A satellite telemetry data feature extraction method and device, equipment and medium

Abstract

The embodiments of the present disclosure disclose a satellite telemetry data feature extraction method and device, equipment and medium, and the satellite telemetry data feature extraction method can comprise: acquiring satellite to-be-processed data; determining prior information related to the to-be-processed data based on historical data, and determining a regression model of the to-be-processed data based on the prior information; and iteratively solving the regression model to obtain target data corresponding to the to-be-processed data.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a method, apparatus, device and medium for extracting features from satellite telemetry data. Background Technology

[0002] Spacecraft telemetry data has enormous potential for utilization because it can provide critical information such as high-precision space environment monitoring, real-time status tracking, and fault diagnosis, thereby supporting the safe operation of space missions and in-depth scientific research. However, due to the massive volume of high-dimensional on-orbit data and the presence of many coupled and complex information, it is difficult to utilize it directly.

[0003] Feature extraction technology not only ensures the normal operation and safety of spacecraft, but also improves data utilization efficiency, supports decision-making and control, and promotes the development of aerospace technology, thus possessing significant practical importance and application value. However, current feature extraction technologies lack accuracy when extracting features from satellite data. Summary of the Invention

[0004] In view of this, the present disclosure aims to provide a method, apparatus, device, and medium for extracting features from satellite telemetry data, which can solve the problem of insufficient accuracy of existing feature extraction technologies for extracting features from satellite data.

[0005] The technical solution of this disclosure embodiment is implemented as follows: In a first aspect, embodiments of this disclosure provide a method for extracting features from satellite telemetry data, including: Acquire satellite data to be processed; Based on historical data, determine prior information related to the data to be processed, and based on the prior information, determine the regression model for the data to be processed; The target data corresponding to the data to be processed is obtained by iteratively solving the regression model.

[0006] Secondly, embodiments of this disclosure provide a satellite telemetry data feature extraction apparatus, including: The data acquisition module is used to acquire satellite data to be processed. The model determination module is used to determine prior information related to the data to be processed based on historical data, and to determine the regression model of the data to be processed based on the prior information; The feature extraction module is used to iteratively solve the regression model to obtain the target data corresponding to the data to be processed.

[0007] Thirdly, embodiments of this disclosure provide an electronic device, the electronic device comprising: a processor and a memory; the processor is configured to execute instructions stored in the memory to implement the satellite telemetry data feature extraction method described in the first aspect.

[0008] Fourthly, embodiments of this disclosure provide a computer storage medium storing at least one instruction, which is executed by a processor to implement the satellite telemetry data feature extraction method as described in the first aspect.

[0009] This disclosure provides a method, apparatus, device, and medium for feature extraction from satellite telemetry data. By combining historical data and prior information to construct a regression model, the accuracy and efficiency of feature extraction are significantly improved. It not only deeply mines the potential information in the data but also optimizes the model's predictive ability through an iterative solution process, making the target data extracted from complex and ever-changing satellite data more accurate and reliable. Attached Figure Description

[0010] Figure 1 A flowchart of a satellite telemetry data feature extraction method provided in this embodiment of the disclosure.

[0011] Figure 2 A schematic diagram of the original feature data provided in the embodiments of this disclosure.

[0012] Figure 3 This is a schematic diagram corresponding to the feature extraction results provided in the embodiments of this disclosure.

[0013] Figure 4 This is a schematic diagram of the original data detection results provided in the embodiments of this disclosure.

[0014] Figure 5 This is a schematic diagram illustrating the detection results of the target data provided in an embodiment of this disclosure.

[0015] Figure 6 A flowchart of another satellite telemetry data feature extraction method provided in this embodiment of the disclosure.

[0016] Figure 7 A flowchart of a satellite detection method provided in an embodiment of this disclosure.

[0017] Figure 8 This is a schematic diagram of a satellite telemetry data feature extraction device provided in an embodiment of the present disclosure.

