Space odor detection system, method, device and storage medium

Abstract

The present application discloses a space odor detection system, method, device and storage medium. It relates to the field of space environment detection. The system includes: an electronic nose system, a space environment detection system and a data transmission system; and a cloud server. The cloud server is configured to: receive an odor data sequence and an environmental data sequence from a satellite; fuse the odor data vector in the odor data sequence with the corresponding environmental data vector in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling moment; use a feature extraction model based on an attention mechanism to generate semantic features corresponding to multiple fused data vectors respectively; and use a pre-set neural network-based odor recognition model to determine the component information corresponding to the space odor collected at each sampling moment based on the semantic features. It solves the problem that space odor detection technology is time-consuming and labor-intensive, has information lags and is costly, and is difficult to achieve real-time monitoring of the space environment.

Description

Technical Field

[0001] The present application relates to the field of space environment detection technology, and in particular to a space odor detection system, method, device and storage medium. Background Art

[0002] Space smell is very important for a comprehensive analysis of the space environment. The space environment is filled with various gas molecules, which mix together to form a unique "space smell". These odor molecules are like fingerprints of the universe, reflecting the physical and chemical properties and biological activity of the space environment, and providing scientists with important information such as the composition of cosmic dust, the role of solar wind, and the distribution of interstellar matter.

[0003] Invention patent CN117929655A provides an odor intensity detection method, device, equipment, storage medium and vehicle, which relates to the field of gas testing. When the gas to be tested in the present application is a volatile colorless organic compound, an initial characteristic value can be extracted from the gas to be tested, and the initial characteristic value is input into the trained convolutional neural network-artificial neural network model, wherein the convolutional neural network can generate a first characteristic value for describing the initial characteristic value, so as to efficiently describe the characteristic values ​​of multiple gases to be tested, and the second characteristic value generated by the artificial neural network describes the association between the first characteristic values, so as to establish an odor threshold relationship between multiple gases to be tested, so as to quickly identify the odor intensity value corresponding to the gas to be tested, so as to realize the output of the objective odor intensity value of the gas to be tested.

[0004] Invention patent CN117312616A discloses an information interaction method and system based on odor detection, which is used to improve the accuracy of the information interaction method based on odor detection. It includes: collecting odor data of a preset area through an odor detection device to obtain an odor data set; constructing a feature map of the odor data set to obtain a target odor feature map; generating an image identifier for the target odor feature map to obtain multiple image identifiers, and transmitting the target odor feature map and multiple image identifiers to a cloud database and matching interactive instructions to obtain multiple instructions to be interacted; collecting information interaction requests sent by target users, and performing data analysis on the information interaction requests to obtain request analysis data; performing instruction matching on multiple instructions to be interacted through request analysis data to obtain at least one target interaction instruction and performing data interaction from the cloud database to obtain data interaction status data.

[0005] Traditional methods of space odor detection mainly rely on astronauts or probes to collect physical samples and bring them back to Earth laboratories for analysis. This process is not only time-consuming and labor-intensive, but also has problems such as sample contamination, information lag and high analysis costs, making it difficult to achieve real-time monitoring of the space environment.

[0006] There is currently no effective solution to the technical problems in existing space odor detection technology, which is time-consuming, labor-intensive, has information delays, and is costly, making it difficult to achieve real-time monitoring of the space environment. Summary of the invention

[0007] The embodiments of the present application provide a space odor detection system, method, device and storage medium to at least solve the technical problems existing in the existing space odor detection technology, which is time-consuming and labor-intensive, has information lag and is costly, making it difficult to achieve real-time monitoring of the space environment.

[0008] According to one aspect of an embodiment of the present application, a space odor detection system is provided, comprising: an electronic nose system deployed on a satellite, a space environment detection system, and a data transmission system; and a cloud server deployed on the ground. The electronic nose system is used to detect space odors in space and generate multi-channel odor data associated with the space odors; the space environment detection system is used to detect the space environment associated with the space odors and generate multi-channel environmental data associated with the space environment; the data transmission system is used to transmit the multi-channel odor data and the multi-channel environmental data to the cloud server. The cloud server is configured to: receive an odor data sequence and an environmental data sequence from a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; fuse the odor data vectors in the odor data sequence with the corresponding environmental data vectors in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling time; generate semantic features corresponding to multiple fused data vectors respectively using a feature extraction model based on an attention mechanism; and determine component information corresponding to the space odor collected at each sampling time based on semantic features using a pre-set neural network-based odor recognition model.

