CN117237804A - Pyrotechnical recognition system and method based on federal learning - Google Patents
Pyrotechnical recognition system and method based on federal learning Download PDFInfo
- Publication number
- CN117237804A CN117237804A CN202311194171.8A CN202311194171A CN117237804A CN 117237804 A CN117237804 A CN 117237804A CN 202311194171 A CN202311194171 A CN 202311194171A CN 117237804 A CN117237804 A CN 117237804A
- Authority
- CN
- China
- Prior art keywords
- smoke
- building
- fire
- identification
- groups
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000000779 smoke Substances 0.000 claims abstract description 383
- 238000012544 monitoring process Methods 0.000 claims abstract description 24
- 238000012216 screening Methods 0.000 claims abstract description 23
- 230000002776 aggregation Effects 0.000 claims abstract description 21
- 238000004220 aggregation Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000013461 design Methods 0.000 claims description 61
- 238000012549 training Methods 0.000 claims description 49
- 238000004364 calculation method Methods 0.000 claims description 22
- 238000010276 construction Methods 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 10
- 230000003993 interaction Effects 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 2
- 238000013527 convolutional neural network Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 12
- 206010000369 Accident Diseases 0.000 abstract description 5
- 230000006872 improvement Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000002411 adverse Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Alarm Systems (AREA)
Abstract
The invention provides a smoke and fire recognition system and a method based on federal learning, which relate to the technical field of data processing, wherein invalid smoke and fire recognition data are obtained by screening smoke and fire recognition building information sets, an initial recognition sub-network is further obtained, H building smoke and fire recognition sub-networks are constructed by optimizing the smoke and fire recognition data, and a target smoke and fire recognition sub-network is generated through federal aggregation; and further optimizing the fire-fighting layout of the smoke-fire identified building after acquiring the smoke-fire risk event set based on the target smoke-fire identified sub-network monitoring. The technical problems that in the prior art, the accuracy and the precision of identifying the smoke and fire of the urban building are poor, so that the timeliness and the effectiveness of identifying the smoke and fire of the urban building on fire accident treatment are improved to a weak degree are solved. The technical effects of obtaining the smoke and fire identification sub-network with high smoke and fire identification precision and identification result reliability and providing guarantee for timely and effective treatment of the smoke and fire risk event of the building are achieved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a smoke and fire identification system and method based on federal learning.
Background
In cities, due to the fact that the accuracy of the current smoke and fire identification system is low, smoke and fire signs on a building cannot be accurately identified, adverse effects are brought to timely treatment and effective control of fire accidents, if the smoke and fire signs of the building cannot be accurately identified, the time for extinguishing actions of smoke and fire can be delayed, the fire is out of control, and further, larger casualties and property loss are caused, and the surrounding environment is threatened. In summary, at the present stage, the accuracy and precision of identifying the smoke and fire of the urban building are poor, so that the effect of improving the timeliness and effectiveness of identifying the smoke and fire of the urban building on fire accident treatment is poor.
Disclosure of Invention
The application provides a smoke and fire identification system and method based on federal learning, which are used for solving the technical problems of weak timeliness and effectiveness improvement effect of urban building smoke and fire accident handling caused by poor precision and accuracy of urban building smoke and fire identification in the prior art.
In view of the foregoing, the present application provides a federal learning-based pyrotechnic identification system and method.
In a first aspect of the application, there is provided a federal learning-based pyrotechnic identification system, the system comprising: the building information interaction module is used for interactively determining a smoke and fire identification building information set based on a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer; the data screening execution module is used for presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value to obtain M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K; the similar building merging module is used for obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained through building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K; the identification network construction module is used for pre-constructing an initial firework identification sub-network, synchronizing the initial firework identification sub-network to the H optimized firework identification buildings and obtaining H initial identification sub-networks; the recognition network training module is used for obtaining H building smoke and fire recognition sub-networks, wherein the H building smoke and fire recognition sub-networks are obtained by performing supervision training on the H initial recognition sub-networks by adopting the H groups of optimized smoke and fire recognition data; the federation execution module is used for federating the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network; the monitoring window setting module is used for presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set; and the fire control layout optimization module is used for optimizing the building fire control layout of the K initial smoke and fire identification buildings based on the smoke and fire risk event set.
In a second aspect of the application, there is provided a federal learning-based pyrotechnic identification method, the method comprising: determining a smoke and fire identification building information set based on interaction of a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer; presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value, and obtaining M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K; obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained through building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K; pre-constructing an initial firework recognition sub-network, and synchronizing the initial firework recognition sub-network to the H optimized firework recognition buildings to obtain H initial recognition sub-networks; obtaining H building smoke and fire identification sub-networks, wherein the H building smoke and fire identification sub-networks are obtained by performing supervision training on the H initial identification sub-networks by adopting the H groups of optimized smoke and fire identification data; federal aggregation is carried out on the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network; presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set; and performing building fire layout optimization of the K initial smoke identification buildings based on the smoke risk event set.
One or more technical schemes provided by the application have at least the following technical effects or advantages:
according to the method provided by the embodiment of the application, the smoke and fire identification building information set is interactively determined based on the preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer; presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value, and obtaining M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K; obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained through building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K; pre-constructing an initial firework recognition sub-network, and synchronizing the initial firework recognition sub-network to the H optimized firework recognition buildings to obtain H initial recognition sub-networks; obtaining H building smoke and fire identification sub-networks, wherein the H building smoke and fire identification sub-networks are obtained by performing supervision training on the H initial identification sub-networks by adopting the H groups of optimized smoke and fire identification data; federal aggregation is carried out on the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network; presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set; and performing building fire layout optimization of the K initial smoke identification buildings based on the smoke risk event set. On the basis of guaranteeing the privacy of smoke and fire identification data of a building, a smoke and fire identification sub-network with high smoke and fire identification precision and identification result reliability is obtained, and a guaranteed technical effect is provided for timely and effectively processing smoke and fire risk events of the building.
