Signal Anti-interference Method for Satellite Remote Sensing Application System

Abstract

The present invention discloses a signal anti-interference method for a satellite remote sensing application system, comprising: using a satellite antenna to receive an electromagnetic signal from a remote sensing satellite and performing preprocessing; performing interference detection on the preprocessed signal, and when the energy of a partial scale is abnormally high, determining that interference exists in the digital signal; combining features extracted from the interference detection into a feature vector, inputting the feature vector into a classification model constructed based on a convolutional neural network to perform interference classification, outputting the probability of each interference type, and selecting the type with the highest probability as the judgment result; based on the output judgment result, using a corresponding adaptive filtering algorithm to suppress the interference signal existing in the digital signal to obtain a purified signal; using a mother wavelet function to perform inverse transformation on the purified signal to reconstruct the signal; and saving the reconstructed signal in the form of a digital file, and transmitting the reconstructed signal to a ground station or a data processing center for subsequent analysis and application.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite communication and signal processing, and in particular to a signal anti-interference method for a satellite remote sensing application system. Background Art

[0002] With the development of remote sensing technology, satellite remote sensing has become an important technical means for environmental monitoring, resource exploration, and disaster warning. Satellite remote sensing systems use high-resolution sensors carried by in-orbit satellites to collect all-round, multi-spectral data on the earth's surface, and use radio frequency bands to transmit remote sensing data to ground stations in real time. In order to ensure the stability and reliability of data transmission, signal reception and processing technologies have also been continuously improved.

[0003] Currently, the processing methods for satellite remote sensing signals mainly include frequency domain processing based on filtering, denoising algorithms in the time domain, and wavelet transform. However, due to the presence of ground radio equipment, interference from other satellite signals, and electromagnetic noise in space, satellite remote sensing signals are often subject to various interferences during transmission, resulting in signal distortion and information loss. In addition, electromagnetic disturbances caused by solar activity may also affect the transmission quality of the signal. Therefore, in order to ensure the accuracy and real-time performance of the data, a more effective signal anti-interference method is urgently needed.

[0004] Existing signal processing technologies, such as simple low-pass filters, band-pass filters, and threshold-based denoising methods, often perform less than ideally in complex electromagnetic environments. The limitations of these traditional methods mainly lie in their lack of sufficient recognition and classification capabilities for multi-source interference, especially when narrowband interference, broadband interference, and random pulse interference coexist. These methods find it difficult to effectively distinguish and suppress different types of interference; in addition, existing filtering methods usually rely on fixed parameter settings and cannot be adaptively adjusted according to dynamic changes in the interference environment, resulting in limited real-time signal processing and anti-interference effects. Summary of the Invention

[0005] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the technical problems solved by the present invention are: the problems of lack of adaptability, insufficient multi-source interference classification capability and incomplete signal recovery in the prior art.

[0008] To solve the above technical problems, the present invention provides the following technical solution: using a satellite antenna to receive an electromagnetic signal from a remote sensing satellite, performing analog-to-digital conversion on the electromagnetic signal to obtain a digital signal;

[0009] Preprocessing the digital signal to eliminate DC offset and high-frequency noise and improve the signal-to-noise ratio;

[0010] Performing interference detection on the pre-processed digital signal, that is, comprehensively analyzing the energy distribution of each scale, and determining that interference exists in the digital signal when the energy of some scales is abnormally high;

[0011] The features extracted from interference detection are combined into a feature vector, which is input into a classification model built based on a convolutional neural network for interference classification. The probability of each interference type is output, and the type with the highest probability is selected as the judgment result.

[0012] According to the output determination result, a corresponding adaptive filtering algorithm is used to suppress interference signals present in the digital signal to obtain a purified signal;

[0013] Using a mother wavelet function, inversely transforming the purified signal, reconstructing the signal, and restoring the original remote sensing information;

[0014] The reconstructed signal is saved in the form of a digital file and transmitted to a ground station or data processing center for subsequent analysis and application.

