Radiation source identification and data processing method and device, storage medium and equipment

Abstract

The invention discloses a radiation source identification and data processing method and device, a storage medium and equipment. The method comprises the following steps: firstly, extracting a target signal feature of a target radiation source signal; inputting the target signal features into a pre-constructed radiation source identification model, and outputting an identification result of the target radiation source signal; wherein the identification result comprises the type and the source of the target radiation source signal, then sending the identification result and the model parameters of the radiation source identification model to an intelligent contract on a block chain platform by using a loose coupling connection model, and then identifying the target radiation source signal through the intelligent contract. And storing the identification result and the model parameters of the radiation source identification model to a block chain platform. Therefore, by integrating the radiation source recognition model obtained by training the deep learning model, the smart contract and the block chain technology, accurate recognition of the target radiation source signal is realized, and an ideal radiation source recognition effect and an ideal data processing effect are achieved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a method, device, storage medium, and equipment for radiation source identification and data processing. Background Art

[0002] With the rapid development of artificial intelligence technology, new breakthroughs and progress have been made in the identification methods of radiation sources. Among them, in the face of a complex and changeable electromagnetic environment, it is particularly important to accurately identify radiation sources and perform corresponding data processing in this complex radio environment.

[0003] Currently, although there are various methods for realizing radiation source identification, these identification methods all have their own disadvantages. For example, the disadvantages of the first identification method based on a traditional spectrum monitoring system are: high cost, large equipment volume, and poor scalability. The processing ability of fixed algorithms is limited, it is difficult to cope with complex and dynamic signal environments, and the update and upgrade are slow. The disadvantages of the second spectrum analysis method based on artificial intelligence are: traditional machine learning algorithms perform poorly in processing complex signal patterns. Lack of real-time data processing ability, relying on a centralized server, network latency and data transmission problems affect real-time performance. The disadvantages of the third radio monitoring solution based on cloud computing are: latency and security problems in the data transmission process affect real-time performance and data privacy. A stable network connection is required, and it may perform poorly in an environment with poor network conditions. The disadvantages of the fourth method of using blockchain technology for identification are: mainly the preliminary application of data storage, low degree of automation in data processing and updating, and the potential of blockchain is not fully utilized. The disadvantages of the fifth method of spectrum monitoring based on the Internet of Things (IoT) are: limited data processing ability, low identification accuracy and inability to perform automatic updates. Strong network dependence may affect the real-time performance and processing efficiency of data, etc. It can be seen that these existing methods for realizing radiation source identification all have relatively large problems, which not only affect the identification effect of radiation sources, but also cannot achieve ideal data processing effects. Summary of the Invention

[0004] The main purpose of the embodiments of this application is to provide a method, device, storage medium, and equipment for radiation source identification and data processing. By integrating deep learning models, smart contracts, and blockchain technology, accurate identification of radiation sources is achieved, and the identification results are uploaded to the blockchain, ensuring the immutability, credible traceability, and cross-organization collaborative application of the identification result data, so as to meet the requirements of various high-trust scenarios, and thus achieve ideal radiation source identification effects and data processing effects.

[0005] The embodiments of this application provide a method for radiation source identification and data processing, including:

[0006] Obtain the target radiation source signal to be recognized; and extract the target signal features of the target radiation source signal;

[0007] Input the target signal features into a pre-constructed radiation source recognition model, and output the recognition result of the target radiation source signal; the recognition result includes the type and source of the target radiation source signal;

[0008] Use a loose coupling connection model to send the recognition result and the model parameters of the radiation source recognition model to a smart contract on the blockchain platform;

[0009] Through the smart contract, store the recognition result and the model parameters of the radiation source recognition model on the blockchain platform.

[0010] In a possible implementation manner, the extracting the target signal features of the target radiation source signal includes:

[0011] Perform time-domain analysis and frequency-domain analysis on the target radiation source signal to extract time-domain features and frequency-domain features;

[0012] Analyze the modulation characteristics of the target radiation source signal to extract modulation features; and use the time-domain features, frequency-domain features, and the modulation features to constitute the target signal features of the target radiation source signal.

[0013] In a possible implementation manner, the radiation source recognition model includes a spatial feature extraction network, a temporal feature extraction network, and a key feature extraction network; the inputting the target signal features into the pre-constructed radiation source recognition model to recognize the type and source of the target radiation source signal as the recognition result includes:

[0014] Perform format conversion on the target signal features to obtain the converted target signal features;

[0015] Input the converted target signal features into the spatial feature processing network, temporal feature processing network, and key feature processing network in the radiation source recognition model to recognize the type and source of the target radiation source signal as the recognition result.

[0016] In a possible implementation manner, the spatial feature extraction network is a deep convolutional neural network DCNN; the temporal feature extraction network is a long short-term memory network LSTM; the key feature processing network is a multi-head self-attention mechanism.

[0017] In a possible implementation manner, the using a loose coupling connection model to send the recognition result and the model parameters of the radiation source recognition model to a smart contract on the blockchain platform includes:

[0018] Using the loose coupling connection model, format conversion is performed on the recognition result and the model parameters of the radiation source recognition model to obtain the converted recognition result and model parameters; the converted recognition result and model parameters meet the input requirements of the smart contract;

[0019] Using the loose coupling connection model, an interface for interacting with the smart contract is called, and according to a preset communication protocol, the converted recognition result and model parameters are sent to the smart contract on the blockchain platform.

[0020] In a possible implementation manner, storing the recognition result and the model parameters of the radiation source recognition model on the blockchain platform through the smart contract includes:

[0021] The smart contract verifies the recognition result and the model parameters of the radiation source recognition model, and combines with the storage mechanism provided by the blockchain to store the verified recognition result and model parameters on the blockchain platform.

[0022] In a possible implementation manner, after storing the recognition result and the model parameters of the radiation source recognition model on the blockchain platform through the smart contract, the method further includes:

[0023] Obtain the model parameters to be updated corresponding to the radiation source recognition model; and use the radiation source recognition model to send the model parameters to be updated to the smart contract;

[0024] Use the smart contract to verify the model parameters to be updated according to a preset verification mechanism to obtain the verified model parameters to be updated;

[0025] Update the verified model parameters to be updated to the blockchain platform for storage; and record the update process of the model parameters to be updated on the blockchain platform for storage.

[0026] In a possible implementation manner, the design process of the smart contract includes the design of the data storage structure, the design of the model upgrade rules, the consensus mechanism, and the security of the contract.

[0027] In a possible implementation manner, the radiation source recognition model is a deep learning model trained by a supervised learning method.

[0028] The embodiment of the present application further provides a radiation source recognition and data processing device, including:

[0029] An extraction unit, configured to obtain a target radiation source signal to be recognized; and extract target signal features of the target radiation source signal;

[0030] An identification unit, configured to input the target signal features into a pre-constructed radiation source identification model and output an identification result of the target radiation source signal; the identification result includes the type and source of the target radiation source signal;

[0031] A sending unit, configured to use a loose coupling connection model to send the identification result and the model parameters of the radiation source identification model to a smart contract on a blockchain platform;

[0032] A storage unit, configured to store the identification result and the model parameters of the radiation source identification model on the blockchain platform through the smart contract.

[0033] In a possible implementation, the extraction unit includes:

[0034] A first analysis subunit, configured to perform time-domain analysis and frequency-domain analysis on the target radiation source signal to extract time-domain features and frequency-domain features;

[0035] A second analysis subunit, configured to analyze the modulation characteristics of the target radiation source signal to extract modulation features; and use the time-domain features, frequency-domain features, and the modulation features to form the target signal features of the target radiation source signal.

[0036] In a possible implementation, the radiation source identification model includes a spatial feature extraction network, a temporal feature extraction network, and a key feature extraction network; the identification unit includes:

[0037] A first conversion subunit, configured to perform format conversion on the target signal features to obtain the converted target signal features;

[0038] An identification subunit, configured to input the converted target signal features into the spatial feature processing network, the temporal feature processing network, and the key feature processing network in the radiation source identification model, and identify the type and source of the target radiation source signal as the identification result.

[0039] In a possible implementation, the spatial feature extraction network is a deep convolutional neural network DCNN; the temporal feature extraction network is a long short-term memory network LSTM; the key feature processing network is a multi-head self-attention mechanism.