Drawings
FIG. 1 is a schematic flow chart of a federal learning-based smoke and fire identification method provided by the application;
FIG. 2 is a schematic flow chart of generating a target pyrotechnic recognition sub-network in the federal learning-based pyrotechnic recognition method provided by the application;
FIG. 3 is a schematic flow chart of optimizing a fire-fighting layout of a building in the federal learning-based smoke and fire identification method provided by the application;
fig. 4 is a schematic structural diagram of a federal learning-based pyrotechnic identification system in accordance with the present application.
Reference numerals illustrate: the system comprises a building information interaction module 1, a data screening execution module 2, a similar building merging module 3, an identification network construction module 4, an identification network training module 5, a federal aggregation execution module 6, a monitoring window setting module 7 and a fire control layout optimization module 8.
Detailed Description
The application provides a smoke and fire identification system and method based on federal learning, which are used for solving the technical problems of weak timeliness and effectiveness improvement effect of urban building smoke and fire accident handling caused by poor precision and accuracy of urban building smoke and fire identification in the prior art. On the basis of guaranteeing the privacy of smoke and fire identification data of a building, a smoke and fire identification sub-network with high smoke and fire identification precision and identification result reliability is obtained, and a guaranteed technical effect is provided for timely and effectively processing smoke and fire risk events of the building.
The technical scheme of the application accords with related regulations on data acquisition, storage, use, processing and the like.
In the following, the technical solutions of the present application will be clearly and completely described with reference to the accompanying drawings, and it should be understood that the described embodiments are only some embodiments of the present application, but not all embodiments of the present application, and that the present application is not limited by the exemplary embodiments described herein. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application. It should be further noted that, for convenience of description, only some, but not all of the drawings related to the present application are shown.
Example 1
As shown in fig. 1, the present application provides a federal learning-based pyrotechnic identification method, the method comprising:
a100, interactively determining a smoke and fire identification building information set based on a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer;
a200, presetting a training data screening threshold, traversing the K groups of smoke and fire identification data based on the training data screening threshold to obtain M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K;
Specifically, in this embodiment, the smoke and fire identification area is an area where the smoke and fire risk identification of the building is planned to be performed together, for example, a local area where the smoke and fire identification area is 10×10km.
And carrying out city division based on a preset smoke and fire identification area to obtain a plurality of actual smoke and fire identification areas, randomly selecting one of the smoke and fire identification areas for building information interaction based on the actual smoke and fire identification areas, and determining a smoke and fire identification building information set, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, K is a positive integer, and the smoke and fire identification data is an image of all smoke and fire events occurring in the history of each initial smoke and fire identification building.
It should be understood that, in order to ensure the accuracy of the output of the identification sub-network when the identification sub-network for identifying the risk of the firework of the building according to the image is trained, there is a quantitative requirement for the data involved in the identification sub-network training, based on which the present embodiment sets the training data screening threshold for evaluating whether the firework identification data satisfies the data amount required for the identification sub-network training corresponding to the initial firework identification data, and the numerical value of the training data screening threshold is not particularly limited in the present embodiment and may be set according to the actual identification sub-network training accuracy requirement.
And traversing the K groups of smoke and fire recognition data based on the training data screening threshold value to obtain M groups of invalid smoke and fire recognition data of which the data volume does not meet the training data screening threshold value, wherein the M groups of invalid smoke and fire recognition data are mapped to M invalid smoke and fire recognition buildings, M is a positive integer smaller than K, and the M invalid smoke and fire recognition buildings cannot independently perform training of recognition sub-networks used for smoke and fire recognition based on the existing data.
A300, obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained by carrying out building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K;
in one embodiment, H optimized smoke-fire identification buildings are obtained, wherein the H optimized smoke-fire identification buildings are mapped to H groups of optimized smoke-fire identification data, the H optimized smoke-fire identification buildings are obtained by performing building similarity analysis on the M invalid smoke-fire identification buildings and the H initial smoke-fire identification buildings, and the method step a300 provided by the present application further comprises:
A310, building a building similarity recognition module in advance;
a320, interactively obtaining K groups of initial building design information, wherein the K groups of initial building design information are mapped with the K initial firework identification buildings one by one;
a330, obtaining M groups of invalid building design information and H groups of initial building design information, wherein the M groups of invalid building design information are obtained by taking M invalid firework identification buildings as references and splitting the K groups of initial building design information;
a340, obtaining M groups of building similarity index sequences, wherein the M groups of building similarity index sequences are obtained by synchronizing the M groups of invalid building design information and the H groups of initial building design information to the building similarity recognition module for building similarity analysis, and each group of building similarity index sequences in the M groups of building similarity index sequences comprises H building similarity indexes;
a350, extracting extremum values based on the M groups of building similarity index sequences to obtain M optimal smoke and fire identification buildings;
a360, correspondingly calling M groups of initial building design information based on the M optimal smoke and fire identification buildings, and virtually merging the M groups of initial building design information with the M groups of invalid building design information to obtain the H optimal smoke and fire identification buildings.
In one embodiment, the building similarity recognition module is pre-built, and the method step a310 provided by the present application further includes:
a311, the building similarity recognition module comprises a building feature recognition unit, a building feature comparison unit and a building similarity calculation unit;
a312, obtaining a plurality of groups of sample building design information;
a313, presetting a building characteristic recognition rule, wherein the building characteristic recognition rule comprises N building characteristic recognition indexes, and N is a positive integer;
a314, carrying out building feature identification marking on the plurality of groups of sample building design information based on the building feature identification rule to obtain a plurality of groups of sample building feature information sets;
a315, constructing the building feature recognition unit based on a CNN neural network, and performing supervision training of the building feature recognition unit by adopting the plurality of groups of sample building design information and the plurality of groups of sample building feature information sets;
a316, pre-constructing a three-dimensional coordinate system in the building feature comparison unit, and performing feature comparison on the building feature information group output by the building feature recognition unit to obtain N building feature similarities, wherein the N building feature similarities are mapped with the N building feature recognition indexes one by one;
A317, pre-configuring building feature weight distribution in the building similarity calculation unit, wherein the building feature weight distribution comprises N building feature index weights, and the N building feature index weights are mapped with the N building feature identification indexes one by one;
and A318, inputting the N building feature similarities into the building similarity calculation unit, and calculating the N building feature index weights in the building similarity calculation unit to obtain a building similarity index.
