A fire investigation and survey system based on cloud computing

Abstract

The present invention discloses a fire investigation and survey system based on cloud computing, comprising an edge computing module, a cloud computing module, a storage module, a data analysis and processing module, an investigation data fusion module and a display processing module, wherein: the cloud computing module is used to perform a second investigation and analysis based on the real-time collected fire scene data and the first investigation and analysis results, so as to obtain the second investigation and analysis results; the data analysis and processing module, data collection and analysis are all processed in parallel by deploying multiple deep learning models on cloud computing nodes, and the investigation and analysis results are recorded as the third investigation and analysis results; the investigation data fusion module is used to perform data fusion on the second investigation and analysis results and the third investigation and analysis results, and output the final investigation and analysis results; the present invention improves the accuracy of fire investigation by performing different deep learning model analyses and fusion algorithm analyses on the data collected at different stages of fire occurrence.

Description

Technical Field

[0001] The present invention belongs to the field of fire investigation and surveying based on cloud computing, and specifically refers to a fire investigation and surveying system based on cloud computing. Background Art

[0002] Fire, as a sudden and destructive natural disaster, poses a serious threat to human society and property safety. With the acceleration of urbanization, the frequency and scope of fires are expanding, and the importance of fire investigation and survey work is becoming increasingly prominent.

[0003] At present, there is little application of intelligent technology in the fire investigation stage when a fire is about to end or after it ends. In particular, in determining the fire point, fire cause and fire spread process, traditional investigation methods are still mainly relied on. This is not only inefficient, but also due to the influence of human factors, it is often difficult to accurately determine the fire point, fire cause and fire spread process, thus affecting the accurate analysis of the fire investigation and the accuracy of the investigation results. Therefore, the research and development and application of intelligent technologies, such as cloud computing and deep learning, to improve the accuracy and efficiency of fire investigations is an issue that needs to be urgently addressed in the field of fire investigation. Summary of the invention

[0004] In order to solve the problems in the above-mentioned prior art that the accurate analysis of fire investigation and the effectiveness of investigation results cannot be achieved, cloud computing and deep learning are not used, and the investigation and analysis before and after the fire is extinguished are not combined, the present invention proposes a fire investigation and survey based on cloud computing to improve the above-mentioned problems.

[0005] The specific application is as follows:

[0006] A fire investigation and survey system based on cloud computing, comprising an edge computing module, a cloud computing module, a storage module, a data analysis and processing module, a survey data fusion module and a display processing module;

[0007] The storage module is used to store the data generated by the fire investigation and survey system;

[0008] The edge computing module is used to perform a first investigation and analysis based on the real-time collected fire scene data, and to package and transmit the real-time collected fire scene data and the first investigation and analysis results to the cloud computing module, wherein the fire scene data is data collected when the fire is about to be extinguished;

[0009] The cloud computing module is used to perform a second investigation and analysis based on the real-time collected fire scene data and the first investigation and analysis results, obtain the second investigation and analysis results, and transmit the second investigation and analysis results to the edge computing module;

[0010] The data analysis and processing module is used to collect data after the fire is extinguished, and to process the collected data after the fire is extinguished by deploying multiple deep learning models on cloud computing nodes, and the processing and analysis results are recorded as the third investigation and analysis results;

[0011] The survey data fusion module is used to fuse the second survey analysis results and the third survey analysis results, and output the final survey analysis results;

[0012] The display processing module is used to obtain the second survey and analysis results from the cloud computing module, the first survey and analysis results from the edge computing module, the third survey and analysis results from the data analysis processing module, and the final survey and analysis results from the survey data fusion module, and display the obtained survey and analysis results.

[0013] Further, the data analysis and processing module includes a data acquisition submodule, a detection submodule, a fire investigation deep learning submodule and a fire investigation pre-configuration submodule;

[0014] The data acquisition submodule is used to obtain data from the scene after the fire through sensors and cameras, the data including data after the fire is extinguished, and convert the data to generate data that can be recognized and analyzed by computer programs, the data including historical acquisition data and real-time acquisition data;

[0015] The data acquisition submodule includes a sensor unit, an image processing unit and a video recording processing unit;

[0016] The sensor unit is used to collect information through sensors after the fire is extinguished, including real-time sensor data and historically collected sensor data;

[0017] The image processing unit is used to process the images collected after the fire is extinguished, including real-time images and historically collected images;

[0018] The video recording processing unit is used to process the video recording collected after the fire is extinguished, including real-time video recording and historically collected video recording;