[0018] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure.

[0019] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0021] Currently, existing solutions for feature extraction from massive high-dimensional satellite telemetry data mainly fall into two categories: machine learning methods and deep learning methods. Machine learning methods, represented by principal component analysis, project telemetry data into a lower-dimensional space through linear transformations, maximizing the variance projected onto these coordinate axes. Deep learning methods, represented by convolutional neural networks, first convert two-dimensional temporal features into three-dimensional image features based on techniques such as time-frequency transformations, and then extract feature maps by sliding convolutional kernels across the input image and performing local weighted summation.

[0022] Machine learning models are highly interpretable, but they suffer from a lack of depth in their learning. This means that machine learning methods sometimes cannot effectively solve feature extraction problems in complex, coupled, and multivariate scenarios.

[0023] Deep learning models have significantly improved learning capabilities compared to machine learning models, but they suffer from insufficient interpretability. This leads to low acceptance of deep learning methods by experts in practical engineering projects, and also results in high hardware requirements and computational costs.

[0024] To address these challenges, this invention proposes a feature extraction method for satellite telemetry data. This method utilizes historical data to determine prior information relevant to the data to be processed and constructs a regression model based on this information, thereby more accurately capturing the key features of the data. By iteratively solving this regression model, this method can effectively extract valuable information from complex satellite data, improving the efficiency and accuracy of data processing, while also enhancing the model's adaptability and predictive ability to data changes. This method is particularly suitable for applications requiring the rapid and accurate extraction of key features from large amounts of remote sensing data.

[0025] The following is combined Figure 1The above-mentioned satellite telemetry data feature extraction method is described, wherein the above-mentioned satellite telemetry data feature extraction method may include steps S110 to S130.

[0026] In step S110, the satellite data to be processed is acquired.

[0027] In one exemplary embodiment of this disclosure, raw satellite data may first be acquired from a satellite receiver. The satellite data may include various telemetry parameters, such as temperature, pressure, velocity, and position information, which are captured by satellite sensors and transmitted back to the ground station in the form of digital signals.

[0028] Next, this raw data is parsed and converted into a processable format, such as converting the data file from binary format to a more easily analyzed format like .csv or .json. The specific format can be customized according to user needs, and will not be elaborated upon in this example implementation.

[0029] During the parsing process, it may be necessary to process information such as data encoding, timestamps, and data units to ensure the accuracy and consistency of the data.

[0030] Next, data filtering is performed, specifically selecting a subset of data relevant to a particular test item from a large amount of raw data. This can be done through data filtering, feature selection, or dimensionality reduction, with the aim of removing irrelevant or redundant information and retaining data useful for subsequent analysis. Specific filtering, feature selection, or dimensionality reduction operations can be found in relevant techniques and will not be elaborated upon here.

[0031] In addition, outlier removal is necessary to ensure data quality. Outliers may be caused by sensor malfunctions, data transmission errors, or other anomalies. These outliers are typically identified and removed using statistical methods (such as z-scores) or thresholds based on historical experience, resulting in the data to be processed.

[0032] In step S120, prior information related to the data to be processed is determined based on historical data, and a regression model for the data to be processed is determined based on the prior information.

[0033] In some examples, the data to be processed may include multiple samples, each of which may contain feature data and noise data. When determining the regression model, prior information related to the data to be processed can first be determined based on historical data. In this example implementation, since the noise information does not include feature information of the image to be processed, the noise data is not processed.

[0034] Specifically, historical data can be collected and analyzed first. This historical data can include telemetry data, environmental monitoring data, and other relevant parameters collected within a preset time period. The preset time can be customized based on user needs, which will not be elaborated here.

[0035] Historical data can be used to extract information that has predictive value for the current data to be processed, such as the statistical characteristics of the data, common trends, and anomalies observed in the past, as the aforementioned prior information.