Specifically, in this embodiment, a building similarity recognition module is pre-constructed, where the building similarity recognition module may perform similarity evaluation of two buildings according to information such as building appearance shapes and building heights of the two buildings, and in this embodiment, the building similarity recognition module is constructed as follows, where the building similarity recognition module includes a building feature recognition unit, a building feature comparison unit, and a building similarity calculation unit.
The building feature recognition unit is used for extracting building features with building similar recognition value on the basis of obtaining the design information of the building. The building feature recognition unit is constructed by obtaining a plurality of sets of sample building design information including, but not limited to, elevation, cross-section, plan.
Building feature recognition rules are preset, wherein the building feature recognition rules comprise N building feature recognition indexes, N is a positive integer, and the building feature recognition indexes comprise, but are not limited to, building appearance shapes, building heights, building volumes and building frames.
And based on the building feature recognition rule, manually marking the building feature recognition of the plurality of groups of sample building design information to obtain a plurality of groups of sample building feature information sets. And constructing the building feature recognition unit based on a CNN (cyclic neural network), and dividing the multiple groups of sample building design information and the multiple groups of sample building feature information sets into a training set, a testing set and a verification set by adopting a data division mode of 8:1:1, so as to perform supervision training, testing and verification of the building feature recognition unit until the output accuracy of the building feature recognition unit meets the requirement of preset accuracy.
And the building feature comparison unit is pre-constructed with a three-dimensional coordinate system for comparing the coincidence degree of the two building identification features as the similarity of the two building identification features, and performs feature comparison on the N groups of building feature information groups output by the building feature identification unit to obtain N building feature similarities, wherein the N building feature similarities are mapped with the N building feature identification indexes one by one.
The building similarity calculation unit is configured to perform comprehensive calculation on the N building feature similarities to obtain overall similarity indexes of two buildings, and in order to improve the reliability of the obtained true problem similarity indexes, in this embodiment, building feature weight distribution is preconfigured in the building similarity calculation unit, where the building feature weight distribution includes N building feature index weights, the N building feature index weights are mapped with the N building feature identification indexes one by one, and in this embodiment, the N building feature index weights are set according to the influence degree and importance of the N building feature identification indexes on the buildings in aesthetic and application science.
And inputting the N building feature similarities into the building similarity calculation unit, and calculating the N building feature index weights in the building similarity calculation unit to obtain a building similarity index.
After the building similarity recognition module is constructed, the embodiment performs building similarity analysis of M invalid firework recognition buildings and H initial firework recognition buildings based on the building similarity recognition module.
It should be appreciated that the present embodiment groups M invalid smoke-recognizing buildings and H initial smoke-recognizing buildings to obtain M x H groups of invalid smoke-recognizing buildings-initial smoke-recognizing buildings without repetition.
And interactively obtaining K groups of initial building design information of the K initial firework identification buildings, splitting the K groups of initial building design information to obtain M groups of invalid building design information and H groups of initial building design information of initial firework identification buildings corresponding to H groups of initial firework identification buildings by taking the M invalid firework identification buildings as references.
And obtaining non-repeated M x H group invalid building design information-initial building design information based on the non-repeated M x H group invalid smoke identification building-initial smoke identification building. And synchronizing M.H groups of non-repeated invalid building design information-initial building design information to the building similarity recognition module one by one to perform building similarity analysis to obtain M groups of building similarity index sequences. It should be understood that each of the M sets of building similarity index sequences includes H building similarity indices, and that the H building similarity indices are ordered from a large to a small similarity index number. And extracting similarity index maximum values based on the M groups of building similarity index sequences to obtain M optimal smoke and fire identification buildings corresponding to the M building similarity index maximum values, wherein the optimal smoke and fire identification buildings are certain initial smoke and fire identification buildings with highest building similarity with invalid smoke and fire identification buildings.
And correspondingly calling M groups of initial building design information based on the M optimal smoke and fire identification buildings, and virtually merging the M groups of initial building design information with the M groups of invalid building design information to obtain H groups of optimal smoke and fire identification data and the H groups of optimal smoke and fire identification buildings.
According to the embodiment, the data of the same building are combined according to the building similarity, so that the technical effects of building and training of an independent smoke and fire recognition sub-network of the building with small auxiliary data quantity are achieved.
A400, pre-constructing an initial smoke and fire identification sub-network, and synchronizing the initial smoke and fire identification sub-network to the H optimized smoke and fire identification buildings to obtain H initial identification sub-networks;
specifically, in this embodiment, the construction of the identification sub-networks for performing independent smoke and fire identification on K initial smoke and fire identification buildings is basically the same, based on which, in this embodiment, the initial smoke and fire identification sub-networks are pre-constructed, and H initial identification sub-networks are obtained by synchronizing the initial smoke and fire identification sub-networks to the building smoke and fire identification management modules of the H optimized smoke and fire identification buildings, and in this embodiment, in the following description, the structure of the initial smoke and fire identification sub-networks is described in detail.
A500, obtaining H building smoke and fire identification sub-networks, wherein the H building smoke and fire identification sub-networks are obtained by performing supervision training on the H initial identification sub-networks by adopting the H groups of optimized smoke and fire identification data;
in one embodiment, H building pyrotechnic identification sub-networks are obtained, wherein the H building pyrotechnic identification sub-networks are obtained by performing supervised training of the H initial identification sub-networks using the H sets of optimized pyrotechnic identification data, and the method step a500 provided by the present application further comprises:
a510, acquiring first optimized smoke and fire identification data and a first initial identification sub-network based on the H groups of optimized smoke and fire identification data and the H initial identification sub-network calls;
a520, the first optimized smoke and fire identification data comprises a plurality of building smoke and fire condition images;
a530, the first initial identification sub-network comprises a decoder module and an encoder module;
a540, carrying out smoke and fire characteristic identification on the smoke and fire condition images of the plurality of buildings to obtain a plurality of groups of sample smoke and fire characteristics;
a550, training the decoder module and the encoder module by adopting the plurality of building smoke and fire condition images and the plurality of groups of sample smoke and fire characteristics to complete the construction of a first building smoke and fire identification sub-network;
And A560, constructing and obtaining the H building smoke and fire identification sub-networks by using the H groups of optimized smoke and fire identification data.
