Fire thermal imaging analysis method and system based on artificial intelligence enhancement

Abstract

The present invention discloses a fire thermal imaging analysis method and system based on artificial intelligence enhancement, comprising: obtaining initial thermal imaging data collected at a fire scene; performing data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data; inputting the target thermal imaging data into a pre-trained target fire thermal imaging analysis model to obtain a fire analysis result. The target fire thermal imaging analysis model can be used to reliably analyze the target thermal imaging data, thereby improving the accuracy and comprehensiveness of thermal imaging data analysis.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a fire thermal imaging analysis method and system based on artificial intelligence enhancement. Background Art

[0002] Fire is one of the most frequent and common disasters that threaten public safety and social development. The application of thermal imaging technology in firefighting scenes is becoming more and more common. However, when a fire occurs, thermal imaging data is often subject to various interferences, making it impossible to perform reliable analysis. Therefore, how to accurately perform fire thermal imaging analysis is still an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to provide a fire thermal imaging analysis method and system based on artificial intelligence enhancement.

[0004] In a first aspect, an embodiment of the present invention provides a fire thermal imaging analysis method based on artificial intelligence enhancement, the method comprising:

[0005] Obtain the initial thermal imaging data collected at the fire scene;

[0006] Performing data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data;

[0007] The target thermal imaging data is input into a pre-trained target fire thermal imaging analysis model to obtain a fire analysis result.

[0008] In a second aspect, an embodiment of the present invention provides a server system, including a server, wherein the server is used to execute any one of the methods in the first aspect.

[0009] Compared with the prior art, the beneficial effects provided by the present invention include: first, initial thermal imaging data collected at the fire scene is acquired, and then the initial thermal imaging data is enhanced to obtain target thermal imaging data, and then the enhanced target thermal imaging data is input into the target fire thermal imaging analysis model. The target fire thermal imaging analysis model can be used to reliably analyze the target thermal imaging data, thereby improving the accuracy and comprehensiveness of the thermal imaging data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0011] Figure 1A schematic diagram of a flow chart of a fire thermal imaging analysis method based on artificial intelligence enhancement provided by an embodiment of the present invention;

[0012] Figure 2 A schematic block diagram of the structure of a server provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0014] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0015] In order to solve the technical problems in the aforementioned background technology, Figure 1 The following is a flow chart of a fire thermal imaging analysis method based on artificial intelligence enhancement provided by an embodiment of the present disclosure. The fire thermal imaging analysis method based on artificial intelligence enhancement is described in detail below. The execution subject of the method may be, for example, a server or a server system, or any other computer device with information processing capabilities, such as Figure 1 As shown, the method comprises the following steps:

[0016] Step S201: Acquire initial thermal imaging data collected at the fire scene.

[0017] In some embodiments, the initial thermal imaging data may be data collected in real time by a thermal imaging sensor in a fire scene.

[0018] Step S202: performing data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data.

[0019] In some embodiments, the purpose of the data enhancement project is to reduce the signal-to-noise ratio of the initial thermal imaging data, such as reducing the thermal imaging objects or data such as high-temperature dust in the initial thermal imaging data that may affect the analysis of the thermal imaging data, and mainly retaining thermal imaging data related to personnel.

[0020] Step S203: inputting the target thermal imaging data into a pre-trained target fire thermal imaging analysis model to obtain a fire analysis result.

[0021] In some embodiments, the fire analysis result includes at least information about whether there are victims in the fire scene. Optionally, the fire analysis result may also include information about the fire point or the spread of the fire, etc., which is not limited in the embodiments of the present disclosure.

[0022] In some embodiments, the target fire thermal imaging analysis model may be a model trained based on information richness, so that the target fire thermal imaging analysis model can extract as much information as possible from the thermal imaging data to the greatest extent possible, thereby ensuring the reliability of the thermal imaging analysis.

[0023] In an embodiment of the present application, first, initial thermal imaging data collected at the fire scene is acquired, and then the initial thermal imaging data is enhanced to obtain target thermal imaging data. The enhanced target thermal imaging data is then input into a target fire thermal imaging analysis model. The target fire thermal imaging analysis model can be used to perform reliable analysis of the target thermal imaging data, thereby improving the accuracy and comprehensiveness of the thermal imaging data analysis.

[0024] In some embodiments, performing data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data may include the following steps:

[0025] (11) The initial thermal imaging data is split to obtain high-energy imaging data in the initial thermal imaging data.

[0026] After obtaining the initial thermal imaging data, since the initial thermal imaging data may contain interference data, such as interference imaging caused by high-temperature dust, the interference data may refer to other thermal imaging data except high-energy imaging data. The high-energy imaging data and the interference data in the initial thermal imaging data may be split to obtain the high-energy imaging data in the initial thermal imaging data; the high-energy imaging data and the interference data in the initial thermal imaging data may also be split to obtain the high-energy imaging data in the initial thermal imaging data. The initial thermal imaging data may be directly processed, the phase information of the initial thermal imaging data may be retained, and better performance may be achieved.

[0027] Optionally, based on the quantitative features of the initial thermal imaging data, the initial thermal imaging data is subjected to mask processing to obtain a mask tensor corresponding to the initial thermal imaging data. Based on the mask tensor corresponding to the initial thermal imaging data, the high-energy imaging data is separated from the initial thermal imaging data.

[0028] The thermal imaging features of different heat sources can be distinguished based on the quantitative features of the initial thermal imaging data, that is, based on the quantitative features of the initial thermal imaging data, and then the mask tensors corresponding to different heat sources are constructed, and the mask tensors contain mask vectors corresponding to each heat source, and the high-energy imaging data is separated from the initial thermal imaging data based on the mask tensors corresponding to the initial thermal imaging data. That is, the thermal imaging data features are converted based on the mask vector corresponding to any heat source, so as to separate the thermal imaging data corresponding to different heat sources in the thermal imaging data.

[0029] (12) quantizing the initial thermal imaging data to obtain quantitative features of the initial thermal imaging data, and generating a data enhancement rate of the high-energy imaging data based on the quantitative features of the initial thermal imaging data.

[0030] The initial thermal imaging data may be quantified to obtain quantitative features of the initial thermal imaging data, wherein the quantitative features of the initial thermal imaging data include the regional area distribution of high-energy imaging data in the initial thermal imaging data, the regional areas of different thermal imaging data in the initial thermal imaging data, and the type of the initial thermal imaging data. The data enhancement rate of the high-energy imaging data is generated based on the quantitative features of the initial thermal imaging data.

[0031] In some embodiments, a method for generating a data enhancement rate for high energy imaging data includes the following operations:

[0032] Operation 1: A region segmentation operation may be performed on the initial thermal imaging data to obtain a plurality of initial thermal imaging data regions, and a region segmentation operation may be performed on the high-energy imaging data to obtain a plurality of high-energy imaging data regions.

