High-rise building fire monitoring system and method based on distributed sensors

Abstract

The invention discloses a high-rise building fire monitoring system and method based on a distributed sensor, and relates to the technical field of intelligent fire fighting, and the method comprises the steps: carrying out the clustering through a K-means algorithm, carrying out the optimization through chaotic mapping, carrying out the feature extraction through SPIHT, carrying out the fuzzy reasoning, optimizing a membership function and a fuzzy rule, and carrying out the correction through a building material attenuation model. And optimization is carried out through an ACO algorithm, a DLOF model is used to generate a DLOF flow light field, an unmanned aerial vehicle inspection path is planned based on the comprehensive risk weight, and the fire spreading probability is monitored. According to the method, the recognition precision of the fire hazard degree is improved through hybrid modeling of fusion of a chaos optimization clustering algorithm, SPIHT infrared feature extraction and fuzzy reasoning, and the accuracy of fire position tracking and the real-time performance of fire spreading trend monitoring are improved through building material attenuation correction, ACO-centroid fusion positioning and DLOF flow light field construction.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fire protection, and particularly to a high-rise building fire monitoring system and method based on distributed sensors. Background Art

[0002] With the acceleration of the urbanization process, the number and scale of high-rise buildings continue to grow, and the fire safety problem becomes increasingly prominent. Traditional fire monitoring systems mainly rely on single sensing devices such as smoke detectors and temperature sensors. Although the technology is mature, there are problems such as limited monitoring range, high false alarm rate, and response delay in complex high-rise building environments. To improve the accuracy and real-time performance of fire monitoring, the introduction of distributed sensors and machine learning algorithms provides a new research direction for fire monitoring. Algorithms such as K-means clustering and fuzzy inference are introduced into the fire monitoring field to improve the accuracy of fire recognition and response speed. Infrared image processing technology (such as the SPIHT algorithm) shows significant advantages in feature extraction, while optimization methods such as ant colony optimization (ACO) algorithm and chaotic mapping provide new ideas for fire source location and path planning in complex environments.

[0003] However, the current related technologies still have deficiencies. The traditional ant colony algorithm (ACO) lacks the integration of the building structure attenuation model when solving the fire location, resulting in systematic deviation in Euclidean distance calculation. In the field of three-dimensional field modeling, existing methods often independently process physical quantities such as temperature field and CO concentration field, lacking a dynamic fusion mechanism based on DLOF, resulting in insufficient spatio-temporal resolution of fire spread prediction. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a high-rise building fire monitoring system and method based on distributed sensors, which solves the problem that the traditional ant colony algorithm (ACO) lacks the integration of the building structure attenuation model when solving the fire location, resulting in systematic deviation in Euclidean distance calculation. In the field of three-dimensional field modeling, existing methods often independently process physical quantities such as temperature field and CO concentration field, lacking a dynamic fusion mechanism based on DLOF, resulting in insufficient spatio-temporal resolution of fire spread prediction.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides a high-rise building fire monitoring system based on distributed sensors, which includes,

[0008] The collection and recognition module is used to collect fixed and infrared image data, cluster the fixed data using the K-means algorithm and optimize it through chaotic mapping, extract features from the infrared image data using SPIHT, generate a multi-dimensional feature vector by combining the optimization results, calculate the probability distribution curvature and weights through the Fisher information matrix, construct a BPNN model and an LSSVM model, optimize them through an optimization algorithm to obtain an initial BPA, and perturb it through chaotic mapping. Then, use fuzzy inference to optimize the membership function and fuzzy rules, calculate the membership degree and convert it into the degree of fire danger;

[0009] The positioning and monitoring module is used to screen the fixed nodes in high-risk areas, generate a subset of high-risk nodes by combining the BPA, calculate the node trust value, correct the Euclidean distance between nodes using the building material attenuation model, optimize the objective function through the ACO algorithm to output the initial fire location coordinates, locate through the centroid method and fuse them with the initial coordinates by weighting to obtain the final fire location. Extract high-risk features within the cube neighborhood, input them into the DLOF model to generate a DLOF stream field, and then fuse it with the temperature field and CO concentration field to obtain a corrected three-dimensional stream field. Plan the UAV inspection path based on the comprehensive risk weight, update the stream field and the fire situation distribution, and monitor the fire spread probability;

[0010] The response and storage module is used to calculate the fire situation level, set response measures, and upload them to the central database for storage.

[0011] As a preferred solution of the high-rise building fire monitoring system based on distributed sensors according to the present invention, among them: the collection of fixed and infrared image data, clustering of the fixed data using the K-means algorithm and optimization through chaotic mapping, feature extraction of the infrared image data using SPIHT, generation of a multi-dimensional feature vector by combining the optimization results, calculation of the probability distribution curvature and weights through the Fisher information matrix, construction of a BPNN model and an LSSVM model, and optimization through an optimization algorithm to obtain an initial BPA, including:

[0012] Collect fixed data through fixed nodes and infrared image data through mobile nodes;

[0013] Cluster the fixed data using the K-means algorithm and extract its cluster centers;

[0014] Generate a Tent chaotic sequence based on the Tent mapping to define the optimization objective function of the K-means algorithm, optimize the cluster centers, calculate the regional weights based on the building perturbation function, and generate weighted cluster centers through multiplication operations;

[0015] Extract the heat source intensity features in the infrared image data using the SPIHT algorithm, combine with the weighted clustering centers to generate a multi-dimensional feature vector, calculate the probability distribution curvature through the Fisher information matrix, quantify it through the Fisher-Rao distance, and calculate the corresponding weight for each vector in the multi-dimensional feature vector to generate a weighted feature vector;

[0016] Construct a BPNN model, optimize it using the GA algorithm, input the weighted feature vector into the optimized BPNN model to generate the BPNN classification probability, construct an LSSVM model, optimize it using the PSO algorithm, input the weighted feature vector into the optimized LSSVM model to generate the LSSVM classification probability, and splice them into the initial BPA.