[0009] According to another aspect of an embodiment of the present application, a space odor detection method is also provided, including: obtaining an odor data sequence and an environment data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; the environment data sequence is generated based on multi-channel environment data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environment data collected at each sampling time constitutes a corresponding environment data vector; fusing the odor data vector in the odor data sequence with the corresponding environment data vector in the environment data sequence to generate multiple fused data vectors corresponding to each sampling time; using a feature extraction model based on an attention mechanism to generate semantic features corresponding to the multiple fused data vectors respectively; and using a pre-set neural network-based odor recognition model to determine, based on the semantic features, component information corresponding to the space odor collected at each sampling time.

[0010] According to another aspect of an embodiment of the present application, a storage medium is further provided, the storage medium including a stored program, wherein the above-described method is executed by a processor when the program is running.

[0011] According to another aspect of an embodiment of the present application, a space odor detection device is also provided, including: a data acquisition module, used to acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector, and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; a data fusion module, used to fuse the odor data vector in the odor data sequence with the corresponding environmental data vector in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling time; a feature extraction module, used to generate semantic features corresponding to multiple fused data vectors respectively using a feature extraction model based on an attention mechanism; and an odor recognition module, used to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features using a pre-set neural network-based odor recognition model.

[0012] According to another aspect of the embodiment of the present application, a space odor detection device is also provided, including: a processor; and a memory, connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector, and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; fusing the odor data vectors in the odor data sequence with the corresponding environmental data vectors in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling time; using a feature extraction model based on an attention mechanism to generate semantic features corresponding to multiple fused data vectors respectively; and using a pre-set neural network-based odor recognition model to determine, based on the semantic features, component information corresponding to the space odor collected at each sampling time.

[0013] In an embodiment of the present application, according to the technical solution of this embodiment, an electronic nose system deployed on a satellite is used to detect the space smell around the satellite in real time, and the multi-channel smell data detected by the electronic nose system is transmitted to a cloud server on the ground. Thus, the cloud server can identify the components of the space smell based on the smell data received from the satellite. In this way, the space smell can be detected in real time, which solves the technical problems of time-consuming and labor-intensive, information-lagging and high cost in the existing space smell detection technology, making it difficult to achieve real-time monitoring of the space environment. In addition, in this embodiment, in order to more accurately identify the space smell, the multi-channel smell data and the multi-channel environment data are fused together to form a fused data vector, so that the fused data vector can more accurately represent the smell information by fusing the data responded by the smell sensor and the environment sensor. And this embodiment inputs the fused data vector into a feature extraction model based on the attention mechanism, and fuses each fused data vector with the fused data vector of the context by means of the attention mechanism to generate a semantic feature vector combined with the context semantics. Thus, the semantic feature vector can reflect the correlation between the smells of each sampling position. Thus, the space smell can be identified more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0015] Figure 1is a schematic diagram of a satellite system for implementing the space odor detection system described in Example 1 of the present application;

[0016] Figure 2 is a module schematic diagram of the space odor detection system according to Example 1 of the present application;

[0017] Figure 3 is a schematic diagram of an electronic nose system of a space odor detection system according to Example 1 of the present application;

[0018] Figure 4 is a schematic diagram of a space environment detection system according to the space odor detection system described in Example 1 of the present application;

[0019] Figure 5 is a schematic diagram of the operation process of the cloud server of the space odor detection system according to Example 1 of the present application;

[0020] Figure 6 is a schematic diagram of the space odor detection system according to Example 1 of the present application performing detection at multiple sampling positions at multiple sampling times;

[0021] Figure 7 This is a schematic diagram of the diffusion of space odors at various sampling locations and their influence on each other;

[0022] Figure 8 is a schematic diagram of generating a semantic feature vector based on a fusion vector according to Embodiment 1 of the present application;

[0023] Fig. 9 It is a schematic diagram of determining the component information of space odor according to the semantic feature vector described in Example 1 of the present application;

[0024] Fig.10 is a schematic diagram of the odor recognition model according to Example 1 of the present application;

[0025] Fig.11 is a flow chart of a method for detecting space odor according to the second aspect of Example 1 of the present application;

[0026] Fig.12 is a schematic diagram of a space odor detection device according to Example 2 of the present application; and

[0027] Fig.13 This is a schematic diagram of the space odor detection device described in Example 3 of the present application. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only embodiments of a part of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Example 1

[0031] Figure 1 FIG. 1 is a schematic diagram showing a satellite system for implementing the space odor detection system according to this embodiment. Figure 1 As shown, the satellite system includes a satellite 100 and a cloud server 200 deployed on the ground. The satellite 100 can communicate with the cloud server 200 through a gateway 300 and a network.

[0032] also, Figure 2 A schematic diagram of a space odor detection system according to this embodiment is further shown. Figure 2 As shown, the space odor detection system described in this embodiment includes: an electronic nose system 110 deployed on a satellite 100, a space environment detection system 120, a data transmission system 130; and a cloud server 200 deployed on the ground.

