A post-facto analysis system for space-borne SAR signals

Abstract

The application relates to a spaceborne SAR signal post-facto fine analysis system and belongs to the field of radar signal reconnaissance; can complete time-frequency analysis and PDW parameter measurement functions; can perform pulse rapid retrieval based on a time scale, realizes fine analysis on a single pulse signal, and includes a fuzzy diagram, a spectral peak degree coefficient and the like; has a radiation source signal sorting function, can manually add a radiation source database, can automatically store a radiation source result obtained through detection and collection into a warehouse, and can sort known radiation sources according to the warehouse; GPU is adopted to accelerate processing of an analysis algorithm, time-frequency analysis speed can reach 1.6GS / s, and PDW measurement speed can reach 80MS / s; the whole processing system is completed in a 2U server disk array all-in-one machine, the system has high integration degree and full functions.

Description

Technical Field

[0001] This invention relates to a post-event fine analysis system for spaceborne SAR signals, belonging to the field of radar signal reconnaissance. Background Technology

[0002] Compared to optical imaging, spaceborne synthetic aperture radar (SAR) can perform high-resolution imaging of targets in all weather conditions and at all times, which is why countries around the world attach great importance to it and have carried out in-depth development. Spaceborne SAR can perform imaging at different resolutions for different detection areas by switching between different imaging modes. Common examples include strip mode, which can image an area on a single sub-strip, spotting mode, which achieves high azimuth resolution imaging of a fixed area on a sub-strip by sacrificing imaging area, and scanning mode, which images multiple sub-strips. The multi-functional and high-precision imaging of spaceborne SAR has posed a significant threat to modern electronic warfare, making electronic countermeasures against it crucial. The primary task of electronic countermeasures is electronic reconnaissance. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a post-hoc fine analysis system for spaceborne SAR signals to solve the problem of large measurement errors in existing electronic reconnaissance receivers when reconnoitering low signal-to-noise ratio SAR signals.

[0004] A post-processing fine analysis system for spaceborne SAR signals includes a radar parameter extraction module, a radiation source signal sorting module, and a single pulse analysis module;

[0005] The radar parameter extraction module includes a time-frequency analysis submodule and a PDW measurement submodule. The time-frequency analysis submodule is used to obtain a time-frequency image within a specified time period. The PDW measurement submodule is used to extract the backbone of the time-frequency image, separate the time-frequency characteristics of the real pulse signal, and thus obtain the PDW measurement result of the signal, specifically including:

[0006] For a linear frequency modulated signal, its frequency f(t) changes linearly with time t, expressed as:

[0007] f(t) = f0 + kt

[0008] The phase can then be expressed as:

[0009]

[0010] The final time-domain expression of the linear frequency modulated signal is:

[0011]

[0012] Where A is the signal amplitude, T is the pulse width, rect is the rectangular window function, f0 is the fixed frequency, Φ0 is the initial phase, and k is the frequency modulation slope, representing the rate of linear frequency change.

[0013] Short-time Fourier transform (SFT) is used to perform synchronous analysis of the signal in the time-frequency domain, resulting in a two-dimensional time-frequency image (STFT). x (t,Ω);

[0014] For discrete signals of linear frequency modulation signals, the short-time Fourier transform is equivalent to first dividing the signal into segments, then windowing each segment, and then taking the Fourier transform of each segment.

[0015] For a fixed moment t0, the frequency with the largest amplitude in the spectrum is taken as the instantaneous frequency of the signal:

[0016] f x (t)=max{STFT x (t, Ω)}

[0017] According to the set threshold Th adapt Accurately extract the real signal from the time-frequency image while filtering out noise; perform two-dimensional detection on the image for signals that exceed the threshold, and accurately obtain the time-domain start and end edges and frequency-domain bandwidth in the PDW measurement results of the linear frequency modulated signal;

[0018] The radiation source signal sorting module includes two processes: sorting known radiation sources and sorting unknown radiation sources.