[0017] As a preferred embodiment of the high-rise building fire monitoring system based on distributed sensors according to the present invention, wherein: the perturbation is performed through chaotic mapping, fuzzy reasoning is used, the membership function and fuzzy rules are optimized, the membership degree is calculated and converted into the fire danger level, including:

[0018] Use the Logistic mapping to generate a Logistic chaotic sequence, obtain the perturbed BPA, and calculate the final BPA;

[0019] Based on fuzzy reasoning, define the input variables and output variables, and set the initial membership function and fuzzy rules;

[0020] Based on the Henon mapping, adjust the parameters of the initial membership function and the weights of the fuzzy rules respectively, and output the optimized fuzzy reasoning;

[0021] Map the multi-dimensional feature vector and the final BPA to the optimized membership function, combine with the fuzzy rules, calculate the membership degree, and convert it into the fire danger level using the centroid method.

[0022] As a preferred embodiment of the high-rise building fire monitoring system based on distributed sensors according to the present invention, wherein: screening the fixed nodes in the high-risk areas, combining with the BPA to generate a high-risk node subset, calculating the node trust value, correcting the Euclidean distance between nodes using the building material attenuation model, and optimizing the objective function through the ACO algorithm to output the initial fire location coordinates, including:

[0023] Extract the grid point coordinates of the fixed nodes at the deformation extreme points deployed in the high fire danger level areas, and splice them with the final BPA to form a fixed node subset;

[0024] The trust value of fixed nodes is obtained by weighted summation to generate a subset of preferred nodes. The Euclidean distance between nodes is obtained through the Euclidean distance formula and incorporated into the subset of preferred nodes. Based on the building material attenuation model, the Euclidean distance between nodes in the subset of preferred nodes is corrected to obtain the fire source distance. The ACO algorithm is initialized, the objective function is defined, and the initial fire location coordinates are output.

[0025] As a preferred solution of the high-rise building fire monitoring system based on distributed sensors according to the present invention, wherein: obtaining the final fire location by centroid positioning and weighted fusion with the initial coordinates, extracting high-risk features within the cubic neighborhood, inputting into the DLOF model to generate the DLOF flow field, and then fusing with the temperature field and CO concentration field to obtain the corrected three-dimensional flow field, planning the UAV inspection path based on the comprehensive risk weight, updating the flow field and fire situation distribution, and monitoring the fire spread probability, including:

[0026] Based on the initial fire location coordinates, the real-time fire location coordinates are obtained by the centroid method, the final fire location coordinates are obtained by weighted fusion, and through cubic neighborhood search, the data in the fixed nodes are screened and spliced into a context feature vector, which is input into the DLOF model to generate the DLOF flow field;

[0027] The optical flow temperature field and CO concentration field are obtained by the optical flow method and weighted fusion is performed to obtain the fixed node flow field;

[0028] The DLOF flow field and the fixed node flow field are weighted-fused to obtain the corrected flow field, and mapping is performed. The comprehensive risk weight of the grid points is calculated, and the greedy algorithm is used to calculate the optimal path for UAV inspection, collect and update the infrared image data and update the fixed data, generate the updated context feature vector, and update the three-dimensional flow field;

[0029] The updated three-dimensional flow field and the updated fixed data are spliced and mapped through IDW interpolation to generate the real-time fire situation distribution;

[0030] Based on the updated flow field and the final BPA, the fire spread probability is calculated.

[0031] As a preferred solution of the high-rise building fire monitoring system based on distributed sensors according to the present invention, wherein: calculating the fire situation level and setting response measures, including:

[0032] The spread speed in the real-time fire situation distribution is extracted, combined with the final BPA of the fire spread probability, and the fire situation level is obtained by weighted fusion, and response measures are set.

[0033] As a preferred solution of the high-rise building fire monitoring system based on distributed sensors according to the present invention, wherein: uploading to the central database for storage, including:

[0034] Upload the collected real-time fire data and response measures to the central database for storage using blockchain technology.

[0035] In a second aspect, the present invention provides a high-rise building fire monitoring method based on distributed sensors, including:

[0036] Collect fixed and infrared image data, use the K-means algorithm to cluster the fixed data and optimize it through chaotic mapping, use SPIHT to extract features from the infrared image data, generate a multi-dimensional feature vector by combining the optimization results, calculate the probability distribution curvature and weights through the Fisher information matrix, construct a BPNN model and an LSSVM model, optimize them through an optimization algorithm to obtain an initial BPA, and perturb it through chaotic mapping, use fuzzy inference to optimize the membership function and fuzzy rules, calculate the membership degree and convert it into the degree of fire hazard;

[0037] Screen the fixed nodes in the high-risk areas, combine the BPA to generate a subset of high-risk nodes, calculate the node trust value, correct the Euclidean distance between nodes using the building material attenuation model, optimize the objective function through the ACO algorithm to output the initial fire location coordinates, locate through the centroid method and fuse them with the initial coordinates by weighting to obtain the final fire location, extract high-risk features within the cube neighborhood, input them into the DLOF model to generate a DLOF flow field, and then fuse it with the temperature field and CO concentration field to obtain a corrected three-dimensional flow field, plan the UAV patrol path based on the comprehensive risk weight, update the flow field and fire situation distribution, and monitor the fire spread probability;

[0038] Calculate the fire situation level and set response measures, and upload them to the central database for storage.