[0033] In an optional embodiment, one initial thermal imaging data region corresponds to one high-energy imaging data region, and the initial thermal imaging data may be segmented to obtain multiple initial thermal imaging data regions, such as segmenting the initial thermal imaging data into K initial thermal imaging data regions, where K is a positive integer greater than or equal to 2. After the initial thermal imaging data is segmented to obtain thermal imaging data in the initial thermal imaging data, the high-energy imaging data may be segmented based on the method for segmenting the initial thermal imaging data to obtain multiple high-energy imaging data regions. Among them, one initial thermal imaging data region corresponds to one high-energy imaging data region, such as the third initial thermal imaging data region in the multiple initial thermal imaging data regions corresponds to the third high-energy imaging data region in the multiple high-energy imaging data regions.

[0034] Operation 2: performing data quantification on each of the multiple initial thermal imaging data regions to obtain a quantitative feature of each initial thermal imaging data region.

[0035] Data quantification is performed on each of the multiple initial thermal imaging data areas to obtain quantitative features of each initial thermal imaging data area. Similarly, the quantitative features of each initial thermal imaging data area include the regional area distribution of high-energy imaging data in the initial thermal imaging data area, the regional areas of different thermal imaging data in the original high-energy imaging data area, and the type of the initial thermal imaging data area.

[0036] Operation 3: generating a data enhancement rate of a corresponding high-energy imaging data region among a plurality of high-energy imaging data regions based on the quantitative features of each initial thermal imaging data region.

[0037] Operation 4: using the data enhancement rate corresponding to each high-energy imaging data region among the multiple high-energy imaging data regions as the data enhancement rate of the high-energy imaging data.

[0038] Then, based on the quantitative characteristics of each initial thermal imaging data area, the data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas is generated, that is, the data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas can be generated based on at least one of the regional area distribution of the high-energy imaging data in the initial thermal imaging data area, the regional area of ​​different thermal imaging data in the initial thermal imaging data area, and the type of the initial thermal imaging data area.

[0039] Optionally, when generating data enhancement rates of corresponding high-energy imaging data regions in multiple high-energy imaging data regions based on the quantitative characteristics of each initial thermal imaging data region, the regional area distribution of high-energy imaging data contained in each initial thermal imaging data region can be determined based on the quantitative characteristics of each initial thermal imaging data region, and the data enhancement rates of corresponding high-energy imaging data regions in multiple high-energy imaging data regions can be generated using the regional area distribution.

[0040] The regional area distribution of high-energy imaging data contained in each initial thermal imaging data region can be determined based on the quantitative characteristics of each initial thermal imaging data region. The regional area distribution refers to the proportion of high-energy imaging data in each initial thermal imaging data region. In response to the more high-energy imaging data in the initial thermal imaging data region, the larger the corresponding regional area distribution is, and in response to the less high-energy imaging data in the initial thermal imaging data region, the smaller the corresponding regional area distribution is. The regional area distribution is used to generate the data enhancement rate of the corresponding high-energy imaging data region in multiple high-energy imaging data regions. The data enhancement rate refers to the signal-to-noise ratio of the high-energy imaging data. In response to the higher signal-to-noise ratio of the high-energy imaging data, the higher the data enhancement rate corresponding to the high-energy imaging data is, and the intensity of the data enhancement engineering for the high-energy imaging data can be reduced; in response to the lower signal-to-noise ratio of the high-energy imaging data, the lower the data enhancement rate of the high-energy imaging data is, and the intensity of the data enhancement engineering for the high-energy imaging data can be increased. The data enhancement rate is used to determine whether it is necessary to divide part of the initial thermal imaging data from the initial thermal imaging data and fuse it into the high-energy imaging data to enhance the performance of the high-energy imaging data. The larger the regional area distribution of high-energy imaging data in the initial thermal imaging data area, the larger the data enhancement rate of the corresponding high-energy imaging data area; the smaller the regional area distribution of high-energy imaging data in the initial thermal imaging data area, the smaller the data enhancement rate of the corresponding high-energy imaging data area. Since the regional areas of high-energy imaging data contained in the thermal imaging data at each position in the initial thermal imaging data are different, that is, the degree of mixed interference data at each position in the initial thermal imaging data is different, the initial thermal imaging data is divided to obtain multiple initial thermal imaging data areas, and then the data enhancement rate of the corresponding high-energy imaging data area is generated based on the quantitative characteristics of each initial thermal imaging data area, so that the data enhancement project can be performed on the high-energy imaging data area more accurately, the signal-to-noise ratio of the target thermal imaging data after the data enhancement project is improved, and the performance of the target thermal imaging data after the data enhancement project is improved.

[0041] Optionally, the data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas can be generated based on the number of different thermal imaging data in the initial thermal imaging data area. In response to the larger number of different thermal imaging data, which indicates that the initial thermal imaging data is more mixed, the data enhancement rate will be smaller; in response to the smaller number of different thermal imaging data, which indicates that the initial thermal imaging data is clean, the data enhancement rate will be larger. The data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas can also be generated based on the type of the initial thermal imaging data area.

[0042] After the data enhancement rate of each high-energy imaging data region in the plurality of high-energy imaging data regions is obtained, the data enhancement rate corresponding to each high-energy imaging data region in the plurality of high-energy imaging data regions is used as the data enhancement rate of the high-energy imaging data.

[0043] Optionally, when determining the regional area distribution of high-energy imaging data contained in each initial thermal imaging data region based on the quantitative characteristics of each initial thermal imaging data region, the regional area of ​​high-energy imaging data contained in each initial thermal imaging data region can be determined based on the quantitative characteristics of each initial thermal imaging data region. Then, the total regional area of ​​the initial thermal imaging data is obtained, and the distribution ratio between the regional area of ​​the high-energy imaging data contained in each initial thermal imaging data region and the total regional area of ​​the initial thermal imaging data is obtained to obtain the regional area distribution of the high-energy imaging data contained in each initial thermal imaging data region.

[0044] The regional area of ​​high-energy imaging data contained in each initial thermal imaging data region can be determined based on the quantitative characteristics of each initial thermal imaging data region, and then the total regional area of ​​the initial thermal imaging data can be obtained. Then, the distribution ratio between the regional area of ​​high-energy imaging data contained in each initial thermal imaging data region and the total regional area of ​​the initial thermal imaging data is calculated, and the regional area distribution of high-energy imaging data contained in each initial thermal imaging data region can be obtained. In this way, the regional area distribution of high-energy imaging data contained in each initial thermal imaging data region can be known more accurately, and then based on the regional area distribution of high-energy imaging data contained in each initial thermal imaging data region, a data enhancement rate is generated, and a data enhancement project is performed on the high-energy imaging data region corresponding to the initial thermal imaging data region, which can improve the signal-to-noise ratio of the target thermal imaging data after the data enhancement project and improve the performance of the target thermal imaging data after the data enhancement project.

[0045] (13) Based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data, a data enhancement project is performed on the high-energy imaging data to obtain data-enhanced target thermal imaging data.

[0046] After obtaining the data enhancement rate of the high-energy imaging data, the data enhancement project can be performed on the high-energy imaging data based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data to obtain data-enhanced target thermal imaging data. In this way, the signal-to-noise ratio of the target thermal imaging data after the data enhancement project can be improved, and the performance damage after the high-energy imaging data is separated from the initial thermal imaging data can be reduced, that is, the performance of the target thermal imaging data after the data enhancement project can be improved.