[0015] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system according to the present invention, obtaining the digital signal includes:

[0016] Using a satellite antenna to receive electromagnetic signals from a remote sensing satellite to obtain analog radio frequency signals;

[0017] Converting the analog radio frequency signal into an intermediate frequency signal through a down converter;

[0018] According to the Nyquist sampling theorem, the intermediate frequency signal is sampled at a sampling frequency not less than twice the highest frequency of the signal to obtain a discrete time signal;

[0019] The sampled analog signal value is converted into a digital value through a quantizer;

[0020] The quantized signal digital value is then encoded into binary data to obtain the digital signal.

[0021] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system described in the present invention, the number of quantization bits converted by the quantizer depends on the resolution of the ADC.

[0022] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system described in the present invention, preprocessing the digital signal includes at least DC offset elimination and high-frequency noise elimination, wherein the DC offset elimination is achieved by calculating the mean of the digital signal and thereby eliminating the DC offset, and the high-frequency noise elimination is achieved by filtering by designing a low-pass filter to thereby eliminate the high-frequency noise.

[0023] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system according to the present invention, the interference detection includes:

[0024] Select mother wavelet function to perform discrete wavelet transform;

[0025] Perform multi-scale decomposition on the preprocessed signal to obtain detail coefficients and approximation coefficients at different scales;

[0026] Calculate the energy of the detail coefficients:

[0027] E j =∑ n |d j (n)| 2

[0028] Set the energy threshold T j , if E j >T j , then it is considered that there is interference on scale j;

[0029] Comprehensively analyzing the energy distribution of each scale, when the energy of a certain scale is abnormally high, it is determined that interference exists in the digital signal;

[0030] Among them, n is the signal number, j is the scale level, d j (n) is the detail coefficient of the jth scale, E j is the energy of the jth scale, T j is the energy threshold of the j-th scale.

[0031] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system according to the present invention, the classification model is constructed based on a convolutional neural network, including:

[0032] Input layer: receives feature vectors, the shape is represented as (N features ,1), where N features is the length of the eigenvector;

[0033] Convolutional layer: Use multiple convolution kernels to convolve the input feature vector to extract high-level features;

[0034] Activation function: Use ReLU function to introduce nonlinearity;

[0035] Pooling layer: Use maximum pooling to reduce feature dimensions and prevent overfitting;

[0036] Fully connected layer: maps the features extracted by the convolutional layer to the classification space;

[0037] Output layer: Use the Softmax activation function to output the probability distribution of each interference type.

[0038] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system described in the present invention, the signal feature vector to be classified is input into the classification model for calculation, the probability of each interference type is output, and the type with the highest probability is selected as the judgment result. The classification model uses the cross entropy loss function as the objective function, and its mathematical expression formula is as follows:

[0039]

[0040] Where C is the number of interference types, y i is the actual label, p i is the probability predicted by the model, and i is the label number.

[0041] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system according to the present invention, the interference types include at least narrowband interference, broadband interference and pulse interference, wherein:

[0042] If it is narrowband interference, select the adaptive notch filter;

[0043] If it is broadband interference, then select the adaptive linear prediction filter;

[0044] If it is pulse interference, select median filter or adaptive threshold filter.

[0045] As a preferred solution of the signal anti-interference method for satellite remote sensing application system described in the present invention, the mother wavelet function is used to perform inverse transformation on the purified signal, reconstruct the signal, and restore the original remote sensing information. The reconstruction formula is as follows:

[0046] x r (n)=∑ j (a j (n)+d j (n))

[0047] Among them, d j (n) is the detail coefficient of the jth scale, a j (n) is the approximate coefficient of the jth scale, n is the signal number, j is the scale level, x r (n) is the reconstructed signal.

[0048] As a preferred solution of the signal anti-interference method for a satellite remote sensing application system according to the present invention, in order to restore the time domain representation of the reconstructed signal and ensure the complete preservation of its effective information, signal correction is required, wherein:

[0049] If the adaptive filtering algorithm used introduces amplitude attenuation, a gain factor is used to correct it;

[0050] If the adaptive filtering algorithm used introduces phase delay, phase compensation technology is used to correct it.