[0019] Pre-sorting clusters the PDW data, classifying it according to radiation source category and removing erroneous PDWs. Main sorting performs secondary analysis on the data in each category based on PRT parameters, using the cumulative difference histogram method to statistically determine the true value of PRT and obtain the inter-pulse modulation type and modulation parameters of the radiation source. Finally, the radiation source sorting results are matched with a known radiation source database. If a match is found, the current radiation source is considered a known radiation source; otherwise, it is considered an unknown radiation source. Before initiating a radiation source signal, the known radiation source database is added as prior information for radiation source sorting. If a signal is sorted as an unknown radiation source, it is added to the database for management.

[0020] The single-pulse analysis module locates the pulse under the set time scale according to the data capacity and performs single-pulse fine analysis, including calculating the fuzzy map and spectral kurtosis coefficient of the signal through pulse compression technology, and obtaining the time-frequency characteristics of the signal by performing time-frequency analysis on the single pulse, extracting the real signal pulse from the noise, and thus calculating its SNR, intra-pulse fluctuation, number of splits, and intra-pulse modulation parameters.

[0021] Preferably, after reading the raw data, the radar parameter extraction module reads the data of the set working path into the GPU cache, and performs a sliding window FFT traversal on the raw data in the cache based on the short-time Fourier transform; wherein, the GPU cache is divided into two parts, ping-pong and ping-pong, and when processing the ping-pong cache, the next batch of data is read into the ping-pong cache; the time-frequency image data generated during the synchronous processing is placed in a separately allocated GPU cache and sampled and displayed on the interface.

[0022] Preferably, the specific method by which the single-pulse analysis module calculates the fuzzy map and spectral kurtosis coefficient of the signal using pulse compression technology includes:

[0023] Pulse compression of a signal is equivalent to matched filtering. If the input signal is s(t), the matched filter can be regarded as a correlator that calculates the autocorrelation function of the input signal. It reaches the maximum output signal-to-noise ratio at time t0, as shown in the following expression, where s0(t) is the output of the matched filter.

[0024]

[0025] Let t0-τ=x, then we have

[0026]

[0027] The ambiguity function is denoted by χ(τ,ξ), where τ represents the difference in delay time between the echo signals of two targets and the transmitted signals, and ξ represents the difference in Doppler frequencies between the two targets. The faster |χ(τ,ξ)| decreases with increasing τ and ξ, the stronger the radar's resolution and the lower the ambiguity. The expression for χ(τ,ξ) is as follows:

[0028]

[0029] It is the time-frequency composite autocorrelation function of the complex envelope of the two target echo signals;

[0030] The spectral kurtosis coefficient of a linear frequency modulated signal includes both the amplitude spectrum and the phase spectrum; for the time-domain signal s(t), performing a Fourier transform yields:

[0031]

[0032] Its amplitude spectrum is as follows:

[0033]

[0034] Its phase spectrum is:

[0035]

[0036] Where C(x) and S(x) denote Fresnel integrals, when the wide product is sufficiently large.

[0037]

[0038] Preferably, the short-time Fourier analysis is performed in parallel within the GPU.

[0039] Preferably, the post-analysis system is implemented using a 2U server disk array all-in-one machine, and the hardware includes a high-speed disk array and a GPU, the former for writing and reading data, and the latter for algorithm implementation.

[0040] Ideally, the time-frequency graph is compressed in the frequency domain. Specifically, the data with fftdot=2048 in the original frequency domain is extracted to extract the maximum value and compressed into 128 channels, while remaining unchanged in the time domain. The maximum amplitude among 2048 / 128=16 frequency points in each channel is taken as the amplitude of the channel, while retaining the frequency corresponding to the maximum amplitude in the channel.

[0041] Preferably, the set threshold Th adapt for:

[0042]

[0043] Where k>1 is the threshold coefficient, δ is the noise amplitude estimate, and δ is the threshold bias.