[0047] Optionally, in response to the data enhancement rate of the target high-energy imaging data area being greater than the first data enhancement rate index and less than the second data enhancement rate index, an initial thermal imaging data sub-area of ​​the target area is demarcated from the target initial thermal imaging data area, and the initial thermal imaging data sub-area and the target high-energy imaging data area are regionally aggregated to obtain a data-enhanced target high-energy imaging data area. The target area is determined based on the regional area distribution of the high-energy imaging data contained in the target initial thermal imaging data area, and the first data enhancement rate index is less than the second data enhancement rate index; in response to the data enhancement rate of the target high-energy imaging data area being greater than or equal to the second data enhancement rate index, the target initial thermal imaging data area is used as a data-enhanced target high-energy imaging data area. Then, the data-enhanced target high-energy imaging data area is associated to obtain the data-enhanced target thermal imaging data.

[0048] Among them, the multiple initial thermal imaging data areas include a target initial thermal imaging data area, the target initial thermal imaging data area is any one of the multiple initial thermal imaging data areas, and the multiple high-energy imaging data areas include a target high-energy imaging data area corresponding to the target initial thermal imaging data area. In response to the data enhancement rate of the target high-energy imaging data area being greater than the first data enhancement rate index and less than the second data enhancement rate index, it is indicated that the target initial thermal imaging data area corresponding to the target high-energy imaging data area contains certain high-energy imaging data and certain interference data, that is, the target initial thermal imaging data area is mixed thermal imaging data. Since the target high-energy imaging data area is data split from the target initial thermal imaging data area, there is a problem of performance damage, so the initial thermal imaging data sub-area of ​​the target area area can be demarcated from the target initial thermal imaging data area, and the initial thermal imaging data sub-area and the target high-energy imaging data area are regionally aggregated to obtain the data enhanced target high-energy imaging data area. In this way, the signal-to-noise ratio of the target high-energy imaging data area can be significantly improved, and the performance damage of the target high-energy imaging data area can be reduced. The target region area is determined based on the regional area distribution of high-energy imaging data contained in the target initial thermal imaging data region, and the first data enhancement rate index is less than the second data enhancement rate index.

[0049] In response to the data enhancement rate of the target high-energy imaging data area being greater than or equal to the second data enhancement rate indicator, it means that the target initial thermal imaging data area corresponding to the target high-energy imaging data area only contains high-energy imaging data. Since there is no interference data in the target initial thermal imaging data area, the target high-energy imaging data area is obtained by splitting from the target initial thermal imaging data area, and the target high-energy imaging data area will also have the problem of performance damage. Therefore, the target initial thermal imaging data area can be used as a target high-energy imaging data area for data enhancement, so that the target high-energy imaging data area for data enhancement will not have the problem of performance damage. After obtaining the target high-energy imaging data area for data enhancement of each high-energy imaging data area in multiple high-energy imaging data areas, the target high-energy imaging data area for data enhancement of each high-energy imaging data area is then associated to obtain the target thermal imaging data for data enhancement.

[0050] Optionally, when generating data enhancement rates of corresponding high-energy imaging data regions in a plurality of high-energy imaging data regions using regional area distribution, in response to the regional area distribution corresponding to the target initial thermal imaging data region being greater than the first regional area distribution index and less than the second regional area distribution index, the first data enhancement rate is determined as the data enhancement rate of the target high-energy imaging data region. The first data enhancement rate is greater than the first data enhancement rate index and less than the second data enhancement rate index. In response to the regional area distribution corresponding to the target initial thermal imaging data region being greater than the second regional area distribution index, the second data enhancement rate is determined as the data enhancement rate of the target high-energy imaging data region, and the second data enhancement rate is greater than or equal to the second data enhancement rate index.

[0051] In which, in response to the area distribution of the target initial thermal imaging data region corresponding to the area distribution index of the first area, and less than the area distribution index of the second area, it is indicated that the target initial thermal imaging data region corresponding to the target high-energy imaging data region contains certain high-energy imaging data and certain interference data, that is, the target initial thermal imaging data region is mixed thermal imaging data. Then the first data enhancement rate can be determined as the data enhancement rate of the target high-energy imaging data region, and the first data enhancement rate is greater than the first data enhancement rate index and less than the second data enhancement rate index. In response to the area distribution of the target initial thermal imaging data region corresponding to the area distribution index of the second area, it is indicated that the target initial thermal imaging data region corresponding to the target high-energy imaging data region only contains high-energy imaging data, and the second data enhancement rate can be determined as the data enhancement rate of the target high-energy imaging data region, and the second data enhancement rate is greater than or equal to the second data enhancement rate index. In response to the area distribution of the target initial thermal imaging data region corresponding to the area distribution index of the first area, it is indicated that the target initial thermal imaging data region does not contain high-energy imaging data, but only contains interference data, and all zero data can be output. In this way, the interference data can be completely removed, and the signal-to-noise ratio of the target thermal imaging data can be improved.

[0052] For example, the first area area distribution index may be 0, and the second area area distribution index may be 1. In response to the area distribution of the corresponding area of ​​the target initial thermal imaging data area being greater than the first area area distribution index and less than the second area area distribution index, it is indicated that the target initial thermal imaging data area contains not only high-energy imaging data but also other thermal imaging data, and is mixed thermal imaging data. Then, the first data enhancement rate may be determined as the data enhancement rate of the target high-energy imaging data area, and the first data enhancement rate may be greater than the first data enhancement rate index and less than the second data enhancement rate index. The index value of the first data enhancement rate may be determined based on the area area distribution of the high-energy imaging data in the target initial thermal imaging data area. For example, the higher the area area distribution of the high-energy imaging data in the target initial thermal imaging data area, the higher the index value of the first data enhancement rate; the lower the area area distribution of the high-energy imaging data in the target initial thermal imaging data area, the lower the index value of the first data enhancement rate. When the data enhancement rate of the target high-energy imaging data area is the second data enhancement rate, it is indicated that the target initial thermal imaging data area only contains high-energy imaging data.

[0053] In some embodiments, the step of enhancing the initial thermal imaging data may specifically include the following operations:

[0054] Operation 1: Perform data splitting on the initial thermal imaging data to obtain high-energy imaging data in the initial thermal imaging data.

[0055] Operation 2: performing a region segmentation operation on the initial thermal imaging data to obtain a plurality of initial thermal imaging data regions, and performing a region segmentation operation on the high-energy imaging data to obtain a plurality of high-energy imaging data regions.

[0056] Operation 3: performing data quantification on each of the multiple initial thermal imaging data regions to obtain a quantitative feature of each initial thermal imaging data region.

[0057] Operation 4: generating a data enhancement rate of a corresponding high-energy imaging data region among a plurality of high-energy imaging data regions based on the quantitative features of each initial thermal imaging data region.

[0058] Operation 5: using the data enhancement rate corresponding to each high-energy imaging data region among the multiple high-energy imaging data regions as the data enhancement rate of the high-energy imaging data.

[0059] Operation 6: Based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data, a data enhancement project is performed on the high-energy imaging data to obtain data-enhanced target thermal imaging data.

[0060] Operation 7: A region segmentation operation may be performed on the initial thermal imaging data to obtain a plurality of initial thermal imaging data regions, and a region segmentation operation may be performed on the high-energy imaging data to obtain a plurality of high-energy imaging data regions.