[0019] By acquiring the corresponding sensor type, camera type, interface type and data format of the sensor unit, the image processing unit and the video recording processing unit, and converting the sensor type, the interface type and the data format into data recognizable by the computer program, the sensor unit controls different types of sensors and uses the sensors to collect data, the image processing unit and the video recording processing unit control different types of cameras and use the cameras to obtain image data or video recording, and the sensors and cameras transmit data through their respective interface types;

[0020] The fire investigation deep learning submodule is used to train the fire investigation deep learning network model based on the data of the acquisition submodule to obtain a fire investigation and survey output model;

[0021] The fire investigation preconfiguration submodule is used to train a preconfigured deep learning network model to obtain a sensor preconfiguration model, a camera preconfiguration model, and a fire trace relationship model, and use the camera preconfiguration model, the sensor preconfiguration model, and the fire trace relationship model for the programmatic collection of fire trace data and the generation of fire trace relationships.

[0022] Furthermore, the types of the sensors specifically include temperature sensors, smoke sensors, heat release rate sensors and CO concentration sensors;

[0023] The data format is used to convert the data collected by the sensor and the camera into a data type format recognizable by a computer program;

[0024] The camera is used to capture images and video data, wherein the images and video data include trace data after the fire is extinguished;

[0025] The historically collected data includes historical data of other fire scenes collected by sensors and cameras;

[0026] The historical data at least includes trace data after the fire is extinguished.

[0027] Furthermore, the interface types include:

[0028] A sensor interface acquisition unit, used to acquire the type of the sensor interface, the sensor interface being used for transmission after collecting data;

[0029] A camera image interface acquisition unit, used to acquire the camera interface type;

[0030] A protocol acquisition unit, used to acquire the communication protocol of the sensor interface and the camera interface;

[0031] A data acquisition unit, configured to acquire the data acquired from the sensor interface and the camera interface by establishing different communication protocols according to the protocol acquisition unit, and store the acquired data in the storage module;

[0032] The fire investigation deep learning submodule includes:

[0033] A model building unit is used to characterize the data of the data acquisition submodule through feature engineering to generate training set data and test set data, train the deep learning network model through the training set data, and obtain a fire investigation and survey output model;

[0034] The model output unit is used to use the data and historical data collected by the data collection submodule in real time as input to the fire investigation and survey output model, and the fire investigation and survey output model outputs the fire point, fire cause and fire spread process.

[0035] Furthermore, the fire investigation pre-configuration submodule includes:

[0036] A sensor pre-configuration unit, configured to train a deep learning network model according to sensor data collected with a type label to obtain the sensor pre-configuration model, wherein the sensor data collected with the type label is subjected to feature extraction and labeling through a deep learning neural network algorithm;

[0037] A camera pre-configuration unit is used to train a deep learning network model according to camera acquisition data with a type label to obtain the camera pre-configuration model, wherein the type label performs feature extraction and labeling on the image acquired by the camera through a deep learning neural network algorithm;

[0038] A fire trace relationship unit is used to construct a fire trace relationship based on the fire trace data collected by the sensor and the camera, train a deep learning network model, and obtain a fire trace relationship model;

[0039] The fire trace relationships include the relationship between the carbonization rate and the direction of fire spread, the relationship between the concrete component peeling trace and the fire temperature, the relationship between the glass thermal cracking trace and the fire temperature, the relationship between the glass thermal softening trace and the fire temperature, the relationship between the glass thermal melting trace and the fire temperature, the relationship between the thermal expansion of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the fire spread change and the distance from the fire ignition point, the relationship between the fire intensity change and the fire cause, and the relationship between the fire intensity change and the distance from the fire ignition point;

[0040] The fire trace data includes a first trace, a second trace, a third trace and a fourth trace;

[0041] The first traces include traces of the fire starting point, traces of the fire spreading, traces of the fire cause and traces of the fire nature;

[0042] The second traces include traces formed by glass, traces formed by metal, traces formed by wood, and traces of flammable liquids;

[0043] The third traces include carbonization traces, ash traces, smoke traces, collapse traces, burning traces, melting traces, discoloration traces, deformation traces, cracking traces, electric heat melting traces, friction traces, separation and displacement traces, and human body burn marks;

[0044] The fourth trace includes a spreading trace and an intensity trace;

[0045] The fire trace relationship is modeled through deep learning of machine learning.