[0044] The present invention has the following beneficial effects:

[0045] 1) This invention has GPU acceleration, which can significantly improve the analysis speed of large amounts of raw data by 20-30 times compared with CPU processing.

[0046] 2) This invention uses a high-speed RAID5 disk array, which greatly speeds up the access speed of raw data, far exceeding the data access speed of non-disk array systems.

[0047] 3) This invention can analyze and display the working parameters of spaceborne SAR from multiple dimensions, and can intelligently analyze the working mode and the detection area.

[0048] 4) This invention has an indexing function, which can quickly locate the original data according to the needs and realize the precise analysis of the signal.

[0049] 5) As a post-hoc detailed analysis system, this invention has higher accuracy and more dimensions than real-time analysis. Attached Figure Description

[0050] Figure 1 This is a functional composition diagram of the spaceborne SAR signal post-processing fine analysis system of the present invention;

[0051] Figure 2This is a flowchart of the radar parameter extraction module of the present invention;

[0052] Figure 3 This is a flowchart of the radiation source signal sorting module of the present invention;

[0053] Figure 4 This is a flowchart of the single-pulse analysis module of the present invention. Detailed Implementation

[0054] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0055] This invention provides a post-event fine analysis system for spaceborne SAR signals, such as... Figure 1 As shown, it includes three modules: radar parameter extraction module, radiation source signal sorting module, and single pulse analysis module.

[0056] The operating parameter settings include time-frequency parameters, coarse detection parameters, PDW parameters, and signal sorting parameters. The time-frequency parameters include the number of points per FFT (default 1024), window type (default Gaussian window), and number of GPU cards (default 1). The coarse detection parameters include the number of channels (default 32) and the step size for finding the maximum / minimum value (default 32). The PDW parameters include the pulse width threshold (default 0), repetition period threshold (default 0), bandwidth threshold (default 0), and sliding window length (default 64). The signal sorting parameters include the sorting frame length (default 50ms), pre-sorting clustering method (including carrier frequency, pulse width, and intra-pulse modulation type), and histogram quantization interval (default 1µs). This invention includes a display and control module, allowing the setting or display of the above parameters within the corresponding modules.

[0057] like Figure 2 As shown, the radar parameter extraction module consists of three parts: time-frequency analysis, PDW measurement, and single-pulse analysis.

[0058] The time-frequency analysis process includes data reading and short-time Fourier transform (SFT) analysis. The system automatically reads data from the defined working path into the GPU cache and performs a sliding-window FFT traversal on the raw data in the cache based on the SFT. The GPU cache is divided into two parts, ping-pong and pong. While processing the ping-pong cache, the system uses multiple threads to read the next batch of data into the pong cache, ensuring that the system can process the pong cache immediately after processing the ping-pong cache. Time-frequency image data generated during synchronous processing is placed in a separate, large GPU cache, and the system samples and displays it on the interface. The time-frequency analysis operation ends after the data traversal is complete.

[0059] The PDW measurement process consists of two parts: coarse detection and signal measurement. After reading in the raw data, coarse detection involves using a time-frequency analysis module to detect within a range of the raw data, determining which data points might contain signals, and tagging them to accelerate signal measurement. Once the raw samples with data are known, the system first initializes the CPU memory based on the coarse detection results, then extracts the signal-containing portions from the disk and copies the data from memory to the video memory. A sliding windowed FFT is performed on the raw signal, and the modulus is calculated. The threshold of the signal in each adaptive segment is determined by the maximum and minimum values ​​of the FFT results for each segment, thereby extracting the true signal mixed with noise. The true signal is then classified and its parameters are coarsely measured. Finally, based on the signal category, fine frequency measurement is performed on either the point-frequency signal or the frequency-coded signal word frequency.

