stripmap synthetic aperture radar (sar) image formation -凯发k8网页登录

synthetic aperture radar imaging

sar generates a two-dimensional (2-d) image. the direction of flight is referred to as the cross-range or azimuth direction. the direction of the antenna boresight (broadside) is orthogonal to the flight path and is referred to as the cross-track or range direction. these two directions provide the basis for the dimensions required to generate an image obtained from the area within the antenna beamwidth through the duration of the data collection window. the cross-track direction is the direction in which pulses are transmitted. this direction provides the slant range to the targets along the flight path. the energy received after reflection off the targets for each pulse must then be processed (for range measurement and resolution). the cross-range or azimuth direction is the direction of the flight path and it is meaningful to process the ensemble of the pulses received over the entire flight path in this direction to achieve the required measurement and resolution. correct focusing in both the directions implies a successful generation of image in the range and cross-range directions. it is a requirement for the antenna beamwidth to be wide enough so that the target is illuminated for a long duration by the beam as the platform moves along its trajectory. this helps provide more phase information. the key terms frequently encountered when working with sar are:

radar configuration

consider a sar radar operating in c-band with a 4 ghz carrier frequency and a signal bandwidth of 50 mhz. this bandwidth yields a range resolution of 3 meters. the radar system collects data orthogonal to the direction of motion of the platform as shown in this figure. the received signal is a delayed replica of the transmitted signal. the delay corresponds in general to the slant range between the target and the platform. for a sar system, the slant range varies over time as the platform traverses a path orthogonal to the direction of antenna beam. this section focuses on defining the parameters for the transmission waveform. the lfm sweep bandwidth can be decided based on the desired range resolution.

set the physical constant for speed of light.

c = physconst('lightspeed');

set the sar center frequency.

fc = 4e9;

set the desired range and cross-range resolution to 3 meters.

rangeresolution = 3;  
crossrangeresolution = 3;

the signal bandwidth is a parameter derived from the desired range resolution.

bw = c/(2*rangeresolution);

in a sar system the prf has dual implications. the prf not only determines the maximum unambiguous range but also serves as the sampling frequency in the cross-range direction. if the prf is too low to achieve a higher unambiguous range, there is a longer duration pulses resulting in fewer pulses in a particular region. at the same time, if the prf is too high, the cross-range sampling is achieved but at the cost of reduced range. therefore, the prf should be less than twice the doppler frequency and should also satisfy the criteria for maximum unambiguous range

prf = 1000; 
aperture = 4;  
tpd = 3*10^-6; 
fs = 120*10^6;

configure the lfm signal of the radar.

waveform = phased.linearfmwaveform('samplerate',fs, 'pulsewidth', tpd, 'prf', prf,...
    'sweepbandwidth', bw);

assume the speed of aircraft is 100 m/s with a flight duration of 4 seconds.

speed = 100;  
flightduration = 4;
radarplatform  = phased.platform('initialposition', [0;-200;500], 'velocity', [0; speed; 0]);
slowtime = 1/prf;
numpulses = flightduration/slowtime  1;
maxrange = 2500;
truncrangesamples = ceil((2*maxrange/c)*fs);
fasttime = (0:1/fs:(truncrangesamples-1)/fs);
% set the reference range for the cross-range processing.
rc = 1000;

configure the sar transmitter and receiver. the antenna looks in the broadside direction orthogonal to the flight direction.

antenna = phased.cosineantennaelement('frequencyrange', [1e9 6e9]);
antennagain = aperture2gain(aperture,c/fc); 
transmitter = phased.transmitter('peakpower', 50e3, 'gain', antennagain);
radiator = phased.radiator('sensor', antenna,'operatingfrequency', fc, 'propagationspeed', c);
collector = phased.collector('sensor', antenna, 'propagationspeed', c,'operatingfrequency', fc);
receiver = phased.receiverpreamp('samplerate', fs, 'noisefigure', 30);

configure the propagation channel.

channel = phased.freespace('propagationspeed', c, 'operatingfrequency', fc,'samplerate', fs,...
    'twowaypropagation', true);

scene configuration

in this example, three static point targets are configured at locations specified below. all targets have a mean rcs value of 1 meter-squared.

targetpos= [800,0,0;1000,0,0; 1300,0,0]'; 
targetvel = [0,0,0;0,0,0; 0,0,0]';
target = phased.radartarget('operatingfrequency', fc, 'meanrcs', [1,1,1]);
pointtargets = phased.platform('initialposition', targetpos,'velocity',targetvel);
% the figure below describes the ground truth based on the target
% locations.
figure(1);h = axes;plot(targetpos(2,1),targetpos(1,1),'*g');hold all;plot(targetpos(2,2),targetpos(1,2),'*r');hold all;plot(targetpos(2,3),targetpos(1,3),'*b');hold off;
set(h,'ydir','reverse');xlim([-10 10]);ylim([700 1500]);
title('ground truth');ylabel('range');xlabel('cross-range');

sar signal simulation

the following section describes how the system operates based on the above configuration. specifically, the section below shows how the data collection is performed for a sar platform. as the platform moves in the cross-range direction, pulses are transmitted and received in directions orthogonal to the flight path. a collection of pulses gives the phase history of the targets lying in the illumination region as the platform moves. the longer the target lies in the illumination region, the better the cross-range resolution for the entire image because the process of range and cross-range focusing is generalized for the entire scene.

