Fire visualization active early warning monitoring system and method based on infrared temperature measurement technology

Abstract

The present invention relates to the field of monitoring and alarm technology, and discloses a fire visualization active early warning monitoring system and method based on infrared temperature measurement technology. The system acquires smoke data by real-time monitoring of smoke, analyzes the smoke data, identifies the type of fire source, and obtains fire source type data. The system can accurately distinguish the type of fire source and identify the cause of the fire source, which is convenient for subsequent firefighting. The fire source type data is combined with real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into a prediction model to predict a fire risk area. Finally, a corresponding alarm mechanism is triggered according to the fire risk area predicted by the prediction model and the type of fire source, so that personnel in the area can be warned to flee the fire area in time. Different alarm mechanisms are set according to risk areas of different levels, which is convenient for guiding personnel outside the risk area to rush to implement firefighting measures in time, so that the monitoring and alarm technology is more efficient and sensitive in fire early warning work.

Description

Technical Field

[0001] The present invention relates to the field of monitoring and alarm technology, and in particular to a fire visualization active early warning monitoring system and method based on infrared temperature measurement technology. Background Art

[0002] Monitoring and alarm technology is widely used in modern society. When monitoring and alarm technology is used in early fire warning work, monitoring equipment monitors and issues alarms at the early stage of a fire to warn people to put out the fire and escape from the fire scene. However, the existing monitoring and alarm technology for large sites (schools, large factories or harbors) has defects in fire warning. Traditional monitoring and alarm technology is difficult to identify and distinguish the type of fire source, and it is difficult to predict the fire spread area based on the type of fire source combined with wind direction and speed data to determine the risk area and issue different alarm mechanisms to warn people in the area to evacuate or put out the fire. This makes the overall efficiency of traditional monitoring and alarm technology low, the alarm mechanism is single and the sensitivity is poor;

[0003] Therefore, the present invention obtains fire smoke data, identifies the type of fire source by analyzing the smoke data, and obtains the identification data to determine the direction of the next experimental operation.

[0004] After identifying the type of fire source, the fire risk area is predicted based on the fire source type combined with the wind direction and wind speed data at the time, and the alarm mechanism is further triggered to conduct relevant research.

[0005] In summary, how to efficiently and sensitively predict and monitor fire conditions and issue alarms is an urgent problem to be solved and optimized in the fire visualization active early warning monitoring system based on infrared temperature measurement technology. Summary of the invention

[0006] The present invention provides a fire visualization active early warning monitoring system based on infrared temperature measurement technology, which solves the technical problems in the related technology that it is difficult to identify and distinguish the types of fire sources, and it is difficult to predict the fire spread area according to the type of fire source combined with wind direction and wind speed data to determine the risk area and issue different alarm mechanisms.

[0007] In order to solve the above technical problems, the present invention provides a fire visualization active early warning monitoring system and method based on infrared temperature measurement technology. The specific technical solutions are as follows:

[0008] In the first aspect, a fire visualization active early warning monitoring system and method based on infrared temperature measurement technology, the method comprises the following steps:

[0009] Monitor smoke in real time to obtain smoke data;

[0010] Analyzing the smoke data and identifying the fire source type to obtain fire source type data;

[0011] The fire source type data is combined with the real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into a prediction model to predict the fire risk area. The prediction model includes:

[0012] Input layer: it receives the preprocessed integrated set;

[0013] Hidden layer: It is set to have two hidden layers with 64 and 32 neurons respectively, and the activation function is used to introduce linear transformation;

[0014] Output layer: it outputs the linear transformation task to obtain the predicted fire risk;

[0015] The fire risk area predicted by the prediction model is combined with the type of fire source to trigger a corresponding alarm mechanism.

[0016] Furthermore, smoke is monitored in real time to obtain smoke data, including:

[0017] Adjust various parameters of the photoelectric smoke sensor;

[0018] Based on the adjusted photoelectric smoke sensor, sensor monitoring data is obtained;

[0019] De-noising the sensor monitoring data, filtering out interference signals, and calibrating and synchronizing the various sensor data so that the various sensor monitoring data are aligned on the same time axis;

[0020] The smoke data is acquired according to the sensor monitoring data.

[0021] Further, the smoke data is analyzed to identify the fire source type to obtain fire source type data, including:

[0022] According to the smoke data Where, C dust is the particle concentration; f is the preset threshold; V smoke It is the voltage output of the photoelectric smoke sensor to obtain smoke concentration data;

[0023] According to the smoke data Where D size is the particle size distribution; f is the preset threshold; V catter The voltage output of the laser scattering sensor is used to obtain the smoke particle size distribution data;

[0024] Based on the smoke concentration data and the smoke particle size distribution data ; In the formula, x is the original eigenvalue; μ is the mean of the feature; σ is the standard deviation of the feature; x' is the standardized eigenvalue after standardization to eliminate the dimensional differences between different features;

[0025] Integrating the standardized smoke data to obtain a comprehensive feature vector;

[0026] The fire source is identified according to the feature vector to obtain fire source type data.

[0027] Further, the fire source is identified according to the feature vector to obtain fire source type data, including:

[0028] Based on the feature vector, feature data is extracted from the feature vector to obtain key feature data;

[0029] Model training is performed according to the key feature data, and the trained model is used to predict the fire source type of the new feature vector;

[0030] The fire source type is predicted based on the model output and combined with the fire source type data pre-defined in the database to determine the specific type of fire source.

[0031] Further, the fire source type data is combined with the real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into the prediction model to predict the fire risk area, including:

[0032] Acquire the real-time wind direction and wind speed data of the day based on the meteorological platform;

[0033] One-hot encoding the fire source type data and the real-time wind direction and wind speed data to obtain encoded data;

[0034] integrating the coded data to obtain an integrated data set;

[0035] The integrated data set is preprocessed to obtain a preprocessed integrated set.