[0039] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the high-rise building fire monitoring method based on distributed sensors as described in the present invention is implemented.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the high-rise building fire monitoring method based on distributed sensors as described in the present invention is implemented.

[0041] The beneficial effects of the present invention are as follows: The present invention performs clustering by using the K-means algorithm, optimizes it through chaotic mapping, extracts features by using SPIHT, uses fuzzy inference, optimizes the membership function and fuzzy rules, corrects it by using the building material attenuation model, optimizes it through the ACO algorithm, generates the DLOF streamer field by using the DLOF model, plans the UAV inspection path based on the comprehensive risk weight, and monitors the fire spread probability; improves the recognition accuracy of the fire danger level, enhances the accuracy of fire location tracking and the real-time performance of fire spread trend monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0043] Figure 1 It is a schematic diagram of a high-rise building fire monitoring system based on distributed sensors in Embodiment 1.

[0044] Figure 2 It is a flowchart of a high-rise building fire monitoring method based on distributed sensors in Embodiment 1.

[0045] Figure 3 It is a schematic diagram of data processing of a high-rise building fire monitoring system based on distributed sensors in Embodiment 1.

[0046] Figure 4 It is a schematic diagram of emergency response and data storage of a high-rise building fire monitoring system based on distributed sensors in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the above-mentioned objects, features, and advantages of the present invention more obvious and understandable, the following will give a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0048] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0049] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that mutually excludes other embodiments.

[0050] Example 1, referring to Figures 1 to 4 , which is the first embodiment of the present invention. This embodiment provides a high-rise building fire monitoring system based on distributed sensors, including the following steps:

[0051] S1. The collection and recognition module is used to collect fixed and infrared image data, cluster the fixed data using the K-means algorithm, optimize it through chaotic mapping, extract the features of the infrared image data using SPIHT, generate a multi-dimensional feature vector by combining the optimization results, calculate the probability distribution curvature and weight through the Fisher information matrix, construct a BPNN model and an LSSVM model, optimize them through an optimization algorithm to obtain an initial BPA, and perturb it through chaotic mapping. Use fuzzy inference to optimize the membership function and fuzzy rules, calculate the membership degree, and convert it into the degree of fire danger;

[0052] Specifically, collecting fixed and infrared image data, clustering the fixed data using the K-means algorithm, optimizing it through chaotic mapping, extracting the features of the infrared image data using SPIHT, generating a multi-dimensional feature vector by combining the optimization results, calculating the probability distribution curvature and weight through the Fisher information matrix, constructing a BPNN model and an LSSVM model, and optimizing them through an optimization algorithm to obtain an initial BPA, including:

[0053] Determine the extreme deformation points of the high-rise building through the Euler-Bernoulli equation, determine the risk area coordinates (grid point coordinates) of the high-rise building through the regularized grid division method and risk hot spot analysis, deploy multi-modal sensors for the extreme deformation points and risk areas, define the multi-modal sensors as fixed nodes, and introduce drones equipped with infrared cameras, and define the drones as mobile nodes;

[0054] The multi-modal sensor includes temperature and humidity, smoke, and CO concentration sensors;

[0055] Collect fixed data through fixed nodes, including temperature and humidity, smoke concentration, and CO concentration data, collect infrared image data through mobile nodes, and perform normalization processing;

[0056] Cluster the fixed data using the K-means algorithm and extract its cluster center;

[0057] Based on the fixed node, generate a Tent chaotic sequence through the Tent mapping, and define the optimization objective function of the K-means algorithm. The formula is:

[0058] ,

[0059] Where is the optimization objective function value, is the fixed data of the th fixed node, is the th centroid, is an indicator variable (indicating 1 if is assigned to the th centroid, otherwise 0), is the number of clusters in the K-means algorithm, is the number of fixed data;

[0060] Optimize the cluster centers using the K-means algorithm to optimize the objective function and generate optimized cluster centers;

[0061] Calculate the regional weights based on the building disturbance function, and the formula is:

[0062] ,

[0063] where, is the th regional weight, is the building disturbance function of the height of the th fixed node, is a control parameter (set based on the dynamic neighborhood radius), is the reference height (set by referring to the building code based on grid division), is the th fixed node height;

[0064] Generate weighted cluster centers through multiplication operation on the optimized cluster centers and regional weights;

[0065] Use the SPIHT algorithm to extract the heat source intensity features in the infrared image data, combine with the weighted cluster centers, perform stitching and execute the normalization process to generate a multi-dimensional feature vector;

[0066] Calculate the probability distribution curvature based on the multi-dimensional feature vector through the Fisher information matrix;

[0067] Quantify based on the probability distribution curvature through the Fisher-Rao distance, and calculate the corresponding weight for each vector in the multi-dimensional feature vector to generate a weighted feature vector, and the formula is:

[0068] ,

[0069] where, is the weight of the th feature vector, is the result of quantification through the Fisher-Rao distance, is the number of features;

[0070] Collect historical fixed data and historical infrared image data to generate historical weighted feature vectors;

[0071] Construct a BPNN model, including an input layer, a hidden layer, and an output layer;

[0072] Input the historical weighted feature vectors into the BPNN model, initialize the weight matrix and bias vector, and generate GA chromosomes through binary coding;

[0073] Based on the historical weighted feature vectors, define the classification error of the GA chromosomes, which is defined as the fitness of the GA chromosomes;