[0060] like Figure 3 As shown, the radiation source signal sorting module includes two main functions: known radiation source sorting and unknown radiation source sorting. The signal sorting process includes pre-sorting and main sorting. Pre-sorting clusters the PDW data, classifying it according to radiation source categories and removing erroneous PDWs. Main sorting performs secondary analysis on the data of each category based on PRT parameters, using the Cumulative Difference Histogram (CDIF) method to statistically analyze the true value of PRT and obtain the inter-pulse modulation type and modulation parameters of the radiation source. Finally, the radiation source sorting results are matched with the known radiation source library. If a match is found, the current radiation source is considered a known radiation source; otherwise, it is considered an unknown radiation source. Before starting the radiation source signal, the known radiation source library can be manually added as prior information for radiation source sorting. If the signal is sorted as an unknown radiation source, it can also be added to the library for management.

[0061] like Figure 4 As shown, single-pulse analysis includes data localization and data analysis workflows. After selecting the capacity or time, since the sampling rate of the raw data is fixed, the capacity can accurately correspond to the time, and data localization can be completed based on the time scale. Then, the analysis capacity can be selected, and the system will read the raw data under the set analysis capacity into the buffer. The time-frequency analysis module will then be used to complete spectrum analysis, time-frequency analysis, and measurements of parameters such as fuzzy maps, spectral kurtosis coefficients, SNR, intrapulse fluctuations, number of splits, and intrapulse modulation parameters.

[0062] This invention utilizes GPU acceleration, enabling rapid and accurate multi-dimensional analysis of raw data at speeds of at least 80 MS / s. Currently, this spaceborne SAR signal post-processing fine analysis system has been successfully applied to SAR signal reconnaissance.

[0063] Example:

[0064] 1. Radar parameter extraction module

[0065] The radar parameter extraction module can quickly perform parallel time-frequency domain analysis and calculation on massive amounts of raw data in blocks, obtaining a time-frequency image within a specified time period. The horizontal axis represents time, the vertical axis represents frequency, and the amplitude represents the intensity of the signal's frequency components. The module then extracts the backbone of the time-frequency image, separating the time-frequency characteristics of the real pulse signal, thereby obtaining the signal's PDW measurement results, including pulse arrival time, pulse width, pulse repetition period, signal center frequency, and signal bandwidth. GPU parallel computing significantly accelerates the time-frequency analysis. The module can display the time-frequency image of the raw data and the PDW measurement results on the display and control module; the specific process includes:

[0066] This radar parameter module is primarily designed for linear frequency modulated (LFM) signals, whose frequency changes linearly with time, and can be expressed as:

[0067] f(t) = f0 + kt

[0068] The phase can then be expressed as:

[0069]

[0070] The final time-domain expression of the linear frequency modulated signal is:

[0071]

[0072] Where A is the signal amplitude, T is the pulse width, rect is the rectangular window function, f0 is the fixed frequency, Φ0 is the initial phase, and k is the frequency modulation slope, representing the rate of linear frequency change.

[0073] Furthermore, the Short-Time Fourier Transform (STFT) is used for synchronous analysis of the signal in the time and frequency domain, obtaining a two-dimensional time-frequency image of the signal. The basic principle is to window the original signal and then perform a Fourier Transform to obtain the STFT. x (t,Ω)

[0074]

[0075] For continuous signals, the principle of STFT is to apply a window function g(τ) to the signal x(τ) at each time t, and then calculate the Fourier transform. Due to the presence of the window function, the interval for calculating the Fourier transform of the original signal is no longer the entire time period, but a small time period centered at time t—that is, the signal within a "short time". Therefore, the spectrum obtained by this method can represent the instantaneous frequency of the signal at time t.

[0076] For discrete signals of linear frequency modulation (LFM), the window length is finite. In this case, the short-time Fourier transform is equivalent to first dividing the signal into segments, then windowing each segment, and then taking the Fourier transform for each segment.