[0036] Further, the prediction model includes:

[0037] Inputting the preprocessed integrated set to the input layer, so that the input layer receives the preprocessed integrated set;

[0038] The hidden layer is based on , where h (l) is the output activation value of the lth layer; W (l) is the weight matrix of the lth layer; h (l-1) is the output of the l-1th layer; b (l) is the bias vector of the lth layer; σ is the activation function, which is defined as: ; is a constant; it sets two hidden layers with 64 and 32 neurons respectively, and introduces linear transformation using activation function;

[0039] According to the linear transformation introduced by the hidden layer: ; Where γ is the fire risk prediction value; W (L) is the weight matrix of the output layer; h (L-1) is the output of the last hidden layer; b (L) is the bias vector of the output layer; the fire risk prediction value is output through the output layer.

[0040] Furthermore, the prediction model also includes:

[0041] Compile the prediction model and measure the difference between the model prediction and the true value to guide the model parameters;

[0042] By optimizing the minimization loss function and iterating the model parameters through multiple rounds, the batch gradient is reduced to update the model parameters;

[0043] According to the updated model parameters, the prediction model is trained to obtain training results;

[0044] A test verification is performed based on the training results to obtain a predicted model after the test.

[0045] Furthermore, according to the fire risk area predicted by the prediction model, the corresponding alarm mechanism is triggered in combination with the fire source type, including:

[0046] Predicting the fire risk area according to the prediction model and capturing the location of the fire point according to the infrared temperature sensor data;

[0047] According to the location of the fire point, with the fire area as the center, four risk areas with different risk levels are planned: within a radius of five meters; within a radius of twenty meters; within a radius of fifty meters and within a radius of one hundred meters;

[0048] Setting different trigger alarm mechanisms according to the four different risk areas;

[0049] According to the alarm mechanism, people in the area are warned, and people outside the area are guided to rush to the area with serious fire in time to implement fire-fighting measures.

[0050] In the third aspect, a fire visualization active early warning monitoring system based on infrared temperature measurement technology includes: the electronic device includes a memory, a processor, and a fire visualization active early warning monitoring method program based on infrared temperature measurement technology stored in the memory and executable on the processor, and the fire visualization active early warning monitoring method program based on infrared temperature measurement technology implements the steps of the fire visualization active early warning monitoring method based on infrared temperature measurement technology when executed by the processor.

[0051] Fourthly, a fire visualization active early warning monitoring system based on infrared temperature measurement technology, which also includes:

[0052] Smoke monitoring module: It is used to monitor smoke in real time to obtain smoke data;

[0053] Fire source identification module: used to analyze the smoke data, identify the fire source type, and obtain fire source type data;

[0054] Prediction area module: used to combine the fire source type data with the real-time wind direction and wind speed data into an integrated set, and input the integrated set into the prediction model to predict the fire risk area;

[0055] Early warning module: It is used to trigger the corresponding alarm mechanism according to the fire risk area predicted by the prediction model and the type of fire source.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The above-mentioned scheme of the present invention obtains smoke data by real-time monitoring of smoke, and then analyzes the smoke data to identify the type of fire source to obtain fire source type data. It can accurately distinguish the type of fire source and identify the cause of the fire, which is convenient for subsequent firefighting. The fire source type data is combined with real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into the prediction model to predict the fire risk area. Finally, the corresponding alarm mechanism is triggered according to the fire risk area predicted by the prediction model and the fire source type, so that people in the area can be warned to flee the fire area in time, and different alarm mechanisms are set according to risk areas of different levels, which is convenient for guiding people outside the risk area to rush to implement firefighting measures in time, making the monitoring alarm technology more efficient and sensitive in fire warning work. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used for further understanding of the present invention and are used to explain the present invention together with the embodiments of the present invention, but do not constitute a limitation of the present invention.

[0059] Figure 1 It is a flow chart of a fire situation visualization active early warning monitoring method based on infrared temperature measurement technology provided by an embodiment of the present invention;

[0060] Figure 2 It is a schematic diagram of a fire visualization active early warning monitoring system based on infrared temperature measurement technology provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0062] In order to better understand the above technical scheme, the technical scheme of the present invention will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification. It should be pointed out that the above steps are only a preferred implementation order. In the specific implementation process, some steps can be replaced without affecting the overall implementation effect. In order to more clearly explain the technical scheme of this application, the following content explains this scheme in a preferred manner.

[0063] like Figure 1 As shown, the fire visualization active early warning monitoring system and method based on infrared temperature measurement technology includes the following steps:

[0064] Step 11, real-time monitoring of smoke to obtain smoke data;

[0065] Step 12, analyzing the smoke data and identifying the fire source type to obtain fire source type data;

[0066] Step 13, combining the fire source type data with the real-time wind direction and wind speed data into an integrated set, and inputting the integrated set into a prediction model to predict the fire risk area, the prediction model includes:

[0067] Input layer: it receives the preprocessed integrated set;

[0068] Hidden layer: It is set to have two hidden layers with 64 and 32 neurons respectively, and the activation function is used to introduce linear transformation;

[0069] Output layer: it outputs the linear transformation task to obtain the predicted fire risk;

[0070] Step 14: triggering a corresponding alarm mechanism based on the fire risk area predicted by the prediction model and the type of fire source.