[0036] Next, prior information extracted from historical data, such as the data's distribution characteristics and patterns of change, is used to construct a regression model. The regression model aims to describe the relationship between the data to be processed and known features.

[0037] When building a model, machine learning techniques can be used, such as choosing an appropriate algorithm (e.g., Lasso regression), determining model parameters, and defining the model structure.

[0038] Then, based on the regression model determined by prior information, for example, if historical data shows periodic changes in the data, the regression model may include periodic terms; if the data shows that certain features are highly correlated with other features, the regression model may include interaction terms of these features.

[0039] Furthermore, the regression model can also be determined based on a convolution dictionary. A convolution dictionary is a set of basic functions used to combine data to be processed through convolution operations to form a sparse representation of the data. The selection and construction of the convolution dictionary can be guided by the characteristics of the data to be processed and how to utilize prior information from historical data. The construction of the convolution dictionary can be customized based on user needs, which will not be elaborated upon in this example implementation.

[0040] Through the steps described above, based on prior information and regression models determined from historical data, valuable features can be effectively extracted from the satellite data to be processed, providing support for subsequent data analysis and decision-making. This method not only improves the accuracy of data processing but also enhances the model's generalization ability and predictive performance by utilizing prior information.

[0041] In step S130, the regression model is iteratively solved to obtain the target data corresponding to the data to be processed.

[0042] In one exemplary implementation of this disclosure, the parameters of the regression model can first be initialized. Before starting the iterative solution, initial values ​​need to be set for the parameters of the regression model. These initial values ​​can be empirical estimates. Initialization is the starting point for the iterative solution and has a significant impact on the final solution result.

[0043] Next, the solution is solved iteratively. In each iteration, based on the current model parameters, the model's predicted value for the data to be processed is calculated and compared with the actual target data. This comparison is usually quantified by a loss function, which measures the difference between the predicted and actual values. Common loss functions include mean squared error (MSE) and cross-entropy loss.

[0044] Then, based on the result of the loss function, the model parameters are updated. This step typically involves gradient descent or a similar optimization algorithm. Gradient descent reduces the value of the loss function by calculating the gradient of the loss function with respect to the model parameters and then adjusting the parameters in the opposite direction of the gradient. The magnitude of the parameter updates is usually controlled by the hyperparameter, the learning rate.

[0045] During the iteration process, it may be necessary to introduce a regularization term to prevent the model from overfitting. The regularization term can be L1 norm (Lasso regularization), L2 norm (Ridge regularization), or other forms. They can penalize the complexity of the model parameters, prompting the model to learn simpler and more generalized features.

[0046] In addition, methods such as momentum and adaptive learning rate (e.g., Adam optimizer) can be used to accelerate convergence while maintaining good stability when dealing with non-convex optimization problems.

[0047] Finally, the iterative solution will stop when a certain termination condition is met, such as the decrease in the loss function being less than a certain threshold, or the number of iterations reaching a preset upper limit. After the iteration process is completed, the obtained model parameters will be used for final feature extraction, thereby obtaining the target data corresponding to the data to be processed.

[0048] The satellite telemetry data feature extraction method in this embodiment significantly improves the accuracy and efficiency of feature extraction by combining historical data and prior information to construct a regression model. It not only deeply mines the potential information in the data but also optimizes the model's predictive ability through an iterative solution process, making the target data extracted from complex and ever-changing satellite data more accurate and reliable.

[0049] In some examples, feature data It is a by This telemetry feature is a column vector composed of samples, and consists of features that can indicate the health status of the satellite's momentum wheel. and irrelevant noise The composition can be represented as:

[0050] Hypothetical features It can be determined by a pre-defined convolution dictionary. Encoding, then the encoding vector It can be estimated using a Lasso regression model of the following form:

[0051] in , , . It is a balance parameter. The sample mean. This represents the sample variance.