[0074] Based on the fitness of the GA chromosomes, perform iterations through roulette wheel selection, crossover, and mutation, and stop when the maximum number of iterations is reached (set based on computational resource limitations), output the optimized weight matrix and optimized bias vector, and obtain the optimized BPNN model;

[0075] Based on the historical fixed data and historical infrared image data, set classification categories through EDA, including fire, smoldering, and no fire;

[0076] Input the weighted feature vectors into the optimized BPNN model to generate BPNN classification probabilities;

[0077] Construct an LSSVM model, including an input layer and an output layer;

[0078] Input the historical weighted feature vectors into the LSSVM model, initialize the Gaussian kernel bandwidth and regularization parameter, and splice them into a hyperparameter group;

[0079] Initialize the particle swarm, define the particles as the hyperparameter group, define the cross-validation as the PSO fitness function, update the velocity and position of the particles through the inertia weight model, stop when the maximum number of iterations is reached, output the global optimal hyperparameter group, and obtain the optimized LSSVM model;

[0080] Input the weighted feature vectors into the optimized LSSVM model to generate LSSVM classification probabilities;

[0081] Normalize the LSSVM classification probabilities and BPNN classification probabilities respectively, and splice them into the initial BPA;

[0082] Quantify based on the LSSVM classification probabilities and BPNN classification probabilities through the Fisher-Rao distance, and obtain the BAP weights through the reciprocal of the quantization results.

[0083] The clustering process is optimized by combining the K-means algorithm with chaotic mapping, effectively overcoming the problem that traditional clustering methods are prone to falling into local optima, enhancing the global nature and stability of data classification, providing more accurate basic data for the rapid identification in the early stage of a fire, being able to identify fire risk areas in advance, using the SPIHT algorithm to extract the heat source intensity features from infrared image data, enabling the system to more clearly identify the location and intensity of the fire source. Especially in a complex high-rise building environment, it can accurately locate the fire point, not only improving the efficiency of image data processing but also ensuring the rapid identification of the occurrence of a fire in a changing environment. The feature quantization and weight calculation are carried out by combining the Fisher information matrix and the Fisher-Rao distance, effectively optimizing the feature vector and making the classification result more accurate. On this basis, the BPNN model and the LSSVM model are respectively used to classify fires, smoldering and no-fire, and through intelligent optimization algorithms, training and parameter adjustment are carried out to finally form a high-precision fire prediction model. Especially in complex scenarios, the combination of LSSVM and BPNN can complement each other's weaknesses and improve the robustness and accuracy of the system.

[0084] Furthermore, through chaotic mapping for perturbation, using fuzzy inference, optimizing the membership function and fuzzy rules, calculating the membership degree and converting it into the degree of fire danger, including:

[0085] Multiply the initial BPA and the BPA weight to generate the weighted BPA;

[0086] Based on empirical rules, set the distribution probability threshold, screen the areas where the LSSVM classification probability or the BPNN classification probability is greater than the classification probability threshold, and set them as high-risk areas. Screen the areas where the LSSVM classification probability and the BPNN classification probability are less than the classification probability threshold, and set them as low-risk areas;

[0087] Based on the high-risk areas and low-risk areas, set the perturbation amplitude, use the Logistic mapping to generate the Logistic chaotic sequence, and perturb the weighted BPA to obtain the perturbed BPA. The formula is:

[0088] ,

[0089] ,

[0090] where, and are the perturbed BPA, where is the BPNN classification probability, is the LSSVM classification probability, is the perturbation amplitude, is the th Logistic chaotic sequence;

[0091] Calculate the perturbed BPA to obtain the final BPA. The formula is:

[0092] B,

[0093] where, is the final BPA, is a subset of classification categories, is a classification category, representing fire, smoldering, and no fire, means no intersection, is a subset of classification categories;

[0094] Based on fuzzy inference, define the fixed data, infrared image data, and perturbed BPA as input variables, and define the classification category and fire risk level (low, medium, and high) as output variables. Set the initial membership function (such as a triangular membership function) and fuzzy rules (for example, if the temperature is high, the smoke concentration is high, and the CO concentration is high, then it is abnormal and judged as a fire);

[0095] Generate a Henon chaotic sequence based on the Henon map, and adjust the parameters of the initial membership function and the weights of the fuzzy rules respectively. Output the optimized fuzzy inference (including the optimized membership function and optimized fuzzy rules). The formula is:

[0096] ,

[0097] ,

[0098] where, is the th membership function, is the th Henon chaotic sequence, is the adjustment factor, is the optimized membership function, is the th optimized fuzzy rule weight, is the th initial fuzzy rule weight, is the parameter of the initial membership function;

[0099] Based on the 30-30-30 rule, verify the fixed data and output the verification result (if the temperature is greater than 30, the humidity is greater than 30, and the CO is greater than 50, the output verification result is 1, otherwise it is 0);

[0100] Adjust the risk threshold for high-risk areas to generate an adjusted threshold. The formula is:

[0101] ,

[0102] ,

[0103] ,

[0104] wherein, 、 and are the adjusted temperature and humidity and CO concentration thresholds respectively, is the adjustment factor;

[0105] Map the multi-dimensional feature vector and the final BPA to the optimized membership function, combine the fuzzy rules, and calculate the membership degree. The formula is:

[0106] ,

[0107] wherein, is the output membership degree, is the number of fuzzy rules, is the smoke concentration, is the final BPA;

[0108] Use the centroid method to convert the membership degree into the fire danger level;

[0109] Screen the fire danger level greater than the low fire danger level threshold as the low fire danger level. Screen the fire danger level greater than or equal to the low fire danger level threshold and less than the high fire danger level threshold as the medium fire danger level. Screen the fire danger level greater than or equal to the high fire danger level threshold as the high fire danger level (the high and low thresholds are both set by ROC curve analysis).