[0046] Furthermore, the edge computing module is used to perform edge computing on the real-time data of the data acquisition unit, and the result of the calculation and analysis is recorded as the first investigation and analysis data, and the first investigation and analysis data is transmitted to the cloud computing module, and the real-time data is the data collected when the fire is about to be extinguished;

[0047] The specific analysis process of the first survey analysis data includes:

[0048] L1: Collect data from the fire scene in real time through sensors and cameras of the data acquisition unit;

[0049] L2: Cleaning, denoising, and formatting preprocessing operations are performed on the collected fire scene data at the edge node;

[0050] L3: Use the neural network learning InceptionV3 algorithm on the edge computing node to analyze data in real time.

[0051] Furthermore, the cloud computing module includes a data management submodule and a second analysis and investigation submodule;

[0052] The data management submodule is used to retrieve historical data under the corresponding fire environment based on the real-time data collected at the fire scene and the first investigation and analysis data results, wherein the historical data includes sensor data, image data and video data;

[0053] The second analysis and investigation submodule is used to perform a secondary investigation and analysis based on the real-time data collected at the fire scene, the first investigation and analysis data results and historical data, including:

[0054] S1: Perform cleaning, denoising, and formatting preprocessing operations on the real-time fire scene data, the first investigation and analysis data results, and the historical data collected at the cloud computing node;

[0055] S2: Use the neural network to learn the YOLOv7 algorithm on the cloud computing node to analyze the data in real time;

[0056] S3: The results of the analysis and investigation are used in a display processing module for display.

[0057] Furthermore, the survey data fusion module calculates the second survey analysis result and the third survey analysis result by using a cosine similarity algorithm. The cosine similarity algorithm is:

[0058]

[0059] where x i ,y i They are the i-th vectors corresponding to the second survey and analysis results and the third survey and analysis results respectively. The calculation result is M, and the threshold is set to Q. If M is greater than Q, it means that the survey and analysis results are correct and the error rate is very low. If M is less than Q, it means that the survey and analysis results are wrong and the error rate is very high. Further investigation and analysis is needed until M is greater than Q and the survey is terminated.

[0060] The beneficial effects of a fire investigation and survey system based on cloud computing of the present invention are as follows:

[0061] The present invention collects the on-site data of the fire that is about to be extinguished in real time through the edge computing module, performs the first investigation and analysis, and packages the real-time collected fire scene data and the first investigation and analysis results to transmit to the cloud computing module. The cloud computing module is used to perform a second investigation and analysis based on the real-time collected fire scene data of the edge computing module and the first investigation and analysis results, and obtain the second investigation and analysis results; the data analysis processing module is used to collect the data after the fire is extinguished, and multiple deep learning models are deployed in the cloud computing node to process the collected data after the fire is extinguished, and the processing and analysis results are recorded as the third investigation and analysis results; the investigation data fusion module is used to fuse the second investigation and analysis results and the third investigation and analysis results, and the final investigation and analysis results are output; the display processing module is used to obtain the first investigation data from the cloud computing module. The present invention obtains the second investigation and analysis results from the edge computing module and the investigation and analysis results from the data analysis and processing module, and displays the obtained investigation and analysis results. The present invention uses cloud computing and deep learning models to obtain data through deep learning models in the fire data acquisition stage. In the fire data analysis stage, the first data analysis of the fire that is about to be extinguished is performed by using a neural network to learn the InceptionV3 algorithm, and the second analysis is performed on the first data analysis combined with historical data by using the network-learned YOLOv7 algorithm. The data after the fire is extinguished is analyzed for the third time through the deep learning model, and the cosine similarity algorithm is used to iteratively perform a fusion analysis on the second analysis results and the third analysis results, and the final investigation results are output. The accuracy of the fire investigation is improved through multiple analyses of deep learning algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a system architecture diagram of a fire investigation and survey system based on cloud computing of the present invention;

[0063] Figure 2 A data analysis and processing module diagram of a fire investigation and survey system based on cloud computing according to the present invention;

[0064] Figure 3This is a data acquisition submodule diagram of a fire investigation and survey system based on cloud computing of the present invention;

[0065] Figure 4 This is an interface type diagram of a fire investigation and survey system based on cloud computing of the present invention;

[0066] Figure 5 A fire investigation deep learning submodule diagram of a fire investigation and survey system based on cloud computing according to the present invention;

[0067] Figure 6 A fire investigation preconfiguration submodule diagram of a fire investigation and survey system based on cloud computing according to the present invention;

[0068] Figure 7 This is a cloud computing module diagram of a fire investigation and survey system based on cloud computing of the present invention.