[0040] In a possible implementation, the sending unit includes:

[0041] A second conversion subunit, configured to use the loose coupling connection model to perform format conversion on the identification result and the model parameters of the radiation source identification model to obtain the converted identification result and model parameters; the converted identification result and model parameters meet the input requirements of the smart contract;

[0042] A sending subunit, configured to use the loose coupling connection model to call an interface for interacting with the smart contract, and send the converted recognition result and model parameters to the smart contract on the blockchain platform according to a preset communication protocol.

[0043] In a possible implementation manner, the storage unit is specifically configured to:

[0044] Verify the recognition result and the model parameters of the radiation source recognition model through the smart contract, and combine with the storage mechanism provided by the blockchain to store the verified recognition result and model parameters on the blockchain platform.

[0045] In a possible implementation manner, the device further includes:

[0046] An obtaining unit, configured to obtain the model parameters to be updated corresponding to the radiation source recognition model; and use the radiation source recognition model to send the model parameters to be updated to the smart contract;

[0047] A verification unit, configured to verify the model parameters to be updated according to a preset verification mechanism through the smart contract to obtain the model parameters to be updated that pass the verification;

[0048] An updating unit, configured to update the model parameters to be updated that pass the verification to the blockchain platform for storage; and record the updating process of the model parameters to be updated on the blockchain platform for storage.

[0049] In a possible implementation manner, the design process of the smart contract includes the design of the data storage structure, the design of the model upgrade rules, the consensus mechanism, and the design of the security of the contract.

[0050] In a possible implementation manner, the radiation source recognition model is a deep learning model trained by means of supervised learning.

[0051] An embodiment of the present application further provides a radiation source recognition and data processing device, including: a processor, a memory, and a system bus;

[0052] The processor and the memory are connected through the system bus;

[0053] The memory is used to store one or more programs, and the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes any one of the implementation manners of the above radiation source recognition and data processing method.

[0054] The embodiments of the present application also provide a computer-readable storage medium, in which instructions are stored. When the instructions run on a terminal device, the terminal device is enabled to execute any implementation manner of the above-mentioned radiation source identification and data processing method.

[0055] The embodiments of the present application also provide a computer program product. When the computer program product runs on a terminal device, the terminal device is enabled to execute any implementation manner of the above-mentioned radiation source identification and data processing method.

[0056] A radiation source identification and data processing method, device, storage medium and equipment provided by the embodiments of the present application first obtain a target radiation source signal to be identified; extract target signal features of the target radiation source signal; then input the target signal features into a pre-constructed radiation source identification model to output an identification result of the target radiation source signal, where the identification result includes the type and source of the target radiation source signal. Then, using a loose coupling connection model, the identification result and the model parameters of the radiation source identification model are sent to a smart contract on a blockchain platform, and further, through the smart contract, the identification result and the model parameters of the radiation source identification model can be stored on the blockchain platform.

[0057] It can be seen that since the present application identifies the target radiation source signal by integrally using a radiation source identification model, a smart contract and blockchain technology obtained by training with a deep learning model, it realizes the accurate identification of the target radiation source signal, and stores the identification result and the model parameters on the blockchain, ensuring the immutability, credible traceability and cross-organization collaborative application of the identification result data and the model parameters, so as to meet the requirements of various high-trust scenarios, and thus achieve an ideal radiation source identification effect and data processing effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0059] Figure 1 It is a schematic flow chart of a radiation source identification and data processing method provided by the embodiments of the present application;

[0060] Figure 2 It is a schematic composition diagram of a radiation source identification and data processing device provided by the embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] Radiation source identification refers to the process of detecting, analyzing, and determining the type, location, or characteristics of the source of radiation by means of technology on the radio signals emitted by the radiation source, and it has broad application prospects in the fields of radio, industrial detection, etc.

[0062] Currently, the methods for realizing radiation source identification usually include the following five:

[0063] The first is the identification method based on traditional spectrum monitoring systems. This traditional spectrum monitoring system mainly relies on hardware devices for the acquisition and analysis of electromagnetic signals. These systems usually include dedicated spectrum analyzers, antennas, and signal processors. Such systems are relatively reliable in terms of performance and accuracy, but they have the disadvantages of high cost, large device volume, and difficulty in expansion. In addition, traditional spectrum monitoring systems usually adopt fixed signal processing algorithms and are difficult to adapt to complex and changing signal environments.

[0064] The second is the spectrum analysis method based on artificial intelligence, specifically using machine learning algorithms for signal classification and anomaly detection. For example, traditional machine learning methods such as support vector machines and decision trees are used for the classification of signal features. These systems can improve the accuracy of signal identification to a certain extent, but in dealing with complex signal patterns and real-time data, they may not perform as well as deep learning models. In addition, these systems usually rely on centralized servers to process data and do not have the ability of distributed data management.

[0065] The third is the radio monitoring solution based on cloud computing, specifically by uploading signal data to the cloud and using powerful computing resources for analysis and storage. This method can process a large amount of data and has high scalability. However, the latency in the data transmission process and the security issues of cloud services may affect real-time performance and data privacy. In addition, cloud computing solutions usually require a stable network connection and may not be applicable to environments with poor network conditions.

[0066] The fourth is the method of using blockchain technology for identification. Specifically, some preliminary spectrum monitoring systems applying blockchain technology attempt to record signal data and monitoring results through blockchain to ensure the immutability of data. However, these systems have not fully realized the integration of smart contract functions. Usually, they only use blockchain for data storage without delving into the dynamic update of the model and the automatic execution of smart contracts. The degree of automation of data processing and update is low, and the potential of blockchain has not been fully exploited.

[0067] The fifth is the method of spectrum monitoring based on the Internet of Things (IoT). Specifically, the spectrum monitoring solution based on the IoT uses a sensor network to conduct real-time monitoring of electromagnetic signals. These systems achieve wide-area coverage and data collection by deploying a large number of sensor nodes. Although IoT systems have high flexibility and real-time performance, in terms of data processing and analysis, they may not be able to provide the accuracy brought by deep learning technologies and the automated updates supported by smart contracts.

[0068] It can be seen that there are relatively large problems with these existing methods for implementing radiation source identification. They not only affect the identification effect of radiation sources but also fail to achieve the ideal data processing effect.

[0069] To address the above defects, the present application provides a method for radiation source identification and data processing. First, obtain the target radiation source signal to be identified; extract the target signal features of the target radiation source signal; then input the target signal features into a pre-constructed radiation source identification model to output the identification result of the target radiation source signal through the model. Among them, the identification result may include the type and source of the target radiation source signal. Then, use the loosely coupled connection model to send the identification result and the model parameters of the radiation source identification model to the smart contract on the blockchain platform. Furthermore, through the smart contract, the identification result and the model parameters of the radiation source identification model can be stored on the blockchain platform.

[0070] It can be seen that when the present application identifies the target radiation source signal, by integrating and utilizing the radiation source identification model, smart contract, and blockchain technology obtained through deep learning model training, it realizes the accurate identification of the target radiation source signal, and stores the identification result and model parameters on the blockchain, ensuring the immutability, credible traceability, and cross-organization collaborative application of the identification result data and model parameters. Thus, it can meet the requirements of various highly trusted scenarios, and further achieve the ideal radiation source identification effect and data processing effect.

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0072] First Embodiment

[0073] See Figure 1 , which is a schematic flowchart of a method for radiation source identification and data processing provided in this embodiment. The method includes the following steps:

[0074] S101: Obtain the target radiation source signal to be recognized; and extract the target signal features of the target radiation source signal.

[0075] In this embodiment, any radiation source recognized using this embodiment is defined as the target radiation source to be recognized, and the signal emitted by it is defined as the target radiation source signal to be recognized. However, this application does not limit the source, content, acquisition method, etc. of the target radiation source signal. For example, the target radiation source signal can be various signals within the radio frequency spectrum, including but not limited to the following types:

[0076] Communication signals: Signals in frequency bands such as cellular communication (e.g., 4G / 5G), Wi-Fi, Bluetooth, satellite communication, etc. Through a radio frequency spectrum monitoring system, these signals can be monitored and recognized in real time.