Specifically, in the present embodiment, the initial pyrotechnic identification sub-network includes a decoder module and an encoder module, and since the H-group optimized pyrotechnic identification data and the training methods of the H initial identification sub-networks have consistency, the present embodiment obtains the first optimized pyrotechnic identification data and the first initial identification sub-network based on the H-group optimized pyrotechnic identification data and the H initial identification sub-network random calls, and performs a representative explanation of the H building pyrotechnic identification sub-network construction method by elaborating a method of constructing the first building pyrotechnic identification sub-network based on the first initial identification sub-network.
The first optimized smoke and fire identification data comprise a plurality of building smoke and fire condition images, the identification processing of smoke and fire image areas is carried out based on manual work, smoke and fire characteristic identification is carried out on the plurality of building smoke and fire condition images, and a plurality of groups of sample smoke and fire characteristics are obtained.
The first initial identification sub-network includes a decoder module and an encoder module;
dividing the multiple building smoke and fire condition images and the multiple groups of sample smoke and fire characteristic identifiers into a training set, a testing set and a verification set, and performing multiple training of the decoder module and the encoder module until the identification precision is higher than a preset value so as to complete the construction of a first building smoke and fire identification sub-network. And similarly, constructing and obtaining the H building smoke and fire identification sub-networks by adopting the H groups of optimized smoke and fire identification data. According to the embodiment, the construction of the smoke and fire identification sub-network of the corresponding building is carried out based on the independent historical data of each building, so that a construction basis is provided for the subsequent use of the target smoke and fire identification sub-network with higher parameter reliability as the smoke and fire intelligent identification monitoring purpose of the H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are used as participants to carry out federal learning by using a federal learning framework.
A600, performing federal aggregation on the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network;
in one embodiment, as shown in fig. 2, the federal aggregation is performed on the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network, and the method step a600 provided by the present application further includes:
a610, extracting parameters of the H building smoke and fire identification subnetworks to obtain H subnetwork parameters;
a620, pre-constructing a federation aggregation algorithm, and carrying out aggregation treatment on the H sub-network parameters based on the federation aggregation algorithm to obtain target global sub-network parameters;
and A630, issuing the target global sub-network parameters to the H building smoke and fire identification sub-networks to update the parameters, and obtaining the target smoke and fire identification sub-networks.
Specifically, in this embodiment, the federal learning framework is a pre-constructed federal aggregation algorithm, where federal is used to perform parameter average calculation on sub-network (model) parameters of multiple participants.
The implementation carries out parameter extraction on the H building smoke and fire identification sub-networks to obtain H sub-network parameters, wherein indexes of each sub-network parameter have consistency, and the same-index aggregation processing and average value calculation of the H sub-network parameters are carried out based on the federal aggregation algorithm to obtain target global sub-network parameters formed by the same-index average value parameter set. And issuing the target global subnetwork parameters to the H building smoke and fire identification subnetworks to update the corresponding parameters to obtain the target smoke and fire identification subnetworks, wherein the identification accuracy of the target smoke and fire identification subnetwork is higher than that of any one of the H building smoke and fire identification subnetworks.
The technical effect of constructing and obtaining the target smoke and fire recognition sub-network with high recognition accuracy on the basis of guaranteeing the data privacy of K groups of smoke and fire recognition data is achieved based on federal learning.
A700, presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set;
in particular, it should be appreciated that the fire fighting equipment within the building should be adaptively adjusted in terms of the occurrence of pyrotechnic events in various areas of the building to enhance the timeliness of the elimination after the identification of the pyrotechnic events.
Based on this, the present embodiment presets a smoke risk monitoring window, and performs collection of occurrence of a smoke event of a building within a certain period of time, for example, the smoke risk monitoring window is used for performing collection records of occurrence positions of smoke events of a building and processing time-consuming time within one year.
And carrying out smoke risk monitoring on the K initial smoke identification buildings based on the target smoke identification sub-network to obtain a smoke risk event set, wherein the smoke risk event set comprises K groups of smoke event position information and smoke event processing time-consuming information.
A800, optimizing the building fire layout of the K initial smoke and fire identification buildings based on the smoke and fire risk event set.
In one embodiment, as shown in fig. 3, the optimizing the fire layout of the K initial smoke-identified buildings based on the set of smoke risk events, the method step a800 provided by the present application further includes:
a810, interactively determining K initial building fire layouts of the K initial firework identification buildings;
a820, interactively obtaining a risk event position information set and a risk event processing time-consuming set by taking the pyrotechnic risk event set as a reference;
a830, optimizing the fire-fighting layout of the K initial smoke and fire identification buildings based on the risk event position information set to obtain K optimized building fire-fighting layouts;
a840, optimizing the fire-fighting tools of the K initial smoke and fire identification buildings based on the risk event processing time-consuming set to obtain K target building fire-fighting layouts;
and A850, adopting the K target building fire-fighting layouts to adjust the K initial building fire-fighting layouts.
Specifically, in the present embodiment, the interactions determine K initial building fire layouts for the K initial pyrotechnic identified buildings. And performing group cancellation on the K groups of pyrotechnic event position information and pyrotechnic event processing time-consuming information to obtain a risk event position information set and a risk event processing time-consuming set.
And carrying out risk event occurrence position feature recognition based on the risk event position information set, carrying out adaptability optimization of the fire control layout of the K initial firework recognition buildings based on the risk event occurrence positions, and carrying out layout optimization adjustment based on consumption equipment so as to obtain the K optimized building fire control layouts without fire control management dead angles currently.
