Research On Key Technology Of Space-borne SAR Satellite Radiation Source Signal Processing

In the surveillance process of space-borne synthetic aperture radar(SAR)satellite,because of low peak power and strong anti-interference,linear frequency modulation(LFM)signal with large time-bandwidth is usually adopted to obtain high-resolution imaging of topography,surface features,large artificial construction and military objectives,which can effectively avoid interception of ground reconnaissance system.To protect our own ground targets from irradiation of SAR satellite emitter signal and interrupt reconnaissance of SAR satellites,ground acquisition systems must intercept,analyze and process SAR satellite emitter signal effectively,in addition,some false echoes are sometime sent back appropriately to deceive satellite reconnaissance.Since the SAR satellites are far away from the ground and electromagnetic environment is complex,the emitter signal passes through a series of medium such as the ionosphere,atmosphere,and clouds,before reaching the ground.In the process of signal propagation,the signal is affected by many disturbances and its strength severely attenuates,thus the intercepted signal in the ground is in much low signal-to-noise ratio(SNR).So,it has become an important research topic in the field of electronic surveillance that how to process and analyze the intercepted space-borne SAR satellite radiation source signal.In this dissertation,relying on the ‘Twelve-Five' National High-tech R&D Program(863 Program)‘Signal Reconnaissance and Processing Technology of Satellite Radiant Source Based on XXXX Analysis',a thorough and detailed research is carried out on the intercepted space-borne SAR satellite radiation source signal.The main contents of this dissertation are summarized as follows:In Chapter 2,short-time Fourier transform of LFM signal with the Gaussian window function is studied.The analysis shows that the short-time spectrum of LFM signal obtained from STFT distributes as a dorsal fin in the time-frequency domain.The approximate frequency range of the really intercepted broadband SAR radiation LFM signal can be determined easily by digital phosphor technology(DPX),but the frequency of noise with different SNR after STFT distributes randomly from 0 to half of the sampling rate.Combining the LFM signal's characteristics of STFT under Gaussian window and the short-time spectrum distribution information of SAR broadband LFM signal acquired through DPX,the algorithm of detection and parameter estimation for space-borne SAR broadband LFM signal is produced.In Chapter 3,to overcome the difficulty in detecting the space-borne SAR pulse LFM signal mixed with spurious clutter and interference under low SNR in time domain or frequency domain,a detection and parameter estimation algorithm based on the time-frequency image enhancement and Hough transform(HT)is proposed.The basic idea is that,1)the time-frequency distribution of space-borne SAR emitter signals is converted into an image,2)digital image processing technology is employed to research and analyze the time-frequency distribution image and 3)thus a new perspective is provided to deal with the space-borne SAR emitter signals.In Chapter 4,aiming at the problem on the adaptive detection and parameter estimation for LFM pulse signal,an algorithm of non-adjacent difference of one-dimensional short-time spectrum is presented to realize adaptive detection and parameter estimation for LFM pulse signal.First,under the Gaussian window with the minimum timewidth-bandwidth product,STFT is implemented on LFM signal to extract the maximum spectrum in each time window.The one-dimensional short-time spectrum is generated by making one-to-one correspondenc between the maximum spectrum and the center of its locating time window.Second,preprocess of median filter is adopted to smooth the one-dimensional short-time spectrum,by which good-performance short-time spectrum less affected by noise is obtained in the LFM pulse period.Next,non-adjacent difference is conducted on the smoothed one-dimensional short-time spectrum,on the basis of which the pseudo differential peaks are removed and the true differential peaks are retained.According to the true differential peaks,the representative point positions are searched out for the rising edge and the falling edge of one-dimensional short-time spectrum.Finally,in view of the relationship between the positions of the true differential peaks and the differential intervals,the start and end time of the LFM pulse signal are determined and further,the chirp line of LFM pulse is extracted and its parameters can be estimated also.In Chapter 5,on the basis of the theoretical foundation of the fractional Fourier transform,spectrum integral is implemented on LFM signal in the fractional Fourier domain(FRFD),which reveals an important feature that the spectrum integral has a minimum value at the optimal rotation angle.This significant discovery expands the characteristic of Fr FT for LFM signal processing.Besides,concerning the process problem of LFM signal with large number of sampling points in FRFD,segmented fractional Fourier transform is used to process LFM signal.In the segmented Fr FT process,the whole big-length LFM signal is split into many small-length LFM signal segments with the same number of sampling points firstly.Secondly,Fr FT is implemented for each small LFM signal respectively,and each two-dimensional spectrum of Fr FT is taken as a frame of image.The sum of the cumulative average of all frames of images in vertical direction is equivalent to the average of the linear superposition of spectrum integral of each small LFM signal.This method enhances the characteristic of spectrum integral effectively,while suppressing the influence of noise.Through extracting the enhanced feature of spectrum integration,parameters of LFM signal with large number of sampling points,such as chirp rate and initial frequency can be estimated,and the time-consuming calculation and the peak search time can be reduced effectively also.In Chapter 6,to solve the problem in estimating the signal-to-noise ratio(SNR)of LFM signal with unknown parameters,an SNR estimation algorithm is presented in this paper,which is based on the Fourier series' fitting of the auto-correlation's main lobe in time domain.First,the LFM signal's auto-correlation is analyzed to deduce the relationship between the SNR and the auto-correlation in the background of the additive white Gaussian noise(AWGN).Second,the main lobe of the LFM signal's auto-correlation is selected by adopting the histogram diversity method.Finally,the Fourier series is used to fit out the main lobe of the auto-correlation and separates the auto-correlation of the LFM signal from that of the noise at the center,through which the SNR can be calculated.Simulations show that,the larger the signal's pulse width is and the narrower the signal's bandwidth is,the more accurate the SNR estimation is in the presented algorithm,through which the SNR estimation outperforms those by adopting the central neighborhood binary average method and the constant-envelop method and the root mean square error(RMSE)is not more than 0.3d B when the SNR varies from-18 d B to 10 d B.When the SNR is from-10 d B to 5d B,SNR estimation reaches the optimal performance.