[0071] In the embodiment of the present invention, the photoelectric smoke sensor is used to monitor the smoke data in the environment in real time through step 11. At this time, the sensor converts the detected sensor into electronic data and transmits it to the data processing center in real time, so that the smoke can be quickly detected at the early stage of the fire, providing the possibility for early intervention, and providing real-time and accurate smoke concentration data to provide a basis for subsequent analysis. Then, the smoke data obtained from step 11 is analyzed through step 12, and the source type of the smoke is identified. The analysis result determines whether the smoke is caused by fire, and the specific type of fire (such as electrical fire, building fire, etc.), so that the fire smoke can be distinguished from other types of smoke (such as tobacco smoke, industrial emissions, etc.), reducing false alarms, and taking more appropriate response measures according to the type of fire source to improve processing efficiency.

[0072] Then, through step 13, the fire source type data is combined with the real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into the prediction model: the integrated data set is received through the input layer of the prediction model, and then the linear transformation of the data features is performed through the double hidden layer (64 and 32 neurons) and the activation function, and finally the predicted fire risk area is output through the output layer. By comprehensively considering various factors, the accuracy of fire risk prediction is improved, and the wind direction and wind speed data can be updated at any time to ensure the timeliness and accuracy of the prediction results; finally, through step 14, according to the output results of the prediction model, the fire risk level of a specific area is evaluated, and according to the fire risk and fire source type, the corresponding alarm system is activated, such as sound alarm, automatic sprinkler, SMS notification, etc., so as to timely issue alarms in areas with high fire risks, reduce casualties and property losses, and select the most appropriate emergency measures according to the type of fire source to improve the pertinence and effectiveness of emergency response.

[0073] In a preferred embodiment of the present invention, the above step 11 may include:

[0074] Step 111, adjusting various parameters of the photoelectric smoke sensor;

[0075] Step 112, obtaining sensor monitoring data based on the adjusted photoelectric smoke sensor;

[0076] Step 113, denoising the sensor monitoring data, filtering out interference signals, and calibrating and synchronizing the sensor data so that the sensor monitoring data of each type are aligned on the same time axis;

[0077] Step 114, obtaining smoke data according to the sensor monitoring data.

[0078] In the embodiment of the invention, the sensitivity of the photoelectric smoke sensor is adjusted to 1.0%-2.0% through step 111 to ensure a quick response to tiny smoke; the sampling frequency is increased to 2-4 times per second (2-4Hz) to ensure a quick response; and the alarm threshold parameter is set to 1.0%-2.0% to ensure a faster alarm, thereby ensuring accurate detection of smoke concentration changes. Through calibration and adjustment, it is ensured that the sensor can accurately reflect the actual temperature and smoke conditions, thereby optimizing the sensor parameters, reducing false alarms caused by environmental changes or noise, and improving the reliability of the system; then through step 112, the adjusted photoelectric smoke sensor starts to monitor the smoke data in the environment in real time, so that the sensor converts the detected temperature and smoke data into electronic signals and transmits them to the data processing unit. The adjusted sensor can provide accurate real-time data to reflect the changes in the current environment, thereby enabling the optimization of sensor parameters to ensure the reliability and accuracy of the data, laying a foundation for subsequent data processing and analysis.

[0079] Then, in step 113, a filtering algorithm is used to remove noise and interference signals in the sensor data, and data calibration is performed by comparing the actual value with the sensor reading to ensure the accuracy of the sensor output data. The sensor data is aligned on the same time axis through a timestamp algorithm. The specific steps are as follows:

[0080] When the smoke data contains smoke particle size data D1 and smoke concentration data D2, and the timestamp of D1 is ; and the timestamp of D2 is The data values ​​are ; and convert all timestamps to a unified format, the conversion formula is: ; At the same time, select a suitable time interval Δt so that: Where, t i is the timestamp, t i+1 is the next timestamp; then the interpolation result is obtained through linear interpolation to align the data in the sensor. The aligned data set is:

[0081] ; In the formula, x i (t i ) is the sensor i at time t i The data value after interpolation is obtained, so after denoising, the data is purer and more accurate, with reduced errors and interference. Through calibration and synchronization, the consistency of the sensor data is ensured, which is convenient for comprehensive analysis.

[0082] Finally, in step 114, temperature data and smoke data related to fire are extracted from the sensor data after de-noising, calibration and synchronization through step 113. These data will be used in the fire detection system to determine whether there is a fire hazard or a fire has occurred. Therefore, accurate temperature and smoke data make fire detection more precise, reduce false alarms and missed alarms, and provide real-time fire-related data to facilitate rapid response and handling of fire situations.

[0083] In a preferred embodiment of the present invention, the above step 12 may include:

[0084] Step 121, according to the smoke data Where, C dust is the particle concentration; f is the preset threshold; V smoke It is the voltage output of the photoelectric smoke sensor to obtain smoke concentration data;

[0085] Step 122, according to the smoke data Where D size is the particle size distribution; f is the preset threshold; V catter The voltage output of the laser scattering sensor is used to obtain the smoke particle size distribution data;

[0086] Step 123, based on the smoke concentration data and the smoke particle size distribution data, ; In the formula, x is the original eigenvalue; μ is the mean of the feature; σ is the standard deviation of the feature; x' is the standardized eigenvalue after standardization to eliminate the dimensional differences between different features;

[0087] Step 124, integrating the standardized smoke data to obtain a comprehensive feature vector;

[0088] Step 125: Identify the fire source according to the feature vector to obtain fire source type data.