[0051] Beneficial effects of the present invention:

[0052] 1. Achieved lossless conversion from analog signals to digital signals, ensuring the accuracy and reliability of subsequent processing;

[0053] 2. It increases the effective information ratio of the signal and reduces the possibility of error accumulation, thereby enhancing the noise resistance of the entire system;

[0054] 3. Accurately locate complex interference through multi-scale analysis, thereby improving the accuracy of interference detection;

[0055] 4. It achieves high-precision classification of interference types and can provide confidence evaluation of classification results through probability output, thereby improving the adaptability of the system in interference suppression;

[0056] 5. It can not only suppress different types of interference in real time, but also dynamically optimize the processing strategy to achieve better anti-interference effect;

[0057] 6. Achieve complete reconstruction of the signal to ensure that filtering processing does not lead to the loss of effective information, thereby improving the accuracy and integrity of remote sensing data. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort. Among them:

[0059] Figure 1 Schematic diagram of the flow of the signal anti-interference method for a satellite remote sensing application system shown in the present invention;

[0060] Figure 2 A comparison diagram of signal waveforms shown in the present invention;

[0061] Figure 3 This is a signal spectrum comparison diagram shown in the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0063] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without making any creative work should fall within the scope of protection of the present invention.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] An embodiment of the present invention proposes a signal anti-interference method for a satellite remote sensing application system. By performing multi-scale analysis on satellite remote sensing signals and combining interference classification and adaptive filtering algorithms using convolutional neural networks, it can accurately classify and dynamically suppress different types of interference signals, solving the problems of lack of adaptability, insufficient multi-source interference classification capabilities, and incomplete signal recovery in existing technologies.

[0066] According to an embodiment of the present invention, Figure 1 The flowchart shown in FIG. 1 is a signal anti-interference method for a satellite remote sensing application system, specifically comprising:

[0067] S1. Use a satellite antenna to receive electromagnetic signals from a remote sensing satellite, perform analog-to-digital conversion on the electromagnetic signals, and obtain digital signals. The following are some of the steps that need to be explained:

[0068] Using a satellite antenna to receive electromagnetic signals from a remote sensing satellite to obtain analog radio frequency signals;

[0069] Convert the analog RF signal into an intermediate frequency signal through a downconverter;

[0070] According to the Nyquist sampling theorem, the intermediate frequency signal is sampled at a sampling frequency not less than twice the highest frequency of the signal to obtain a discrete time signal;

[0071] The sampled analog signal value is converted into a digital value through a quantizer;

[0072] The quantized signal digital value is then encoded into binary data to obtain a digital signal.

[0073] As an example, the number of quantization bits converted by the quantizer depends on the resolution of the ADC (eg, 12 bits, 16 bits).

[0074] As an example, according to the Nyquist sampling theorem, the sampling frequency f is not less than twice the highest frequency of the signal. s For the intermediate frequency signal s IF (t) is sampled, and the mathematical expression formula of the discrete time signal s[n] is as follows:

[0075] s[n]=s IF (nT s )

[0076] in, is the sampling period, n=0,1,2,…,N-1, f s is the sampling frequency, and N is the total number of sampling points.

[0077] It should be noted that through the above processing, the received electromagnetic signal is converted into a digital form that can be used for digital signal processing, providing a basis for subsequent processing operations, that is, efficiently and accurately digitizing the analog signal transmitted by the satellite, ensuring that the received signal can be further analyzed and processed in the digital domain.

[0078] S2. Preprocess the digital signal to eliminate DC offset and high-frequency noise, thereby improving the signal-to-noise ratio. It should be noted that preprocessing the digital signal at least includes eliminating DC offset and high-frequency noise, wherein:

[0079] DC offset cancellation is achieved by calculating the mean of the digital signal. The mathematical formula for calculating the mean value (DC component) of the signal is as follows:

[0080]

[0081] The mathematical expression for eliminating DC offset is as follows:

[0082] x dc (n) = x(n) - μ

[0083] Signal x after removing DC offset dc (n) more stable;

[0084] High-frequency noise elimination is achieved by designing a low-pass filter to filter out the high-frequency noise. Specifically, by designing a low-pass filter: select a finite impulse response (FIR) or infinite impulse response (IIR) filter, set the cutoff frequency f c As the upper limit of the signal bandwidth, use the impulse response h(n) of the filter to calculate x dc (n) Perform convolution:

[0085]

[0086] Where M is the filter length, μ is the signal mean, and x dc (n) is the signal after eliminating DC offset, x p (n) is the preprocessed signal, and h(n) is the filter impulse response. c is the filter cutoff frequency.