[0110] By performing multiplication operations on the initial BPA and BPA weights to generate weighted BPA and using Logistic chaotic mapping for perturbation, the accuracy and diversity of the system's fire risk assessment for different regions can be enhanced, the randomness of the system can be improved, making the fire risk assessment more robust, effectively avoiding the local optimum problem caused by overly regular data. By setting the thresholds of the LSSVM and BPNN classification probabilities, different regions can be divided into high-risk or low-risk regions, accurately defining the dangerous areas of the fire and ensuring that only the truly high-risk areas are identified. This not only improves the accuracy of the monitoring system but also effectively reduces the waste of system resources, ensuring focused attention and resource allocation for high-risk regions. By using the fixed data, infrared image data, and perturbed BPA as input variables and combining the initial membership function and fuzzy rules, the system can adjust the membership degree according to the actual situation and calculate the fire risk level. The optimized membership function and fuzzy rules can more accurately reflect the possibility and danger level of the fire occurrence, thus improving the efficiency of fire monitoring and response, enabling the fire monitoring system to handle fuzziness and uncertainty, and enhancing its adaptability and accuracy in complex environments. By using the Henon chaotic mapping to adjust the parameters of the initial membership function and the weights of the fuzzy rules, the fuzzy reasoning process can be further optimized, enhancing the flexibility of the system under different environmental conditions. Through the fuzzy reasoning rules optimized by chaos, more refined fire risk assessment can be achieved, thereby improving the response speed and decision-making quality of the entire monitoring system.

[0111] S2. A positioning and monitoring module, which is used to screen the fixed nodes in high-risk areas, generate a high-risk node subset by combining with BPA, calculate the node trust value, correct the Euclidean distance between nodes using the building material attenuation model, optimize the objective function through the ACO algorithm to output the initial fire location coordinates, obtain the final fire location by centroid positioning and weighted fusion with the initial coordinates, extract high-risk features within the cube neighborhood, input them into the DLOF model to generate a DLOF flow field, and then fuse it with the temperature field and CO concentration field to obtain a corrected three-dimensional flow field, plan the UAV patrol path based on the comprehensive risk weight, update the flow field and fire situation distribution, and monitor the fire spread probability;

[0112] Specifically, screening the fixed nodes in high-risk areas, generating a high-risk node subset by combining with BPA, calculating the node trust value, and correcting the Euclidean distance between nodes using the building material attenuation model, and optimizing the objective function through the ACO algorithm to output the initial fire location coordinates, including:

[0113] Extract the grid point coordinates of the fixed nodes at the deformation extreme points deployed in high-fire-risk areas, and splice them into a fixed node subset by combining with the final BPA;

[0114] Extract the historical positioning stability of fixed nodes (collect historical data and calculate the positioning error of fixed nodes through root mean square error) and the remaining energy percentage (receive the current energy and initial energy through a digital interface and calculate the ratio), and obtain the trust value of fixed nodes through weighted summation;

[0115] Filter out the fixed nodes whose trust values of fixed nodes are greater than the trust value threshold (set based on historical data analysis) to generate a subset of preferred nodes;

[0116] Based on the subset of preferred nodes, obtain the Euclidean distance between nodes through the Euclidean distance formula and incorporate it into the subset of preferred nodes;

[0117] Based on the building material attenuation model, correct the Euclidean distance between nodes in the subset of preferred nodes to obtain the fire source distance. The formula is:

[0118] ,

[0119] where, is the fire source distance of the th fixed node, is a constant, is the path loss factor, is the signal strength of the th fixed node (obtained based on RSSI), is the wall attenuation coefficient, is the th fixed node's wall thickness (obtained based on the BIM database);

[0120] Based on the fire source distance and the grid point coordinates of the fixed nodes in the subset of preferred nodes, initialize the ACO algorithm and define the objective function. The formula is:

[0121] ,

[0122] where, is the objective function value, is the trust value of the th fixed node, is the th fixed node's estimated fire source distance (obtained based on signal strength), is the number of fixed nodes;

[0123] Update the objective function value through the pheromone update mechanism and stop when the maximum number of iterations is reached (set based on experimental verification), and output the initial fire location coordinates.

[0124] By screening the fixed nodes deployed in high fire-risk areas, it is possible to ensure that the monitoring system conducts efficient monitoring of key areas. Using BPA to generate a subset of high-risk nodes helps optimize the aggregation and selection of information, thus providing more accurate fire warnings. The calculation of trust values is based on a weighted summation method of historical positioning errors and the percentage of remaining energy, which can accurately evaluate the reliability of each node during the fire monitoring process. The Euclidean distance formula provides a simple and effective way to quantify the spatial relationship between nodes. Combining with the building material attenuation model for correction can eliminate signal errors caused by building walls, obstacles, etc. The ACO algorithm is applied to fire source location. By optimizing the objective function, the fire source location can be accurately calculated, improving the accuracy of fire location estimation. Integrating fire source location with the UAV patrol path planning can achieve comprehensive monitoring and real-time update of fire information, helping to improve the emergency response efficiency and reduce the losses caused by fire spread.

[0125] Furthermore, the final fire location is obtained by centroid positioning and weighted fusion with the initial coordinates. High-risk features are extracted within the cubic neighborhood, and the input DLOF model generates the DLOF flow field. Then, it is fused with the temperature field and CO concentration field to obtain the corrected three-dimensional flow field. Based on the comprehensive risk weight, the UAV patrol path is planned, the flow field and fire situation distribution are updated, and the fire spread probability is monitored, including:

[0126] Based on the initial fire location coordinates, the UAV is used to collect real-time infrared image data and extract the real-time heat source intensity features, and the real-time fire location coordinates are obtained by the centroid method;

[0127] For the initial fire location coordinates and the real-time fire location coordinates, the final fire location coordinates are obtained through weighted fusion;

[0128] Based on the final fire location coordinates, through cubic neighborhood search, a cubic region with a radius of N is defined, the high-risk areas in this region are extracted, and the data in the fixed nodes are screened (setting the fixed node data threshold based on statistical analysis and screening the data greater than the fixed node data threshold), and then stitched into a context feature vector;

[0129] Build a DLOF model, including an input layer, a core calculation layer, and an output layer;

[0130] Use the CFD fire simulation dataset of NIST to train the DLOF model;

[0131] Input the context feature vector into the DLOF model, output the DLOF value, set the DLOF value threshold based on the fixed threshold method, and generate the DLOF flow field through binarization;

[0132] Extract the temperature and CO concentration data of the fixed nodes, and obtain the temperature field and CO concentration field through the central difference method;

[0133] Based on the temperature field and the CO concentration field, the optical flow temperature field and the CO concentration field are obtained by using the optical flow method respectively, and the weighted fusion is carried out to obtain the fixed-node optical flow field;

[0134] The DLOF optical flow field and the fixed-node optical flow field are weighted and fused to obtain the corrected optical flow field;

[0135] The corrected optical flow field is mapped to generate a three-dimensional optical flow field, and the formula is:

[0136] ,

[0137] where, is the three-dimensional optical flow field, is the coordinate mapping function, are the grid point coordinates of the high-risk area, is the corrected optical flow field;

[0138] Based on the fire danger level and the three-dimensional optical flow field, the comprehensive risk weight of the grid points is calculated, and the formula is:

[0139] ,

[0140] where, is the comprehensive risk weight of the th grid point, is the risk weight of the th grid point (obtained by weighted summation based on the final BPA, temperature, smoke concentration, and CO concentration data), is the optical flow influence coefficient;

[0141] Based on the comprehensive risk weight, the greedy algorithm is used to calculate the optimal path for the drone inspection, and the drone is used to collect and update the infrared image data, the fixed data is collected and updated through the fixed nodes, and the updated context feature vector is generated by screening;

[0142] Based on the updated infrared image data and the updated context feature vector, the three-dimensional optical flow field is updated, and the formula is:

[0143] ,

[0144] where, is the updated optical flow field, are the DLOF model parameters, is the time of the updated infrared image data, is of the updated infrared image data, is the updated context feature vector;

[0145] The updated 3D flow field of fire and the updated fixed data are spliced, and mapped through IDW interpolation to generate the real-time fire distribution;

[0146] Based on the updated flow field of fire and the final BPA, calculate the fire spread probability. The formula is:

[0147] ,

[0148] where, is the fire spread probability, is the weighting coefficient, is the final BPA.

[0149] By combining the centroid method with the real-time infrared image data collected by the UAV and the initial fire location, the exact location of the fire can be dynamically calculated, effectively eliminating the uncertainty of the fire location caused by measurement errors, sensor biases, etc., improving the accuracy of fire location. The DLOF model can effectively describe the spatial variation of the fire by dynamically tracking the temperature field and the CO concentration field. By weighted fusion of the initial fire location coordinates and the real-time fire location coordinates, historical data and real-time data can be effectively combined to improve the accuracy of fire location. Using the cube neighborhood search technology, high-risk areas can be screened out from the spatial data and further analyzed based on the risk weights, helping the system to take targeted emergency measures in the early stage of the fire. By weighted fusion of the DLOF flow field of fire and the fixed-node flow field of fire, the generated 3D flow field of fire can intuitively reflect the dynamic process of fire occurrence and spread, not only providing support for the real-time update of the fire location, but also providing a data basis for the planning of the UAV inspection path, ensuring that the inspection path can cover high-risk areas and improving the inspection efficiency. The inspection path planning based on the comprehensive risk weights and the greedy algorithm improves the inspection efficiency of the UAV in the complex fire environment. By calculating the real-time fire spread probability, the trend of fire spread can be predicted, providing important decision-making support for emergency response and resource allocation.

[0150] S3. The response storage module is used to calculate the fire level and set response measures, and upload them to the central database for storage;

[0151] Specifically, calculating the fire level and setting response measures includes:

[0152] Extract the spread speed in the real-time fire distribution, combine it with the final BPA of the fire spread probability, and obtain the fire level through weighted fusion;

[0153] Set the low and high thresholds of the fire situation level based on the analysis of fire situation data. Screen the fire situation levels less than the low threshold of the fire situation level and set them as low levels. Screen the fire situation levels greater than or equal to the low threshold of the fire situation level and less than the high threshold of the fire situation level and set them as medium levels. Screen the fire situation levels greater than or equal to the high threshold of the fire situation level and set them as high levels;

[0154] Based on the classification of the fire situation level, set response measures, including increasing the sampling frequency of fixed nodes in high-risk areas for low levels, collecting real-time fire situation data, and reporting it to the building management center. For medium levels, trigger an alarm in high-risk areas and activate the local area fire sprinkler. For high levels, trigger a building-wide alarm and start the building-wide fire sprinkler.

[0155] The weighted fusion of the fire spread speed and BPA helps to more accurately predict the development of the fire, can give an accurate prediction of the fire danger level in a shorter time, significantly improve the accuracy of early fire warning. The classification of the fire situation level can help the emergency response system take different actions according to the actual risk of the fire. Setting response measures at different fire situation levels can effectively improve the efficiency and accuracy of fire emergency handling, and significantly enhance the fire safety of buildings.