[0089] In an embodiment of the invention, in step 121, smoke data is collected by a sensor, including information on suspended particles in the air, and data on smoke concentration, that is, the number of smoke particles in a unit volume of air, is extracted. Therefore, by extracting concentration data, air quality and pollution level can be more accurately evaluated, thereby providing real-time smoke concentration data, which is helpful for timely response to emergencies such as fires; then in step 122, the smoke data collected by the sensor in step 121 is used to simultaneously obtain particle size distribution information, so that not only the distribution of particles of different sizes can be understood, but also the source and nature of smoke can be better analyzed, so that smoke particles of different sizes are helpful for the analysis of fire source types, and obtaining particle size distribution data is helpful for more accurate evaluation of fire source types; then in step 123, different features (such as concentration and particle size) can be compared in the same analysis model, and standardization processing is required to make the data have the same dimension or range, thereby eliminating dimensional differences, so that different features can be processed in the same analysis framework, avoiding deviations caused by different scales, and standardization processing helps to improve the performance of the machine learning model, making it converge faster, and improving prediction accuracy.

[0090] Through step 124, the standardized smoke concentration data and particle size distribution data are integrated to generate a vector containing all feature information, namely, a comprehensive feature vector. Therefore, the comprehensive feature vector contains all key information, which is helpful for comprehensive analysis and evaluation of smoke conditions, and integrates multiple features into one vector, which simplifies the data processing and analysis process and improves efficiency. Finally, through step 125, the fire source type is identified according to the comprehensive feature vector, and the identification type model is used to predict new data, thereby providing rich information based on the high-dimensional comprehensive feature vector, improving the accuracy of fire source identification, and being able to quickly identify the fire source type, helping to take corresponding fire prevention measures in time to reduce losses and hazards.

[0091] Specifically in another preferred embodiment, step 125 further includes:

[0092] Step 1251, based on the feature vector, extracting feature data from the feature vector to obtain key feature data;

[0093] Step 1252, performing model training according to the key feature data, and using the trained model to predict the fire source type of the new feature vector;

[0094] Step 1253, predicting the fire source type based on the model output and combining it with the fire source type data pre-defined in the database, to determine the specific type of the fire source.

[0095] In this embodiment, the most influential features for fire source type prediction are extracted from the comprehensive feature vector in step 1251. This usually involves feature selection or feature extraction techniques, such as PCA (principal component analysis), LDA (linear discriminant analysis), information gain, etc. These techniques help reduce the dimension of features while retaining the most representative key feature data, and by extracting key features, reduce the complexity and dimension of data, so that the selection of key features can reduce noise and improve the accuracy and generalization ability of the model.

[0096] Then, the prediction model is trained in step 1252 by using the key feature data extracted in step 1251. Commonly used models include support vector machine (SVM), random forest, gradient boosting tree, neural network, etc.;

[0097] The training process specifically includes:

[0098] Data segmentation: Divide the data set into a training set and a validation set. When training the model, use historical data for verification and testing, which helps to tune the model parameters in the next step and reduce overfitting or underfitting problems.

[0099] Hyperparameter tuning: Select the best model parameters through cross-validation;

[0100] Model evaluation: Evaluate model performance on the validation set and select the optimal model so that the trained model can process new input data in real time, quickly predict the type of fire source, and improve emergency response speed.

[0101] The trained model is used to predict the fire source type of the new feature vector. The new feature vector is input, and the model outputs the prediction result (fire source type), so that the fire source type can be automatically identified and classified, improving the intelligence level of the fire warning system, and using machine learning algorithms to improve the accuracy and robustness of the prediction so that it can accurately map the input features to the output fire source type. The trained model is then applied to the new feature vector to predict the fire source type.

[0102] Finally, after the model outputs the prediction results of the fire source type in step 1253, the system compares these results with the pre-defined fire source type data in the database. The database contains detailed descriptions and classification standards of known fire source types to ensure the accuracy and consistency of the prediction results. Through comparison and confirmation, the specific type of the fire source is finally determined, and the prediction results are ensured to be accurate by comparing with the standard fire source type data in the database, so that the system can not only predict the fire source type, but also provide more detailed information about the fire source, such as the possible degree of danger, countermeasures, etc. At the same time, using the historical data accumulated in the database, the system can continuously improve and update the model and enhance the recognition ability.

[0103] By judging the type of fire source (wood and paper, fuel and gasoline, electrical fire, metal fire, and chemical fire), the type of ignition material is different, so the speed of fire spread in a certain period of time is also different. Therefore, it is convenient for subsequent prediction models to predict fire risk areas in different time periods and issue alarms in time.

[0104] In a preferred embodiment of the present invention, the above step 13 may include:

[0105] Step 131, obtaining real-time wind direction and wind speed data of the day based on the meteorological platform;

[0106] Step 132, performing one-hot encoding on the fire source type data and the real-time wind direction and wind speed data to obtain encoded data;

[0107] Step 133, integrating the encoded data to obtain an integrated data set;

[0108] Step 134 , preprocessing the integrated data set to obtain a preprocessed integrated set.

[0109] In an embodiment of the present invention, real-time wind direction and wind speed data are obtained from a meteorological platform through step 131. A meteorological platform usually provides high-precision meteorological data. When it is necessary to obtain real-time wind direction and wind speed data of the day from the meteorological platform, the meteorological platform usually provides an API interface, obtains relevant data through HTTP requests, and then uses API calls to obtain meteorological data. The system extracts wind direction and wind speed by parsing the JSON or XML format data returned by the API. At this time, the wind direction data obtained is usually expressed as degrees (0-360 degrees), and the wind speed data is expressed as speed units (such as meters per second). The wind direction and wind speed information of a specific area can be obtained in real time. By obtaining real-time data, the accuracy of fire source monitoring and prediction can be ensured, reflecting the actual conditions of the current environment, and real-time acquisition of wind direction and wind speed data helps to dynamically analyze the diffusion path and speed of the fire source, so as to better formulate countermeasures.