[0087] It is not difficult to understand that DC offset and high-frequency noise will affect the baseline and spectral characteristics of the signal, resulting in error accumulation and signal distortion. This embodiment significantly improves the quality and stability of the signal by removing DC offset and high-frequency noise, and improves the signal-to-noise ratio of the signal, providing a clear signal basis for subsequent interference detection and classification.

[0088] S3. Perform interference detection on the pre-processed digital signal, that is, comprehensively analyze the energy distribution of each scale. When the energy of some scales is abnormally high, it is determined that there is interference in the digital signal. It should be noted that this step includes:

[0089] Wavelet transform:

[0090] Select mother wavelet function to perform discrete wavelet transform;

[0091] Perform multi-scale decomposition on the preprocessed signal to obtain detail coefficients and approximate coefficients at different scales; Feature extraction:

[0092] Calculate the energy of the detail coefficients:

[0093] E j =∑ n |d j (n)| 2

[0094] Anomaly detection: setting energy threshold T j , if E j >T j , then it is considered that there is interference on scale j; Interference existence judgment:

[0095] By integrating the energy distribution of each scale, when the energy of a certain scale is abnormally high, it is determined that there is interference in the digital signal;

[0096] Among them, n is the signal number, j is the scale level, d j (n) is the detail coefficient of the jth scale, E j is the energy of the jth scale, T j is the energy threshold of the j-th scale.

[0097] This embodiment introduces a multi-scale analysis method to deeply explore and detect the local characteristics of the signal, thereby identifying interference signals in different frequency bands or time periods. The analysis of energy distribution allows the system to observe the signal at different frequency scales, so that it can accurately locate and identify the occurrence of interference, thereby improving the detection sensitivity, so that even low-energy narrowband interference or random pulse interference can be detected.

[0098] S4. Combine the features extracted from the interference detection into a feature vector, input it into a classification model built based on a convolutional neural network to perform interference classification, output the probability of each interference type, and select the type with the highest probability as the judgment result. Among them, it is necessary to explain the following in this step:

[0099] The detail coefficient d extracted in step S3 j (n), energy E j , scale j features are combined into feature vector F;

[0100] Build a classification model based on convolutional neural network, including:

[0101] Input layer: receives feature vectors, the shape is represented as (N features ,1), where N features is the length of the eigenvector;

[0102] Convolutional layer: Use multiple convolution kernels to convolve the input feature vector to extract high-level features;

[0103] Activation function: Use ReLU function to introduce nonlinearity;

[0104] Pooling layer: Use maximum pooling to reduce feature dimensions and prevent overfitting;

[0105] Fully connected layer: maps the features extracted by the convolutional layer to the classification space;

[0106] Output layer: Use the Softmax activation function to output the probability distribution of each interference type.

[0107] As an example, the signal feature vector to be classified is input into the classification model for calculation, the probability of each interference type is output, and the type with the highest probability is selected as the judgment result. The classification model uses the cross entropy loss function as the objective function, and its mathematical expression formula is as follows:

[0108]

[0109] Where C is the number of interference types, y i is the actual label, p i is the probability predicted by the model, and i is the label number.

[0110] As an example, the interference type includes at least narrowband interference, broadband interference, and pulse interference.

[0111] As an example, an optimization algorithm (such as Adam or SGD) is used to minimize the loss function and adjust the network weights to optimize the training of the classification model. Validation is required during the training process to prevent overfitting.

[0112] It should be noted that convolutional neural networks have powerful feature extraction and pattern recognition capabilities. They classify different types of interference through input feature vectors, identify them efficiently and accurately based on the characteristics of the interference, and achieve accurate identification of a variety of complex interference signals.

[0113] S5. Based on the output judgment result, a corresponding adaptive filtering algorithm is used to suppress the interference signal in the digital signal to obtain a purified signal. Among them, it should be noted that:

[0114] If it is narrowband interference (such as frequency-specific sinusoidal interference), select the adaptive notch filter;

[0115] If it is broadband interference (such as noise with a wide bandwidth), an adaptive linear prediction filter is selected;

[0116] If it is pulse interference (such as short pulses that appear randomly), select a median filter or an adaptive threshold filter.