[0033] Among them, reference Figure 2 As shown, the electronic nose system 110 is used to detect space odor in space and generate multi-channel odor data associated with the space odor. The space environment detection system 120 is used to detect environmental parameters in the space environment and generate multi-channel environmental data related to the space environment. The data transmission system 130 is used to transmit the multi-channel odor data and the multi-channel environmental data to the cloud server 200. The multi-channel odor data and the multi-channel environmental data will be described in detail later.

[0034] In addition, further reference Figure 2 As shown, the cloud server 200 includes a data preprocessing module 210, a data feature extraction module 220, an odor recognition module 230, and a space odor statistics module 240. Among them, the data preprocessing module 210 is used to preprocess the multi-channel odor data and the multi-channel environmental data, wherein the preprocessing operation is described in detail later. The data feature extraction module 220 is used to extract features based on the multi-channel odor data and the multi-channel environmental data. The odor recognition module 230 is used to identify the various components in the space odor according to the data features extracted by the data feature extraction module 220. The space odor statistics module 240 is used to count the space odor situation in the orbit of the satellite 100 according to the recognition result of the odor recognition module 230.

[0035] in, Figure 3 A schematic diagram of the electronic nose system 110 is further shown, referring to Figure 3 As shown, the electronic nose system 110 includes a space gas collection unit, a microfluidic enrichment chip and a nanosensor array. The space gas collection unit is used to collect rarefied gas in space. The microfluidic enrichment chip is used to achieve efficient capture, collection and transportation of odor molecules. The nanosensor array includes a plurality of sensor modules composed of sensitive materials, sensors and odor signal preprocessing modules. The sensitive materials and sensors are used to generate corresponding odor signals in response to the concentration of corresponding substances in the space odor. The odor signal preprocessing module is used to preprocess the odor signal, including filtering out noise and performing analog-to-digital conversion. Thus, during the operation of the satellite 100, the electronic nose system 110 can detect the space odor around the satellite in real time.

[0036] Therefore, multiple sensor modules can generate multi-channel odor data at the same sampling time, and different channels correspond to different sensor modules. Figure 3 As shown, in this embodiment, there are a total of I For example, at a sampling moment, the electronic nose system 110 generates a multi-channel odor data, which can be represented by vector A To express, that is A =[ a 1, a 2, a 3, ..., a I ] T .

[0037] also, Figure 4 Further illustrating a schematic diagram of the space environment detection system 120, refer to Figure 4 As shown, the space environment detection system 120 includes a plurality of environment sensors 1~ J, and environmental sensor 1~ J Connected environmental signal preprocessing module 1~ J Among them, the environmental sensor is used to detect the environmental parameters of the satellite environment, where environmental sensor 1~ J For example, different types of environmental sensors are used to detect different types of environmental parameters, such as temperature, pressure, and radiation, so as to generate different types of environmental signals. The environmental signal preprocessing module is used to preprocess the environmental signal, including filtering out noise and performing analog-to-digital conversion. Thus, during the operation of the satellite 100, the space environment detection system 120 can detect the environmental parameters around the satellite in real time.

[0038] Therefore, the space environment detection system 120 can generate multi-channel environmental data at the same sampling time, and different channels correspond to different types of environmental sensors. Figure 4 As shown, in this embodiment, there are a total of J Thus, at a sampling moment, the space environment detection system 120 can generate a multi-channel environment data, which can be expressed as vector B To express, that is B =[ b 1, b 2, b 3, ..., b J ] T .

[0039] Further, Figure 5 A flowchart showing the operation of the cloud server 200 is shown, referring to Figure 5 As shown, the cloud server 200 is configured to execute the following process:

[0040] S502: receiving an odor data sequence and an environment data sequence from a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by the electronic nose system at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environment data sequence is generated based on multi-channel environment data collected by the space environment detection system at multiple sampling times, and the multi-channel environment data collected at each sampling time constitutes a corresponding environment data vector;

[0041] S504: Fusing the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment;

[0042] S506: Generate semantic features corresponding to the multiple fused data vectors respectively using a feature extraction model based on an attention mechanism; and

[0043] S508: Using a preset neural network-based odor recognition model, based on semantic features, determine component information corresponding to the space odor collected at each sampling time.

[0044] Specifically, the cloud server 200 can receive the odor data sequence from the satellite 100 through the gateway station 300 and the network. SA ={ A 1~ A K},in A k =[ a k,1 , a k,2 , ..., a k,I ] T is the multi-channel odor data in vector form, where k =1~ K .vector A k Indicated in k At sampling moments, the electronic nose system 110 collects I channels of multi-channel odor data. a k,i Indicates k The sampling time, i The sensor module outputs the i channels of odor data, where i =1~ I .