[0077] Radar signals: Electromagnetic wave signals generated by military or civilian radar systems. They generally contain specific modulation characteristics that can be extracted and further recognized in spectrum analysis.

[0078] Interference signals: Include electromagnetic interference signals generated by illegal signal sources, electronic warfare (such as jammers), etc. These signals may affect legal communication and navigation systems.

[0079] Remote sensing signals: Signals generated by satellite remote sensing devices and ground sensors, commonly used in fields such as meteorological monitoring and environmental monitoring.

[0080] Moreover, the target radiation source signal at different locations (such as ground radiation sources like communication base stations, mobile radiation sources like mobile phones, special radiation sources like radars, and interference sources like interference devices) can be obtained through a Software Defined Radio (SDR) device or other electromagnetic spectrum monitoring. Among them, the software radio device becomes an ideal choice due to its flexibility and programmability, and can cover a wide spectrum range from low frequency to high frequency. The SDR can receive electromagnetic signals of various frequencies through an antenna and convert these analog signals into digital signals. A key challenge in the acquisition process is to capture a variety of different target radiation source signals, such as Wi-Fi, Bluetooth, radar, etc. This requires the device to have high sensitivity and a wide dynamic range to ensure that both weak signals and strong signals can be reliably captured and processed.

[0081] Next, after obtaining the target radiation source signal, in order to improve the recognition effect, existing or future signal preprocessing methods (including but not limited to operations such as filtering, denoising, normalization, etc.) can be further used to clean and optimize the target radiation source signal to improve the signal quality.

[0082] Among them, filtering is an important step to remove unnecessary frequency components in the target radiation source signal and retain the desired signal. Common filtering methods include low-pass filtering, high-pass filtering, and band-pass filtering. The purpose of denoising is to reduce the useless noise components in the target radiation source signal (such as random noise caused by environmental noise, equipment errors, or other interference sources), and improve the signal quality. Common denoising methods include mean filtering, wavelet transform denoising, and adaptive noise cancellation, etc. The normalization operation mainly adjusts the signal amplitude so that signals with different amplitudes have a unified scale, thus being comparable and facilitating subsequent processing steps and model training, etc.

[0083] On this basis, after obtaining the target radiation source signal (or the preprocessed target radiation source signal), in order to improve the recognition effect of the target radiation source signal, further target signal features of the target radiation source signal (or the preprocessed target radiation source signal) can be extracted as signal features that help with subsequent radiation source recognition, for use in performing subsequent step S102.

[0084] Specifically, an optional implementation method is that after obtaining the target radiation source signal (or the preprocessed target radiation source signal), time-domain analysis and frequency-domain analysis can be first performed on the target radiation source signal (or the preprocessed target radiation source signal) to extract time-domain features (including but not limited to the amplitude, phase, envelope, etc. of the signal) and frequency-domain features (including but not limited to spectral energy distribution, bandwidth, center frequency, etc.). Then, the modulation characteristics of the target radiation source signal are analyzed to extract modulation features, and the time-domain features, frequency-domain features, and modulation features are used to form the target signal features of the target radiation source signal.

[0085] In this implementation method, it should be noted that the purpose of target signal feature extraction is to transform the original signal (i.e., the target radiation source signal (or the preprocessed target radiation source signal)) into a more meaningful form to assist in the analysis and model processing of subsequent steps. The extraction of target signal features can be carried out in combination with time-domain analysis and frequency-domain analysis according to different signal types and application requirements. The following is the specific implementation process and example of feature extraction.

[0086] Specific implementation process of feature extraction:

[0087] (1) Time-domain analysis:

[0088] Time-domain analysis refers to directly analyzing the changes in the signal on the time axis. Common time-domain features include the amplitude, waveform, duration, peak value, etc. of the signal. These features can reveal the basic characteristics of the signal and are applicable to certain specific signal types.

[0089] Common time-domain feature extraction methods: mean and variance, peak value and kurtosis, etc.

[0090] Autocorrelation function: It represents the correlation between a signal and its own delay, which helps to identify periodic signals (such as radar signals).

[0091] Signal duration and waveform: Analyzing the duration and waveform characteristics of a signal helps to distinguish between pulsed signals and continuous signals.

[0092] Example:

[0093] For Wi-Fi signals, time-domain characteristics can be used to judge the stability and changes of the signal by analyzing its transmission duration, power fluctuation, pulse width, etc.

[0094] For radar signals, different types of radars (short-range or long-range radars) can be distinguished by the peak value, duration, and length of the signal, etc.

[0095] (2) Frequency-domain analysis:

[0096] Frequency-domain analysis is to transform a signal from the time domain to the frequency domain and display the frequency components of the signal in the form of a spectrum. Commonly used frequency-domain analysis methods include Fourier transform and wavelet transform, which can provide the frequency characteristics of the signal, such as the bandwidth, frequency distribution, harmonics, etc. of the signal.

[0097] Commonly used frequency-domain feature extraction methods: power spectral density, center frequency and bandwidth, etc.

[0098] Spectrum characteristics: such as peak frequency, frequency offset, etc., which help to further distinguish different radiation source signals.

[0099] Frequency-amplitude relationship: It is used to analyze the frequency response of a signal and distinguish the characteristics of different signals.

[0100] Example:

[0101] For Bluetooth signals, their spectrum characteristics usually have a definite carrier frequency and a small bandwidth, which are very different from other broadband signals (such as Wi-Fi signals).

[0102] For radar signals, their center frequency and bandwidth can be extracted. The spectrum of a radar is usually relatively concentrated and has a large bandwidth.

[0103] (3) Hybrid time-domain and frequency-domain analysis:

[0104] Hybrid time-domain and frequency-domain analysis methods (such as wavelet transform, Hilbert transform, etc.) are used to obtain more refined signal characteristics.

[0105] Commonly used methods: wavelet transform, Hilbert-Huang transform, etc.

[0106] Example:

[0107] For burst signals in a communication system (such as burst radio signals), wavelet transform can help extract the signal change characteristics within a short time, revealing the periodicity and frequency components of the emergency event.

[0108] For radar signals, through time-frequency analysis, its rapidly changing frequency characteristics and instantaneous frequency can be captured more precisely.

[0109] The extraction of modulation features is also an important step in the feature extraction process, especially when identifying complex communication signals (such as radiation source signals with diverse modulation modes). Modulation features help distinguish different signal sources, especially when the modulation methods and signal bandwidths are different. And the extraction of modulation features usually analyzes the modulation method of the target radiation source signal (or the preprocessed target radiation source signal) based on the already extracted time-domain / frequency-domain features, and further classifies and identifies the details of the target radiation source signal (or the preprocessed target radiation source signal).

[0110] The specific implementation process of modulation features:

[0111] The core purpose of modulation feature extraction is to identify the modulation type of the signal by analyzing the modulation method of the signal. The modulation method refers to the way the signal carries information during transmission. Common modulation methods include: amplitude modulation, frequency modulation, phase modulation, sine wave modulation, quadrature amplitude modulation, and pulse modulation.

[0112] The extraction steps of modulation features:

[0113] Step 1: Signal segmentation and time-frequency analysis: When extracting modulation features, it is usually first necessary to divide the signal into multiple small segments according to a time window and perform short-time Fourier transform or wavelet transform to obtain the time-frequency characteristics of the signal. The spectrum of the signal changes with time. Therefore, time-frequency analysis helps to reveal the modulation changes of the signal in different time periods.

[0114] Step 2: Extract modulation features: The following methods can be used to extract modulation features:

[0115] Carrier frequency and phase change: By analyzing the carrier frequency change and phase change of the signal, the modulation method of the signal can be extracted. For example, a phase modulation signal will exhibit obvious phase jump characteristics.

[0116] Frequency offset and amplitude fluctuation: For frequency modulation signals, modulation features can be extracted through the changes in frequency offset and amplitude; for amplitude modulation signals, the modulation depth (amplitude change) is usually the key feature.

[0117] Envelope analysis: The envelope of a modulated signal usually reflects the modulation depth and type of the signal. Through envelope analysis, the amplitude modulation characteristics of the signal can be extracted to further identify the modulation mode.

[0118] Carrier frequency and modulation bandwidth: For sinusoidal modulation (such as ASK, FSK, PSK), analyzing the bandwidth of the signal and the change of the carrier frequency can help identify the specific modulation method.