[0077] For a fixed moment t0, the frequency with the largest amplitude in the spectrum is taken as the instantaneous frequency of the signal.

[0078]

[0079] To reduce computational burden and improve system analysis speed, the time-frequency graph is compressed in the frequency domain, thus reducing computational load. For example, the original data with fftdot = 2048 (number of fft points) in the frequency domain is extracted by maximum value extraction and compressed into 128 channels, while remaining unchanged in the time domain. For each channel, the maximum amplitude among 2048 / 128 = 16 frequency points is taken as the channel amplitude, while retaining the frequency corresponding to the maximum amplitude within the channel. Channelization reduces subsequent computational load while preserving frequency information.

[0080] To accurately extract the true signal from the time-frequency image, an adaptive thresholding method is used to filter out noise. The noise is estimated to be... Then the threshold Th adapt Set to:

[0081]

[0082] Where k>1 is the threshold coefficient, Here, δ represents the noise amplitude estimate, and δ is the threshold bias. The purpose of the threshold bias is to artificially add a value independent of noise but relevant to the system to the noise floor, preventing abnormal over-threshold conditions such as simulation testing or out-of-band receiver noise and signal amplitude. Since the exact locations of signal and noise cannot be determined for sampling and noise parameter calculation during actual measurements, it is essential to ensure that the entire noise portion is sampled for mean calculation, rather than the signal portion; otherwise, the threshold value would be too high. Noise estimation is determined using a "segmented, smaller value" approach.

[0083] By performing two-dimensional detection on the image of the signal that has exceeded the threshold, the time-domain start and end edges and frequency-domain bandwidth of the PDW measurement results of the linear frequency modulated signal can be accurately obtained.

[0084] 2. Radiation source signal sorting module

[0085] The measurement results from the PDW measurement module are statistically analyzed in units of sorting frame length to obtain statistical parameters of the radiation sources, including inter-pulse time-domain modulation parameters (such as inter-pulse modulation type, including fixed repetition rate, staggered repetition rate, and slip repetition rate; fixed repetition rate period value; staggered repetition rate period value; slip repetition rate maximum repetition period and maximum repetition period, etc.) and inter-pulse carrier frequency modulation parameters (such as carrier frequency modulation type, including spot frequency, step frequency, and agile frequency, etc.; frequency step size, etc.). Radiation source libraries can be manually added, and the identification results of current radiation sources can be managed by adding them to the library. After the system has a known radiation source library, it can be distributed to subsystems for known radiation source sorting. During the measurement process, the module displays the real-time radiation source information on the display and control interface. After all measurements are completed, the results are stored in the current working directory as a CSV file.

[0086] The signal sorting section employs the Cumulative Difference Histogram (CDIF) method. A detection threshold is set. First, the first-level histogram of the input pulse descriptor word (PRT) is calculated. If a PRT exceeds the threshold, its TOA (Transmission of Aspects) is used as the pulse retrieval threshold for that PRT and then removed. The search continues for other PRT values. If no PRTs exceeding the threshold are found in the first-level histogram, the second-level histogram is calculated, and the above steps are repeated until the true value of the PRT is obtained.

[0087] The system describes radar radiation sources using three-dimensional features and manages them in a MySQL database. This includes basic information (radar name, radiation intensity, radar type, etc.), frequency sets (spot frequency, stepped frequency, frequency diversity), and waveform sets (fixed repetition frequency, staggered repetition frequency, jittered repetition frequency). A single frequency set can correspond to several waveform sets. After signal sorting, the data is compared with the radiation source database information, and the database is updated accordingly.

[0088] 3. Single Pulse Analysis Module

[0089] When the system sampling rate remains constant, the storage rate of the original data is fixed. Therefore, based on the data capacity, the pulse under the set time scale can be quickly located and a single pulse fine analysis can be performed. The signal-to-noise ratio of the pulse signal can be improved by pulse compression technology, and the fuzzy map, spectral kurtosis coefficient, etc. of the signal can be calculated. Time-frequency analysis of the single pulse can obtain the time-frequency characteristics of the signal, extract the real signal pulse from the noise, and thus calculate its SNR, intrapulse fluctuation, number of splits, intrapulse modulation parameters, etc.