[0045] In addition, the cloud server 200 can receive the environmental data sequence from the satellite 100 through the gateway station 300 and the network. SB ={ B 1~ B K},in B k =[ b k,1 , b k,2 , ..., b k,J ] T is the multi-channel environment data in vector form, where k =1~ K .vector B k Indicated in k At sampling moments, the space environment detection system 120 collects J channels of multi-channel environment data. b k,j Indicatesk The sampling time, j The environmental sensor outputs j Channels of environmental data, j =1~ J .

[0046] Thus, the cloud server 200 receives the odor data sequence from the satellite 100 A 1~ A K and environmental data series B 1~ B K (S502).

[0047] In addition, further reference Figure 6 As shown, p 1~ p K It indicates that the satellite 100 is moving along the arrows, respectively with the sampling time 1~ K That is, in this embodiment, since the satellite 100 is in continuous motion, it is at different sampling positions at different sampling times. A 1~ A K Corresponding to each sampling position on the orbit of satellite 100 p 1~ p K Similarly, the environmental data sequence B 1~ B K Corresponding to each sampling position p 1~ p K Environmental information of the place.

[0048] Then, the data preprocessing module 210 of the cloud server 200 converts each odor data vector in the odor data sequence into A 1~ A K , and the corresponding environmental data vector in the environmental data sequence B 1~ B K Fusion is performed to obtain the corresponding fused data vector C 1~ C K (S504).

[0049] in, .

[0050] in, M = I +J , c 1,1 ~ c 1,I = a 1,1 ~ a 1,I , c 1,I+1 ~ c 1,M = b 1,1 ~ b 1,J .

[0051] Likewise, ,

[0052] And so on. .

[0053] After conducting experiments, the inventors found that even if the odor is of the same component, the odor sensor responds to different signals under different environmental conditions (such as temperature, radiation, etc.). Therefore, in this embodiment, in order to more accurately identify the space odor, the multi-channel odor data and the multi-channel environmental data are fused together to form a fused data vector C 1~ C K , so that the fusion data vector C 1~ C K Can represent odor information more accurately.

[0054] Then, the data feature extraction module 220 of the cloud server 200 fuses the data vector C 1~ C K Input the pre-trained feature extraction model based on the attention mechanism to obtain the fused data vector C 1~ C K The corresponding semantic feature vector F 1~ F K (S506).

[0055] Specifically, see Figure 7 As shown, at each sampling location p 1~ p K , the smell is also spreading to other sampling locations, as shown by the arrows. p 1~ p KThe smell is actually affected by the smell at other locations. p 1~ p K The smell of the odor is also correlated with the smell of other locations. Figure 8 As shown, by fusing the data vector C 1~ C K Input to the feature extraction model based on the attention mechanism, and use the attention mechanism to integrate each fused data vector C 1~ C K Fuse with the context fusion data vector to generate a semantic feature vector that combines contextual semantics F 1~ F K Thus, the semantic feature vector F 1~ F K Can reflect each sampling location p 1~ p K This enables the subsequent recognition step to more accurately identify space odors.

[0056] Then, refer to Fig. 9 As shown, the odor recognition module 230 of the cloud server 200 uses a pre-trained neural network-based odor recognition model (such as MLP) based on the feature vector F 1~ F K Determine the time interval between each sampling K The component information corresponding to the collected space odor G 1~ G K (S508).

[0057] Specifically, refer to Fig. 9 As shown, the odor recognition model may be, for example, an MLP-based odor recognition model, wherein the output layer of the MLP may include, for example N neurons, and connected to the Softmax classifier. N is the number of known odor components, so the output of the Softmax classifier can correspond to the content of each odor component. F 1~ F K The odor recognition model outputs the corresponding component information vector G 1~ G K .

[0058] in, .

[0059] in, G 1 is the component information of the space odor corresponding to the first sampling time and the first sampling position, where g 1,1 , g 1,2 ,..., g 1,N Corresponding to odor component 1~ N The content of g 1,1 , g 1,2 ,..., g 1,N ≤1, and g 1,1 + g 1,2 +...+ g 1,N =1.

[0060] Similarly, ; ......

[0061] .

[0062] Right now, ,in k =1~ K .

[0063] in, G k It is with k Sampling time and k The component information vector of the space odor corresponding to the sampling position, where g k,1 , g k,2 ,..., g k,N Corresponding to odor component 1~ N The content of g k,1 , g k,2 ,..., g k,N ≤1, and g k,1 + g k,2 +...+ g k,N =1.

[0064] As described in the background technology, traditional methods of space odor detection mainly rely on astronauts or probes to collect physical samples and bring them back to Earth laboratories for analysis. This process is not only time-consuming and labor-intensive, but also has problems such as sample contamination, information lag and high analysis costs, making it difficult to achieve real-time monitoring of the space environment.

[0065] In view of this, according to the technical solution of this embodiment, the electronic nose system deployed on the satellite is used to detect the space smell around the satellite in real time, and the multi-channel smell data detected by the electronic nose system is transmitted to the cloud server on the ground. Therefore, the cloud server can identify the components of the space smell based on the smell data received from the satellite. In this way, the space smell can be detected in real time, which solves the technical problems of time-consuming and labor-intensive, information-lagging and high cost in the existing space smell detection technology, making it difficult to achieve real-time monitoring of the space environment. In addition, in this embodiment, in order to more accurately identify the space smell, the multi-channel smell data and the multi-channel environment data are fused together to form a fused data vector, so that the fused data vector can more accurately represent the smell information by fusing the data responded by the smell sensor and the environment sensor. And this embodiment inputs the fused data vector into the feature extraction model based on the attention mechanism, and fuses each fused data vector with the fused data vector of the context by means of the attention mechanism to generate a semantic feature vector combined with the context semantics. Thus, the semantic feature vector can reflect the correlation between the smells at each sampling position. So that the space smell can be identified more accurately.

[0066] Optionally, the operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism includes:

[0067] Generate and fuse multiple data vectors separately C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ;

[0068] For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K :

[0069] (1)

[0070] (2)

[0071] (3)

[0072] in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The rate at which the influence of decays relative to the temporal distance.

[0073] Specifically, according to this embodiment, in order to determine the fusion data vector C k The semantic feature vector of F k , first use formula (3) to calculate each fused data vector C x Relative to the fused data vector C k Correlation coefficient h k,x , that is, calculation h k,1 ~ h k,K .

[0074] Then, using formula (2) hk,1 ~ h k,K Calculate the softmax function to get the correlation coefficient h k,1 ~ h k,K The corresponding weight w k,1 ~ w k,K .

[0075] Then, use formula (3) to calculate the fused data vector C k The semantic feature vector of F k .

[0076] As Figure 7 As shown, at each sampling location p 1~ p K , the smell is also spreading to other sampling locations, as shown by the arrows. p 1~ p K The smell is actually affected by the smell at other locations. p 1~ p K The smell of the sample is also correlated with the smell of other locations. p 1~ p K The influence of the smell of a location on the smell of other locations is gradually attenuated with the distance in time and space. In this embodiment, in the fused data vector sequence C 1~ C K In the above equation, the farther the distance between two fused data vectors, the smaller their influence on each other.

[0077] If the traditional feature extraction method based on attention mechanism is used, since it only extracts the fusion data vector C 1~ C K The semantic features between the two fused data vectors are ignored, and the fact that the influence of each other between the two fused data vectors decays with the distance is ignored. Therefore, the extracted semantic feature vector F k It cannot truly reflect the correlation between each fused data vector.

[0078] In view of this, this application introduces the timing related term ,because exist x =k takes the value "1" when | x - k As | increases, its value is infinitely close to 0. Therefore, the time series correlation term can more accurately reflect the time series correlation between each fused data vector.

[0079] Therefore, this embodiment adopts the attention mechanism algorithm after introducing the time-related terms, which can take into account the time-related correlation and semantic correlation between the fused data vectors at the same time, so that the extracted semantic features can more accurately reflect the components of the space odor at the corresponding position.

[0080] Optionally, the operation of using a preset neural network-based odor recognition model to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features includes: using the odor recognition model to determine the component information corresponding to the space odor collected at each sampling time based on the semantic feature vector F k , generate the corresponding component information vector, where each element of the component information vector is used to represent the corresponding component in the semantic feature vector F k The corresponding content in the space smell.

[0081] Specifically, refer to Fig.10 As shown, according to this embodiment, the odor recognition module 230 converts the semantic feature vector F k Input to the neural network-based odor recognition model, where reference Fig.10 As shown in Figure 1, the odor recognition model includes MLP (Multi-layer Perceptron) and Softmax classifier. Thus, the Softmax classifier outputs a semantic feature vector F k Corresponding N Dimensional component information vector G k Representation and semantic feature vector F k The content of each component in the corresponding space smell.

[0082] In this way, for each semantic feature vector F k Generate the corresponding G k , so that in this way, the component content corresponding to the space odor at various locations can be identified.

[0083] Optionally, the cloud server 200 is further configured to: perform statistics on the component distribution of the space odor in the satellite orbit based on the determined component information corresponding to the space odor collected at each sampling time.

[0084] Specifically, when determining the satellite 100 at each sampling time and each sampling position p 1~ p K The corresponding component information vector of space smell G 1~ G K Afterwards, the space odor statistics module 240 can be used to calculate the component information vector G 1~ G K For each sampling position on the satellite 100 orbit p 1~ p K The component contents of space smell are counted.

[0085] For example, the space odor statistics module 240 can count different components at each sampling location. p 1~ p K The changes in the content of each component at each sampling location can be plotted p 1~ p K The distribution curve of the content.

[0086] In addition, the space odor statistics module 240 can also perform a spatial odor statistics analysis on the same sampling location (e.g. p 1) The component contents of space odor at different times are counted, so that for any sampling location, the distribution of the content of each component at different sampling times can be counted.

[0087] Therefore, through the technical solution of the present application, it is possible to detect the space odor around the satellite 100 in real time, and to collect statistics on the distribution and changes of the space odor in the orbit of the satellite 100 based on the historical data of the space odor detected by the satellite 100.

[0088] In addition, according to a second aspect of this embodiment, a method for detecting space odor is provided, the method comprising: Figure 1 and Figure 2 The cloud server 200 shown in is implemented. Fig.11 A schematic diagram showing the process of the method is shown in FIG. Fig.11 As shown, the method includes:

[0089] S1102: Acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector;

[0090] S1104: Fusing the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment;

[0091] S1106: Generate semantic features corresponding to the multiple fused data vectors respectively using a feature extraction model based on an attention mechanism; and

[0092] S1108: Using a pre-set neural network-based odor recognition model, based on semantic features, determine component information corresponding to the space odor collected at each sampling time.

[0093] Optionally, the operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism includes:

[0094] Generate and fuse multiple data vectors separately C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ;as well as

[0095] For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K :

[0096] ;

[0097] ;

[0098] ,

[0099] in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The rate at which the influence of decays relative to the temporal distance.

[0100] Optionally, the operation of using a preset neural network-based odor recognition model to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features includes: using the odor recognition model to determine the component information corresponding to the space odor collected at each sampling time based on the semantic feature vector F k , generate the corresponding component information vector, where each element of the component information vector is used to represent the corresponding component in the semantic feature vector F k The corresponding content in the space smell.

[0101] Optionally, the method further includes: based on the determined component information corresponding to the space odor collected at each sampling moment, performing statistics on the component distribution of the space odor on the satellite orbit.

[0102] In addition, reference Figure 1 As shown, according to the third aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0103] Thus, according to this embodiment, the electronic nose system deployed on the satellite is used to detect the space smell around the satellite in real time, and the multi-channel smell data detected by the electronic nose system is transmitted to the cloud server on the ground. Thus, the cloud server can identify the components of the space smell based on the smell data received from the satellite. Thus, in this way, the space smell can be detected in real time, which solves the technical problems of time-consuming and labor-intensive, information-lagging and high cost in the existing space smell detection technology, making it difficult to achieve real-time monitoring of the space environment. In addition, in this embodiment, in order to more accurately identify the space smell, the multi-channel smell data and the multi-channel environment data are fused together to form a fused data vector, so that the fused data vector can more accurately represent the smell information by fusing the data responded by the smell sensor and the environment sensor. And this embodiment inputs the fused data vector into the feature extraction model based on the attention mechanism, and fuses each fused data vector with the fused data vector of the context by means of the attention mechanism to generate a semantic feature vector combined with the context semantics. Thus, the semantic feature vector can reflect the correlation between the smells of each sampling position. Thus, the space smell can be identified more accurately.

[0104] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0105] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0106] Example 2

[0107] Fig.12 The space odor detection device 1200 according to this embodiment is shown, and the space odor detection device 1200 corresponds to the method described in the second aspect of embodiment 1. Fig.12As shown, the space odor detection device 1200 includes: a data acquisition module 1210, which is used to acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector, and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; a data fusion module 1220, which is used to fuse the odor data vector in the odor data sequence with the corresponding environmental data vector in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling time; a feature extraction module 1230, which is used to generate semantic features corresponding to multiple fused data vectors respectively using a feature extraction model based on an attention mechanism; and an odor recognition module 1240, which is used to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features using a pre-set neural network-based odor recognition model.

[0108] Optionally, the feature extraction module 1230 includes: a vector generation submodule for generating vectors corresponding to the plurality of fusion data C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ; and a semantic feature vector generation submodule for each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K :

[0109] ;

[0110] ;

[0111] ,

[0112] in x =1~ K ; w k,x Represents the fused data vectorC x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The rate at which the influence of decays relative to the temporal distance.

[0113] Optionally, the odor recognition module 1240 includes: an odor recognition submodule for using the odor recognition model to identify the odor according to the semantic feature vector F k , generating a corresponding component information vector, wherein each element of the component information vector is used to represent the corresponding component in relation to the semantic feature vector F k The corresponding content in the space smell.