[0069] Among them, 1-edge computing module, 2-cloud computing module, 21-data management submodule, 22-second analysis and investigation submodule, 3-storage module, 4-data analysis and processing module, 401-sensor interface acquisition unit, 402-camera image interface acquisition unit, 403-protocol acquisition unit, 404-collected data acquisition unit, 41-data acquisition submodule, 411-sensor unit, 412-image processing unit, 413-video recording processing unit, 42-detection submodule, 43-fire investigation deep learning submodule, 431-model building unit, 432-model output unit, 44-fire investigation pre-configuration submodule, 441-sensor pre-configuration unit, 442-camera pre-configuration unit, 443-fire trace relationship unit, 444-execution pre-configuration unit, 5-investigation data fusion module, 6-display processing module. DETAILED DESCRIPTION

[0070] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] Example 1

[0072] First, the technical terms involved in the embodiments of the present application are explained.

[0073] (1) Cloud computing

[0074] Cloud computing is a service model that provides on-demand access to computing resources (such as servers, storage, databases, networks, software, analysis, etc.) through the Internet. Instead of purchasing and maintaining physical hardware, users rent these resources through cloud service providers (CSPs). The core advantages of cloud computing include flexibility, scalability, cost-effectiveness, and high availability;

[0075] (2) Edge computing

[0076] Edge computing is a distributed computing architecture that transfers data processing and analysis tasks from centralized data centers to the edge of the network, close to the data source. This architecture can reduce data transmission latency, improve response speed, and reduce bandwidth consumption.

[0077] Figure 1 This is a system architecture diagram of a fire investigation and survey system based on cloud computing in the present invention.

[0078] The system includes an edge computing module 1, a cloud computing module 2, a storage module 3, a data analysis and processing module 4, a survey data fusion module 5 and a display processing module 6. Figures 2 to 7 ;

[0079] The storage module 3 is used to store the data generated by the fire investigation and survey system;

[0080] The edge computing module 1 is used to perform a first investigation and analysis based on the real-time collected fire scene data, and to package and transmit the real-time collected fire scene data and the first investigation and analysis results to the cloud computing module, wherein the fire scene data is data collected when the fire is about to be extinguished;

[0081] The cloud computing module 2 is used to perform a second investigation and analysis based on the real-time collected fire scene data and the first investigation and analysis results, obtain the second investigation and analysis results, and transmit the second investigation and analysis results to the edge computing module;

[0082] The data analysis and processing module 4 is used to collect data after the fire is extinguished, and to process the collected data after the fire is extinguished by deploying multiple deep learning models on the cloud computing nodes, and the processing and analysis results are recorded as the third investigation and analysis results;

[0083] The survey data fusion module 5 is used to fuse the second survey analysis results and the third survey analysis results, and output the final survey analysis results;

[0084] The display processing module 6 is used to obtain the second survey and analysis results from the cloud computing module 2, the first survey and analysis results from the edge computing module 1, the third survey and analysis results from the data analysis processing module 4, and the final survey and analysis results from the survey data fusion module 5, and display the obtained survey and analysis results.

[0085] Further, the data analysis and processing module 4 includes a data acquisition submodule 41, a detection submodule 42, a fire investigation deep learning submodule 43 and a fire investigation pre-configuration submodule 44;

[0086] The data acquisition submodule 41 is used to acquire data from the scene after the fire through sensors and cameras, the data including data after the fire is extinguished, and convert the data to generate data that can be recognized and analyzed by a computer program, the data including historical acquisition data and real-time acquisition data;

[0087] The data acquisition submodule 41 includes a sensor unit 411, an image processing unit 412 and a video recording processing unit 413;

[0088] The sensor unit 411 is used to collect information through sensors after the fire is extinguished, including real-time sensor data and historical sensor data;

[0089] The image processing unit 412 is used to process the images collected after the fire is extinguished, including real-time images and historical collected images;

[0090] The video recording processing unit 413 is used to process the video recording collected after the fire is extinguished, including real-time video recording and historical video recording;

[0091] By acquiring the corresponding sensor type, camera type, interface type and data format of the sensor unit 411, the image processing unit 412 and the video recording processing unit 413, and converting the sensor type, the interface type and the data format into data recognizable by the computer program, the sensor unit controls different types of sensors and uses the sensors to collect data, the image processing unit and the video recording processing unit control different types of cameras and use the cameras to obtain image data or video recording, and the sensors and cameras transmit data through their respective interface types;