[0110] Then, in step 132, the fire source type data and the real-time wind direction and wind speed data are one-hot encoded, and the categorical data are converted into a binary vector through one-hot encoding, and each categorical value is represented by a unique binary vector, which specifically includes: converting the categorical variable into a binary vector, and each category is represented as an independent bit; then, the wind direction data can be divided into several categories (such as north wind, east wind, south wind, west wind, etc.), and each category is encoded as a binary vector; finally, the wind speed data can also be segmented according to the range (such as low wind speed, medium wind speed, high wind speed), and one-hot encoding is performed;

[0111] Fire source type one-hot encoding:

[0112] ; In the formula, n is the number of fire source types, The fire source type is ; One-hot encoding of wind direction:

[0113] ; In the formula, Wind direction ; One-hot encoding of wind speed range: ; Where, wind speed range for ; It converts different types of data into a unified encoding format to facilitate subsequent model processing. At the same time, one-hot encoding can avoid the potential sequence in the data that misleads the model. For example, there is no natural order relationship between fire source types.

[0114] Then, in step 133, the fire source type data, wind direction data and wind speed data after one-hot encoding are integrated together to form a comprehensive feature vector, and then the data from different sources are spliced ​​to form a unified data set for subsequent analysis; the data with different characteristics are spliced ​​together: ; In the formula, ⊕ represents the concatenation operation of the vector; during the integration process, ensure that data from different sources can be processed in a unified framework, so that the integrated data set contains all necessary information, which is convenient for unified processing and analysis. Through data integration, the inconsistency in data source and format is eliminated, and the efficiency and accuracy of data processing are improved. At the same time, all relevant features are integrated into one data set, which is convenient for unified processing and analysis, providing comprehensive information and improving the prediction ability of the model; finally, the integrated data set is preprocessed through step 134. Common preprocessing steps include data standardization: scaling the data to the same range, and common methods include standardization and normalization; missing value processing: filling or deleting missing values, and common methods include mean filling, median filling, interpolation, etc.; and outlier processing: detecting and processing outliers, and common methods include box plot method, Z-score method, etc. Preprocessing is a key step in machine learning, usually including data cleaning, normalization, denoising and other operations. By processing the integrated data set, problems such as missing values, outliers and differences between different dimensions in the data are solved. Preprocessed data is more suitable for model training and prediction. Preprocessing can ensure the quality of input data, reduce the impact of noise and outliers on the model, and improve the training effect of the model. Preprocessing can make different features in the data set on the same scale, thereby improving the convergence speed and prediction accuracy of the model, and ensure the reliability of the data by removing or correcting errors and incomplete parts in the data.

[0115] In a preferred embodiment of the present invention, the above step 134 may include:

[0116] Step 1341, inputting the preprocessed integrated set to the input layer, so that the input layer receives the preprocessed integrated set;

[0117] Step 1342, the hidden layer is based on , where h (l) is the output activation value of the lth layer; W (l) is the weight matrix of the lth layer; h (l-1) is the output of layer l - 1; b (l) is the bias vector of the lth layer; σ is the activation function, which is defined as: ; is a constant; it sets two hidden layers with 64 and 32 neurons respectively, and introduces linear transformation using activation function;

[0118] Step 1343, according to the linear transformation introduced by the hidden layer: ; Where γ is the fire risk prediction value; W (L) is the weight matrix of the output layer; h (L-1)is the output of the last hidden layer; b (L) is the bias vector of the output layer; the fire risk prediction value is output through the output layer.

[0119] In the embodiment of the present invention, the preprocessed data set is input into the input layer of the neural network through step 1341. The input layer is the first layer of the neural network, which is used to receive the preprocessed integrated set as the input feature, and then each neuron in the input layer corresponds to a feature in the preprocessed data set. The number of neurons in the input layer is equal to the number of features of the preprocessed integrated set. The main function of the input layer is to receive external data and pass it to the next layer (hidden layer). By directly passing the data of the preprocessed integrated set to the neurons in the input layer, it can be ensured that the preprocessed integrated data set can smoothly enter the neural network for further processing, and the input layer maps each feature in the preprocessed data to the input layer. The neurons emitted to the neural network enable each feature to be processed in the model, and at the same time, the pre-processed data can be effectively received and processed, ensuring that the data can smoothly enter the neural network for further processing, directly corresponding to the features of the pre-processed data, and easy to implement and understand; then, two hidden layers are configured through the neural network of step 1342, and the hidden layer is responsible for feature extraction and nonlinear transformation of the input data, extracting high-order features, and enhancing the expression ability of the model through the superposition of multiple neurons and layers. The first hidden layer is set to contain 64 neurons, and the input is the feature vector output by the input layer; the second hidden layer contains 32 neurons, and the input is the output of the first hidden layer;

[0120] The output of the first hidden layer H 1， As the input of the second hidden layer; the output of the second hidden layer H 2， As the final feature representation: ; Where X is the input feature vector, W1 and W2 weight matrices, b i

[0121] is the bias vector, f iis the activation function; the neurons of each hidden layer are linearly transformed through weighting and bias, and then nonlinear transformation is introduced through activation function (such as ReLU). Activation function can help the model capture complex patterns and nonlinear relationships in the data, so the multi-layer structure and a large number of neurons of the hidden layer can capture complex feature patterns and improve the prediction ability of the model, and the nonlinearity introduced by the activation function enables the model to learn and represent more complex relationships; finally, the output of the hidden layer is passed to the output layer through step 1343. The output layer is usually composed of one or more neurons. According to the fire risk area prediction, the number of neurons and activation function of the output layer will be different. The output layer performs the final linear transformation and activation function processing on the output of the hidden layer to generate the predicted value of the fire risk, and generates the final prediction result through the output layer, providing the specific numerical value or classification of the fire risk, and the output layer designed according to the specific prediction task can directly output the required prediction indicators, such as the probability of fire risk, and finally the prediction result of the output layer can be directly used for decision support, such as the formulation of fire warning and response strategy. At this time, the model can give a fire risk prediction based on the input characteristics, and improve the accuracy and response speed of the fire warning system.