[0119] Modulation depth and symbol rate: For phase modulation and frequency modulation, analyzing the symbol rate and modulation depth can accurately determine the modulation method of the signal.

[0120] Step 3: Feature vectorization: Convert the extracted modulation features into feature vectors, which are used as the input of subsequent machine learning models (such as support vector machines, etc.) or rule engines for the identification and classification of radiation sources.

[0121] The reason for extracting modulation features to form the target signal features of the target radiation source signal is as follows: First, different modulation methods have a significant impact on the manifestation of the signal. Extracting modulation features can help identify different signal source types, such as distinguishing different communication signals like Wi-Fi, radar, satellite communication, Bluetooth, etc. Second, modulation features are a key feature in signal identification, which helps to classify signal sources more accurately. Especially when facing complex signals, the accuracy can be further improved through modulation features. Third, in a complex electromagnetic environment, extracting modulation features helps to improve the adaptability of the model to different signals, especially in the case of interference or multipath propagation.

[0122] S102: Input the target signal features into a pre-constructed radiation source identification model, and output the identification result of the target radiation source signal; where the identification result includes the type and source of the target radiation source signal.

[0123] In this embodiment, after extracting the target signal features (such as time-domain features, frequency-domain features, and modulation features) of the target radiation source signal (or the preprocessed target radiation source signal) through step S101, in order to improve the identification effect of the target radiation source signal (or the preprocessed target radiation source signal), the target signal features can further be input into a pre-constructed radiation source identification model to obtain the identification result corresponding to the target radiation source signal. Among them, the identification result can include, but is not limited to, the type and source of the target radiation source signal, etc., for performing the subsequent step S103.

[0124] Among them, the radiation source recognition model is a deep learning model trained through supervised learning, so that the deep learning technology of the model can be used to analyze the target signal features (such as time-domain features, frequency-domain features, and modulation features), and more accurately identify the type and source of the target radiation source signal as the recognition result.

[0125] Specifically, an optional implementation manner is that when the radiation source recognition model includes a spatial feature extraction network, a temporal feature extraction network, and a key feature extraction network, the implementation process of "inputting the target signal features into the pre-constructed radiation source recognition model to identify the type and source of the target radiation source signal as the recognition result" in this step S102 may include the following steps S1021-S1022:

[0126] S1021: Perform format conversion on the target signal features to obtain the converted target signal features.

[0127] In this implementation manner, for the target signal features composed of time-domain features, frequency-domain features, and modulation features, format conversion needs to be performed first, that is, convert them into a format suitable for input into the deep learning model (such as feature vectors or multi-dimensional tensors, etc.) to obtain the converted target signal features for performing the subsequent step S1022.

[0128] In practical applications, the converted target signal features may include various types of features, such as time series data, spectral images, or modulation patterns. These data will be used as the input of the radiation source recognition model trained by the deep learning model for further analysis.

[0129] Among them, the feature vector refers to combining the extracted time-domain features, frequency-domain features, and modulation features into a one-dimensional vector, and each feature value corresponds to an element in the vector. The multi-dimensional tensor refers to converting the time-frequency analysis result of the signal (such as a spectrogram) into a two-dimensional or three-dimensional tensor and passing it to the radiation source recognition model trained by the deep learning model for analysis.

[0130] S1022: Input the converted target signal features into the spatial feature processing network, temporal feature processing network, and key feature processing network in the radiation source recognition model to identify the type and source of the target radiation source signal as the recognition result.

[0131] In this implementation manner, after obtaining the converted target signal features through step S1021, the converted target signal features can be further input into the spatial feature processing network, temporal feature processing network, and key feature processing network in the radiation source recognition model to identify the type (such as Wi-Fi, Bluetooth, radar signal, satellite communication signal, etc.) and source (such as device ID (e.g., MAC address of a Wi-Fi device or positioning information of a radar signal), geographical location of the signal source, communication frequency band, etc.) of the target radiation source signal as the recognition result.

[0132] Among them, the signal type may include Wi-Fi signals, Bluetooth signals, radar signals, etc., and the signal source refers to the device or location identifier where the signal is emitted, such as the MAC address of the device, the GPS coordinates of the signal source, etc. The identified signal type and source will be used as part of the recognition result and further used for spectrum monitoring and interference detection. At the same time, relevant data (such as device ID, signal type, etc.) will be stored in the blockchain through a smart contract in subsequent steps to ensure the security and transparency of the data.

[0133] It should be noted that the specific composition structures of the spatial feature extraction network, temporal feature extraction network, and key feature extraction network included in the radiation source recognition model of this application are not limited and can be selected according to actual situations and empirical values. For example, the spatial feature extraction network can be a Deep Convolutional Neural Networks (DCNN) or a Convolutional Neural Network (CNN); the temporal feature extraction network can be a Recurrent Neural Network (RNN) such as a Long Short-Term Memory (LSTM) or a Gated Recurrent Unit (GRU); the key feature processing network can be a Multi-Head Attention mechanism, etc.

[0134] Among them, CNN (or DCNN) is very effective in processing data with spatial features and is particularly suitable for processing spectral images or other forms of two-dimensional data. Through layer-by-layer convolution operations, CNN can automatically extract and learn local features in the input data, summarize these features layer by layer, and finally generate classification results. For spectral images or other two-dimensional feature maps, CNN can effectively capture the spatial frequency distribution and pattern features of signals. In contrast, RNNs (such as LSTM, GRU) are more suitable for processing sequential data, especially time series data. When processing the time-domain features of electromagnetic signals, RNNs can remember and analyze the relationships between adjacent time points before and after in the sequence. This gives RNNs an obvious advantage when processing feature data with time dependence. For example, when identifying signals with a rapidly changing frequency, RNNs can capture the dynamic features of the signal changing over time.

[0135] Moreover, in order to overcome the problem of vanishing gradients that traditional RNNs (such as LSTM, GRU) may encounter when processing long sequence data, this application adopts a supervised learning training method to train a radiation source recognition model. Among them, in order to train the radiation source recognition model, first of all, a labeled dataset is required, including a large number of radiation source signals of known types and their corresponding labels. During the training process, the model gradually adjusts its parameters by minimizing the loss function so as to output the correct radiation source type when given input features. After training, the model can predict new input feature data to identify the type of radiation source. The evaluation and optimization of model performance are key steps to ensure the accuracy of the radiation source recognition model. Commonly used evaluation metrics include accuracy, recall, etc. By comparing the output of the model with the actual labels, the recognition performance of the model can be quantified. If the recognition accuracy is insufficient, the model performance can be further improved by increasing the amount of training data, adjusting the model structure, or performing parameter optimization. In addition, in practical applications, in order to cope with new types of radiation sources or changes in signal features, the model also needs to be retrained and updated regularly, which can also be achieved through the on-chain upgrade mechanism of smart contracts in subsequent steps. The radiation source recognition model can effectively map complex electromagnetic signal features to specific radiation source types, providing a basic guarantee for subsequent on-chain data storage and model upgrade.

[0136] Specifically, during the training process, each input data (signal feature) of the model has a clear label (type of the signal). The training process includes the input of data, the calculation of the loss function, and the optimization process. Among them, the input data is the electromagnetic signal data after preprocessing and feature extraction. For CNN, this is usually a multi-dimensional tensor (such as time-frequency graph, spectrogram, etc.); for RNN, the input may be a sequence of time-series data. For example, if CNN is used, the input data is usually two-dimensional image data, such as the spectrogram, time-frequency graph of the signal, etc. Suppose there is a spectrogram, and the input shape is 10x100 (10 time windows, 100 frequency points). If RNN is used, the input data can be a one-dimensional sequence, such as the feature sequence of the signal. Suppose the sequence has 100 time-moment data, and each time moment has 10 features, the input shape is (100, 10). The target output is the class label corresponding to each signal. For example, if identifying the signal source as Wi-Fi, Bluetooth, radar, or other types of devices, then the output is a classification label. For example, the label for Wi-Fi is 0, the label for Bluetooth is 1, the label for radar is 2, and so on.

[0137] The training process includes:

[0138] Forward propagation: During the training process, the input data undergoes forward propagation through the neural network. Neurons in each layer calculate an output value, and these output values will ultimately obtain the prediction result of the model through the last layer.