[0117] As an example, an adaptive notch filter (for narrowband interference) has the following structure:

[0118] Design a digital notch filter to filter out the interference frequency f0. The filter transfer function is:

[0119]

[0120] Adaptive adjustment: If the interference frequency f0 changes over time, an adaptive algorithm is used to estimate f0 in real time and adjust the filter parameters;

[0121] Adaptive linear prediction filter (for broadband interference), using the LMS algorithm:

[0122] Initialization: Set the filter order M and initialize the weights w(0) = [0, 0, ..., 0] T ;

[0123] Iterative update: For each time point, input vector:

[0124] x(n)=[x p (n),x p (n-1),…,x p (n-M+1)] T

[0125] Filter output:

[0126] y(n)=w T (n)x(n)

[0127] Error signal: e(n) = d(n) - y(n)

[0128] Where d(n) is the desired signal, which can be taken as x p (n) estimated value;

[0129] Weight update: is the step size factor, satisfying Adaptive median filter (for pulse interference), filtering process:

[0130] Set the window size W (e.g. W=5);

[0131] For each sample signal, take W samples in its neighborhood:

[0132] {x p (nk)∣k=-(W-1) / 2,…,(W-1) / 2}

[0133] Calculate the median of these samples as the filter output:

[0134] x s (n) = median{x p (nk)}

[0135] Get the signal after interference suppression: filter output x s (n) is the signal after interference suppression;

[0136] Where f0 is the interference frequency, H(z) is the filter transfer function, w(n) is the filter weight vector, x(n) is the input signal vector, y(n) is the filter output, and e(n) is the error signal. is the step size factor, x s (n) is the signal after interference suppression.

[0137] It should be noted that an adaptive filtering algorithm is used based on the classification results, and different suppression strategies are adopted for different types of interference. The adaptive filter can adjust parameters according to the real-time changes of signals and interference, thereby achieving effective suppression of interference. Compared with filters with fixed parameters, adaptive filtering has greater flexibility and adaptability, and can provide more accurate signal purification effects in complex interference environments.

[0138] S6. Use the mother wavelet function to perform an inverse transform on the purified signal, reconstruct the signal, and restore the original remote sensing information. It should be noted that the reconstruction formula for this step is as follows:

[0139] x r (n)=∑ j (a j (n)+d j (n))

[0140] Among them, d j (n) is the detail coefficient of the jth scale, a j (n) is the approximate coefficient of the jth scale, n is the signal number, j is the scale level, x r (n) is the reconstructed signal.

[0141] In an optional embodiment, in order to restore the time domain representation of the reconstructed signal and ensure that its effective information is completely preserved, signal correction is required, wherein:

[0142] If the adaptive filtering algorithm used introduces amplitude attenuation, a gain factor is used to correct it;

[0143] If the adaptive filtering algorithm used introduces phase delay, phase compensation technology is used to correct it.

[0144] Preferably, this step uses the mother wavelet function to perform an inverse transform on the purified signal, reconstructing the filtered signal into an approximate form of the original signal, and the inverse transform restores the filtered signal from the feature space to the time domain, thereby retaining more signal detail information.

[0145] S7. Save the reconstructed signal in the form of a digital file and transmit it to a ground station or data processing center for subsequent analysis and application.

[0146] As an example, the output form of the reconstructed signal at least includes a data file (including the reconstructed signal data) and a graphic file (displaying a comparison chart before and after signal processing).

[0147] For example, when the data is saved in digital file format, the file format includes at least MAT file and CSV file.

[0148] In an optional embodiment, the reconstructed signal is transmitted to a ground station or a data processing center for subsequent analysis and application, such as generating signal waveforms and spectrum diagrams, visually displaying the processing effects, or performing feature extraction and information analysis on the reconstructed signal.

[0149] It should be further explained that the traditional method cannot dynamically adjust the processing strategy according to the interference type, which may cause signal distortion or interference not to be effectively suppressed, and the filter with fixed parameters is difficult to cope with complex and changeable interference environments; in order to verify the effectiveness of the anti-interference method provided in this embodiment in anti-interference of satellite remote sensing signals, this embodiment compares the performance differences between this technical solution and the traditional signal processing method under the same conditions by comparison with the traditional method, and verifies the effect of the method of the present invention in anti-interference ability, signal recovery accuracy and signal-to-noise ratio improvement.