Acquiring risk event processing time consumption of corresponding position features of each risk event based on a risk time position information set, performing consumption equipment increase and decrease processing on corresponding positions of the K optimized building fire layouts based on the risk event processing time consumption set, completing fire tool optimization, and acquiring the K target building fire layouts without the shortage of fire equipment tools; and adopting the K target building fire-fighting layouts to adjust the K initial building fire-fighting layouts.
According to the embodiment, the fire control layout of the building is adaptively adjusted according to the occurrence positions of the smoke and fire risk events of the multiple buildings and the time-consuming processing conditions, so that the technical effect of improving the safety level of the building in the area is achieved.
In one embodiment, the method steps provided by the application further comprise:
a910, presetting a data updating threshold;
A920, dynamically updating the target firework identification sub-network based on the data updating threshold value to obtain an updated firework identification sub-network;
a930, synchronizing the updated smoke and fire identification sub-network to the building smoke and fire identification management module of the K initial smoke and fire identification buildings in multiple rounds.
In particular, it should be appreciated that over time, real world conditions may change, which may result in reduced performance of the recognition sub-network, based on which the present implementation may utilize new data to correct errors and flaws in the recognition sub-network by periodically updating and retraining the recognition sub-network, improving the accuracy and robustness of the recognition sub-network.
In this embodiment, the data update period is a period of data acquisition, for example, the data update threshold is 12 months. And obtaining K groups of newly-added smoke identification data of the K initial smoke identification buildings based on the data updating threshold value.
And C, adopting K groups of newly-added smoke and fire identification data to perform identification sub-network construction and federal learning by adopting the same method as the steps A100-A700 so as to dynamically update the target smoke and fire identification sub-network, obtaining an updated smoke and fire identification sub-network, synchronizing the updated smoke and fire identification sub-network to the building smoke and fire identification management modules of the K initial smoke and fire identification buildings for a plurality of times, wherein the building smoke and fire identification management modules are arranged in the K initial smoke and fire identification buildings and are connected with building monitoring of the K initial smoke and fire identification buildings so as to perform real-time dynamic identification of building smoke and fire. The technical effect of improving the recognition accuracy of the firework recognition sub-network is achieved.
Example two
Based on the same inventive concept as the federal learning-based pyrotechnic identification method in the previous embodiments, as shown in fig. 4, the present application provides a federal learning-based pyrotechnic identification system, wherein the system includes:
the building information interaction module 1 is used for interactively determining a smoke and fire identification building information set based on a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer;
the data screening execution module 2 is used for presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value, and obtaining M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K;
a similar building merging module 3, configured to obtain H optimized smoke and fire identification buildings, where the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, and the H optimized smoke and fire identification buildings are obtained by performing building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, where H is a positive integer, and h+m=k;
The recognition network construction module 4 is used for pre-constructing an initial firework recognition sub-network, synchronizing the initial firework recognition sub-network to the H optimized firework recognition buildings and obtaining H initial recognition sub-networks;
the recognition network training module 5 is used for obtaining H building smoke and fire recognition sub-networks, wherein the H building smoke and fire recognition sub-networks are obtained by performing supervision training on the H initial recognition sub-networks by adopting the H groups of optimized smoke and fire recognition data;
the federation execution module 6 is used for federating the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network;
the monitoring window setting module 7 is used for presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set;
a fire layout optimization module 8 for performing a building fire layout optimization of the K initial pyrotechnic identified buildings based on the set of pyrotechnic risk events.
In one embodiment, the system further comprises:
the similarity recognition execution unit is used for pre-constructing a building similarity recognition module;
The design information interaction unit is used for interactively obtaining K groups of initial building design information, wherein the K groups of initial building design information are mapped with the K initial firework identification buildings one by one;
the design information splitting unit is used for obtaining M groups of invalid building design information and H groups of initial building design information, wherein the M groups of invalid building design information are obtained by taking the M invalid firework identification buildings as references and splitting the K groups of initial building design information;
the similar building identification unit is used for obtaining M groups of building similarity index sequences, wherein the M groups of building similarity index sequences are obtained by synchronizing the M groups of invalid building design information and the H groups of initial building design information to the building similarity identification module for building similarity analysis, and each group of building similarity index sequences in the M groups of building similarity index sequences comprises H building similarity indexes;
the similarity sequence extraction unit is used for extracting extremum values based on the M groups of building similarity index sequences to obtain M optimal smoke and fire identification buildings;
and the building virtual merging unit is used for correspondingly calling M groups of initial building design information based on the M optimal smoke and fire identification buildings, and carrying out building virtual merging on the M groups of initial building design information and the M groups of invalid building design information to obtain the H optimal smoke and fire identification buildings.
In one embodiment, the system further comprises:
the module constitution setting unit is used for the building similarity recognition module to comprise a building feature recognition unit, a building feature comparison unit and a building similarity calculation unit;
the sample data obtaining unit is used for obtaining a plurality of groups of sample building design information;
the building feature recognition system comprises a recognition rule setting unit and a recognition rule setting unit, wherein the recognition rule setting unit is used for presetting a building feature recognition rule, the building feature recognition rule comprises N building feature recognition indexes, and N is a positive integer;
the building feature marking unit is used for marking building feature identification on the plurality of groups of sample building design information based on the building feature identification rule to obtain a plurality of groups of sample building feature information sets;
the feature recognition training unit is used for constructing the building feature recognition unit based on a CNN neural network, and performing supervision training of the building feature recognition unit by adopting the plurality of groups of sample building design information and the plurality of groups of sample building feature information sets;
the feature comparison construction unit is used for pre-constructing a three-dimensional coordinate system in the building feature comparison unit, and carrying out feature comparison on the building feature information group output by the building feature recognition unit to obtain N building feature similarities, wherein the N building feature similarities are mapped with the N building feature recognition indexes one by one;
The feature weight distribution unit is used for pre-configuring building feature weight distribution in the building similarity calculation unit, wherein the building feature weight distribution comprises N building feature index weights, and the N building feature index weights are mapped with the N building feature identification indexes one by one;
and the similarity index calculation unit is used for inputting the N building feature similarities into the building similarity calculation unit, and calculating the N building feature index weights in the building similarity calculation unit to obtain a building similarity index.