[0114] Optionally, the device also includes a statistical module for performing statistics on the component distribution of the space odor on the satellite orbit based on the determined component information corresponding to the space odor collected at each sampling time.

[0115] Thus, according to this embodiment, the electronic nose system deployed on the satellite is used to detect the space smell around the satellite in real time, and the multi-channel smell data detected by the electronic nose system is transmitted to the cloud server on the ground. Thus, the cloud server can identify the components of the space smell based on the smell data received from the satellite. Thus, in this way, the space smell can be detected in real time, which solves the technical problems of time-consuming and labor-intensive, information-lagging and high cost in the existing space smell detection technology, making it difficult to achieve real-time monitoring of the space environment. In addition, in this embodiment, in order to more accurately identify the space smell, the multi-channel smell data and the multi-channel environment data are fused together to form a fused data vector, so that the fused data vector can more accurately represent the smell information by fusing the data responded by the smell sensor and the environment sensor. And this embodiment inputs the fused data vector into the feature extraction model based on the attention mechanism, and fuses each fused data vector with the fused data vector of the context by means of the attention mechanism to generate a semantic feature vector combined with the context semantics. Thus, the semantic feature vector can reflect the correlation between the smells of each sampling position. Thus, the space smell can be identified more accurately.

[0116] Example 3

[0117] Fig.13 FIG. 1 shows a space odor detection device 1300 according to this embodiment, which corresponds to the method described in the second aspect of Embodiment 1. Fig.13 As shown, the space odor detection device 1300 includes: a processor 1310; and a memory 1320, which is connected to the processor 1310 and is used to provide the processor 1310 with instructions for processing the following processing steps: obtaining an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by an electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector, and the environmental data sequence is generated based on multi-channel environmental data collected by a space environment detection system on the satellite at multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; fusing the odor data vector in the odor data sequence with the corresponding environmental data vector in the environmental data sequence to generate multiple fused data vectors corresponding to each sampling time; using a feature extraction model based on an attention mechanism to generate semantic features corresponding to multiple fused data vectors respectively; and using a pre-set neural network-based odor recognition model to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features.

[0118] Optionally, the operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism includes:

[0119] Respectively generate the multiple fused data vectors C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ;as well as

[0120] For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K :

[0121] ;

[0122] ;

[0123] ,

[0124] in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The rate at which the influence of decays relative to the temporal distance.

[0125] Optionally, the operation of determining the component information corresponding to the space odor collected at each sampling moment based on the semantic features by using a preset neural network-based odor recognition model includes:

[0126] Using the odor recognition model, according to the semantic feature vector F k , generating a corresponding component information vector, wherein each element of the component information vector is used to represent the corresponding component in relation to the semantic feature vector F k The corresponding content in the space smell.

[0127] Optionally, the memory is also used to provide the processor with instructions for processing the following processing steps: based on the determined component information corresponding to the space odor collected at each sampling time, statistically analyzing the component distribution of the space odor on the satellite orbit.

[0128] Thus, according to this embodiment, the electronic nose system deployed on the satellite is used to detect the space smell around the satellite in real time, and the multi-channel smell data detected by the electronic nose system is transmitted to the cloud server on the ground. Thus, the cloud server can identify the components of the space smell based on the smell data received from the satellite. Thus, in this way, the space smell can be detected in real time, which solves the technical problems of time-consuming and labor-intensive, information-lagging and high cost in the existing space smell detection technology, making it difficult to achieve real-time monitoring of the space environment. In addition, in this embodiment, in order to more accurately identify the space smell, the multi-channel smell data and the multi-channel environment data are fused together to form a fused data vector, so that the fused data vector can more accurately represent the smell information by fusing the data responded by the smell sensor and the environment sensor. And this embodiment inputs the fused data vector into the feature extraction model based on the attention mechanism, and fuses each fused data vector with the fused data vector of the context by means of the attention mechanism to generate a semantic feature vector combined with the context semantics. Thus, the semantic feature vector can reflect the correlation between the smells of each sampling position. Thus, the space smell can be identified more accurately.

[0129] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0130] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0135] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A space odor detection system, characterized in that: include: An electronic nose system (110), a space environment detection system (120) and a data transmission system (130) deployed on a satellite (100); and a cloud server (200) deployed on the ground, wherein The electronic nose system (110) is used to detect space odor in space and generate multi-channel odor data associated with the space odor; The space environment detection system (120) is used to detect the space environment associated with the space odor and generate multi-channel environmental data associated with the space environment; The data transmission system (130) is used to transmit the multi-channel odor data and the multi-channel environmental data to the cloud server (200), and The cloud server (200) is configured to: receiving an odor data sequence and an environmental data sequence from the satellite (100), wherein the odor data sequence is generated based on multi-channel odor data collected by the electronic nose system (110) at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by the space environment detection system (120) at the multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; Fusing the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment; Generate semantic features corresponding to the multiple fused data vectors respectively by using a feature extraction model based on an attention mechanism; as well as Using a preset neural network-based odor recognition model, based on the semantic features, component information corresponding to the space odor collected at each sampling moment is determined, and wherein The operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism comprises: Generate and fuse multiple data vectors separately C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ; as well as For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K : ； ； ， in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The influence of is attenuated relative to the temporal distance. d k represents the dimension of the key vector.