[0061] In the embodiment of the present application, the initial thermal imaging data may be subjected to a region segmentation operation to obtain a plurality of initial thermal imaging data regions, such as dividing the initial thermal imaging data into K initial thermal imaging data regions, where K is a positive integer greater than or equal to 2. After the initial thermal imaging data is subjected to data splitting to obtain thermal imaging data in the initial thermal imaging data, the high-energy imaging data may be segmented based on the method for segmenting the initial thermal imaging data to obtain a plurality of high-energy imaging data regions. Among them, one initial thermal imaging data region corresponds to one high-energy imaging data region.

[0062] Data quantification is performed on each of the multiple initial thermal imaging data areas to obtain quantitative features of each initial thermal imaging data area. Similarly, the quantitative features of each initial thermal imaging data area include the regional area distribution of high-energy imaging data in the initial thermal imaging data area, the regional areas of different thermal imaging data in the original high-energy imaging data area, and the type of the initial thermal imaging data area.

[0063] Then, based on the quantitative characteristics of each initial thermal imaging data area, the data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas is generated, that is, the data enhancement rate of the corresponding high-energy imaging data area in the multiple high-energy imaging data areas can be generated based on at least one of the regional area distribution of the high-energy imaging data in the initial thermal imaging data area, the regional area of ​​different thermal imaging data in the initial thermal imaging data area, and the type of the initial thermal imaging data area.

[0064] After obtaining the data enhancement rate of the high-energy imaging data, the data enhancement project can be performed on the high-energy imaging data based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data to obtain data-enhanced target thermal imaging data. In this way, the signal-to-noise ratio of the target thermal imaging data after the data enhancement project can be improved, and the performance damage after the high-energy imaging data is separated from the initial thermal imaging data can be reduced, that is, the performance of the target thermal imaging data after the data enhancement project can be improved.

[0065] In one embodiment, the training of the target fire thermal imaging analysis model may include the following steps:

[0066] (1) Obtain a sample thermal imaging dataset.

[0067] The sample thermal imaging data set includes a plurality of thermal imaging data for training a fire thermal imaging analysis model.

[0068] (2) Inputting the sample thermal imaging data set into the first thermal imaging data analysis model to update the model parameters, thereby obtaining a second thermal imaging data analysis model, wherein the second thermal imaging data analysis model includes a first data analysis module and a first collaborative analysis module.

[0069] Among them, the first thermal imaging data analysis model may refer to a thermal imaging data analysis model with model parameters initialized. The thermal imaging data analysis model is a model during training. The first thermal imaging data analysis model includes an initial data analysis module and an initial collaborative analysis module. The initial data analysis module refers to a data analysis module with parameter index values ​​initialized. The data analysis module is used to perform data analysis and obtain feature vectors in thermal imaging data. During training, the data analysis module is the main network model for training. The initial collaborative analysis module refers to a collaborative analysis module with parameter index values ​​initialized. The collaborative analysis module is also used for data analysis when used. During training, the collaborative analysis module is used to assist in the training of the data analysis module. The second thermal imaging data analysis model refers to a thermal imaging data analysis model obtained after the parameters of the first thermal imaging data analysis model are updated. The first data analysis module refers to the data analysis module after the parameter index values ​​are updated. The first collaborative analysis module refers to the collaborative analysis module after the parameter index values ​​are updated.

[0070] Specifically, a sample thermal imaging data set is input into a first thermal imaging data analysis model, which performs feature extraction through an initial data analysis module and an initial collaborative analysis module, respectively, and determines the loss of the feature extraction results. The loss of the feature extraction results is then used to reversely update the parameter index value of the first thermal imaging data analysis model to obtain a second thermal imaging data analysis model, which includes the updated first data analysis module and the first collaborative analysis module.

[0071] In one embodiment, a sample thermal imaging data set can be obtained, and the sample thermal imaging data set is subjected to multiple rounds of parameter updates, and all training samples are processed once for each round of parameter updates. In each round of parameters, the sample thermal imaging data set can be divided into multiple batches of sample subsets, and in each batch training, the sample subset is obtained to update the model parameters. That is, the sample subset is input into the first thermal imaging data analysis model, and feature extraction is performed through the initial data analysis module and the initial collaborative analysis module respectively, and the loss of the feature extraction result corresponding to the sample subset is determined, and then the loss of the feature extraction result corresponding to the sample subset is used to reversely update the parameter index value of the first thermal imaging data analysis model to obtain the second thermal imaging data analysis model.

[0072] (3) Determine the data analysis richness of the analysis path in the first data analysis module, and extract the analysis path to be optimized from the analysis path of the first data analysis module based on the data analysis richness.

[0073] Among them, the analysis path can refer to a channel, and the data analysis module can include multiple analysis paths. The analysis path is a filter. Different analysis paths correspond to different parameter index values, which are used to extract features from the input data using the parameter index values. The analysis path in the first data analysis module refers to the channel output in the first data analysis module. The data analysis richness is used to indicate the activation level of the analysis path in the corresponding first data analysis module. The higher the activation level, the higher the richness of the feature values ​​output by the corresponding analysis path, and the analysis path is a valid analysis path. Conversely, the lower the activation level, the lower the richness of the feature values ​​output by the corresponding analysis path, and the analysis path may be an invalid analysis path. The analysis path to be optimized refers to an analysis path with low data analysis richness, which has no substantial effect on data analysis and is an invalid analysis path. The parameter index value corresponding to the analysis path to be optimized needs to be optimized through the first collaborative analysis module.

[0074] Specifically, the parameter index value of the analysis path in the first data analysis module can be used to determine the data analysis richness corresponding to each analysis path in the data analysis module. Then, the data analysis richness corresponding to each analysis path is used to extract the analysis paths with lower data analysis richness from the analysis paths of the first data analysis module to obtain the analysis paths to be optimized. In one embodiment, the analysis paths in the first data analysis module can be sorted from small to large according to the data analysis richness corresponding to the analysis paths, and then a certain number of analysis paths in the first data analysis module can be selected in sequence from small to large to obtain the analysis paths to be optimized. Among them, the number of selections can be set according to demand.

[0075] (4) Determine the collaborative analysis richness of the analysis path in the first collaborative analysis module, and extract a benchmark analysis path from the analysis path in the first collaborative analysis module based on the collaborative analysis richness.

[0076] The collaborative analysis richness refers to the activation degree of the analysis path in the first collaborative analysis module. The benchmark analysis path refers to the analysis path with a higher activation degree in the first collaborative analysis module. The parameter index value of the benchmark analysis path can be used to optimize the parameter index value of the analysis path to be optimized.

[0077] Specifically, the parameter index value of the analysis path in the first collaborative analysis module can be used to determine the collaborative analysis richness corresponding to each analysis path in the collaborative analysis module. Then, the collaborative analysis richness corresponding to each analysis path is used to extract the analysis path with higher collaborative analysis richness from the analysis path of the first collaborative analysis module to obtain the benchmark analysis path.

[0078] In one embodiment, the analysis paths in the first data analysis module may be sorted from large to small according to the data analysis richness corresponding to the analysis paths, and then a certain number of analysis paths in the first collaborative analysis module may be selected in order from large to small to obtain various benchmark analysis paths. The number of selections may be set according to requirements and may be the same as or different from the number of analysis paths to be optimized.