In one embodiment, the system further comprises:
the information calling execution unit is used for calling and obtaining first optimized smoke and fire identification data and a first initial identification sub-network based on the H groups of optimized smoke and fire identification data and the H initial identification sub-networks;
a smoke and fire image obtaining unit, configured to include a plurality of building smoke and fire situation images in the first optimized smoke and fire identification data;
the module composition obtaining unit is used for the first initial identification sub-network to comprise a decoder module and an encoder module;
the smoke and fire characteristic identification unit is used for carrying out smoke and fire characteristic identification on the smoke and fire condition images of the plurality of buildings to obtain a plurality of groups of sample smoke and fire characteristics;
The identification network construction unit is used for training the decoder module and the encoder module by adopting the plurality of building smoke and fire condition images and the plurality of groups of sample smoke and fire characteristics to finish the construction of a first building smoke and fire identification sub-network;
and the network set generating unit is used for constructing and obtaining the H building smoke and fire identification sub-networks by adopting the H groups of optimized smoke and fire identification data.
In one embodiment, the system further comprises:
the parameter extraction execution unit is used for extracting parameters of the H building smoke and fire identification sub-networks to obtain H sub-network parameters;
the aggregation algorithm construction unit is used for pre-constructing a federation aggregation algorithm and carrying out aggregation treatment on the H sub-network parameters based on the federation aggregation algorithm to obtain target global sub-network parameters;
and the parameter updating execution unit is used for issuing the target global sub-network parameters to the H building smoke and fire identification sub-networks to update the parameters so as to obtain the target smoke and fire identification sub-networks.
In one embodiment, the system further comprises:
the fire control layout calling unit is used for interactively determining K initial building fire control layouts of the K initial firework identification buildings;
The risk information calling unit is used for interactively obtaining a risk event position information set and a risk event processing time-consuming set by taking the pyrotechnic risk event set as a reference;
the fire control layout optimizing unit is used for optimizing the fire control layout of the K initial smoke and fire identification buildings based on the risk event position information set to obtain K optimized building fire control layouts;
the fire-fighting tool optimizing unit is used for optimizing the fire-fighting tools of the K initial firework identification buildings based on the risk event processing time-consuming set to obtain K target building fire-fighting layouts;
and the fire control layout adjusting unit is used for adjusting the K initial building fire control layouts by adopting the K target building fire control layouts.
In one embodiment, the system further comprises:
an update threshold setting unit for presetting a data update threshold;
the information updating execution unit is used for dynamically updating the target firework identification sub-network based on the data updating threshold value to obtain an updated firework identification sub-network;
and the management module updating unit is used for synchronizing the updated smoke and fire identification sub-network to the building smoke and fire identification management modules of the K initial smoke and fire identification buildings in a plurality of rounds.
Any of the methods or steps described above may be stored as computer instructions or programs in various non-limiting types of computer memories, and identified by various non-limiting types of computer processors, thereby implementing any of the methods or steps described above.
Based on the above-mentioned embodiments of the present invention, any improvements and modifications to the present invention without departing from the principles of the present invention should fall within the scope of the present invention.
Claims (8)
1. A federal learning-based pyrotechnic identification system, the system comprising:
the building information interaction module is used for interactively determining a smoke and fire identification building information set based on a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer;
the data screening execution module is used for presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value to obtain M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K;
The similar building merging module is used for obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained through building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K;
the identification network construction module is used for pre-constructing an initial firework identification sub-network, synchronizing the initial firework identification sub-network to the H optimized firework identification buildings and obtaining H initial identification sub-networks;
the recognition network training module is used for obtaining H building smoke and fire recognition sub-networks, wherein the H building smoke and fire recognition sub-networks are obtained by performing supervision training on the H initial recognition sub-networks by adopting the H groups of optimized smoke and fire recognition data;
the federation execution module is used for federating the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network;
the monitoring window setting module is used for presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set;
And the fire control layout optimization module is used for optimizing the building fire control layout of the K initial smoke and fire identification buildings based on the smoke and fire risk event set.
2. The system of claim 1, wherein H optimized pyrotechnic identification buildings are obtained, wherein the H optimized pyrotechnic identification buildings are mapped to H sets of optimized pyrotechnic identification data, the H optimized pyrotechnic identification buildings are obtained by building similarity analysis of the M invalid pyrotechnic identification buildings and H initial pyrotechnic identification buildings, the system further comprising:
the similarity recognition execution unit is used for pre-constructing a building similarity recognition module;
the design information interaction unit is used for interactively obtaining K groups of initial building design information, wherein the K groups of initial building design information are mapped with the K initial firework identification buildings one by one;
the design information splitting unit is used for obtaining M groups of invalid building design information and H groups of initial building design information, wherein the M groups of invalid building design information are obtained by taking the M invalid firework identification buildings as references and splitting the K groups of initial building design information;
the similar building identification unit is used for obtaining M groups of building similarity index sequences, wherein the M groups of building similarity index sequences are obtained by synchronizing the M groups of invalid building design information and the H groups of initial building design information to the building similarity identification module for building similarity analysis, and each group of building similarity index sequences in the M groups of building similarity index sequences comprises H building similarity indexes;
The similarity sequence extraction unit is used for extracting extremum values based on the M groups of building similarity index sequences to obtain M optimal smoke and fire identification buildings;
and the building virtual merging unit is used for correspondingly calling M groups of initial building design information based on the M optimal smoke and fire identification buildings, and carrying out building virtual merging on the M groups of initial building design information and the M groups of invalid building design information to obtain the H optimal smoke and fire identification buildings.