[0150]

Experimental subjects

[0151] The method of the present invention includes the complete processing flow of signal preprocessing, interference detection, interference classification, adaptive interference suppression and signal reconstruction as shown above;

[0152] Traditional method: uses fixed-parameter digital filters (such as low-pass filters or band-pass filters) for signal processing, and does not include interference detection and classification steps.

[0153]

Experimental conditions

[0154] Hardware environment:

[0155] Computer configuration: CPU: Intel Core i7-9700K;

[0156] Memory: 16GB RAM

[0157] Operating system: Windows 10 64-bit;

[0158] Software environment:

[0159] MATLAB version: MATLAB R2021b;

[0160] Toolboxes: Signal Processing Toolbox, Wavelet Toolbox, and Deep Learning Toolbox;

[0161] Dataset:

[0162] Simulated signal data: In order to control experimental variables, MATLAB was used to generate simulated satellite remote sensing signals and interference signals;

[0163] Signal parameters:

[0164] Sampling frequency f s :1MHz;

[0165] Signal duration T: 1 second;

[0166] Total number of sampling points N: N = f s ×T=1×10 6 ;

[0167] Signal composition: original signal x true (n);

[0168] Simulate satellite remote sensing signals using a 50kHz sine wave:

[0169] x true (n) = Asin(2πf0nT s )

[0170] Among them, A=1 is the signal amplitude, f0=50kHz is the signal frequency, T s =1 / f s =1μs is the sampling period, n=0,1,2,…,N-1,

[0171] Interference signal: narrowband interference x narrow (n), a sinusoidal interference signal with a frequency of 100kHz:

[0172] x narrow (n) = A n sin(2πf n nT s )

[0173] Among them, A n =0.5 is the interference signal amplitude, f n =100kHz is the interference signal frequency;

[0174] Pulse interference x pulse (n): short pulses appearing at random positions, with an amplitude of 1 and a duration of one sampling point, for a total of 1000 pulses;

[0175] Gaussian white noise noise(n): Gaussian white noise with a mean of 0 and a standard deviation of 0.1;

[0176] Received signal x(n): Superposition of original signal, interference signal and noise:

[0177] x(n)=x true (n)+x narrow (n)+x pulse (n)+noise(n);

[0178]

Experimental process

[0179] The process of the present invention is as follows:

[0180] Signal Generation:

[0181] Generate time vector: t = nT s ,n=0,1,2,…,N-1;

[0182] Generate the original signal: x true (n) = sin(2π×50×10 3 ×t);

[0183] Generate narrowband interference: x narrow (n) = 0.5sin(2π×100×10 3 ×t);

[0184] Generate pulse interference x pulse (n): Randomly select 1000 sampling points and assign them a value of 1, and the rest are 0;

[0185] Generate Gaussian white noise noise(n): noise(n)~N(0,σ 2 ),σ=0.1;

[0186] Generate received signal x(n): x(n) = x true (n)+x narrow (n)+x pulse (n)+noise(n); signal preprocessing:

[0187] Eliminate DC offset:

[0188] High-frequency noise filtering: Design a low-pass filter using a 6th-order Butterworth low-pass filter with a cutoff frequency of 150 kHz (normalized frequency of 0.3).

[0189] Filtering x p (n) = filter(b lp ,a lp ,x dc (n)), where b lp and a lp is the filter coefficient; interference detection:

[0190] Perform a 5-layer discrete wavelet transform (DWT):

[0191] Select Daubechies 4 ('db4') as the mother wavelet function and obtain the detail coefficient d j (n) and the approximate coefficient a j (n;

[0192] Feature extraction: Calculate the energy E of each layer’s detail coefficient j :E j =∑ n |(n)| 2 ;

[0193] Interference existence judgment: set threshold T j , if E j >T j , then it is considered that there is interference in this layer;

[0194] Interference classification:

[0195] Feature construction: The detail coefficient energy and scale j are used as the feature vector F;

[0196] Load a pre-trained classification model: The model has been trained on a signal dataset containing various interference types, including narrowband interference, pulse interference, and broadband noise;

[0197] Interference classification: Input the feature vector F to the classification model and output the interference type;

[0198] Interference Suppression:

[0199] Select the filter based on the interference type: if narrowband interference and pulse interference are detected, an adaptive notch filter and a median filter are used respectively;