[0079] (5) Optimizing the parameter index values ​​of the analysis path to be optimized based on the parameter index values ​​of the benchmark analysis path to obtain the optimized index values ​​of the analysis path to be optimized, obtaining a target data analysis module based on the optimized index values ​​of the analysis path to be optimized, and obtaining a target thermal imaging data analysis model based on the target data analysis module.

[0080] The optimization index value of the analysis path to be optimized refers to the parameter index value of the analysis path to be optimized obtained by optimizing the parameter index value of the benchmark analysis path. The target data analysis module refers to the data analysis module after optimizing the parameter index value of the analysis path to be optimized using the parameter index value of the benchmark analysis path. The target thermal imaging data analysis model refers to the model after optimizing the parameter index value of the analysis path to be optimized in the data analysis module using the parameter index value of the benchmark analysis path in the collaborative analysis module.

[0081] Specifically, the parameter index value of the benchmark analysis path and the parameter index value of the analysis path to be optimized may be weighted and summed to obtain the optimization index value of the analysis path to be optimized.

[0082] In one embodiment, the parameter index value of the benchmark analysis path can also be used to optimize the parameter index value of the analysis path to be optimized to obtain the optimization index value of the analysis path to be optimized. For example, the parameter index value of the benchmark analysis path can be used to replace the parameter index value of the analysis path to be optimized to obtain the optimization index value of the analysis path to be optimized.

[0083] (6) Determine whether the training is completed. In response to the optimization stop threshold being met, execute operation 2a. In response to the optimization stop threshold not being met, execute operation 2b, and return to step (2) for iterative execution.

[0084] Among them, judging whether the training meets the stopping optimization threshold may include that the loss of model training reaches a preset loss value or the parameter index value of the model no longer changes or the loss of model training remains unchanged, etc. For example, the loss of model training has not decreased for 10 consecutive rounds.

[0085] Operation 2b, using the target thermal imaging data analysis model as the first thermal imaging data analysis model, and returning to the step of inputting the sample thermal imaging data set into the first thermal imaging data analysis model to update the model parameters, thereby obtaining the second thermal imaging data analysis model.

[0086] Operation 2a, obtaining a target data analysis model, and obtaining a target fire thermal imaging analysis model according to a data analysis module in the target data analysis model. The target fire thermal imaging analysis model is used to perform data analysis on the input thermal imaging data.

[0087] The target data analysis model refers to a trained data analysis model. The target fire thermal imaging analysis model refers to a model that removes the collaborative analysis module and only retains the data analysis module.

[0088] Specifically, when the training is not completed, the target thermal imaging data analysis model can be used as the first thermal imaging data analysis model for parameter update, and in response to meeting the stop optimization threshold, the second thermal imaging data analysis model obtained by the last parameter update is used as the target data analysis model for training. Then, the collaborative analysis module is deleted from the target data analysis model, and the retained data analysis module is obtained to obtain the target fire thermal imaging analysis model.

[0089] In one embodiment, after obtaining the second thermal imaging data analysis model, it is also possible to determine whether the training is completed, that is, before determining the data analysis richness of the analysis path in the first data analysis module and extracting the analysis path to be optimized from the analysis path of the first data analysis module based on the data analysis richness, it is determined whether the model training is completed. When it is determined that the training is completed, the target data analysis model is obtained, and the target fire thermal imaging analysis model is obtained based on the data analysis module in the target data analysis model. When it is determined that the training is not completed, the collaborative analysis richness of the analysis path in the first collaborative analysis module is determined, and the benchmark analysis path is extracted from the analysis path in the first collaborative analysis module based on the collaborative analysis richness; the parameter index value of the analysis path to be optimized is optimized based on the parameter index value of the benchmark analysis path to obtain the optimized index value of the analysis path to be optimized, the target data analysis module is obtained based on the optimized index value of the analysis path to be optimized, the target thermal imaging data analysis model is obtained based on the target data analysis module, the target thermal imaging data analysis model is used as the first thermal imaging data analysis model, and the step (2) is returned.

[0090] In the above embodiment, after training the first thermal imaging data analysis model, the analysis path to be optimized is extracted from the analysis path of the first data analysis module using the data analysis richness of the analysis path in the first data analysis module, and the benchmark analysis path is extracted from the analysis path of the first collaborative analysis module using the collaborative analysis richness of the analysis path in the first collaborative analysis module. Then, the parameter index value of the analysis path to be optimized is optimized according to the parameter index value of the benchmark analysis path to obtain the optimization index value of the analysis path to be optimized, the target data analysis module is obtained according to the optimization index value of the analysis path to be optimized, the target thermal imaging data analysis model is obtained according to the target data analysis module, and then the target thermal imaging data analysis model is used as the first thermal imaging data analysis model until the optimization stop threshold is met, and the target data analysis model is obtained, and the target fire thermal imaging analysis model is obtained according to the data analysis module in the target data analysis model. Since the parameter index values ​​of the analysis path to be optimized in the data analysis module are optimized using the benchmark analysis path parameters in the collaborative analysis module, the data analysis module can extract features with rich thermal imaging information, and then the target fire thermal imaging analysis model is obtained using the data analysis module, so that the thermal imaging features extracted by the target fire thermal imaging analysis model have improved information richness, thereby improving the accuracy of thermal imaging data analysis.

[0091] In one embodiment, determining the data analysis richness of the analysis path in the first data analysis module, and extracting the analysis path to be optimized from the analysis path of the first data analysis module according to the data analysis richness, includes:

[0092] Operation 1: Use the parameter index values ​​corresponding to each analysis path in the first data analysis module to evaluate the expected value of information volume, and obtain the data analysis richness corresponding to each analysis path in the first data analysis module.

[0093] Among them, data analysis richness refers to the degree of activation determined by using the parameter indicator value corresponding to the analysis path in the data analysis module. The degree of activation is used to measure the importance of the corresponding analysis path. The higher the degree of activation, the more rich information the features extracted from the analysis path have.

[0094] Specifically, each parameter indicator value corresponding to each analysis path in the first data analysis module is used to evaluate the expected value of the amount of information, and the determined expected value of the amount of information, such as information entropy, is directly used as the data analysis richness corresponding to each analysis path in the first data analysis module.

[0095] Operation 2: Compare each data analysis richness with a preset data analysis richness index, and extract analysis paths that do not reach the preset data analysis richness index to obtain each analysis path to be optimized.

[0096] The preset data analysis richness index refers to a pre-set data analysis richness index. When it is lower than the preset data analysis richness index, it means that the feature richness extracted by the analysis path is low, the information content is small, and it may be an invalid analysis path. The analysis path to be optimized refers to the analysis path whose parameter index value needs to be updated.

[0097] Specifically, each data analysis richness is compared with a preset data analysis richness index, and analysis paths that do not reach the preset data analysis richness index are extracted as analysis paths to be optimized.