[0139] Calculating the loss function: The loss function is a function that measures the difference between the model output and the actual label. In supervised learning, the goal of the model is to minimize this loss function. The commonly used loss function is the cross-entropy loss function, which is particularly suitable for classification tasks.

[0140] Calculation of cross-entropy loss: The cross-entropy loss function is used to measure the difference between the predicted probability and the actual label, and is particularly suitable for classification problems. If the predicted probability differs more from the actual label, the loss value is larger. The goal of training is to minimize the value of the loss function, so that the model's prediction is as close as possible to the true label.

[0141] Backward propagation: After forward propagation, the value of the loss function is calculated. Next, the backpropagation algorithm is used to calculate the gradient of each parameter (weights and biases of neurons), and update the model parameters according to these gradients to reduce the value of the loss function.

[0142] Parameter update: Use an optimization algorithm (such as gradient descent method) to update the weights of the model. Common optimization algorithms include: stochastic gradient descent, Adam optimizer, etc.

[0143] The optimization process minimizes the loss function by adjusting the model parameters, ultimately enabling the model to accurately predict the type of radiation source.

[0144] Among them, an example of minimizing the loss function can be: Assume the cross-entropy loss function is adopted. When the true label of the input signal is Wi-Fi (the label is 0), and the probability output by the model is [0.7, 0.2, 0.1] (that is, the probability that the model predicts Wi-Fi is 70%, the probability of Bluetooth is 20%, and the probability of radar is 10%).

[0145] S103: Use the loose coupling connection model to send the recognition result and the model parameters of the radiation source recognition model to the smart contract on the blockchain platform.

[0146] In this embodiment, after obtaining the recognition result of the target radiation source signal (including but not limited to the type and source of the target radiation source signal, etc.) by using the radiation source recognition model in step S102, further, the loose coupling connection model can be used to convert the formats of the recognition result and the model parameters of the radiation source recognition model (such as neural network weights, optimization algorithm parameters, etc.) to obtain the converted recognition result and model parameters. Among them, the converted recognition result and model parameters meet the input requirements of the smart contract. Then, use the loose coupling connection model to call the interface for interacting with the smart contract, and send the converted recognition result and model parameters to the smart contract on the blockchain platform according to the preset communication protocol for subsequent step S104.

[0147] S104: Store the recognition result and the model parameters of the radiation source recognition model on the blockchain platform through the smart contract.

[0148] In this embodiment, after using the loose coupling connection model in step S103 to send the recognition result and the model parameters of the radiation source recognition model to the smart contract on the blockchain platform, further, the smart contract can verify the recognition result and the model parameters of the radiation source recognition model, and combine the storage mechanism provided by the blockchain to store the verified recognition result and model parameters on the blockchain platform, so as to achieve the immutability, credible traceability and cross-organization collaborative application of the result data, meet the requirements of various highly trusted scenarios, and achieve the ideal radiation source recognition effect and data processing effect.

[0149] On this basis, in order to further improve the recognition effect, the continuous optimization and adaptation to new challenges of the radiation source recognition model can also be ensured by means of on-chain upgrade of the model chain. Specifically, an optional implementation method is as follows: First, the model parameters to be updated corresponding to the radiation source recognition model can be obtained; and the model parameters to be updated are sent to the smart contract by using the radiation source recognition model; then, the smart contract is used to verify the model parameters to be updated according to the preset verification mechanism to obtain the model parameters to be updated that pass the verification; then, the model parameters to be updated that pass the verification are updated and stored on the blockchain platform; and the update process of the model parameters to be updated is recorded and stored on the blockchain platform, so that the on-chain verification and upgrade of the model parameters are realized by using the smart contract, ensuring that the model can adapt to the changing environment and data while maintaining high performance, thereby improving the recognition effect of the model.

[0150] It should be noted that in the whole process of radiation source recognition and data on-chain, the loose coupling connection model plays a bridging role. Through the API or communication protocol, the recognition results generated by the radiation source recognition model are effectively docked with the smart contract on the blockchain platform, realizing data on-chain and the update of the model in the subsequent steps. Therefore, through the loose coupling connection model, the scalability, flexibility and independence among various parts of the entire recognition system can be ensured, avoiding the impact of changes or failures in a single module on the stable operation of the entire recognition system.

[0151] Specifically, the action process of the loose coupling connection model can specifically include:

[0152] (1) Data interaction:

[0153] Preset communication protocol: The loose coupling connection model may interact with the radiation source recognition model and the smart contract through standardized preset communication protocols (such as Hypertext Transfer Protocol (HTTP), Message Queuing Telemetry Transport (MQTT), etc.). The radiation source recognition model can transfer the recognition results (such as signal type, frequency band, radiation source location, etc.) to the loose coupling connection model through these interfaces. For example, the radiation source recognition model can send the recognition results through an API request, and the interface design will ensure the unified data format and dock with the data format required by the smart contract.

[0154] Among them, the API design and implementation are the core parts of the loosely coupled connection model. The API (Application Programming Interface) provides a standardized interface, enabling the radiation source identification model to structure and send the identified data to the smart contract. The API needs to be highly general and extensible to adapt to different types of data and smart contract requirements. When designing the API, standards such as the RESTful architecture or GraphQL are usually adopted to ensure the flexibility and usability of the interface. During the implementation process, the API should support multiple data formats (such as JSON, XML) and communication protocols (such as HTTP) to ensure efficient data transmission between different modules.

[0155] The selection and configuration of the preset communication protocol are also important links in the loosely coupled connection model. The communication protocol is used to define the rules and processes for data transmission between various parts of the system. In this embodiment, the communication protocol needs to ensure the security, speed, and reliability of the data transmission process from the radiation source identification model to the smart contract. Common protocol options include HTTP / HTTPS, MQTT, etc. Among them, HTTP / HTTPS is often used to transmit the identification results to the blockchain platform due to its wide use and security. For scenarios with high real-time requirements, a message queue-based protocol (such as MQTT) can be considered to support asynchronous communication and message distribution, thereby improving the system's response speed and processing capacity.

[0156] It should be noted that the smart contract is a part of the blockchain platform, but its function is more focused on automated and programmed operations. The smart contract is an automated script or protocol running on the blockchain, and its main functions include: First, automatic execution, that is, the smart contract will automatically execute specific operations when specific conditions are met. For example, when the radiation source identification model sends data to the smart contract, the smart contract can automatically write the data into the blockchain or trigger model updates according to certain conditions. Second, logical processing, that is, the smart contract not only stores data, but also contains business logic and rules, such as verifying the validity of the identification results, controlling data flow, and triggering subsequent operations. The smart contract can also update, verify, or adjust model parameters. Third, control and management, that is, the smart contract can manage various operations, such as controlling access permissions to data, automatically updating model parameters, and triggering blockchain events. Therefore, the blockchain platform is mainly responsible for data storage and decentralized management, ensuring the immutability and transparency of data. The smart contract is responsible for implementing decentralized automated operations, automatically executing predefined rules and tasks, and operating on data processing and updates.

[0157] For example: Suppose the radiation source identification model detects a new Wi-Fi signal and transmits its identification results (such as signal type, frequency band, location, etc.) to the smart contract through the API. After receiving the identification results, the smart contract will verify the validity of the data and automatically decide whether to write the data into the blockchain according to the rules. If certain conditions are met (such as the data being valid and the signal strength exceeding a certain threshold), the smart contract will trigger relevant operations. Under the operation of the smart contract, the data will be written into the blockchain platform for permanent storage. This process ensures the immutability and transparency of the data, and the record is available for subsequent query and auditing.

[0158] In short, the smart contract is mainly responsible for programming and executing business logic, while the blockchain platform is responsible for storing and managing data. The two cooperate to ensure the automation, transparency, and immutability of the system.

[0159] (2) Data format conversion:

[0160] The loosely coupled connection model also needs to appropriately convert the data to ensure that the data can meet the input requirements of the smart contract platform. This can include, but is not limited to, data encoding and data verification. Among them, data encoding refers to converting the identification results (such as frequency band, signal strength, radiation source type, etc.) into JSON format, Protobuf, or other data formats suitable for storage and transmission. Data verification refers to verifying the integrity and validity of the data to ensure that the data conforms to the preset rules (for example, whether the signal strength is within the acceptable range and whether it meets the standards for radiation source identification).