[0156] Furthermore, upload it to the central database for storage, including:

[0157] Upload the collected real-time fire situation data and response measures to the central database and store them through blockchain technology.

[0158] Uploading the fire data and response measures to the central database and storing them through blockchain technology not only ensures the long-term storage and immutability of the data, but also provides convenience for post-fire data analysis and traceability. The decentralized feature of blockchain technology ensures the transparency and credibility of the data, preventing the risks of human tampering and data loss.

[0159] This embodiment also provides a high-rise building fire monitoring method based on distributed sensors, including:

[0160] Collect fixed and infrared image data, use the K-means algorithm to cluster the fixed data and optimize it through chaotic mapping, use SPIHT to extract features from the infrared image data, generate a multi-dimensional feature vector by combining the optimization results, calculate the probability distribution curvature and weight through the Fisher information matrix, construct a BPNN model and an LSSVM model, optimize them through an optimization algorithm to obtain an initial BPA, and perturb it through chaotic mapping, use fuzzy reasoning, optimize the membership function and fuzzy rules, calculate the membership degree and convert it into the degree of fire danger;

[0161] Screen fixed nodes in high-risk areas, combine with BPA to generate a subset of high-risk nodes, calculate node trust values, use the building material attenuation model to correct the Euclidean distance between nodes, optimize the objective function through the ACO algorithm to output the initial fire location coordinates, locate through the centroid method and fuse them with the initial coordinates with weights to obtain the final fire location, extract high-risk features within the cube neighborhood, input the DLOF model to generate the DLOF light flow field, and then fuse it with the temperature field and CO concentration field to obtain the corrected three-dimensional light flow field, plan the UAV inspection path based on the comprehensive risk weight, update the light flow field and fire situation distribution, and monitor the fire spread probability;

[0162] Calculate the fire situation level and set response measures, and upload them to the central database for storage.

[0163] This embodiment also provides a computer device, which is applicable to the situation of a high-rise building fire monitoring system based on distributed sensors, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the high-rise building fire monitoring method based on distributed sensors as proposed in the above embodiment.

[0164] This computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, operator network, NFC (Near Field Communication) or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0165] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for monitoring high-rise building fires based on distributed sensors as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0166] In summary, the present invention performs clustering using the K-means algorithm, optimizes it through chaotic mapping, extracts features using SPIHT, uses fuzzy inference, optimizes the membership function and fuzzy rules, corrects using the building material attenuation model, optimizes through the ACO algorithm, generates the DLOF stream field using the DLOF model, plans the UAV inspection path based on the comprehensive risk weight, and monitors the fire spread probability; improves the recognition accuracy of the fire danger level, enhances the accuracy of fire location tracking, and the real-time performance of fire spread trend monitoring.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A high-rise building fire monitoring system based on distributed sensors, characterized in that: Including A collection and recognition module, which is used to collect fixed and infrared image data, cluster the fixed data using the K-means algorithm and optimize it through chaotic mapping, extract the features of the infrared image data using SPIHT, generate a multi-dimensional feature vector by combining the optimization results, calculate the probability distribution curvature and weights through the Fisher information matrix, construct a BPNN model and an LSSVM model, optimize them through an optimization algorithm to obtain an initial BPA, and perturb it through chaotic mapping, use fuzzy inference, optimize the membership function and fuzzy rules, calculate the membership degree and convert it into the degree of fire hazard; A positioning and monitoring module, which is used to screen the fixed nodes in the high-risk area, generate a subset of high-risk nodes by combining the BPA, calculate the node trust value, correct the Euclidean distance between nodes using the building material attenuation model, optimize the objective function through the ACO algorithm to output the initial fire location coordinates, locate through the centroid method and fuse them with the initial coordinates by weighting to obtain the final fire location, extract high-risk features within the cube neighborhood, input them into the DLOF model to generate a DLOF streamer field, and then fuse it with the temperature field and CO concentration field to obtain a corrected three-dimensional streamer field, plan the UAV patrol path based on the comprehensive risk weight, update the streamer field and the fire situation distribution, and monitor the fire spread probability; A response and storage module, which is used to calculate the fire situation level and set response measures, and upload them to the central database for storage.

2. The high-rise building fire monitoring system based on distributed sensors according to claim 1, wherein: The collection of fixed and infrared image data, clustering the fixed data using the K-means algorithm and optimizing it through chaotic mapping, extracting the features of the infrared image data using SPIHT, generating a multi-dimensional feature vector by combining the optimization results, calculating the probability distribution curvature and weights through the Fisher information matrix, constructing a BPNN model and an LSSVM model, and optimizing them through an optimization algorithm to obtain an initial BPA, includes: Collecting fixed data through fixed nodes and infrared image data through mobile nodes; Clustering the fixed data using the K-means algorithm and extracting its cluster centers; Generating a Tent chaotic sequence based on the Tent mapping to define the optimization objective function of the K-means algorithm, optimizing the cluster centers, calculating the regional weights based on the building disturbance function, and generating weighted cluster centers through multiplication; Using the SPIHT algorithm to extract the heat source intensity features in the infrared image data, combining with the weighted cluster centers to generate a multi-dimensional feature vector, calculating the probability distribution curvature through the Fisher information matrix, quantifying it through the Fisher-Rao distance, and calculating the corresponding weights for each vector in the multi-dimensional feature vector to generate a weighted feature vector; Constructing a BPNN model, optimizing it using the GA algorithm, inputting the weighted feature vector into the optimized BPNN model to generate the BPNN classification probability, constructing an LSSVM model, optimizing it using the PSO algorithm, inputting the weighted feature vector into the optimized LSSVM model to generate the LSSVM classification probability, and splicing them into an initial BPA.