[0122] In a preferred embodiment of the present invention, the above step 13 further includes:

[0123] Step 135, compile the prediction model and measure the difference between the model prediction and the true value to guide the model parameters;

[0124] Step 136, by optimizing the minimization loss function and iterating the model parameters through multiple rounds to make batch gradients descend to update the model parameters;

[0125] Step 137, performing prediction training on the prediction model according to the updated model parameters to obtain training results;

[0126] Step 138, performing testing and verification based on the training results to obtain a predicted model after testing.

[0127] In an embodiment of the present invention, through step 135, in the neural network framework, the compiled model is compiled by specifying the optimizer, loss function and evaluation index of the model; the loss function is a measure of the difference between the model prediction value and the true value:

[0128] ;

[0129] Where N is the number of samples, y is the true label, and γ is the predicted value; the optimizer is used to update the model parameters, and commonly used optimizers include stochastic gradient descent (SGD), Adam, RMSprop, etc.; evaluation indicators; used to evaluate model performance, such as accuracy, precision, recall, etc. After compiling the model, the loss function is calculated using the training data to measure the difference between the predicted value and the true value, so that the model can optimize the target and evaluation criteria, provide a basis for the subsequent training process, and can measure the difference between the predicted value and the true value, which helps to guide the adjustment and optimization of the model parameters; then through step 136, the model parameters are optimized by minimizing the loss function, so that the prediction result of the model is as close to the true value as possible, and the model parameters are updated using the gradient descent algorithm. The gradient descent algorithm adjusts the model parameters along the negative direction of the gradient according to the gradient of the loss function to reduce the value of the loss function, and calculates the gradient and updates the parameters by using all the training data at once: ; In the formula, θ is the model parameter, η is the learning rate, and ∇θL(γ,y) is the gradient of the loss function with respect to the parameter; thus, through gradient descent and iterative optimization, the model parameters are continuously adjusted, so that the model performance is gradually improved, and batch gradient descent can stably update the parameters, which is suitable for training large data sets.

[0130] Using the model parameters updated in step 136, multiple rounds of iterative training are performed on the training data in step 137 to continuously optimize the model. In each round of training (i.e., each epoch), the model uses the current parameters to predict the training data, calculates the error based on the loss function, updates the parameters through gradient descent, and records the loss value and evaluation index value of each round of training, such as accuracy, precision, etc., and then evaluates the learning effect and convergence of the model based on the training results. Thus, through multiple rounds of iterative training, the model continuously learns and adjusts to improve the prediction performance, so that the training results provide the performance of the model on the training data, providing a reference for subsequent model evaluation and adjustment. Finally, after the training is completed, the model is verified using an independent test data set in step 138. The test data set is used to evaluate the generalization ability of the model, that is, its performance on unseen data. The loss value and evaluation index value on the test data are compared with the results of the training data. According to the test results, the performance and stability of the model are evaluated, and it is checked whether there is overfitting (the training data performs well, but the test data performs poorly) or underfitting (both the training data and the test data perform poorly). At this time, through test verification, we can understand the performance of the model in real scenarios and ensure the generalization ability of the model, so as to evaluate the performance of the model and provide a basis for further optimization and deployment of the model.

[0131] In a preferred embodiment of the present invention, the above step 14 may include:

[0132] Step 141, predicting the fire risk area according to the prediction model, and capturing the location of the fire point according to the infrared temperature sensor data;

[0133] Step 142, according to the location of the fire point, with the fire area as the center, four risk areas of different risk levels are planned: within a radius of five meters; within a radius of twenty meters; within a radius of fifty meters and within a radius of one hundred meters;

[0134] Step 143, setting different trigger alarm mechanisms according to the four different risk areas;

[0135] Step 144, alerting people in the area according to the alarm mechanism, and guiding people outside the area to rush to the serious fire area in time to implement fire-fighting measures.

[0136] In the embodiment of the present invention, the fire risk of a specific area is predicted by using the trained prediction model through step 141. The model evaluates the possibility of fire occurrence and spread within a certain time (the time interval is ten minutes, and the risk area model needs to re-predict the risk area every ten minutes) through input data (such as wind direction and wind speed data, historical fire data, environmental data, etc.). At the same time, the infrared temperature sensor monitors the temperature change of the area in real time. When an abnormally high temperature is detected, the specific location of the ignition point is determined and displayed to the firefighters for reference through a visualization device, so that the fire risk can be predicted in advance, which is helpful to take preventive measures and reduce the possibility of fire. The infrared temperature sensor can quickly capture the location of the ignition point and provide an accurate target for subsequent fire fighting. Then, after determining the location of the ignition point, four concentric circular areas are divided according to different radii (5 meters, 20 meters, 50 meters, and 100 meters) with the ignition area as the center through step 142. Each area represents a different fire risk level. As the distance increases, the risk level gradually decreases. Clearly dividing the risk level area facilitates the adoption of different response strategies. At the same time, resources can be preferentially allocated to high-risk areas to improve fire fighting efficiency and safety.