[0098] In one embodiment, after comparing each data analysis richness with a preset data analysis richness index and extracting analysis paths that do not reach the preset data analysis richness index to obtain each analysis path to be optimized, the method further includes:

[0099] The number of paths to be optimized for each analysis path to be optimized is determined, and in response to the number of paths to be optimized being higher than or equal to or lower than a preset number index, the second thermal imaging data analysis model is used as the first thermal imaging data analysis model, and the sample thermal imaging data set is input into the first thermal imaging data analysis model to update the model parameters to obtain the second thermal imaging data analysis model. In response to the number of paths to be optimized being higher than or equal to a preset number index, the step of determining the collaborative analysis richness of the analysis paths in the first collaborative analysis module is performed, and the step of extracting a benchmark analysis path from the analysis paths in the first collaborative analysis module according to the collaborative analysis richness is performed.

[0100] The preset number index refers to the number of analysis paths that are preset to update the parameter index values ​​of the analysis paths to be optimized.

[0101] Specifically, the number of paths to be optimized of each analysis path to be optimized is determined, and whether to update the parameter index value in the analysis path to be optimized is triggered according to the number of paths to be optimized. That is, the number of paths to be optimized is compared with the preset number index. In response to the number of paths to be optimized being higher than or equal to or lower than the preset number index, the parameter index value of the analysis path to be optimized is not updated, and the process directly returns to step (2) for execution. In response to the number of paths to be optimized being higher than or equal to the preset number index, it indicates that the update of the parameter index value in the analysis path to be optimized is triggered, and step (4) is executed at this time.

[0102] In the above embodiment, the parameter index value in the analysis path to be optimized is updated only when it is determined that the number of paths to be optimized is greater than or equal to the preset number index, which can improve the accuracy of updating the parameter index value of the analysis path to be optimized and improve the training efficiency.

[0103] In one embodiment, determining the collaborative analysis richness of the analysis path in the first collaborative analysis module, and extracting a benchmark analysis path from the analysis path in the first collaborative analysis module according to the collaborative analysis richness, includes:

[0104] Operation 1: Use the parameter index values ​​corresponding to each analysis path in the first collaborative analysis module to evaluate the expected value of the amount of information, and obtain the collaborative analysis richness corresponding to each analysis path in the first collaborative analysis module, where the collaborative analysis richness refers to the degree of activation determined by using the parameter index values ​​corresponding to the analysis paths in the collaborative analysis module.

[0105] Specifically, each parameter indicator value corresponding to each analysis path in the first collaborative analysis module is used to evaluate the expected value of information volume, and the determined expected value of information volume is directly used as the collaborative analysis richness corresponding to each analysis path in the first collaborative analysis module.

[0106] Operation 2: Compare each collaborative analysis richness with the preset collaborative analysis richness index, and extract the analysis path that reaches the preset data analysis richness index to obtain each benchmark analysis path.

[0107] The preset collaborative analysis richness index refers to a preset collaborative analysis richness threshold. When it is greater than the preset data analysis richness index, it indicates that the features extracted by the analysis path have a high amount of information. The benchmark analysis path refers to the analysis path in the collaborative analysis module that needs to be used when updating the optimized analysis path.

[0108] Specifically, each collaborative analysis richness is compared with a preset collaborative analysis richness index, and an analysis path that reaches the preset data analysis richness index is extracted, that is, an analysis path that can extract a relatively large amount of information is selected to obtain each benchmark analysis path.

[0109] In the above embodiment, the expected value of the information amount of the analysis path is determined to obtain the activation degree of the analysis path, and then the analysis path to be optimized in the data analysis module and the benchmark analysis path in the collaborative analysis module are determined, thereby improving the accuracy of the analysis path to be optimized and the benchmark analysis path.

[0110] In one embodiment, the parameter index value of the analysis path to be optimized is optimized according to the parameter index value of the benchmark analysis path to obtain the optimization index value of the analysis path to be optimized, including:

[0111] Operation 1: Determine the number of reference paths of the reference analysis path, and obtain the number of paths to be optimized of the analysis path to be optimized.

[0112] Operation 2: Compare the number of benchmark paths with the number of paths to be optimized to obtain quantitative relationship information, and determine the benchmark analysis path of the number of paths to be optimized from the benchmark analysis path of the benchmark path number based on the quantitative relationship information.

[0113] Specifically, it is necessary to determine the benchmark analysis paths that are consistent with the number of analysis paths to be optimized from the benchmark analysis paths, so that the parameter index values ​​of the analysis paths can be updated one by one. At this time, the number of benchmark paths of the benchmark analysis paths is determined, and the number of paths to be optimized of the analysis paths to be optimized is obtained, and the number of benchmark paths is compared with the number of paths to be optimized to obtain quantitative relationship information, and the benchmark analysis paths with the number of paths to be optimized are determined from the benchmark analysis paths with the number of benchmark paths according to the quantitative relationship information.

[0114] Operation 3: weighted summation is performed based on the parameter index value of the benchmark analysis path of the number of paths to be optimized and the parameter index value of the analysis path to be optimized of the number of paths to be optimized, to obtain the optimization parameter index value of the analysis path to be optimized of the number of paths to be optimized.

[0115] Specifically, a preset weight is obtained, and the parameter index value of the benchmark analysis path and the corresponding parameter index value of the analysis path to be optimized are weighted by using the weight, and the weighted sum is determined to obtain the optimized parameter index value of the analysis path to be optimized. The optimized parameter index value of each analysis path to be optimized is determined in turn, and finally the optimized parameter index value of the analysis path to be optimized of the number of paths to be optimized is obtained.

[0116] In one embodiment, determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the benchmark path number according to the quantity relationship information includes:

[0117] In response to the fact that the number of paths to be optimized is greater than the number of reference paths, a quantitative offset between the number of paths to be optimized and the number of reference paths is determined, and the reference analysis paths with the quantitative offset are extracted in a disorderly manner from the reference analysis paths of the number of paths to be optimized; and the reference analysis paths of the number of paths to be optimized are obtained based on the reference path number and the reference analysis paths with the quantitative offset.

[0118] Specifically, when it is determined that the number of paths to be optimized is greater than the number of benchmark paths, that is, the number of benchmark analysis paths is small, at this time, the numerical offset between the number of paths to be optimized and the number of benchmark paths is determined, and the benchmark analysis paths of the number of paths to be optimized are randomly and non-repetitively extracted from the benchmark analysis paths of the number of paths to be optimized, and the benchmark analysis paths of the number of paths to be optimized are obtained by summing them up with the existing benchmark analysis paths, thereby improving the accuracy of the obtained benchmark analysis paths.

[0119] In one embodiment, a benchmark analysis path of the number of paths to be optimized is determined from a benchmark analysis path of the number of paths according to the quantitative relationship information, including: in response to the number of benchmark paths being equal to the number of paths to be optimized, directly using the benchmark analysis path of the number of paths as the benchmark analysis path of the number of paths to be optimized.

[0120] In one embodiment, determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the benchmark path number according to the quantity relationship information includes:

[0121] Operation 1: In response to the number of paths to be optimized being less than the number of reference paths, obtaining analysis paths other than the analysis paths to be optimized from the analysis paths of the first data analysis module to obtain at least one candidate analysis path.

[0122] The candidate analysis path refers to an analysis path in the analysis path of the first data analysis module that is higher than a preset data analysis richness index.