[0200] Design an adaptive notch filter:

[0201] Notch frequency: f0 = 100kHz;

[0202] Normalized angular frequency: ω0=2πf0 / f s ;

[0203] Filter coefficient calculation: b = [1, -2cos(ω0), 1], a = [1];

[0204] Apply notch filter: x notch (n)=filter(b,a,x p (n));

[0205] Apply a median filter: A median filter with a window size of 5:

[0206] x s (n) = medfilt1(x notch (n),5);

[0207] Signal reconstruction:

[0208] Perform inverse wavelet transform (IDWT): x t (n) = waverec(c, l, 'db4'), where c and l are the wavelet coefficients and their corresponding lengths;

[0209] Signal correction: Since the filter may introduce amplitude and phase distortion, correction processing is performed.

[0210] Traditional method processing flow:

[0211] Design a fixed-parameter low-pass filter: a 6th-order Butterworth low-pass filter with a cutoff frequency of 80 kHz (normalized frequency of 0.16);

[0212] Filtering: x traditional (n) = filter(b t ,a t ,x(n)), where b t and a t is the filter coefficient.

[0213]

Experimental results

[0214] Calculate the mean squared error (MSE):

[0215]

[0216] The method of the present invention: MSE proposed =0.005;

[0217] Traditional method: MSE traditional =0.02;

[0218] Signal-to-noise ratio (SNR) calculation:

[0219]

[0220] The method of the present invention: SNR proposed =23dB;

[0221] Traditional method: SNR traditional =14dB;

[0222] Draw waveforms and spectrograms to compare the performance of the original signal, the signal containing interference, and the processed signal in the time domain and frequency domain.

[0223] Reference Figure 2 ,from Figure 2 It can be seen intuitively in the waveform of the received signal with interference that the signal contains obvious interference and noise, and the waveform is distorted; however, in the signal processed by the method of the present invention, the interference is effectively suppressed, the signal waveform is highly consistent with the original signal, and the signal is well recovered; however, in the signal processed by the traditional method, the interference is not completely eliminated, part of the interference still exists, the signal waveform is distorted, and the signal is still distorted.

[0224] Reference Figure 3 , it can be seen intuitively that Figure 3 The upper left corner is the spectrum of the original signal, which shows a clear peak at 50kHz, representing the main frequency of the original signal; Figure 3The upper right corner shows the spectrum of the received signal with interference. In addition to the main frequency of 50kHz, there is a clear interference peak at 100kHz. Due to the presence of pulse interference and noise, the spectrum noise floor is high. Figure 3 The lower left corner shows the signal spectrum after processing by the method of the present invention. The interference peak at 100kHz is effectively suppressed, the spectrum noise floor is reduced, and the main frequency of 50kHz remains intact. Figure 3 The lower right corner shows the signal spectrum after traditional processing. Although the interference is suppressed to a certain extent, the main frequency of 50kHz is also weakened, and the signal amplitude decreases significantly.

[0225] It should be noted that traditional methods cannot effectively suppress narrowband interference of specific frequencies and may weaken the effective components of the signal. They lack specificity and have poor processing effects, and cannot meet the anti-interference requirements in complex electromagnetic environments. The method of the present invention performs well in suppressing narrowband interference and pulse interference, and the signal quality is significantly improved. The method of the present invention can dynamically adjust the processing strategy according to the interference type. While suppressing interference, it retains the effective information of the signal as much as possible, thereby improving the signal-to-noise ratio and quality of the signal.

[0226] The optimization training of the aforementioned classification model and the signal correction method for reconstructing the signal can be performed using methods and means in the existing technology and will not be described in detail in this example.