[0092] The fire investigation deep learning submodule 43 is used to train the fire investigation deep learning network model according to the data of the data acquisition submodule 41 to obtain a fire investigation and survey output model;

[0093] The fire investigation preconfiguration submodule 44 is used to train a preconfigured deep learning network model to obtain a sensor preconfiguration model, a camera preconfiguration model, and a fire trace relationship model, and use the camera preconfiguration model, the sensor preconfiguration model, and the fire trace relationship model for programmatic collection of fire trace data and generation of fire trace relationships.

[0094] Furthermore, the types of the sensors specifically include temperature sensors, smoke sensors, heat release rate sensors and CO concentration sensors;

[0095] The data format is used to convert the data collected by the sensor and the camera into a data type format recognizable by a computer program;

[0096] The camera is used to capture images and video data, wherein the images and video data include trace data after the fire is extinguished;

[0097] The historically collected data includes historical data of other fire scenes collected by sensors and cameras;

[0098] The historical data at least includes trace data after the fire is extinguished.

[0099] Furthermore, the interface types include:

[0100] A sensor interface acquisition unit 401 is used to acquire the type of the sensor interface, and the sensor interface is used for transmission after collecting data;

[0101] A camera image interface acquisition unit 402, used to acquire the camera interface type;

[0102] The protocol acquisition unit 403 is used to acquire the communication protocol of the sensor interface and the camera interface;

[0103] A data acquisition unit 404 is used to establish different communication protocols according to the protocol acquisition unit 403 to acquire the data collected by the sensor interface and the camera interface, and store the collected data in the storage module 3;

[0104] The fire investigation deep learning submodule 43 includes:

[0105] A model building unit 431 is used to characterize the data of the data acquisition submodule 41 through feature engineering, generate training set data and test set data, train a deep learning network model through the training set data, and obtain a fire investigation and survey output model;

[0106] The model output unit 432 is used to use the data and historical data collected in real time by the data collection submodule 41 as input to the fire investigation and survey output model, and the fire investigation and survey output model outputs the fire point, fire cause and fire spread process.

[0107] Furthermore, the fire investigation pre-configuration submodule 44 includes:

[0108] The sensor pre-configuration unit 441 is used to train a deep learning network model according to the sensor collection data with type labels to obtain the sensor pre-configuration model, wherein the sensor collection data with type labels is subjected to feature extraction and labeling through a deep learning neural network algorithm;

[0109] The camera pre-configuration unit 442 is used to train a deep learning network model according to the camera collection data with type labels to obtain the camera pre-configuration model, wherein the type labels are used to extract features and label images collected by the camera through a deep learning neural network algorithm;

[0110] The fire trace relationship unit 443 is used to construct a fire trace relationship based on the fire trace data collected by the sensor and the camera, train the deep learning network model, and obtain a fire trace relationship model;

[0111] The fire trace relationships include the relationship between the carbonization rate and the direction of fire spread, the relationship between the concrete component peeling trace and the fire temperature, the relationship between the glass thermal cracking trace and the fire temperature, the relationship between the glass thermal softening trace and the fire temperature, the relationship between the glass thermal melting trace and the fire temperature, the relationship between the thermal expansion of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the fire spread change and the distance from the fire ignition point, the relationship between the fire intensity change and the fire cause, and the relationship between the fire intensity change and the distance from the fire ignition point;

[0112] The fire trace data includes a first trace, a second trace, a third trace and a fourth trace;

[0113] The first traces include traces of the fire starting point, traces of the fire spreading, traces of the fire cause and traces of the fire nature;

[0114] The second traces include traces formed by glass, traces formed by metal, traces formed by wood, and traces of flammable liquids;

[0115] The third traces include carbonization traces, ash traces, smoke traces, collapse traces, burning traces, melting traces, discoloration traces, deformation traces, cracking traces, electric heat melting traces, friction traces, separation and displacement traces, and human body burn marks;

[0116] The fourth trace includes a spreading trace and an intensity trace;

[0117] The fire trace relationship is modeled through deep learning of machine learning.

[0118] Furthermore, the edge computing module 1 is used to perform edge computing on the real-time data of the data acquisition unit 404, and the result of the calculation and analysis is recorded as the first investigation and analysis data, and the first investigation and analysis data is transmitted to the cloud computing module 2, and the real-time data is the data collected when the fire is about to be extinguished;

[0119] The specific analysis process of the first survey analysis data includes:

[0120] L1: Collecting data of the fire scene in real time through sensors and cameras of the data acquisition unit 404;

[0121] L2: Cleaning, denoising, and formatting preprocessing operations are performed on the collected fire scene data at the edge node;

[0122] L3: Use the neural network learning InceptionV3 algorithm on the edge computing node to analyze data in real time.

[0123] Furthermore, the cloud computing module 2 includes a data management submodule 21 and a second analysis and investigation submodule 22;

[0124] The data management submodule 21 is used to retrieve historical data under the corresponding fire environment based on the real-time data collected at the fire scene and the first investigation and analysis data results, wherein the historical data includes sensor data, image data and video data;

[0125] The second analysis and investigation submodule 22 is used to perform a secondary investigation and analysis based on the real-time data collected at the fire scene, the first investigation and analysis data results and historical data, including:

[0126] S1: Perform cleaning, denoising, and formatting preprocessing operations on the real-time fire scene data, the first investigation and analysis data results, and the historical data collected at the cloud computing node;

[0127] S2: Use the neural network to learn the YOLOv7 algorithm on the cloud computing node to analyze the data in real time;

[0128] S3: The results of the analysis and investigation are used by the display processing module 6 for display.

[0129] Furthermore, the survey data fusion module 5 calculates the second survey analysis result and the third survey analysis result by using a cosine similarity algorithm. The cosine similarity algorithm is:

[0130]

[0131] where x i ,y i They are the i-th vectors corresponding to the second survey and analysis results and the third survey and analysis results respectively. The calculation result is M, and the threshold is set to Q. If M is greater than Q, it means that the survey and analysis results are correct and the error rate is very low. If M is less than Q, it means that the survey and analysis results are wrong and the error rate is very high. Further investigation and analysis is needed until M is greater than Q and the survey is terminated.

[0132] It should be noted that this embodiment can realize automatic identification and investigation of the fire point, fire cause and fire spread process, overcoming the shortcomings of manual identification methods such as high labor intensity and low work efficiency; cloud computing and deep learning methods can be used for investigation only by collecting on-site data, and the investigation process is simple and easy to implement.

[0133] The beneficial effects of a fire investigation and survey system based on cloud computing of the present invention are as follows:

[0134] The present invention collects the on-site data of the fire that is about to be extinguished in real time through the edge computing module, performs the first investigation and analysis, and packages the real-time collected fire scene data and the first investigation and analysis results to transmit to the cloud computing module. The cloud computing module is used to perform a second investigation and analysis based on the real-time collected fire scene data of the edge computing module and the first investigation and analysis results, and obtain the second investigation and analysis results; the data analysis processing module is used to collect the data after the fire is extinguished, and multiple deep learning models are deployed in the cloud computing node to process the collected data after the fire is extinguished, and the processing and analysis results are recorded as the third investigation and analysis results; the investigation data fusion module is used to fuse the second investigation and analysis results and the third investigation and analysis results, and the final investigation and analysis results are output; the display processing module is used to obtain the first investigation data from the cloud computing module. The present invention obtains the second investigation and analysis results from the edge computing module and the investigation and analysis results from the data analysis and processing module, and displays the obtained investigation and analysis results. The present invention uses cloud computing and deep learning models to obtain data through deep learning models in the fire data acquisition stage. In the fire data analysis stage, the first data analysis of the fire that is about to be extinguished is performed by using a neural network to learn the InceptionV3 algorithm, and the second analysis is performed on the first data analysis combined with historical data by using the network-learned YOLOv7 algorithm. The data after the fire is extinguished is analyzed for the third time through the deep learning model, and the cosine similarity algorithm is used to iteratively perform a fusion analysis on the second analysis results and the third analysis results, and the final investigation results are output. The accuracy of the fire investigation is improved through multiple analyses of deep learning algorithms.

[0135] The present invention and its embodiments are described above, which is not restrictive. What is shown in the accompanying drawings is only one of the embodiments of the present invention, and the actual content is not limited to this. In short, if ordinary technicians in this field are inspired by it and do not deviate from the purpose of the invention, they can creatively design structural methods and embodiments similar to the technical scheme, which should all fall within the scope of protection of the present invention.