[0052] Then, prior information can be introduced into the Lasso regression model to improve it, resulting in the regression model of this disclosure. Specifically, the regression model of this disclosure can be:

[0053] in It is the Hadamard product, which is the product of corresponding elements in a matrix. Features representing expert experience information related to the working state and mode of the momentum wheel. This is a penalty item used to punish undesirable information that hinders the accurate reconstruction process. and Similarly, since these are balancing parameters, λ1 and λ2 can be obtained through grid search and joint cross-validation.

[0054] In some examples, the data to be processed is encoded layer by layer based on the balancing parameters, the convolution dictionary, and the prior information, and the encoding vector of each layer is searched, and the encoding vector of the Lth layer is used as the regression model; The encoding vector of the i-th layer is determined by the encoding vector of the (i-1)-th layer and the convolution dictionary of the i-th layer, and the encoding vector of the first layer is determined by the feature data and the convolution dictionary of the first layer. Where i is greater than or equal to 2 and less than or equal to L.

[0055] Specifically,

[0056] in Represents the convolution dictionary for each coding layer. This represents the features of expert experience information at each coding layer. This represents the penalty factor for each coding layer. and These represent the threshold values ​​for the principal component information regularization term and the auxiliary information regularization term, respectively. This represents the number of coding layers.

[0057] Based on the above, we can obtain the... The encoding vectors of each encoding layer are:

[0058] in Dictionary arrive The product of. and Each represents one of them Dictionary arrive The product of. and They represent the first The balance parameter between the principal component information regularization term and the auxiliary information regularization term can be used to... The encoding vectors of each encoding layer are used as the regression model described above.

[0059] After obtaining the above regression model, iterative solutions can be applied to iterate through the model. Specifically, for a given regression model... Layer-based improved algorithm unfolds the network, defining , , Then the first The encoding vectors of each encoding layer can be simplified to:

[0060] Among them when hour, ;when hour, .

[0061] Classical Lasso regression models are typically solved using an iterative threshold shrinkage algorithm. The iterative process first calculates the intermediate variables. Then obtain Substitute into the proximal operator to calculate .

[0062]

[0063] in express The first gradient, is the Lipschitz constant.

[0064] For a component with two parts and nonsmooth objective function Similarly, optimization can be achieved using similar methods:

[0065] in express The generalized gradient.

[0066] When the number of coding layers is The number of iterations is The improved algorithm can be iteratively solved by unfolding the network in the following way:

[0067] in For the improved threshold function, .

[0068] Improved threshold function This is called a weighted smooth soft thresholding function, which provides a smoother and more flexible thresholding process, avoiding hard switching in signal processing. This method can reduce artifacts in certain situations. The specific design scheme is as follows:

[0069] in This is the function to be processed by the threshold function. It is a with Vectors of the same dimension represent the weights of each component. It is a smoothing parameter. Among them, The larger the value, the less smooth the surface; conversely, The smaller the value, the smoother the surface. For example... The value range can be [0.1, 10].

[0070] In summary, the algorithm designed based on the improved objective function and threshold function expands the network's output results. Can be used as a feature The feature extraction results are the target data in this application.

[0071] In some examples, during the iterative solution of the regression model described above, the number of unrollings and the number of encoding layers can be determined as follows: specifically, a parameter search algorithm can be used to learn the optimal combination of parameters among the number of training epochs, the number of encoding layers, and the number of unrollings. Before trying each set of hyperparameters, historical hyperparameters can be obtained first, and then the hyperparameters can be adjusted based on the evaluation results of the historical hyperparameters.

[0072] First, a black-box objective function can be constructed as a standard for evaluating the model's performance, and the standardized mutual information between features and the target can be selected as the evaluation metric. Next, the upper and lower bounds of each hyperparameter that needs to be debugged (number of encoding layers, number of unfoldings) are given to determine their respective search ranges. Finally, the Bayesian optimization object is instantiated and the number of training rounds is determined.