[0227] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A signal anti-interference method for a satellite remote sensing application system, characterized in that: include: Using a satellite antenna to receive an electromagnetic signal from a remote sensing satellite, performing analog-to-digital conversion on the electromagnetic signal to obtain a digital signal; Preprocessing the digital signal to eliminate DC offset and high-frequency noise and improve the signal-to-noise ratio; Performing interference detection on the pre-processed digital signal, that is, comprehensively analyzing the energy distribution of each scale, and determining that interference exists in the digital signal when the energy of some scales is abnormally high; Performing the interference detection includes: Select mother wavelet function to perform discrete wavelet transform; Perform multi-scale decomposition on the preprocessed signal to obtain detail coefficients and approximation coefficients at different scales; Calculate the energy of the detail coefficients: E j =∑ n |d j (n)| 2 Set the energy threshold T j , if E j >T j , then it is considered that there is interference on scale j; Comprehensively analyzing the energy distribution of each scale, when the energy of a certain scale is abnormally high, it is determined that interference exists in the digital signal; Among them, n is the signal number, j is the scale level, d j (n) is the detail coefficient of the jth scale, E j is the energy of the jth scale, T j is the energy threshold of the j-th scale; The detail coefficient d extracted from the interference detection j (n), energy E j , scale j features are combined into a feature vector, which is input into the classification model built based on the convolutional neural network for interference classification, and the probability of each interference type is output. The type with the largest probability is selected as the judgment result; The signal feature vector to be classified is input into the classification model for calculation, and the probability of each interference type is output. The type with the largest probability is selected as the judgment result. The classification model uses the cross entropy loss function as the objective function, and its mathematical expression formula is as follows: Where C is the number of interference types, y i is the actual label, p i is the probability predicted by the model, i is the label number; Based on the output determination result, a corresponding adaptive filtering algorithm is used to suppress the interference signal present in the digital signal to obtain a purified signal; the interference type includes at least narrowband interference, broadband interference and pulse interference, wherein: If it is narrowband interference, select the adaptive notch filter; If it is broadband interference, then select the adaptive linear prediction filter; If it is pulse interference, select median filter or adaptive threshold filter; The mother wavelet function is used to perform an inverse transform on the purified signal to reconstruct the signal and restore the original remote sensing information. The reconstruction formula is as follows: x r (n)=∑ j (a j (n)+d j (n)) Among them, d j (n) is the detail coefficient of the jth scale, a j (n) is the approximate coefficient of the jth scale, n is the signal number, j is the scale level, x r (n) is the reconstructed signal; The reconstructed signal is saved in the form of a digital file and transmitted to a ground station or data processing center for subsequent analysis and application.

2. The signal anti-interference method for a satellite remote sensing application system according to claim 1, characterized in that: Obtaining the digital signal, comprising: Using a satellite antenna to receive electromagnetic signals from a remote sensing satellite to obtain analog radio frequency signals; Converting the analog radio frequency signal into an intermediate frequency signal through a down converter; According to the Nyquist sampling theorem, the intermediate frequency signal is sampled at a sampling frequency not less than twice the highest frequency of the signal to obtain a discrete time signal; The sampled analog signal value is converted into a digital value through a quantizer; The quantized signal digital value is then encoded into binary data to obtain the digital signal.

3. The signal anti-interference method for a satellite remote sensing application system according to claim 2, characterized in that: The number of quantization bits converted by the quantizer depends on the resolution of the ADC.

4. The signal anti-interference method for a satellite remote sensing application system according to claim 1 or 2, characterized in that: The preprocessing of the digital signal includes at least DC offset elimination and high-frequency noise elimination, wherein the DC offset elimination is achieved by calculating the mean of the digital signal to eliminate the DC offset, and the high-frequency noise elimination is achieved by designing a low-pass filter to perform filtering processing to eliminate the high-frequency noise.

5. The signal anti-interference method for a satellite remote sensing application system according to claim 1, characterized in that: The classification model is constructed based on the convolutional neural network, including: Input layer: receives feature vectors, the shape is represented as (N features ,1), where N features is the length of the eigenvector; Convolutional layer: Use multiple convolution kernels to convolve the input feature vector to extract high-level features; Activation function: Use ReLU function to introduce nonlinearity; Pooling layer: Use maximum pooling to reduce feature dimensions and prevent overfitting; Fully connected layer: maps the features extracted by the convolutional layer to the classification space; Output layer: Use the Softmax activation function to output the probability distribution of each interference type.

6. The signal anti-interference method for a satellite remote sensing application system according to claim 1, characterized in that: In order to restore the time domain representation of the reconstructed signal and ensure that its effective information is completely preserved, signal correction is required, wherein: If the adaptive filtering algorithm used introduces amplitude attenuation, a gain factor is used to correct it; If the adaptive filtering algorithm used introduces phase delay, phase compensation technology is used to correct it.

Images