[0137] Then, different alarm mechanisms are designed according to the characteristics of different risk level areas through step 143. For example, the high-risk area within a radius of five meters can be set to flash a red warning light accompanied by an emergency alarm whistle, and at the same time trigger emergency fire extinguishing measures; the medium-risk area within a radius of twenty meters can be set to flash a yellow warning light to warn and require enhanced monitoring; the low-risk area within a radius of fifty meters can be set to flash a blue warning light to warn personnel to remain vigilant and have capable people cooperate in firefighting, and finally, in the low-risk area within one hundred meters, a green warning light can be set to flash to remind and suggest that regional personnel take preventive measures, and capable people actively rush to the medium and heavy risk areas to implement firefighting. Therefore, according to the risk level, different alarm mechanisms are set to ensure that high-risk areas can be responded to in a timely manner, so as to avoid the risk area personnel not having time to evacuate in time without knowing it after the fire spreads, and to enable the fire point area to be extinguished in time; finally, once the alarm mechanism is triggered in step 144, the system will send warning information to the personnel in the area according to the risk level to guide them to take corresponding safety measures. At the same time, the system will also notify firefighters outside the area and guide them to quickly rush to the area with serious fire to carry out fire-fighting operations. By warning people in the area in advance, it can ensure that they can evacuate or take self-rescue measures in time to reduce casualties. It also guides firefighters outside the area to quickly reach the fire point, control the fire in time, and reduce fire losses, thereby ensuring that the fire emergency response is orderly and efficient and improving the overall effectiveness of fire-fighting work.

[0138] like Figure 2 As shown, the present invention provides a fire visualization active early warning monitoring system 20 based on infrared temperature measurement technology, which includes the following steps:

[0139] Smoke monitoring module 201: used to monitor smoke in real time to obtain smoke data;

[0140] Fire source identification module 202: used to analyze the smoke data, identify the fire source type, and obtain fire source type data;

[0141] Prediction area module 203: It is used to combine the fire source type data with the real-time wind direction and wind speed data to form an integrated set, and input the integrated set into the prediction model to predict the fire risk area. The prediction model includes:

[0142] Input layer: it receives the preprocessed integrated set;

[0143] Hidden layer: It is set to have two hidden layers with 64 and 32 neurons respectively, and the activation function is used to introduce linear transformation;

[0144] Output layer: it outputs the linear transformation task to obtain the predicted fire risk;

[0145] Early warning module 204: It is used to trigger a corresponding alarm mechanism based on the fire risk area predicted by the prediction model and the type of fire source.

[0146] In a preferred embodiment of the present invention, a photoelectric smoke sensor is used in the smoke monitoring module 201 to continuously monitor the smoke data in the environment, and the detected smoke data is converted into an electronic signal, and transmitted to the fire source identification module through a data interface, so that the smoke data can be monitored in real time and continuously, providing accurate data support, and the smoke data can be quickly detected, and abnormalities can be discovered in time in the early stages of a fire; then the fire source identification module 202 receives the smoke data from the smoke monitoring module for analysis, identifies the smoke characteristics, and determines the source of the smoke, such as kitchen fumes, industrial smoke, fire smoke, etc., based on the analysis results, and generates fire source type data, so that different types of smoke sources can be effectively distinguished, false alarms can be reduced, system reliability can be improved, and key fire source type information can be provided for subsequent fire risk prediction.

[0147] The fire source type data generated by the fire source identification module is then combined with the real-time wind direction and wind speed data through the prediction area module 203 to form an integrated set, and the integrated set data is preprocessed so that it can be input into the prediction model, thereby improving the model complexity and expression ability. Finally, the prediction result, i.e., the fire risk area, is output through the prediction model, so that a more accurate fire risk area prediction is provided by comprehensively considering multiple factors (fire source type, wind direction, wind speed), and the data is updated in real time to promptly reflect environmental changes and provide the latest risk assessment; finally, according to the fire risk area data provided by the early warning module 204 through the prediction area module, the risk level is evaluated, and combined with the fire source type, the corresponding alarm mechanism is triggered, such as sound alarm, SMS notification, automatic fire extinguishing system, etc., so as to issue an alarm before the fire occurs or in the early stage to reduce fire losses. At the same time, according to different fire source types, targeted alarm measures are triggered to improve response efficiency.

[0148] An embodiment of the present invention further provides an electronic device according to a fire visualization active early warning monitoring system and method based on infrared temperature measurement technology, including a memory, a processor, and a fire visualization active early warning monitoring method program based on infrared temperature measurement technology stored in the memory and executable on the processor. When the fire visualization active early warning monitoring method program based on infrared temperature measurement technology is executed by the processor, the steps of the fire visualization active early warning monitoring method based on infrared temperature measurement technology are implemented. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0149] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0150] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0151] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0152] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0153] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0155] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical disks.

[0156] In addition, it should be pointed out that in the apparatus and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. Moreover, the steps of performing the above series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order, and some steps can be performed in parallel or independently of each other. For those of ordinary skill in the art, it can be understood that all or any steps or components of the method and apparatus of the present invention can be implemented in hardware, firmware, software or a combination thereof in any computing device (including processors, storage media, etc.) or a network of computing devices, which can be achieved by those of ordinary skill in the art using basic programming skills after reading the description of the present invention.

[0157] Therefore, the purpose of the present invention can also be achieved by running a program or a group of programs on any computing device. The computing device can be a well-known general device. Therefore, the purpose of the present invention can also be achieved by simply providing a program product containing a program code that implements the method or device. That is to say, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any well-known storage medium or any storage medium developed in the future. It should also be pointed out that in the device and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. In addition, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order. Some steps can be performed in parallel or independently of each other.