3. The system of claim 2, wherein the building similarity identification module is pre-constructed, the system further comprising:
the module constitution setting unit is used for the building similarity recognition module to comprise a building feature recognition unit, a building feature comparison unit and a building similarity calculation unit;
the sample data obtaining unit is used for obtaining a plurality of groups of sample building design information;
the building feature recognition system comprises a recognition rule setting unit and a recognition rule setting unit, wherein the recognition rule setting unit is used for presetting a building feature recognition rule, the building feature recognition rule comprises N building feature recognition indexes, and N is a positive integer;
the building feature marking unit is used for marking building feature identification on the plurality of groups of sample building design information based on the building feature identification rule to obtain a plurality of groups of sample building feature information sets;
The feature recognition training unit is used for constructing the building feature recognition unit based on a CNN neural network, and performing supervision training of the building feature recognition unit by adopting the plurality of groups of sample building design information and the plurality of groups of sample building feature information sets;
the feature comparison construction unit is used for pre-constructing a three-dimensional coordinate system in the building feature comparison unit, and carrying out feature comparison on the building feature information group output by the building feature recognition unit to obtain N building feature similarities, wherein the N building feature similarities are mapped with the N building feature recognition indexes one by one;
the feature weight distribution unit is used for pre-configuring building feature weight distribution in the building similarity calculation unit, wherein the building feature weight distribution comprises N building feature index weights, and the N building feature index weights are mapped with the N building feature identification indexes one by one;
and the similarity index calculation unit is used for inputting the N building feature similarities into the building similarity calculation unit, and calculating the N building feature index weights in the building similarity calculation unit to obtain a building similarity index.
4. The system of claim 1, wherein H building pyrotechnic identification sub-networks are obtained, wherein the H building pyrotechnic identification sub-networks are obtained by supervised training of the H initial identification sub-networks with the H sets of optimized pyrotechnic identification data, the system further comprising:
the information calling execution unit is used for calling and obtaining first optimized smoke and fire identification data and a first initial identification sub-network based on the H groups of optimized smoke and fire identification data and the H initial identification sub-networks;
a smoke and fire image obtaining unit, configured to include a plurality of building smoke and fire situation images in the first optimized smoke and fire identification data;
the module composition obtaining unit is used for the first initial identification sub-network to comprise a decoder module and an encoder module;
the smoke and fire characteristic identification unit is used for carrying out smoke and fire characteristic identification on the smoke and fire condition images of the plurality of buildings to obtain a plurality of groups of sample smoke and fire characteristics;
the identification network construction unit is used for training the decoder module and the encoder module by adopting the plurality of building smoke and fire condition images and the plurality of groups of sample smoke and fire characteristics to finish the construction of a first building smoke and fire identification sub-network;
And the network set generating unit is used for constructing and obtaining the H building smoke and fire identification sub-networks by adopting the H groups of optimized smoke and fire identification data.
5. The system of claim 4, wherein the H building pyrotechnic identification sub-networks are federally aggregated to generate a target pyrotechnic identification sub-network, the system further comprising:
the parameter extraction execution unit is used for extracting parameters of the H building smoke and fire identification sub-networks to obtain H sub-network parameters;
the aggregation algorithm construction unit is used for pre-constructing a federation aggregation algorithm and carrying out aggregation treatment on the H sub-network parameters based on the federation aggregation algorithm to obtain target global sub-network parameters;
and the parameter updating execution unit is used for issuing the target global sub-network parameters to the H building smoke and fire identification sub-networks to update the parameters so as to obtain the target smoke and fire identification sub-networks.
6. The system of claim 1, wherein building fire layout optimization of the K initial pyrotechnic identified buildings is performed based on the set of pyrotechnic risk events, the system further comprising:
the fire control layout calling unit is used for interactively determining K initial building fire control layouts of the K initial firework identification buildings;
The risk information calling unit is used for interactively obtaining a risk event position information set and a risk event processing time-consuming set by taking the pyrotechnic risk event set as a reference;
the fire control layout optimizing unit is used for optimizing the fire control layout of the K initial smoke and fire identification buildings based on the risk event position information set to obtain K optimized building fire control layouts;
the fire-fighting tool optimizing unit is used for optimizing the fire-fighting tools of the K initial firework identification buildings based on the risk event processing time-consuming set to obtain K target building fire-fighting layouts;
and the fire control layout adjusting unit is used for adjusting the K initial building fire control layouts by adopting the K target building fire control layouts.
7. The system of claim 1, wherein the system further comprises:
an update threshold setting unit for presetting a data update threshold;
the information updating execution unit is used for dynamically updating the target firework identification sub-network based on the data updating threshold value to obtain an updated firework identification sub-network;
and the management module updating unit is used for synchronizing the updated smoke and fire identification sub-network to the building smoke and fire identification management modules of the K initial smoke and fire identification buildings in a plurality of rounds.