Claims

1. A fire investigation and survey system based on cloud computing, characterized in that: It includes an edge computing module (1), a cloud computing module (2), a storage module (3), a data analysis and processing module (4), a survey data fusion module (5) and a display processing module (6); The storage module (3) is used to store data generated by the fire investigation and survey system; The edge computing module (1) is used to perform a first investigation and analysis based on real-time collected fire scene data, and to package and transmit the real-time collected fire scene data and the first investigation and analysis results to the cloud computing module, wherein the fire scene data is data collected when the fire is about to be extinguished; The cloud computing module (2) is used to perform a second investigation and analysis based on the real-time collected fire scene data and the first investigation and analysis results, obtain the second investigation and analysis results, and transmit the second investigation and analysis results to the edge computing module; The data analysis and processing module (4) is used to collect data after the fire is extinguished, and to process the collected data after the fire is extinguished by deploying multiple deep learning models on the cloud computing nodes, and the processing and analysis results are recorded as the third investigation and analysis results; The survey data fusion module (5) is used to fuse the second survey analysis result and the third survey analysis result, and output the final survey analysis result; The display processing module (6) is used to obtain the second survey and analysis results from the cloud computing module (2), the first survey and analysis results from the edge computing module (1), the third survey and analysis results from the data analysis processing module (4), and the final survey and analysis results from the survey data fusion module (5), and to display the obtained survey and analysis results.

2. A cloud computing-based fire investigation and survey system according to claim 1, characterized in that: The data analysis and processing module (4) comprises a data acquisition submodule (41), a detection submodule (42), a fire investigation deep learning submodule (43) and a fire investigation pre-configuration submodule (44); The data acquisition submodule (41) is used to acquire data on the scene after the fire through sensors and cameras, the data including data after the fire is extinguished, and convert the data to generate data that can be recognized and analyzed by a computer program, the data including historically collected data and real-time collected data; The data acquisition submodule (41) comprises a sensor unit (411), an image processing unit (412) and a video recording processing unit (413); The sensor unit (411) is used to collect data through sensors after the fire is extinguished, including real-time sensor data and historically collected sensor data; The image processing unit (412) is used to process images collected after the fire is extinguished, including real-time images and historically collected images; The video recording processing unit (413) is used to process the video recording collected after the fire is extinguished, including real-time video recording and historical video recording; By acquiring the corresponding sensor type, camera type, interface type and data format of the sensor unit (411), the image processing unit (412) and the video recording processing unit (413), and converting the sensor type, the interface type and the data format into data recognizable by the computer program, the sensor unit controls different types of sensors and uses the sensors to collect data, the image processing unit and the video recording processing unit control different types of cameras and use the cameras to obtain image data or video recording, and the sensors and cameras transmit data through their respective interface types; The fire investigation deep learning submodule (43) is used to train a fire investigation deep learning network model based on the data of the data acquisition submodule (41) to obtain a fire investigation and survey output model; The fire investigation preconfiguration submodule (44) is used to train a preconfigured deep learning network model to obtain a sensor preconfiguration model, a camera preconfiguration model and a fire trace relationship model, and use the camera preconfiguration model, the sensor preconfiguration model and the fire trace relationship model for programmatic collection of fire trace data and generation of fire trace relationships.

3. A cloud computing-based fire investigation and survey system according to claim 2, characterized in that: The types of sensors specifically include temperature sensors, smoke sensors, heat release rate sensors and CO concentration sensors; The data format is used to convert the data collected by the sensor and the camera into a data type format that can be recognized by the computer program; The camera is used to capture images and video data, wherein the images and video data include trace data after the fire is extinguished; The historically collected data includes historical data of other fire scenes collected by sensors and cameras; The historical data at least includes trace data after the fire is extinguished.

4. A cloud computing-based fire investigation and survey system according to claim 3, characterized in that: The interface types include: A sensor interface acquisition unit (401) is used to acquire the type of the sensor interface, the sensor interface being used for transmission after collecting data; A camera image interface acquisition unit (402), used to acquire the camera interface type; A protocol acquisition unit (403), used for acquiring the communication protocol of the sensor interface and the camera interface; A collected data acquisition unit (404), used for acquiring the collected data of the sensor interface and the camera interface according to different communication protocols established by the protocol acquisition unit (403), and storing the collected data in the storage module (3); The fire investigation deep learning submodule (43) comprises: A model building unit (431) is used to characterize the data of the data acquisition submodule (41) through feature engineering to generate training set data and test set data, train a deep learning network model through the training set data, and obtain a fire investigation and survey output model; The model output unit (432) is used to use the data and historical data collected in real time by the data collection submodule (41) as input to a fire investigation and survey output model, wherein the fire investigation and survey output model outputs the fire point, fire cause and fire spread process.