[0090] Pulse compression of a signal is equivalent to matched filtering. If the input signal is s(t), the matched filter can be regarded as a correlator that calculates the autocorrelation function of the input signal. It reaches the maximum output signal-to-noise ratio at time t0, as shown in the following expression, where s0(t) is the output of the matched filter.

[0091]

[0092] Let t0-τ=x, then we have

[0093]

[0094] The ambiguity function is denoted by χ(τ,ξ), where τ represents the difference in delay between the echo signals of two targets and the transmitted signals, and ξ represents the difference in Doppler frequencies between the two targets. The faster |χ(τ,ξ)| decreases with increasing τ and ξ, the stronger the radar's resolution and the lower the ambiguity. The expression for χ(τ,ξ) is as follows:

[0095]

[0096] It is the time-frequency composite autocorrelation function of the complex envelope of the two target echo signals.

[0097] The spectral kurtosis coefficients of a linear frequency modulated (LFM) signal include both the amplitude and phase spectra. Taking a Fourier transform of the time-domain signal s(t), we obtain:

[0098]

[0099] Its amplitude spectrum is

[0100]

[0101] Its phase spectrum is

[0102]

[0103] Where C(x) and S(x) denote Fresnel integrals, when the wide product is sufficiently large.

[0104]

[0105]

[0106] The basic workflow of a post-processing fine analysis system for spaceborne SAR signals according to the present invention is as follows: The operator selects time-frequency analysis to perform coarse analysis on the raw data of a single task, achieving coarse signal detection; then, PDW measurement is selected to measure the parameters of the data after time-frequency analysis; finally, the signal sorting function is selected to obtain the working parameters of the spaceborne SAR. Single-pulse analysis is independent of the above process and can independently achieve fine analysis of signals under a specified time and capacity of large amounts of raw data. The operator needs to configure the current capacity or time, and the system will process it in real time to obtain the spectrum, time-frequency image, and other fine analysis results of the current signal.

[0107] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A post-event fine analysis system for spaceborne SAR signals, characterized in that, It includes a radar parameter extraction module, a radiation source signal sorting module, and a single pulse analysis module; The radar parameter extraction module includes a time-frequency analysis submodule and a PDW measurement submodule; the time-frequency analysis submodule is used to obtain time-frequency images within a specified time period; The PDW measurement submodule is used to extract the backbone of the time-frequency image, separate the time-frequency features of the real pulse signal, and thus obtain the PDW measurement result of the signal, specifically including: For a linear frequency modulated signal, its frequency Over time Linear change, expressed as: The phase can then be expressed as: The final time-domain expression of the linear frequency modulated signal is: in, A For signal amplitude, T The pulse width. rect For rectangular window functions, For a fixed frequency, This is the initial phase; k The frequency modulation slope represents the rate of linear change of frequency. Short-time Fourier transform is used to perform synchronous analysis of the signal in the time and frequency domain, resulting in a two-dimensional time-frequency image of the signal. ; For discrete signals of linear frequency modulated signals, the short-time Fourier transform is equivalent to first dividing the signal into segments, then windowing each segment, and then taking the Fourier transform of each segment. For a fixed moment t 0, take the frequency with the largest amplitude in the spectrum as the instantaneous frequency of the signal: According to the set threshold Accurately extract the real signal from the time-frequency image while filtering out noise; perform two-dimensional detection on the image for signals that exceed the threshold, and accurately obtain the time-domain start and end edges and frequency-domain bandwidth in the PDW measurement results of the linear frequency modulated signal; The radiation source signal sorting module includes two processes: sorting known radiation sources and sorting unknown radiation sources. Pre-sorting clusters the PDW data, classifying it according to radiation source category and removing erroneous PDWs. Main sorting performs secondary analysis on the data in each category based on PRT parameters, using the cumulative difference histogram method to statistically determine the true value of PRT and obtain the inter-pulse modulation type and modulation parameters of the radiation source. Finally, the radiation source sorting results are matched with a known radiation source database. If a match is found, the current radiation source is considered a known radiation source; otherwise, it is considered an unknown radiation source. Before initiating a radiation source signal, the known radiation source database is added as prior information for radiation source sorting. If a signal is sorted as an unknown radiation source, it is added to the database for management. The single-pulse analysis module locates the pulse under the set time scale according to the data capacity and performs single-pulse fine analysis, including calculating the fuzzy map and spectral kurtosis coefficient of the signal through pulse compression technology, and obtaining the time-frequency characteristics of the signal by performing time-frequency analysis on the single pulse, extracting the real signal pulse from the noise, and thus calculating its SNR, intra-pulse fluctuation, number of splits, and intra-pulse modulation parameters.