[0161] (3) Data on-chain:

[0162] The core task of data on-chain is to store the processed and verified identification results on the blockchain. The loosely coupled connection model submits the data (such as identification results, timestamp, frequency band, etc.) to the blockchain network by calling the smart contract interface. The smart contract will perform data storage operations on the chain and generate immutable transaction records.

[0163] It should be noted that data uploading to the blockchain is one of the key functions of the loose coupling connection model. After the radiation source recognition result generated by the recognition model is transmitted to the smart contract through the API or communication protocol, data uploading operation needs to be carried out. This process usually includes steps such as data format conversion, signature authentication, and interaction with the blockchain. Format conversion is to convert the recognition result into a data format suitable for blockchain storage, such as the commonly used ABI format on Ethereum. Signature authentication is used to ensure the credibility of the data source and prevent the data from being tampered with during transmission. Finally, through interaction with the blockchain node, the data is written into the blockchain to complete the data uploading operation. This process is usually achieved by calling the smart contract of the blockchain platform, and the smart contract verifies the legality of the data according to the predefined rules and stores it on the chain.

[0164] (4) Trigger model update:

[0165] After the smart contract completes data uploading to the blockchain, the loose coupling connection model can also call another interface to trigger model update. The model update process may be initiated by triggering certain conditions (such as the addition of new recognition results, signal type changes, etc.). For example, when it is found that a certain type of radiation source (such as Wi-Fi) appears frequently, the loose coupling connection model can trigger the training of a new model or fine-tune the existing model to enhance the recognition accuracy.

[0166] Among them, the updated model includes two contents: one is the update of model parameters. For example, when using new data to train the radiation source recognition model, the weights, biases, and other parameters of the model will change. Whenever the recognition result changes or new data arrives, the loose coupling connection module can trigger the smart contract to update these parameters. The updated model can be retrained and deployed to the recognition model to improve the recognition accuracy and adaptability. The other is model version management, that is, each time the model is updated, the version number of the new model needs to be recorded. Through the smart contract, these version information will be uploaded to the chain to ensure that each version can be traced, and different versions of the model can be managed in parallel.

[0167] That is to say, while uploading data to the blockchain, the model update function also needs to be realized through the loose coupling connection model. When the radiation source recognition model detects a decline in model performance or needs to be upgraded, the parameters of the new model will be transmitted to the smart contract through the API. Under the management of the smart contract, the new model parameters will be verified and stored on the blockchain, and then the model can be dynamically updated according to the needs. This process ensures the transparency and traceability of model upgrades, and at the same time, using the decentralized characteristics of the blockchain, it ensures the security and credibility of the model update process.

[0168] In addition, to ensure the stability and efficiency of the overall recognition system, monitoring and logging are an essential part of the loosely coupled connection model. It is necessary to monitor the status of data transmission, the success rate of API calls, and the process of model updates in real time, and retain detailed information on key operations through logging. These monitoring and log data not only help troubleshoot problems and optimize system performance but also provide important audit information for recognition security. The loosely coupled connection model enables seamless connection between the radiation source recognition model and the smart contract, allowing recognition data to be reliably uploaded to the chain and providing support for subsequent model updates. This loosely coupled design ensures high scalability and flexibility of the system, enabling different modules to be independently upgraded and expanded while maintaining the overall consistency and stability of the system.

[0169] It should also be noted that, to execute the above steps S103 - S104, it is necessary to accurately design the smart contract in advance, as it directly relates to the reliability of data upload to the chain, the security of model upgrades, and the automation level of the system. Through the smart contract, key tasks such as data storage, verification, and model updates can be automatically executed without manual intervention. An optional implementation method is that the design process of the smart contract can include, but is not limited to, the design of data storage structures, the design of model upgrade rules, consensus mechanisms, and the security of the contract. The specific explanations are as follows:

[0170] (1) The design of data storage is the basic part of the smart contract design. When storing radiation source recognition results and model parameters on the blockchain, it is necessary to consider the data format, storage location, and access method. Generally, the two main ways to store data on the blockchain are event logs and contract state variables. Event logs are suitable for storing historical data, such as recognition results, for subsequent query and traceability. State variables are used to store current active data, such as the latest model parameters. The smart contract needs to define the structure and format of the data, such as using structs to organize complex data types, and implement data indexing and storage through methods such as mapping or arrays. To ensure the integrity and security of the data, the smart contract also needs to encrypt or hash the stored data to prevent tampering.

[0171] (2) The design of model upgrade rules is one of the core functions of smart contracts. When the radiation source identification model needs to be updated, the parameters of the new model must be verified and deployed through the smart contract. To ensure the transparency and security of the model upgrade process, clear upgrade rules need to be defined in the smart contract. First, the smart contract should include a model verification mechanism, such as using digital signatures to verify the source and integrity of the new model. Second, the contract needs to define the storage and invocation methods of model parameters, such as storing the new model parameters in state variables and providing interfaces for the radiation source identification module to call. In addition, to address potential errors or attacks, the contract can also design a Revert Mechanism. When the new model fails verification or deployment, it can automatically roll back to the previous version to ensure the stability of the radiation source identification and data uploading functions.

[0172] (3) The selection and implementation of the consensus mechanism are important steps to ensure the effective operation of smart contracts. The consensus mechanism determines how nodes on the blockchain reach an agreement on the correctness of data. On public blockchains, common consensus mechanisms include Proof of Work (PoW), Proof of Stake (PoS), etc., while in consortium blockchains, mechanisms such as Byzantine Fault Tolerance (BFT) may be used. Regardless of the consensus mechanism adopted, the speed, energy consumption, and security of consensus need to be considered in smart contract design. For example, in an Ethereum-based system, the contract can use the PoS mechanism to speed up transaction confirmation and reduce energy consumption.

[0173] (4) Security design is an aspect that cannot be ignored in smart contract development. Since smart contracts cannot be changed once deployed, sufficient security considerations must be taken during the design phase to prevent potential vulnerabilities and attacks. Common security measures include reentrancy attack protection, overflow checks, and access control. For example, a reentrancy lock should be added to the contract function to prevent reentrancy attacks, and overflow and underflow checks should be performed through the SafeMath library to ensure the security of integer operations. In addition, access control should be set for operations related to model updates and data storage in the contract, allowing only authorized users or contracts to call these critical functions to prevent malicious users from tampering with data or maliciously upgrading the model.

[0174] Finally, the deployment and testing of smart contracts are important steps to ensure their correctness and stability. Before deploying the contract, sufficient testing needs to be carried out on the test network (such as Ropsten and Rinkeby of Ethereum) to simulate various possible usage scenarios, discover and fix potential problems. After testing, the smart contract can be deployed to the main network and monitored and maintained through tools such as blockchain browsers to ensure that the contract performs well in actual operation.

[0175] In addition, the on-chain data storage achieved through the above step 104 is a key link to ensure data security and credibility. By storing the radiation source identification results and model parameters on the blockchain, the decentralized, immutable, and transparent characteristics of the blockchain can be utilized to ensure the integrity and credibility of the data.

[0176] This step specifically involves the design and implementation of aspects such as data format conversion, data encryption and signature, storage mechanism selection, and data access control. Data format conversion is the first step in on-chain data storage. After the radiation source recognition model generates data, this data needs to be converted into a format suitable for blockchain storage. For example, the recognition results usually exist in the form of structured data, which may include information such as the type of radiation source, signal strength, frequency range, etc. This information needs to be serialized into a standard format, such as JSON, for on-chain storage. Model parameters are usually high-dimensional floating-point matrices and need to be further compressed and encoded to reduce the storage space occupied on the blockchain. To ensure data consistency and readability, the data format conversion process should strictly follow the predefined standards and specifications. Next, data encryption and signature are important steps to ensure data security. Before storing data on the blockchain, the data needs to be encrypted to prevent unauthorized access during data transmission and storage. Common encryption algorithms include symmetric encryption (such as AES) and asymmetric encryption (such as RSA). For relatively sensitive data such as model parameters, asymmetric encryption is usually used because it can ensure that only authorized parties holding the correct private key can decrypt and use this data. In addition, to verify the authenticity of the data, digital signatures are also applied to the data. This process usually involves the data generator signing the data with its private key, and then storing the public key on the chain so that other nodes can verify the integrity of the data and the credibility of its source. The selection of the storage mechanism is the core of on-chain data storage. When storing data on the blockchain, there are various storage methods to choose from, common ones including contract state variables, event logs, and distributed storage (such as IPFS), etc. Contract state variables are suitable for storing small-scale, structured data, such as the current model parameters or the latest recognition results. These data can be directly stored in the smart contract and accessed and updated through the state management mechanism of the smart contract. For large-scale historical data or relatively large model parameters, event logs or distributed storage can be considered. Event logs can record all changes in recognition results and are easy to retrieve and verify. Distributed storage (such as IPFS) is suitable for storing large files. The blockchain only stores the hash value of the file, thus achieving efficient data management and search, while reducing the storage burden on the blockchain. Data access control is another key link to ensure the security of on-chain data. Although the blockchain has the characteristics of being publicly transparent, certain sensitive data should not be publicly available to all users. Therefore, an access control mechanism needs to be added in the design of the smart contract to ensure that only authorized users can access or modify specific data. This can be achieved through various methods, such as role-based access control, multi-signature, and permission management lists, etc. Through these mechanisms, who has the right to read, write, or update the data on the chain can be strictly controlled, preventing data leakage or unauthorized tampering.

[0177] Finally, the integrity and traceability of data are the ultimate goals of on-chain data storage. The immutable feature of the blockchain ensures that all stored data cannot be arbitrarily modified once written. Through the blockchain's historical record function, the source and change process of any data can be traced, thus ensuring the credibility of the data. For example, when verifying the result of a radiation source identification, detailed information such as the generation time, generation device, and model version used at that time can be queried through the blockchain. This provides strong support for data auditing and security management and also enhances the overall transparency and credibility of the on-chain process.

[0178] In addition, when using smart contracts for on-chain model upgrades, the specific implementation process can include steps such as uploading new model parameters, designing a verification mechanism (which can be understood as for verifying new model parameters), deploying and activating the model, and managing update records.

[0179] Among them, the upload of new model parameters is the starting step of on-chain upgrades. When new model parameters are obtained, they need to be uploaded to the blockchain through a predefined interface. This is usually done by the radiation source identification model sending the model parameters to the smart contract via an API. To optimize storage and transmission efficiency, the new model parameters should be compressed and encoded. For example, the model weight matrix can be compressed into a binary format to reduce the on-chain storage space requirement. During the upload process, the contract usually saves the hash value of the parameters for use in subsequent verification processes.

[0180] The verification of new model parameters is a crucial link in ensuring model quality and security. Strict verification rules need to be defined in the smart contract to check whether the uploaded model parameters meet the expectations. These verification steps include checking the correctness of the parameter format, verifying the digital signature to ensure the legitimacy of the parameter source, and performing a reasonableness check on the parameters through predefined standards. For example, the contract can ensure that the parameters have not been tampered with during transmission by comparing the hash value of the uploaded parameters with the hash value stored on the chain.

[0181] Among them, hash value comparison is usually a means used to ensure data integrity and prevent tampering during data transmission. The process of calculating and comparing the hash values of model parameters can be as follows: Before uploading the model parameters to the blockchain, it is first necessary to calculate the hash value of the model parameters (for example, using hash algorithms such as SHA-256). The hash value is the digital fingerprint of the model parameter content and is unique. Before uploading, calculate the hash value of the model first, and then store this hash value in the smart contract. When the model parameters are first uploaded to the blockchain, the hash value (instead of the actual model data) will be stored on the blockchain. In this way, the data in the blockchain only contains the hash value and does not store the complete model parameters. When a new model version (or new model parameters) is uploaded, it may first calculate the hash value of the uploaded model version or parameters, and then compare it with the hash value pre-stored on the blockchain. If the uploaded model version or parameters match the predetermined model version or parameters, it indicates that the model data has not been tampered with. If the hash values match, it indicates that the transmitted data is consistent with the original data, and the uploaded new model version or parameters are legal, and the subsequent model upgrade and update process can continue. If they do not match, it means that the data may have been tampered with during transmission, and the subsequent model upgrade and update process cannot be carried out.

[0182] In addition, the smart contract can also set an approval mechanism, requiring multiple authorized nodes to verify the new model version or parameters to further improve the credibility of the model parameters.

[0183] The deployment and activation of the model are the core links of on-chain upgrade. The verified new model parameters need to be deployed into the smart contract. The smart contract updates the new model parameters to the on-chain storage and updates the current model version stored in the contract. The deployment process usually includes writing the new model parameters into the state variables of the contract and adjusting the model running environment to ensure that the new model can be recognized and called. To ensure the smooth activation of the new model, the contract can set a phased testing and evaluation mechanism to ensure that the performance of the new model in actual operation meets the expectations.

[0184] The management of update records is the key to ensuring the transparency and traceability of the overall process of on-chain model upgrade. Each model upgrade needs to record detailed upgrade information, including upgrade time, uploader, model version, verification results, etc. These records will be stored on the blockchain as part of the smart contract to ensure that the process of each upgrade is public and auditable. Such records not only help to track the historical changes of the model but also provide important debugging information when problems occur.

[0185] In addition, automation and governance mechanisms can also serve as additional features during the in-chain model upgrade process. Upgrade conditions that trigger automatically can be preset during the design of smart contracts, such as automatically triggering an update request when the model performance degrades. In addition, smart contracts can introduce governance mechanisms that allow specific users (such as administrators) to review and vote on model updates, thereby increasing the flexibility and governance capabilities of the recognition system. Such mechanisms ensure that model updates are not only automated but also in line with the actual requirements and strategies of the recognition system.

[0186] Finally, upgraded monitoring and evaluation can also be an important part of ensuring the continuous optimization of the new model. After deploying the new model, its performance needs to be continuously monitored to detect its performance in the actual environment. This includes evaluating indicators such as the accuracy, response time, and computing resource consumption of the model, and further adjusting and optimizing the model based on the monitoring results. If it is found that the new model performs poorly in actual applications, it is possible to quickly roll back to the previous model version through a smart contract to ensure the stability and reliability of the function.

[0187] Thus, through the steps of in-chain model upgrade, it can be ensured that the radiation source recognition model always remains in the optimal state and adapts to environmental changes and new challenges. Moreover, the use of smart contracts not only improves the automation and security of the upgrade process but also guarantees the transparency and traceability of the entire upgrade process.

[0188] In summary, for a method for radiation source recognition and data processing provided in this embodiment, first, a target radiation source signal to be recognized is obtained; then the target signal features of the target radiation source signal are extracted; then the target signal features are input into a pre-constructed radiation source recognition model to output the recognition result of the target radiation source signal, where the recognition result includes the type and source of the target radiation source signal. Then, using a loosely coupled connection model, the recognition result and the model parameters of the radiation source recognition model are sent to a smart contract on a blockchain platform, and further through the smart contract, the recognition result and the model parameters of the radiation source recognition model can be stored on the blockchain platform.

[0189] It can be seen that since this application uses a radiation source recognition model, a smart contract, and blockchain technology trained by integrating deep learning models to recognize the target radiation source signal, it realizes the accurate recognition of the target radiation source signal, and stores the recognition result and model parameters on the blockchain, ensuring the immutability, credible traceability, and cross-organization collaborative application of the recognition result data and model parameters, thereby meeting the requirements of various highly trusted scenarios and achieving an ideal radiation source recognition effect and data processing effect.

[0190] Second Embodiment

[0191] This embodiment will introduce a radiation source identification and data processing device. For related content, please refer to the above method embodiment.

[0192] See Figure 2 , which is a schematic diagram of the composition of a radiation source identification and data processing device provided in this embodiment. The device includes:

[0193] An extraction unit 201, configured to obtain a target radiation source signal to be identified; and extract target signal features of the target radiation source signal;

[0194] An identification unit 202, configured to input the target signal features into a pre-constructed radiation source identification model, and output an identification result of the target radiation source signal; the identification result includes the type and source of the target radiation source signal;

[0195] A sending unit 203, configured to use a loose coupling connection model to send the identification result and the model parameters of the radiation source identification model to a smart contract on a blockchain platform;

[0196] A storage unit 204, configured to store the identification result and the model parameters of the radiation source identification model to the blockchain platform through the smart contract.

[0197] In an implementation manner of this embodiment, the extraction unit 201 includes:

[0198] A first analysis subunit, configured to perform time-domain analysis and frequency-domain analysis on the target radiation source signal to extract time-domain features and frequency-domain features;

[0199] A second analysis subunit, configured to analyze the modulation characteristics of the target radiation source signal to extract modulation features; and use the time-domain features, frequency-domain features, and the modulation features to constitute the target signal features of the target radiation source signal.

[0200] In an implementation manner of this embodiment, the radiation source identification model includes a spatial feature extraction network, a temporal feature extraction network, and a key feature extraction network; the identification unit 202 includes:

[0201] A first conversion subunit, configured to perform format conversion on the target signal features to obtain converted target signal features;

[0202] An identification subunit, configured to input the converted target signal features into the spatial feature processing network, temporal feature processing network, and key feature processing network in the radiation source identification model, and identify the type and source of the target radiation source signal as an identification result.

[0203] In one implementation of this embodiment, the spatial feature extraction network is a deep convolutional neural network DCNN; the temporal feature extraction network is a long short-term memory network LSTM; the key feature processing network is a multi-head self-attention mechanism.

[0204] In one implementation of this embodiment, the sending unit 203 includes:

[0205] A second conversion subunit, configured to use the loose coupling connection model to perform format conversion on the recognition result and the model parameters of the radiation source recognition model, so as to obtain the converted recognition result and model parameters; the converted recognition result and model parameters meet the input requirements of the smart contract;

[0206] A sending subunit, configured to use the loose coupling connection model to call an interface for interacting with the smart contract, and send the converted recognition result and model parameters to the smart contract on the blockchain platform according to a preset communication protocol.

[0207] In one implementation of this embodiment, the storage unit 204 is specifically configured to:

[0208] Verify the recognition result and the model parameters of the radiation source recognition model through the smart contract, and combine the storage mechanism provided by the blockchain to store the verified recognition result and model parameters on the blockchain platform.

[0209] In one implementation of this embodiment, the device further includes:

[0210] An acquisition unit, configured to acquire the model parameters to be updated corresponding to the radiation source recognition model; and use the radiation source recognition model to send the model parameters to be updated to the smart contract;

[0211] A verification unit, configured to use the smart contract to verify the model parameters to be updated according to a preset verification mechanism, so as to obtain the model parameters to be updated that pass the verification;

[0212] An update unit, configured to update the model parameters to be updated that pass the verification to the blockchain platform for storage; and record the update process of the model parameters to be updated on the blockchain platform for storage.

[0213] In one implementation of this embodiment, the design process of the smart contract includes the design of the data storage structure, the design of the model upgrade rules, the consensus mechanism, and the security of the contract.

[0214] In one implementation of this embodiment, the radiation source recognition model is a deep learning model trained by means of supervised learning.

[0215] Furthermore, an embodiment of the present application further provides a radiation source identification and data processing device, including: a processor, a memory, and a system bus;

[0216] The processor and the memory are connected through the system bus;

[0217] The memory is used to store one or more programs, and the one or more programs include instructions that, when executed by the processor, cause the processor to execute any implementation method of the above-mentioned radiation source identification and data processing method.

[0218] Furthermore, an embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored, and when the instructions are run on a terminal device, the terminal device is caused to execute any implementation method of the above-mentioned radiation source identification and data processing method.

[0219] Furthermore, an embodiment of the present application further provides a computer program product, which, when run on a terminal device, causes the terminal device to execute any implementation method of the above-mentioned radiation source identification and data processing method.

[0220] From the description of the above embodiments, those skilled in the art can clearly understand that all or part of the steps in the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in each embodiment or some parts of the embodiments of the present application.

[0221] It should be noted that the embodiments in this specification are described in a progressive manner, and the key point of each embodiment is to describe the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0222] It should also be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0223] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying a radiation source and processing data, characterized in that, Including: Obtain a target radiation source signal to be recognized; and extract target signal features of the target radiation source signal; Input the target signal features into a pre-constructed radiation source recognition model, and output an identification result of the target radiation source signal; the identification result includes the type and source of the target radiation source signal; Use a loose coupling connection model to send the identification result and the model parameters of the radiation source recognition model to a smart contract on a blockchain platform; Through the smart contract, store the identification result and the model parameters of the radiation source recognition model on the blockchain platform.

2. The method according to claim 1, wherein The extracting the target signal features of the target radiation source signal includes: Perform time-domain analysis and frequency-domain analysis on the target radiation source signal to extract time-domain features and frequency-domain features; Analyze the modulation characteristics of the target radiation source signal to extract modulation features; and use the time-domain features, frequency-domain features, and the modulation features to constitute the target signal features of the target radiation source signal.

3. The method according to claim 1, characterized in that, The radiation source recognition model includes a spatial feature extraction network, a temporal feature extraction network, and a key feature extraction network; The inputting the target signal features into a pre-constructed radiation source recognition model to identify the type and source of the target radiation source signal as an identification result includes: Perform format conversion on the target signal features to obtain converted target signal features; Input the converted target signal features into the spatial feature processing network, temporal feature processing network, and key feature processing network in the radiation source recognition model to identify the type and source of the target radiation source signal as an identification result.

4. The method according to claim 3, wherein The spatial feature extraction network is a deep convolutional neural network DCNN; the temporal feature extraction network is a long short-term memory network LSTM; the key feature processing network is a multi-head self-attention mechanism.

5. The method according to claim 1, characterized in that, The using a loose coupling connection model to send the identification result and the model parameters of the radiation source recognition model to a smart contract on a blockchain platform includes: Use the loose coupling connection model to perform format conversion on the identification result and the model parameters of the radiation source recognition model to obtain converted identification result and model parameters; the converted identification result and model parameters meet the input requirements of the smart contract; Use the loose coupling connection model to call an interface for interacting with the smart contract, and send the converted identification result and model parameters to the smart contract on the blockchain platform according to a preset communication protocol.

6. The method according to claim 1, wherein The through the smart contract, storing the identification result and the model parameters of the radiation source recognition model on the blockchain platform includes: Verify the identification result and the model parameters of the radiation source recognition model through the smart contract, and combine with the storage mechanism provided by the blockchain to store the verified identification result and model parameters on the blockchain platform.

7. The method according to claim 1, wherein After the through the smart contract, storing the identification result and the model parameters of the radiation source recognition model on the blockchain platform, the method further includes: Obtain the model parameters to be updated corresponding to the radiation source identification model; and use the radiation source identification model to send the model parameters to be updated to the smart contract; Use the smart contract to verify the model parameters to be updated according to a preset verification mechanism to obtain the model parameters to be updated that pass the verification; Update the model parameters to be updated that pass the verification to the blockchain platform for storage; and record the update process of the model parameters to be updated on the blockchain platform for storage.

8. The method according to any one of claims 1-7, characterized in that The design process of the smart contract includes the design of the data storage structure, the design of the model upgrade rules, the consensus mechanism, and the design of the security of the contract.

9. The method according to any one of claims 1-7, characterized in that, The radiation source identification model is a deep learning model trained by means of supervised learning.

10. A radiation source identification and data processing device, characterized in that, It includes: An extraction unit, configured to obtain a target radiation source signal to be identified; and extract target signal features of the target radiation source signal; An identification unit, configured to input the target signal features into a pre-constructed radiation source identification model and output an identification result of the target radiation source signal; the identification result includes the type and source of the target radiation source signal; A sending unit, configured to use a loosely coupled connection model to send the identification result and the model parameters of the radiation source identification model to a smart contract on the blockchain platform; A storage unit, configured to store the identification result and the model parameters of the radiation source identification model on the blockchain platform through the smart contract.

11. A radiation source identification and data processing device, characterized in that, It includes: A processor, a memory, and a system bus; The processor and the memory are connected through the system bus; The memory is configured to store one or more programs, and the one or more programs include instructions that, when executed by the processor, cause the processor to execute the method according to any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, Instructions are stored in the computer-readable storage medium, and when the instructions run on the terminal device, the terminal device is caused to execute the method according to any one of claims 1-9.

Images