3. The high-rise building fire monitoring system based on distributed sensors according to claim 2, wherein: The perturbation by chaotic mapping, using fuzzy inference, optimizing membership functions and fuzzy rules, calculating membership degrees and converting them into the degree of fire danger, includes: Using the Logistic mapping to generate a Logistic chaotic sequence, obtaining the perturbed BPA, and calculating the final BPA; Based on fuzzy inference, defining input variables and output variables, and setting initial membership functions and fuzzy rules; Based on the Henon mapping, adjusting the parameters of the initial membership functions and the weights of the fuzzy rules respectively, and outputting the optimized fuzzy inference; Mapping the multi-dimensional feature vector and the final BPA to the optimized membership functions, combining the fuzzy rules, calculating the membership degrees, and converting them into the degree of fire danger using the centroid method.

4. The high-rise building fire monitoring system based on distributed sensors according to claim 3, characterized in that: The screening of fixed nodes in high-risk areas, combining the BPA to generate a subset of high-risk nodes, calculating the node trust values, correcting the Euclidean distance between nodes using the building material attenuation model, and optimizing the objective function through the ACO algorithm to output the initial fire location coordinates, includes: Extracting the grid point coordinates of the fixed nodes at the extreme deformation points deployed in the high fire danger degree area, and splicing them into a subset of fixed nodes in combination with the final BPA; Obtaining the fixed node trust values through weighted summation, generating a subset of preferred nodes, obtaining the Euclidean distance between nodes through the Euclidean distance formula, and incorporating it into the subset of preferred nodes; correcting the Euclidean distance between nodes in the subset of preferred nodes based on the building material attenuation model to obtain the fire source distance, initializing the ACO algorithm, defining the objective function, and outputting the initial fire location coordinates.

5. The high-rise building fire monitoring system based on distributed sensors according to claim 4, wherein: The positioning by the centroid method and weighted fusion with the initial coordinates to obtain the final fire location, extracting high-risk features within the cube neighborhood, inputting them into the DLOF model to generate a DLOF light flow field, and then fusing it with the temperature field and CO concentration field to obtain a corrected three-dimensional light flow field, planning the UAV patrol path based on the comprehensive risk weight, updating the light flow field and the fire situation distribution, and monitoring the fire spread probability, includes: Based on the initial fire location coordinates, obtaining the real-time fire location coordinates by the centroid method, obtaining the final fire location coordinates by weighted fusion, and screening the data in the fixed nodes through cube neighborhood search, splicing them into a context feature vector, and inputting it into the DLOF model to generate a DLOF light flow field; Using the optical flow method to obtain the optical flow temperature field and CO concentration field, and performing weighted fusion to obtain the fixed node light flow field; Weightedly fusing the DLOF light flow field and the fixed node light flow field to obtain a corrected light flow field, and performing mapping, calculating the comprehensive risk weight of the grid points, using the greedy algorithm to calculate the optimal path for the UAV patrol, collecting and updating the infrared image data and the updated fixed data, generating an updated context feature vector, and updating the three-dimensional light flow field; Splicing the updated three-dimensional light flow field and the updated fixed data, and performing mapping through IDW interpolation to generate the real-time fire situation distribution; Calculating the fire spread probability based on the updated light flow field and the final BPA.

6. The high-rise building fire monitoring system based on distributed sensors according to claim 5, characterized in that: The calculating of the fire situation level and setting of response measures, includes: Extracting the spread speed in the real-time fire situation distribution, combining the final BPA of the fire spread probability, obtaining the fire situation level through weighted fusion, and setting response measures.

7. The high-rise building fire monitoring system based on distributed sensors according to claim 6, characterized in that: The uploading to the central database for storage includes: Uploading the collected real-time fire data and response measures to the central database for storage through blockchain technology.

8. A method for monitoring high-rise building fires based on distributed sensors, based on the high-rise building fire monitoring system based on distributed sensors according to any one of claims 1 to 7, characterized in that: Including,[[]]END]] Collecting fixed and infrared image data, clustering the fixed data using the K-means algorithm and optimizing it through chaotic mapping, extracting features of the infrared image data using SPIHT, generating a multi-dimensional feature vector by combining the optimization results, calculating the probability distribution curvature and weights through the Fisher information matrix, constructing a BPNN model and an LSSVM model, optimizing them through an optimization algorithm to obtain an initial BPA, perturbing it through chaotic mapping, using fuzzy inference to optimize the membership function and fuzzy rules, calculating the membership degree and converting it into the fire danger level; Screening the fixed nodes in the high-risk areas, generating a high-risk node subset by combining the BPA, calculating the node trust value, correcting the Euclidean distance between nodes using the building material attenuation model, optimizing the objective function through the ACO algorithm to output the initial fire location coordinates, locating through the centroid method and weighted fusion with the initial coordinates to obtain the final fire location, extracting high-risk features within the cube neighborhood, inputting them into the DLOF model to generate a DLOF flow field, and then fusing it with the temperature field and CO concentration field to obtain a corrected three-dimensional flow field, planning the UAV patrol path based on the comprehensive risk weight, updating the flow field and fire situation distribution, and monitoring the fire spread probability; Calculating the fire level and setting response measures, and uploading them to the central database for storage.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the high-rise building fire monitoring method based on distributed sensors described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the high-rise building fire monitoring method based on distributed sensors described in claim 8.

Images