2. The post-event fine analysis system for spaceborne SAR signals as described in claim 1, characterized in that, After reading the raw data, the radar parameter extraction module reads the data of the set working path into the GPU cache and performs a sliding window FFT traversal on the raw data in the cache based on the short-time Fourier transform. The GPU cache is divided into two parts, ping-pong and ping-pong. When processing the ping-pong cache, the next batch of data is read into the ping-pong cache. The time-frequency image data generated during the synchronous processing is placed in a separately allocated GPU cache and sampled and displayed on the interface.

3. The post-event fine analysis system for spaceborne SAR signals as described in claim 1, characterized in that, The specific methods used by the single-pulse analysis module to calculate the fuzzy map and spectral kurtosis coefficient of the signal through pulse compression technology include: Pulse compression of the signal is equivalent to matched filtering. If the input signal is... A matched filter can be viewed as a correlator that calculates the autocorrelation function of the input signal. t The maximum output signal-to-noise ratio is reached at time 0, as shown in the following expression. The output of the matched filter; make Then there is For fuzzy functions, use It means that among them This represents the difference in time between the echo signals of two targets and the transmitted signals. This represents the difference in Doppler frequencies between two targets; when along with and The faster the increase and decrease, the stronger the radar's resolution and the lower the ambiguity. The expression is as follows: It is the time-frequency composite autocorrelation function of the complex envelope of the two target echo signals; The spectral kurtosis coefficient of a linear frequency modulated signal includes both the amplitude spectrum and the phase spectrum; for time-domain signals... s(t) Performing a Fourier transform on it yields: Its amplitude spectrum is as follows: Its phase spectrum is: in, and This represents the Fresnel integral, where the width product is sufficiently large. 。 4. The post-event fine analysis system for spaceborne SAR signals as described in claim 2, characterized in that, The short-time Fourier analysis is performed in parallel on the GPU.

5. The post-event fine analysis system for spaceborne SAR signals as described in claim 1, characterized in that, The post-analysis system is implemented using a 2U server disk array all-in-one machine. The hardware includes a high-speed disk array and a GPU. The former is used for data writing and reading, and the latter is used for algorithm implementation.

6. The post-event fine analysis system for spaceborne SAR signals as described in claim 1, characterized in that, The time-frequency graph is compressed in the frequency domain. Specifically, the data with fftdot=2048 in the original frequency domain is extracted to extract the maximum value and compressed into 128 channels, while remaining unchanged in the time domain. The maximum amplitude among 2048 / 128=16 frequency points in each channel is taken as the amplitude of the channel, while the frequency corresponding to the maximum amplitude in the channel is retained.

7. The post-event fine analysis system for spaceborne SAR signals as described in claim 1, characterized in that, The set threshold for: in, For threshold coefficients, This is an estimate of the noise amplitude. This is the threshold bias.

Images