[0158] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A fire situation visualization active early warning monitoring method based on infrared temperature measurement technology, characterized in that: The method comprises the following steps: Monitor smoke in real time to obtain smoke data; Analyzing the smoke data and identifying the fire source type to obtain fire source type data; The step of analyzing the smoke data and identifying the fire source type to obtain fire source type data includes: According to the smoke data ; Where Cdust is the particle concentration; f is the preset threshold; Vsmoke is the voltage output of the photoelectric smoke sensor to obtain smoke concentration data; According to the smoke data ; In the formula, Dsize is the particle size distribution; f is the preset threshold; Vcatter is the voltage output of the laser scattering sensor; to obtain the smoke particle size distribution data; Based on the smoke concentration data and the smoke particle size distribution data ; In the formula, x is the original eigenvalue; μ is the mean of the feature; σ is the standard deviation of the feature; x' is the standardized eigenvalue after standardization to eliminate the dimensional differences between different features; Integrating the standardized smoke data to obtain a comprehensive feature vector; Identify the fire source according to the feature vector to obtain fire source type data; The fire source type data is combined with the real-time wind direction and wind speed data to form an integrated set, and the integrated set is input into a prediction model to predict the fire risk area. The prediction model includes: Input layer: it receives the preprocessed integrated set; Hidden layer: It is set to have two hidden layers with 64 and 32 neurons respectively, and the activation function is used to introduce linear transformation; Output layer: it outputs the linear transformation task to obtain the predicted fire risk; The fire risk area predicted by the prediction model is combined with the type of fire source to trigger a corresponding alarm mechanism.

2. The fire visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 1 is characterized in that: Real-time smoke monitoring to obtain smoke data, including: Adjust various parameters of the photoelectric smoke sensor; Based on the adjusted photoelectric smoke sensor, sensor monitoring data is obtained; De-noising the sensor monitoring data, filtering out interference signals, and calibrating and synchronizing the various sensor data so that the various sensor monitoring data are aligned on the same time axis; The smoke data is acquired according to the sensor monitoring data.

3. The fire situation visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 2 is characterized in that: The fire source is identified according to the feature vector to obtain fire source type data, including: Based on the feature vector, feature data is extracted from the feature vector to obtain key feature data; Model training is performed according to the key feature data, and the trained model is used to predict the fire source type of the new feature vector; The fire source type is predicted based on the model output and combined with the fire source type data pre-defined in the database to determine the specific type of fire source.

4. The fire situation visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 3 is characterized in that: The fire source type data is combined with real-time wind direction and wind speed data to form an integrated set, which is input into a prediction model to predict fire risk areas, including: Obtain the real-time wind direction and wind speed data of the day based on the meteorological platform; One-hot encoding the fire source type data and the real-time wind direction and wind speed data to obtain encoded data; integrating the coded data to obtain an integrated data set; The integrated data set is preprocessed to obtain a preprocessed integrated set.

5. The fire situation visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 4 is characterized in that: Predictive models include: Inputting the preprocessed integrated set to the input layer, so that the input layer receives the preprocessed integrated set; The hidden layer is based on , where h (l) is the output activation value of the lth layer; W (l) is the weight matrix of the lth layer; h (l-1) is the output of layer l - 1; b (l) is the bias vector of the lth layer; σ is the activation function, which is defined as: ; is a constant; it sets two hidden layers with 64 and 32 neurons respectively, and introduces linear transformation using activation function; According to the linear transformation introduced by the hidden layer: ; Where γ is the fire risk prediction value; W (L) is the weight matrix of the output layer; h (L-1) is the output of the last hidden layer; b (L) is the bias vector of the output layer; the fire risk prediction value is output through the output layer.

6. The fire situation visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 5 is characterized in that: The forecasting model also includes: Compile the prediction model and measure the difference between the model prediction and the true value to guide the model parameters; By optimizing and minimizing the loss function, and iterating the model parameters through multiple rounds, the batch gradient is reduced to update the model parameters; According to the updated model parameters, the prediction model is trained to obtain training results; A test verification is performed based on the training results to obtain a predicted model after the test.

7. The fire visualization active early warning monitoring method based on infrared temperature measurement technology according to claim 6 is characterized in that: According to the fire risk area predicted by the prediction model, the corresponding alarm mechanism is triggered in combination with the fire source type, including: Predicting the fire risk area according to the prediction model and capturing the location of the fire point according to the infrared temperature sensor data; According to the location of the fire point, with the fire area as the center, four risk areas with different risk levels are planned: within a radius of five meters; within a radius of twenty meters; within a radius of fifty meters and within a radius of one hundred meters; Set up different trigger alarm mechanisms according to four different risk areas; According to the alarm mechanism, people in the area are warned, and people outside the area are guided to rush to the area with serious fire in time to implement fire-fighting measures.

8. Fire visualization active early warning monitoring system based on infrared temperature measurement technology, characterized by: The system includes a memory, a processor, and a fire visualization active early warning monitoring method program based on infrared temperature measurement technology stored in the memory and executable on the processor. When the fire visualization active early warning monitoring method program based on infrared temperature measurement technology is executed by the processor, the steps of the fire visualization active early warning monitoring method based on infrared temperature measurement technology as described in any one of claims 1 to 7 are implemented.

9. The fire visualization active early warning monitoring system based on infrared temperature measurement technology according to claim 8 is characterized in that: The system includes: Smoke monitoring module: It is used to monitor smoke in real time to obtain smoke data; Fire source identification module: used to analyze the smoke data, identify the fire source type, and obtain fire source type data; Prediction area module: used to combine the fire source type data with the real-time wind direction and wind speed data into an integrated set, and input the integrated set into the prediction model to predict the fire risk area; Early warning module: It is used to trigger the corresponding alarm mechanism according to the fire risk area predicted by the prediction model and the type of fire source.

Images