[0073] Optionally, the range of possible values ​​for the number of encoding layers can be set to (2, 4), the range of possible values ​​for the number of unrollings can be set to (2, 5), and the range of possible values ​​for the number of training rounds can be set to (1, 200). After Bayesian optimization, the parameter combination with the highest evaluation index is: number of training rounds = 176, number of encoding layers L = 4, and number of unrollings K = 4.

[0074] In some examples, refer to Figure 2 and Figure 3 The simulation duration was set to 250 minutes. A multi-input single-output simulation dataset was constructed, and an improved algorithm with pre-configured parameters was used to expand the network to extract information from the input features. It was found that the complexity of the obtained target data was less than that of the original image, and it could represent all the feature information of the original image.

[0075] In some examples, we consider telemetry data from the momentum wheel component of a real on-orbit satellite's attitude and orbit control subsystem over one year, totaling 8779 samples. The first 4000 samples are used as the training set, and the last 4779 samples are used as the test set. The original telemetry features include momentum wheel rotational speed and momentum wheel housing temperature, etc.

[0076] After determining the number of unfolding times and the number of encoding layers, feature extraction can be performed on the data to be processed based on the expression function of the target data to achieve high-precision anomaly detection.

[0077] The extracted new features are used as input to the subsequent anomaly detection model, and the anomaly detection results are as follows: Figure 4 and Figure 5 As shown in the figure, the features extracted by expanding the network using the improved algorithm can more accurately reflect abnormal information, and have higher anomaly detection accuracy when using the same model for detection.

[0078] The satellite telemetry data feature extraction method provided in this example significantly improves the accuracy and efficiency of feature extraction by combining historical data and prior information to construct a regression model. It not only delves deeper into the potential information within the data but also optimizes the model's predictive capabilities through an iterative solution process, resulting in more accurate and reliable target data extracted from complex and ever-changing satellite data.

[0079] In some examples, refer to Figure 6 The above-mentioned method for extracting features from satellite telemetry data will be described, specifically including steps S610 to S670.

[0080] In step S610, satellite data is acquired; In step S620, the satellite data is parsed, filtered, outlier removed, and completed. In step S630, historical data is acquired and prior information is determined based on the historical data; In step S640, a regression model for the data to be processed is determined based on prior information; In step S650, the regression model is modeled using multi-level sparse coding to update the regression model; In step S660, hyperparameter optimization determines the parameters of the regression model; In step S670, the regression model is iteratively solved to obtain the target data.

[0081] It should be noted that the details of each step mentioned above have already been explained in detail, so they will not be repeated here.

[0082] Reference Figure 7 As shown, the above method can also be used to detect satellites. When used to detect satellites, it may include steps S710 to S740.

[0083] In step S710, the satellite data to be processed is acquired.

[0084] In step S720, prior information related to the data to be processed is determined based on historical data, and a regression model for the data to be processed is determined based on the prior information.

[0085] In step S730, the regression model is iteratively solved to obtain the target data corresponding to the data to be processed.

[0086] It should be noted that the process of obtaining the target data mentioned above can be referred to the description of steps S110 to S130, and will not be repeated here.

[0087] In step S740, the above is detected based on the target data to obtain the detection result.

[0088] In this example implementation, after obtaining the target data, standard data corresponding to the target data can be acquired. The standard data characterizes data that enables the satellite to function normally, or data that maintains normal operation in a scenario detected by the satellite. Then, the target data can be compared with the standard data. If the target data is equal to the standard data within a threshold range, the detection result is considered normal. If the target data is not equal to the standard data within the threshold range, the detection result is determined to be abnormal.

[0089] In some examples, an alarm signal can be issued when the detection result is abnormal, so that the user can be informed and take action.

[0090] Furthermore, this disclosure also provides a satellite telemetry data feature extraction device, referring to... Figure 8 As shown, the satellite telemetry data feature extraction device 800 may include a data acquisition module 810, a model determination module 820, and a feature extraction module 830. Wherein: The data acquisition module 810 can be used to acquire satellite data to be processed.