[0123] Specifically, in response to the number of paths to be optimized being less than the number of reference paths, it is indicated that the paths to be optimized with the number of paths to be optimized need to be selected from the reference analysis paths. At this time, the analysis paths other than the paths to be optimized are obtained from the analysis paths of the first data analysis module to obtain at least one candidate analysis path.

[0124] Operation 2: Determine information residual between the benchmark analysis paths of the benchmark path number and at least one candidate analysis path, and obtain analysis deviations between each benchmark analysis path and at least one candidate analysis path.

[0125] The information residual information refers to the residual of the expected value of the information between the parameter index value of the benchmark analysis path and the parameter index value of the candidate analysis path, which can be, for example, cross entropy or KL divergence, etc.

[0126] Specifically, the parameter index value of the baseline analysis path and the parameter index value of the candidate analysis path are used to determine the information amount residual information between the baseline analysis path and each candidate analysis path, and the respective information amount residual information corresponding to the baseline analysis path is obtained, and the respective information amount residual information is directly used as the analysis deviation between the baseline analysis path and at least one candidate analysis path.

[0127] Operation 3: Determine the target analysis deviation corresponding to each benchmark analysis path from each analysis deviation, and select the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the target analysis deviation.

[0128] The target analysis deviation refers to the minimum analysis deviation, which is used to characterize the analysis deviation between the benchmark analysis path and the candidate analysis path.

[0129] Specifically, each analysis deviation is compared, and the smallest analysis deviation is selected as the target analysis deviation corresponding to the benchmark analysis path. Then, the target analysis deviations corresponding to each benchmark analysis path are sorted from large to small, and then the benchmark analysis paths with the number of paths to be optimized are selected from large to small, thereby improving the accuracy of the obtained benchmark analysis paths.

[0130] In one embodiment, determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the benchmark path number according to the quantity relationship information includes:

[0131] In response to the number of paths to be optimized being less than the number of reference paths, the reference analysis paths of the number of paths to be optimized are extracted out of order from the reference analysis paths of the number of reference paths.

[0132] Specifically, when the number of paths to be optimized is less than the number of reference paths, the reference analysis paths of the reference paths can be directly extracted in random order from the reference analysis paths of the reference paths, which can improve efficiency. On the other hand, an embodiment of the present invention also provides a fire thermal imaging analysis system or server system based on artificial intelligence enhancement, the system includes at least one server, which can be provided as Figure 2 The server 100 shown may be used to execute some or all of the steps involved in the above method embodiments, such as obtaining data or models from other computers or other servers in the fire thermal imaging analysis system enhanced based on artificial intelligence, but is not limited to this and will not be elaborated here.

[0133] The embodiment of the present invention provides a server 100, which includes a processor and a non-volatile memory storing a certain machine instruction. When the certain machine instruction is executed by the processor, the server 100 executes the aforementioned fire thermal imaging analysis method based on artificial intelligence enhancement. Figure 2 As shown, Figure 2 The server 100 is a block diagram of a structure of a server 100 provided in an embodiment of the present invention. The server 100 includes a memory 111, a processor 112, and a communication unit 113. To achieve data transmission or interaction, the memory 111, the processor 112, and the communication unit 113 are electrically connected to each other directly or indirectly. For example, one or more communication buses or signal lines can be used to achieve electrical connection between these components.

[0134] For illustrative purposes, the foregoing characterizations are made with reference to specific embodiments. However, the above illustrative discussions are not intended to be exhaustive or to limit the disclosure to the precise form disclosed. Based on the above teachings, numerous modifications and variations are possible. These embodiments are selected and characterized in order to best illustrate the principles of the disclosure and its practical application, so that those skilled in the art can best utilize the disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. A fire thermal imaging analysis method based on artificial intelligence enhancement, characterized in that: The method comprises: Obtain the initial thermal imaging data collected at the fire scene; Performing data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data; Inputting the target thermal imaging data into a pre-trained target fire thermal imaging analysis model to obtain a fire analysis result; The training of the target fire thermal imaging analysis model includes: Get a sample thermal imaging dataset; Inputting the sample thermal imaging data set into a first thermal imaging data analysis model to update model parameters, thereby obtaining a second thermal imaging data analysis model, wherein the second thermal imaging data analysis model includes a first data analysis module and a first collaborative analysis module; Determine the data analysis richness of the analysis path in the first data analysis module, and extract the analysis path to be optimized from the analysis path of the first data analysis module according to the data analysis richness; Determine the collaborative analysis richness of the analysis paths in the first collaborative analysis module, and extract a benchmark analysis path from the analysis paths in the first collaborative analysis module according to the collaborative analysis richness; Optimizing the parameter index value of the analysis path to be optimized according to the parameter index value of the benchmark analysis path to obtain the optimization index value of the analysis path to be optimized, obtaining a target data analysis module according to the optimization index value of the analysis path to be optimized, and obtaining a target thermal imaging data analysis model according to the target data analysis module; The target thermal imaging data analysis model is used as the first thermal imaging data analysis model, and the step of inputting the sample thermal imaging data set into the first thermal imaging data analysis model to update the model parameters to obtain the second thermal imaging data analysis model is returned until the optimization stop threshold is met to obtain the target data analysis model, and the target fire thermal imaging analysis model is obtained according to the data analysis module in the target data analysis model, and the target fire thermal imaging analysis model is used to perform data analysis on the input thermal imaging data.

2. The method according to claim 1, characterized in that: The performing of data enhancement engineering on the initial thermal imaging data to obtain target thermal imaging data includes: Performing data splitting on the initial thermal imaging data to obtain high-energy imaging data in the initial thermal imaging data; quantizing the initial thermal imaging data to obtain quantitative features of the initial thermal imaging data, and generating a data enhancement rate of the high-energy imaging data based on the quantitative features of the initial thermal imaging data; Based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data, a data enhancement project is performed on the high-energy imaging data to obtain data-enhanced target thermal imaging data.