5. A cloud computing-based fire investigation and survey system according to claim 4, characterized in that: The fire investigation pre-configuration submodule (44) comprises: A sensor pre-configuration unit (441) is used to train a deep learning network model based on sensor collection data with type labels to obtain the sensor pre-configuration model, wherein the sensor collection data with type labels is subjected to feature extraction and labeling through a deep learning neural network algorithm; A camera pre-configuration unit (442) is used to train a deep learning network model based on camera acquisition data with type labels to obtain the camera pre-configuration model, wherein the image acquired by the camera is subjected to feature extraction and labeling through a deep learning neural network algorithm with type labels; A fire trace relationship unit (443) is used to construct a fire trace relationship based on the fire trace data collected by the sensor and the camera, train a deep learning network model, and obtain a fire trace relationship model; The fire trace relationships include the relationship between the carbonization rate and the direction of fire spread, the relationship between the concrete component peeling trace and the fire temperature, the relationship between the glass thermal cracking trace and the fire temperature, the relationship between the glass thermal softening trace and the fire temperature, the relationship between the glass thermal melting trace and the fire temperature, the relationship between the thermal expansion of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the thermal change of metal objects and the fire temperature, the relationship between the fire spread change and the distance from the fire ignition point, the relationship between the fire intensity change and the fire cause, and the relationship between the fire intensity change and the distance from the fire ignition point; The fire trace data includes a first trace, a second trace, a third trace and a fourth trace; The first traces include traces of the fire starting point, traces of the fire spreading, traces of the fire cause and traces of the fire nature; The second traces include traces formed by glass, traces formed by metal, traces formed by wood, and traces of flammable liquids; The third traces include carbonization traces, ash traces, smoke traces, collapse traces, burning traces, melting traces, discoloration traces, deformation traces, cracking traces, electric heat melting traces, friction traces, separation and displacement traces, and human body burn marks; The fourth trace includes a spreading trace and an intensity trace; The fire trace relationship is modeled through deep learning of machine learning.

6. A cloud computing-based fire investigation and survey system according to claim 5, characterized in that: The edge computing module (1) is used to perform edge computing on the real-time data of the data acquisition unit (404), and the result of the computing and analysis is recorded as first investigation and analysis data, and the first investigation and analysis data is transmitted to the cloud computing module (2), and the real-time data is the data collected when the fire is about to be extinguished; The specific analysis process of the first survey analysis data includes: L1: Collecting data of the fire scene in real time through sensors and cameras of the data acquisition unit (404); L2: Cleaning, denoising, and formatting preprocessing operations are performed on the collected fire scene data at the edge node; L3: Use the neural network learning InceptionV3 algorithm on the edge computing node to analyze data in real time.

7. The cloud computing-based fire investigation and survey system according to claim 6, characterized in that: The cloud computing module (2) comprises a data management submodule (21) and a second analysis and investigation submodule (22); The data management submodule (21) is used to retrieve historical data under the corresponding fire environment based on the acquired real-time data collected at the fire scene and the first investigation and analysis data results, wherein the historical data includes sensor data, image data and video data; The second analysis and investigation submodule (22) is used to perform a secondary investigation and analysis based on the acquired real-time data collected at the fire scene, the first investigation and analysis data results and historical data, including: S1: Perform cleaning, denoising, and formatting preprocessing operations on the real-time fire scene data, the first investigation and analysis data results, and the historical data collected at the cloud computing node; S2: Use the neural network to learn the YOLOv7 algorithm on the cloud computing node to analyze the data in real time; S3: The results of the analysis and investigation are used for display in the display processing module (6).

8. The cloud computing-based fire investigation and survey system according to claim 7, characterized in that: The survey data fusion module (5) calculates the second survey analysis result and the third survey analysis result by using a cosine similarity algorithm. The cosine similarity algorithm is: where x i ,y i They are the i-th vectors corresponding to the second survey and analysis results and the third survey and analysis results respectively. The calculation result is M, and the threshold is set to Q. If M is greater than Q, it means that the survey and analysis results are correct and the error rate is very low. If M is less than Q, it means that the survey and analysis results are wrong and the error rate is very high. Further investigation and analysis is needed until M is greater than Q and the survey is terminated.

Images