[0091] The model determination module 820 can be used to determine prior information related to the data to be processed based on historical data, and to determine the regression model of the data to be processed based on the prior information; The feature extraction module 830 can be used to iteratively solve the regression model to obtain the target data corresponding to the data to be processed.

[0092] In some examples, the data to be processed includes multiple samples, each of which may include feature data and noise data. The model determination module 820 can also obtain the convolution dictionary corresponding to the feature data and determine the regression model of the data to be processed based on the convolution dictionary and prior information.

[0093] In some examples, the model determination module 820 can also be used to determine the sample mean and sample variance of the samples, and determine the balancing parameters based on the sample variance and sample mean; and determine the regression model for the data to be processed based on the balancing parameters and the convolution dictionary.

[0094] In some examples, the model determination module 820 can also be used to encode the data to be processed layer by layer based on the balancing parameters, the convolution dictionary, and prior information, and search for the encoding vector of each layer, and use the encoding vector of the Lth layer as the regression model; wherein, the encoding vector of the ith layer is determined based on the encoding vector of the (i-1)th layer and the convolution dictionary of the ith layer, and the encoding vector of the first layer is determined by the feature data and the convolution dictionary of the first layer; wherein, i is greater than or equal to 2 and less than or equal to L.

[0095] In some examples, the feature extraction module 830 can also be used to obtain smoothing parameters and determine a smoothing threshold function based on the smoothing parameters and feature data; and iteratively solve the regression model based on the smoothing threshold function to obtain the target data corresponding to the data to be processed.

[0096] In some examples, the data acquisition module 810 can also be used to acquire satellite data, parse and filter the satellite data to determine the initial satellite data corresponding to the test item; and remove outliers from the satellite data to obtain the data to be processed.

[0097] It should be understood that the above-described device embodiments are merely illustrative, and the device disclosed herein can be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, integrated into another system, or some features may be ignored or not executed.

[0098] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this disclosure can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0099] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0100] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMSDD). Based on this understanding, the technical solution of this disclosure, 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. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0101] Please refer to Figure 9 This diagram illustrates a structural block diagram of an electronic device provided in an exemplary embodiment of this disclosure. In some examples, the electronic device may be at least one of devices such as a smartphone, smartwatch, desktop computer, laptop, virtual reality terminal, augmented reality terminal, wireless terminal, and laptop computer. The electronic device has communication capabilities and can access wired or wireless networks. The term "electronic device" can refer to one of multiple terminals; those skilled in the art will understand that the number of terminals may be more or less. It is understood that the electronic device undertakes the computational and processing work of the technical solution of this disclosure, and the embodiments of this disclosure do not limit this aspect.

[0102] like Figure 9 As shown, the electronic device 900 may include at least one processor 910, a memory 920, and a communication interface 930.

[0103] The memory 920 is used to store programs. Specifically, the program may include program code, which includes computer operation instructions.

[0104] The memory 920 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0105] The processor 910 executes computer execution instructions stored in the memory 920 to implement the satellite telemetry data feature extraction method described in the foregoing method embodiments. The processor 910 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this disclosure.

[0106] The electronic device 900 may also include a communication interface 930, through which it can communicate and interact with external devices. In specific implementations, if the communication interface 930, memory 920, and processor 910 are implemented independently, they can be interconnected via a bus to complete communication. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc., but this does not imply that there is only one bus or one type of bus.

[0107] Optionally, in a specific implementation, if the communication interface 930, memory 920 and processor 910 are integrated on a single chip, then the communication interface 930, memory 920 and processor 910 can communicate through an internal interface.

[0108] This disclosure also provides a computer-readable storage medium, which may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory, a random access memory, a disk, or an optical disk. Specifically, the computer-readable storage medium stores program instructions, which are used for the satellite telemetry data feature extraction method in the above embodiments.