8. A method of smoke identification based on federal learning, the method comprising:
determining a smoke and fire identification building information set based on interaction of a preset smoke and fire identification area, wherein the smoke and fire identification building information set comprises K initial smoke and fire identification buildings and K groups of smoke and fire identification data, and K is a positive integer;
presetting a training data screening threshold value, traversing the K groups of smoke and fire identification data based on the training data screening threshold value, and obtaining M groups of invalid smoke and fire identification data, wherein the M groups of invalid smoke and fire identification data are mapped to M invalid smoke and fire identification buildings, and M is a positive integer smaller than K;
obtaining H optimized smoke and fire identification buildings, wherein the H optimized smoke and fire identification buildings are mapped to H groups of optimized smoke and fire identification data, the H optimized smoke and fire identification buildings are obtained through building similarity analysis on the M invalid smoke and fire identification buildings and H initial smoke and fire identification buildings, H is a positive integer, and H+M=K;
pre-constructing an initial firework recognition sub-network, and synchronizing the initial firework recognition sub-network to the H optimized firework recognition buildings to obtain H initial recognition sub-networks;
obtaining H building smoke and fire identification sub-networks, wherein the H building smoke and fire identification sub-networks are obtained by performing supervision training on the H initial identification sub-networks by adopting the H groups of optimized smoke and fire identification data;
Federal aggregation is carried out on the H building smoke and fire identification sub-networks to generate a target smoke and fire identification sub-network;
presetting a smoke and fire risk monitoring window, and carrying out smoke and fire risk monitoring on the K initial smoke and fire identification buildings based on the target smoke and fire identification sub-network to obtain a smoke and fire risk event set;
and performing building fire layout optimization of the K initial smoke identification buildings based on the smoke risk event set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311194171.8A CN117237804B (en) | 2023-09-15 | 2023-09-15 | Pyrotechnical recognition system and method based on federal learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311194171.8A CN117237804B (en) | 2023-09-15 | 2023-09-15 | Pyrotechnical recognition system and method based on federal learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117237804A true CN117237804A (en) | 2023-12-15 |
| CN117237804B CN117237804B (en) | 2024-02-13 |
Family
ID=89097887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311194171.8A Active CN117237804B (en) | 2023-09-15 | 2023-09-15 | Pyrotechnical recognition system and method based on federal learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117237804B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109068272A (en) * | 2018-08-30 | 2018-12-21 | 北京三快在线科技有限公司 | Similar users recognition methods, device, equipment and readable storage medium storing program for executing |
| CN111131771A (en) * | 2019-12-12 | 2020-05-08 | 中国科学院深圳先进技术研究院 | Video monitoring system |
| CN111666856A (en) * | 2020-05-29 | 2020-09-15 | 武汉大学 | High-resolution single-polarization SAR image building target detection method based on structural characteristics |
| WO2021189906A1 (en) * | 2020-10-20 | 2021-09-30 | 平安科技（深圳）有限公司 | Target detection method and apparatus based on federated learning, and device and storage medium |
| CN116307138A (en) * | 2023-03-01 | 2023-06-23 | 中国科学院上海微系统与信息技术研究所 | Electrical fire early warning method, device, terminal and medium based on federal learning |
| CN116612595A (en) * | 2023-02-15 | 2023-08-18 | 南京林业大学 | Forest fire prediction method and system based on federal learning |
-
2023
- 2023-09-15 CN CN202311194171.8A patent/CN117237804B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109068272A (en) * | 2018-08-30 | 2018-12-21 | 北京三快在线科技有限公司 | Similar users recognition methods, device, equipment and readable storage medium storing program for executing |
| CN111131771A (en) * | 2019-12-12 | 2020-05-08 | 中国科学院深圳先进技术研究院 | Video monitoring system |
| CN111666856A (en) * | 2020-05-29 | 2020-09-15 | 武汉大学 | High-resolution single-polarization SAR image building target detection method based on structural characteristics |
| WO2021189906A1 (en) * | 2020-10-20 | 2021-09-30 | 平安科技（深圳）有限公司 | Target detection method and apparatus based on federated learning, and device and storage medium |
| CN116612595A (en) * | 2023-02-15 | 2023-08-18 | 南京林业大学 | Forest fire prediction method and system based on federal learning |
| CN116307138A (en) * | 2023-03-01 | 2023-06-23 | 中国科学院上海微系统与信息技术研究所 | Electrical fire early warning method, device, terminal and medium based on federal learning |
Non-Patent Citations (4)
| Title |
|---|
| RAED ABDEL SATER 等: "A Federated Learning Approach to Anomaly Detection in Smart Buildings", 《ACM TRANSACTIONS ON INTERNET OF THINGS》, vol. 2, no. 4, pages 1 - 23, XP055919068, DOI: 10.1145/3467981 * |
| SHARNIL PANDYA 等: "Federated learning for smart cities:A comprehensive survey", 《SUSTAINABLE ENERGY TECHNOLOGIES AND ASSESSMENTS》, pages 1 - 13 * |
| YUE HU 等: "Distributed Fire Detection and Localization Model Using Federated Learning", 《MATHEMATICS》, pages 1 - 19 * |
| 杨帆 等: "基于联邦学习的火灾图像检测算法", 《陕西科技大学学报》, vol. 40, no. 6, pages 184 - 191 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117237804B (en) | 2024-02-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Soft fuzzy rough sets and its application in decision making | |
| CN110222406B (en) | Unmanned aerial vehicle autonomous capacity assessment method based on task stage complexity | |
| CN106950945B (en) | A kind of fault detection method based on dimension changeable type independent component analysis model | |
| CN103793853A (en) | Overhead power transmission line running state assessment method based on bidirectional Bayesian network | |
| CN107967487A (en) | A kind of colliding data fusion method based on evidence distance and uncertainty | |
| Wang et al. | Research on fault diagnosis expert system based on the neural network and the fault tree technology | |
| CN108830335A (en) | Photovoltaic plant fault early warning method and system | |
| CN107632590A (en) | A kind of bottom event sort method based on priority | |
| CN117237804B (en) | Pyrotechnical recognition system and method based on federal learning | |
| CN111047476A (en) | Dam structure safety monitoring accurate prediction method and system based on RBF neural network | |
| CN104793492B (en) | A kind of indoor environmental characteristic extracting method based on entropy and gray relation grades | |
| CN113468724A (en) | Digital twin system simulation method and device for airport aircraft landing guidance | |
| CN108761263A (en) | A kind of fault diagnosis system based on evidence theory | |
| CN103197564B (en) | Digitizing man-machine interface monitor unit quantity optimization method and system | |
| CN113449966B (en) | Gypsum board equipment inspection method and system | |
| CN111143774B (en) | Power load prediction method and device based on influence factor multi-state model | |
| CN106529725A (en) | Gas outburst prediction method based on firefly algorithm and SOM network | |
| CN108345975A (en) | A kind of oil depot Failure risk evaluation method and device | |
| Bhinge et al. | An Intelligent Machine Monitoring System Using Gaussian Process Regression for Energy Prediction | |
| Peng et al. | Research conflict problems of DS evidence and its application in multi-sensor information fusion technology | |
| CN117575319B (en) | Method and system for evaluating urban flood disaster grade | |
| Wen et al. | Detection of Research Front Topic Based on Data of NSF Artificial Intelligence Project | |
| PENG et al. | Evaluation of Unmanned Ground Equipment Based on Improved Cloud Model Method | |
| Zhihong et al. | The fast estimation model of project cost based on BP neural network | |
| Xu et al. | Human factors analysis to Yangtze River water traffic accidents based on fault tree |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |