The feature F4′ is obtained. Formula (10) indicates that the feature F4 is calculated through the pooling layer and the 5th convolution layer to obtain the feature F5. Formula (11) indicates that the feature F5 is calculated through the branch convolution layer to obtain the feature F5′. Formula (12) indicates that the feature F5 is calculated through the 6th convolution layer to obtain the feature F6. Formula (13) indicates that the features F6, F2′, F3′, F4′ and F5′ are calculated through the upsampling layer and the 6th, 7th, 8th and 9th convolution layers and downsampling layers, and the two-dimensional features are flattened into a one-dimensional vector to obtain the output vector V. output ; g represents the convolution operation of the convolution layer, z represents the activation function ReLU, P represents the average pooling operation of the pooling layer, Up represents the transposed convolution operation of the upsampling layer, and R represents the average pooling operation and flattens the two-dimensional matrix into a one-dimensional vector operation; Among them, the second sub-neural network, the third sub-neural network and the fourth sub-neural network all use fully connected neural networks with the same network structure, which are used to analyze the smoke concentration data, the dangerous gas concentration data and the basic parameters of the target building respectively. The input size of the above fully connected neural network is 32×1, with a total of 5 hidden layers, and the number of neurons in each hidden layer is 6144. The activation function of the hidden layer neurons uses the Sigmoid function; the number of neurons in the output layer is 588, and the formula is as follows: Out=f(a(f(a(f(a(f(a(f(a(x)))))))))))(14) a(x)=Wx+b(15) Wherein, formula (14) indicates that smoke concentration data, hazardous gas concentration data or basic parameters of target buildings are analyzed by fully connected neural network; formula (15) that is, a(x) represents the linear calculation of hidden layer; W is the weight matrix; b is the bias; f(*) is the activation function Sigmoid function of hidden layer neurons, and Out is the output of fully connected neural network; The input size of the second prediction subnetwork is 588×1, with 5 hidden layers, 6144 neurons in the hidden layers, and the activation function of the neurons in the hidden layers is the Sigmoid function; the number of neurons in the output layer is 4; the formula is as follows: Out 等级 =f(a(f(a(f(a(f(a(f(a(Out attention )))))))))) (18) a(x)=Wx+b (19) In the formula, Out attention represents the output of the attention mechanism network; Formula (18) represents the prediction of the fire risk level by the output of the attention mechanism network through the fully connected neural network; Out 等级 represents the predicted fire risk level; formula (19) that is, a(x) represents the linear calculation of a hidden layer, W is the weight matrix, b is the bias; f(*) is the activation function Sigmoid function of the hidden layer neuron.

4. A device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method for dynamic assessment of fire risk based on Internet of Things monitoring as described in any one of claims 1 to 2 is implemented.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for dynamic assessment of fire risk based on Internet of Things monitoring as described in any one of claims 1 to 2 is implemented.

Images