[0109] This disclosure also provides a computer program product including computer instructions stored in a computer-readable storage medium; a processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the satellite telemetry data feature extraction method of the above embodiments.

[0110] Those skilled in the art will recognize that the functions described in the embodiments of this disclosure in one or more of the foregoing examples can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0111] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0112] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention applied herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not claimed herein.

[0113] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for feature extraction from satellite telemetry data, characterized in that, include: Acquire satellite data to be processed, the data to be processed including multiple samples, each sample including feature data and noise data; Based on historical data, determine prior information related to the data to be processed, and based on the prior information, determine the regression model for the data to be processed; Obtain the convolution dictionary corresponding to the feature data, and determine the encoding vector according to the following formula: ； in, This represents the feature data. This represents the convolution dictionary. This represents the encoded vector. This represents the encoding vector that minimizes the objective function of the regression model. Represents the Hadamard product. Features representing expert experience information related to the operating state and mode of the satellite momentum wheel. Indicates a penalty item. and Represents the balance parameters. Describes the norm 1. Represents the L2 norm; Based on the balancing parameters, the convolution dictionary, and the prior information, the data to be processed is encoded layer by layer. The encoding vector of each layer is searched, and the first layer is then... The layer's encoding vector serves as the regression model; Among them, the The encoding vector of the layer is based on the first The encoding vector of layer and the first The convolutional dictionary of the layers is determined, and the encoding vector of the first layer is determined based on the feature data and the convolutional dictionary of the first layer. Greater than or equal to 2 and less than or equal to ; The regression model is iteratively solved to obtain the target data corresponding to the data to be processed, including: Obtain the smoothing parameters, and determine the smoothing threshold function based on the smoothing parameters and the feature data according to the following formula: ； in, This represents the variable to be processed by the smoothing threshold function. Indicates and Weight vectors of the same dimension Indicates the threshold parameter. This represents the smoothing parameter. Represents a symbolic function; The regression model is iteratively solved based on the smoothing threshold function to obtain the target data corresponding to the data to be processed.

2. The method according to claim 1, characterized in that, The determination of the regression model includes: Determine the sample mean and sample variance of the sample, and determine the balance parameters based on the sample variance and the sample mean; The regression model for the data to be processed is determined based on the balancing parameters and the convolution dictionary.

3. The method according to claim 1, characterized in that, Acquire satellite data to be processed, including: Acquire satellite data, and parse and filter the satellite data to determine the initial satellite data corresponding to the test item; Outliers are removed from the satellite data to obtain the data to be processed.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: The satellite is detected based on the target data to obtain the detection results.

5. A satellite telemetry data feature extraction device, characterized in that, The satellite telemetry data feature extraction device is configured to perform the satellite telemetry data feature extraction method as described in claim 1, and the satellite telemetry data feature extraction device includes: The data acquisition module is used to acquire satellite data to be processed, which includes multiple samples, each of which includes feature data and noise data; The model determination module is used to determine prior information related to the data to be processed based on historical data, and to determine the regression model of the data to be processed based on the prior information. The step of determining the regression model of the data to be processed based on the prior information includes: obtaining a convolution dictionary corresponding to the feature data, and determining the regression model of the data to be processed based on the convolution dictionary and the prior information. The feature extraction module is used to iteratively solve the regression model to obtain the target data corresponding to the data to be processed. The iterative solution of the regression model to obtain the target data corresponding to the data to be processed includes: obtaining a smoothing parameter, and determining a smoothing threshold function based on the smoothing parameter and the feature data; and iteratively solving the regression model based on the smoothing threshold function to obtain the target data corresponding to the data to be processed.

6. An electronic device, characterized in that, The electronic device includes a processor and a memory; the processor is used to execute instructions stored in the memory to implement the satellite telemetry data feature extraction method as described in any one of claims 1 to 3.

7. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is executed by a processor to implement the satellite telemetry data feature extraction method as described in any one of claims 1 to 3.

Images