2. The system according to claim 1, characterized in that The operation of determining the component information corresponding to the space odor collected at each sampling time based on the semantic features by using a preset neural network-based odor recognition model includes: Using the odor recognition model, according to the semantic feature vector F k , generating a corresponding component information vector, wherein each element of the component information vector is used to represent the corresponding component in relation to the semantic feature vector F k The corresponding content in the space smell.

3. The system according to claim 1, characterized in that The cloud server (200) is also configured to: Based on the determined component information corresponding to the space odor collected at each sampling time, statistics are taken on the component distribution of the space odor on the orbit of the satellite (100).

4. A method for detecting space odor, characterized in that: include: Acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by the electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by the space environment detection system on the satellite at the multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; Fusing the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment; Generate semantic features corresponding to the multiple fused data vectors respectively by using a feature extraction model based on an attention mechanism; as well as Using a preset neural network-based odor recognition model, based on the semantic features, component information corresponding to the space odor collected at each sampling moment is determined, and wherein The operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism comprises: Generate and fuse multiple data vectors separately C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ; as well as For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K : ； ； ， in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The influence of is attenuated relative to the temporal distance. d k represents the dimension of the key vector.

5. The method according to claim 4, characterized in that The operation of determining the component information corresponding to the space odor collected at each sampling time based on the semantic features by using a preset neural network-based odor recognition model includes: Using the odor recognition model, according to the semantic feature vector F k , generating a corresponding component information vector, wherein each element of the component information vector is used to represent the corresponding component in relation to the semantic feature vector F k The corresponding content in the space smell.

6. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is executed, the method according to claim 4 or 5 is executed by a processor.

7. A space odor detection device, characterized in that: include: A data acquisition module, used to acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by the electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by the space environment detection system on the satellite at the multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; a data fusion module, used to fuse the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment; a feature extraction module, configured to generate semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism; and The odor recognition module is used to determine the component information corresponding to the space odor collected at each sampling time based on the semantic features using a preset neural network-based odor recognition model, and wherein The feature extraction module includes: a vector generation submodule for generating a plurality of fusion data vectors C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ;as well as Semantic feature vector generation submodule is used for each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K : ； ； ， in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The influence of is attenuated relative to the temporal distance. d k represents the dimension of the key vector.

8. A space odor detection device, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Acquire an odor data sequence and an environmental data sequence transmitted by a satellite, wherein the odor data sequence is generated based on multi-channel odor data collected by the electronic nose system on the satellite at multiple sampling times, and the multi-channel odor data collected at each sampling time constitutes a corresponding odor data vector; and the environmental data sequence is generated based on multi-channel environmental data collected by the space environment detection system on the satellite at the multiple sampling times, and the multi-channel environmental data collected at each sampling time constitutes a corresponding environmental data vector; Fusing the odor data vectors in the odor data sequence with the corresponding environment data vectors in the environment data sequence to generate a plurality of fused data vectors corresponding to each sampling moment; Generate semantic features corresponding to the multiple fused data vectors respectively by using a feature extraction model based on an attention mechanism; as well as Using a preset neural network-based odor recognition model, based on the semantic features, component information corresponding to the space odor collected at each sampling moment is determined, and wherein The operation of generating semantic features respectively corresponding to the plurality of fused data vectors by using a feature extraction model based on an attention mechanism comprises: Generate and fuse multiple data vectors separately C 1~ C K The corresponding key vector K 1~ K K , query vector Q 1~ Q K and the value vector V 1~ V K ; as well as For each fused data vector C k , calculated by the following formula C k The corresponding semantic feature vector F k ,in k =1~ K : ； ； ， in x =1~ K ; w k,x Represents the fused data vector C x Relative to the fused data vector C k The weight of h k,x Represents the fused data vector C x The key vector K x With the fused data vector C k The query vector Q k The correlation coefficient between is a time-series related term used to represent the fused data vector C x and the fused data vector C k The correlation in time series, where arccot ( ) is the inverse cotangent function, β is the attenuation factor, used to represent the fused data vector C x To fusion data vector C k The influence of is attenuated relative to the temporal distance. d k represents the dimension of the key vector.

Images