3. The method according to claim 2, characterized in that The step of quantizing the initial thermal imaging data to obtain the quantitative features of the initial thermal imaging data, and generating the data enhancement rate of the high-energy imaging data based on the quantitative features of the initial thermal imaging data includes: Performing a region segmentation operation on the initial thermal imaging data to obtain a plurality of initial thermal imaging data regions, and performing a region segmentation operation on the high-energy imaging data to obtain a plurality of high-energy imaging data regions, wherein one initial thermal imaging data region corresponds to one high-energy imaging data region; quantifying each of the multiple initial thermal imaging data regions to obtain a quantitative feature of each of the initial thermal imaging data regions; Generating data enhancement rates of corresponding high-energy imaging data regions in the plurality of high-energy imaging data regions based on the quantitative features of each of the initial thermal imaging data regions; using the data enhancement rate corresponding to each high-energy imaging data region among the plurality of high-energy imaging data regions as the data enhancement rate of the high-energy imaging data; The step of generating the data enhancement rate of the corresponding high-energy imaging data region among the plurality of high-energy imaging data regions based on the quantitative features of each of the initial thermal imaging data regions comprises: Determining the regional area distribution of the high-energy imaging data contained in each of the initial thermal imaging data regions based on the quantitative characteristics of each of the initial thermal imaging data regions; Generate data enhancement rates of corresponding high-energy imaging data regions in the plurality of high-energy imaging data regions by using the region area distribution; The determining, based on the quantitative characteristics of each of the initial thermal imaging data regions, the regional area distribution of the high-energy imaging data contained in each of the initial thermal imaging data regions comprises: Determining the area of ​​the high-energy imaging data contained in each of the initial thermal imaging data regions based on the quantitative characteristics of each of the initial thermal imaging data regions; Acquire the total area of ​​the initial thermal imaging data; Obtaining a distribution ratio between the area of ​​the high-energy imaging data contained in each of the initial thermal imaging data areas and the total area of ​​the initial thermal imaging data, to obtain an area distribution of the high-energy imaging data contained in each of the initial thermal imaging data areas; The multiple initial thermal imaging data regions include a target initial thermal imaging data region, and the multiple high-energy imaging data regions include a target high-energy imaging data region corresponding to the target initial thermal imaging data region; The step of performing a data enhancement project on the high-energy imaging data based on the data enhancement rate of the high-energy imaging data and the initial thermal imaging data to obtain data-enhanced target thermal imaging data includes: In response to the data enhancement rate of the target high-energy imaging data region being greater than a first data enhancement rate indicator and less than a second data enhancement rate indicator, an initial thermal imaging data sub-region of the target region area is demarcated from the target initial thermal imaging data region, and the initial thermal imaging data sub-region and the target high-energy imaging data region are regionally aggregated to obtain a data-enhanced target high-energy imaging data region; the target region area is determined based on the regional area distribution of the high-energy imaging data contained in the target initial thermal imaging data region, and the first data enhancement rate indicator is less than the second data enhancement rate indicator; In response to the data enhancement rate of the target high-energy imaging data region being greater than or equal to the second data enhancement rate indicator, taking the target initial thermal imaging data region as the target high-energy imaging data region for data enhancement; The data-enhanced target high-energy imaging data region is associated to obtain data-enhanced target thermal imaging data.

4. The method according to claim 3, characterized in that The step of using the area distribution to generate the data enhancement rate of the corresponding high energy imaging data region among the plurality of high energy imaging data regions comprises: In response to the area distribution corresponding to the target initial thermal imaging data area being greater than a first area distribution index and less than a second area distribution index, determining a first data enhancement rate as a data enhancement rate of the target high-energy imaging data area; the first data enhancement rate being greater than the first data enhancement rate index and less than the second data enhancement rate index; In response to the area distribution of the region corresponding to the target initial thermal imaging data region being greater than the second region area distribution index, the second data enhancement rate is determined as the data enhancement rate of the target high-energy imaging data region; the second data enhancement rate is greater than or equal to the second data enhancement rate index.

5. The method according to claim 2, characterized in that: The step of performing data splitting on the initial thermal imaging data to obtain high-energy imaging data in the initial thermal imaging data includes: Based on the quantitative features of the initial thermal imaging data, performing mask processing on the initial thermal imaging data to obtain a mask tensor corresponding to the initial thermal imaging data; The high-energy imaging data is obtained by splitting the initial thermal imaging data based on the mask tensor corresponding to the initial thermal imaging data.

6. The method according to claim 1, characterized in that The step of determining the data analysis richness of the analysis path in the first data analysis module, and extracting the analysis path to be optimized from the analysis path of the first data analysis module according to the data analysis richness, comprises: Using the parameter index values ​​corresponding to each analysis path in the first data analysis module to evaluate the expected value of information volume, the data analysis richness corresponding to each analysis path in the first data analysis module is obtained; Compare each data analysis richness with a preset data analysis richness index, and extract analysis paths that do not reach the preset data analysis richness index to obtain each analysis path to be optimized; After comparing each data analysis richness with a preset data analysis richness index and extracting analysis paths that do not reach the preset data analysis richness index to obtain each analysis path to be optimized, the method further includes: Determining the number of paths to be optimized of each of the analysis paths to be optimized, and in response to the number of paths to be optimized being higher than or equal to or lower than a preset number indicator, using the second thermal imaging data analysis model as the first thermal imaging data analysis model, and returning to inputting the sample thermal imaging data set into the first thermal imaging data analysis model to update model parameters, thereby obtaining a second thermal imaging data analysis model; In response to the number of paths to be optimized being greater than or equal to the preset number indicator, performing the step of determining the collaborative analysis richness of the analysis paths in the first collaborative analysis module, and extracting a benchmark analysis path from the analysis paths in the first collaborative analysis module according to the collaborative analysis richness; The determining of the collaborative analysis richness of the analysis path in the first collaborative analysis module, and extracting a benchmark analysis path from the analysis path in the first collaborative analysis module according to the collaborative analysis richness, comprises: Using the parameter index values ​​corresponding to each analysis path in the first collaborative analysis module to evaluate the expected value of information volume, and obtaining the collaborative analysis richness corresponding to each analysis path in the first collaborative analysis module; The richness of each collaborative analysis is compared with the preset collaborative analysis richness index, and the analysis path that reaches the preset data analysis richness index is extracted to obtain each benchmark analysis path.

7. The method according to claim 1, characterized in that The step of optimizing the parameter index value of the analysis path to be optimized according to the parameter index value of the benchmark analysis path to obtain the optimization index value of the analysis path to be optimized includes: Determine the number of reference paths of the reference analysis path, and obtain the number of paths to be optimized of the analysis path to be optimized; Comparing the number of reference paths with the number of paths to be optimized to obtain quantitative relationship information, and determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of reference paths according to the quantitative relationship information; Performing weighted summation based on the parameter index value of the benchmark analysis path of the number of paths to be optimized and the parameter index value of the analysis path to be optimized of the number of paths to be optimized, to obtain the optimization parameter index value of the analysis path to be optimized of the number of paths to be optimized; Determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the quantitative relationship information includes: In response to the number of paths to be optimized being greater than the number of reference paths, determining a number offset between the number of paths to be optimized and the number of reference paths, and extracting the reference analysis paths with the number offset out of order from the reference analysis paths of the number of paths to be optimized; Obtaining a benchmark analysis path of the number of paths to be optimized according to the benchmark path number and the benchmark analysis path of the number offset; Determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the quantitative relationship information includes: In response to the reference path number being equal to the path number to be optimized, the reference analysis path of the reference path number is used as the reference analysis path of the path number to be optimized.

8. The method according to claim 7, characterized in that Determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the quantitative relationship information includes: In response to the number of paths to be optimized being less than the number of reference paths, acquiring analysis paths other than the analysis paths to be optimized from the analysis paths of the first data analysis module to obtain at least one candidate analysis path; Determine information residual of the amount of information between the benchmark analysis paths of the benchmark path number and the at least one candidate analysis path, and obtain analysis deviations between each benchmark analysis path and the at least one candidate analysis path; Determine a target analysis deviation corresponding to each benchmark analysis path from each analysis deviation, and select a benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the target analysis deviation; Determining the benchmark analysis path of the number of paths to be optimized from the benchmark analysis paths of the number of benchmark paths according to the quantitative relationship information includes: In response to the number of paths to be optimized being less than the number of reference paths, the reference analysis paths of the number of paths to be optimized are extracted in a disordered manner from the reference analysis paths of the number of reference paths.

9. A server system, characterized in that: The method comprises a server, wherein the server is used to execute the method according to any one of claims 1 to 8.

Images