% define the broadside angle
refangle = zeros(1,size(targetpos,2));
rxsig = zeros(truncrangesamples,numpulses);
for ii = 1:numpulses
    % update radar platform and target position
    [radarpos, radarvel] = radarplatform(slowtime);
    [targetpos,targetvel] = pointtargets(slowtime);
    
    % get the range and angle to the point targets
    [targetrange, targetangle] = rangeangle(targetpos, radarpos);
    
    % generate the lfm pulse
    sig = waveform();
    % use only the pulse length that will cover the targets.
    sig = sig(1:truncrangesamples);
    
    % transmit the pulse
    sig = transmitter(sig);
    
    % define no tilting of beam in azimuth direction
    targetangle(1,:) = refangle;
    
    % radiate the pulse towards the targets
    sig = radiator(sig, targetangle);
    
    % propagate the pulse to the point targets in free space
    sig = channel(sig, radarpos, targetpos, radarvel, targetvel);
    
    % reflect the pulse off the targets
    sig = target(sig);
    
    % collect the reflected pulses at the antenna
    sig = collector(sig, targetangle);
    
    % receive the signal  
    rxsig(:,ii) = receiver(sig);
    
end

visualize the received signal. the received signal can now be visualized as a collection of multiple pulses transmitted in the cross-range direction. the plots show the real part of the signal for the three targets. the range and cross-range chirps can be seen clearly. the target responses can be seen as overlapping as the pulse-width is kept longer to maintain average power.

imagesc(real(rxsig));title('sar raw data')
xlabel('cross-range samples')
ylabel('range samples')

perform range compression. each row of the received signal, which contains all the information from each pulse, can be matched filtered to get the dechirped or range compressed signal.

pulsecompression = phased.rangeresponse('rangemethod', 'matched filter', 'propagationspeed', c, 'samplerate', fs);
matchingcoeff = getmatchedfilter(waveform);
[cdata, rnggrid] = pulsecompression(rxsig, matchingcoeff);

this figure shows the response after matched filtering has been performed on the received signal. the phase histories of the three targets are clearly visible along the cross-range direction and range focusing has been achieved.

imagesc(real(cdata));title('sar range compressed data')
xlabel('cross-range samples')
ylabel('range samples')

perform azimuth compression. there are multiple techniques to process the cross-range data and get the final image from the sar raw data once range compression has been achieved. in essence, range compression helps achieve resolution in the fast-time or range direction and the resolution in the cross-range direction is achieved by azimuth or cross-range compression. two such techniques are the range migration algorithm and the back-projection algorithm, which are demonstrated in this example.

rma_processed = helperrangemigration(cdata,fasttime,fc,fs,prf,speed,numpulses,c,rc);
bpa_processed = helperbackprojection(cdata,rnggrid,fasttime,fc,fs,prf,speed,crossrangeresolution,c);

visualize the final sar image. plot the focused sar image using the range migration algorithm and the approximate back projection algorithm. only a section of the image formed via the range migration algorithm is shown to accurately point the location of the targets.

the range migration and accurate form of the backprojection algorithm as shown by [2] and [3] provides theoretical resolution in the cross-track as well as along-track direction. since the back-projection used here is of the approximate form, the spread in azimuth direction is evident in case of the back-projection, whereas the data processed via range migration algorithm shows that theoretical resolution is achieved.

figure(1);
imagesc((abs((rma_processed(1700:2300,600:1400).'))));
title('sar data focused using range migration algorithm ')
xlabel('cross-range samples')
ylabel('range samples')

figure(2)
imagesc((abs(bpa_processed(600:1400,1700:2300))));
title('sar data focused using back-projection algorithm ')
xlabel('cross-range samples')
ylabel('range samples')

summary

this example shows how to develop sar processing leveraging a lfm signal in an airborne data collection scenario. the example also shows how to generate an image from the received signal via range migration and approximate form of the back-projection algorithms.

references

appendix

range migration algorithm

function azcompresseddata = helperrangemigration(sigdata,fasttime,fc,fs,prf,speed,numpulses,c,rc)

this function demonstrates the range migration algorithm for imaging the side-looking synthetic aperture radar. the pulse compressed synthetic aperture data is considered in this algorithm.

set the range frequency span.

frequencyrange = linspace(fc-fs/2,fc fs/2,length(fasttime));
krange = 2*(2*pi*frequencyrange)/c;

set the cross-range wavenumber.

kaz = 2*pi*linspace(-prf/2,prf/2,numpulses)./speed;

generate a matrix of the cross-range wavenumbers to match the size of the received two-dimensional sar signal.

kazimuth = kaz.';
kx = krange.^2-kazimuth.^2;

set the final wavenumber to achieve azimuth focusing.

kx = sqrt(kx.*(kx > 0));
kfinal = exp(1i*kx.*rc);

perform a two-dimensional fft on the range compressed signal.

sdata =fftshift(fft(fftshift(fft(sigdata,[],1),1),[],2),2);

perform bulk compression to get the azimuth compression at the reference range. perform filtering of the 2-d fft signal with the new cross-range wavenumber to achieve complete focusing at the reference range and as a by-product, partial focusing of targets not lying at the reference range.

fsmpol = (sdata.').*kfinal;

perform stolt interpolation to achieve focusing for targets that are not lying at the reference range.

stoltpol = fsmpol;
for i = 1:size((fsmpol),1)
    stoltpol(i,:) = interp1(kx(i,:),fsmpol(i,:),krange(1,:));
end
stoltpol(isnan(stoltpol)) = 1e-30;
stoltpol = stoltpol.*exp(-1i*krange.*rc);
azcompresseddata = ifft2(stoltpol);
end

back-projection algorithm

function data = helperbackprojection(sigdata,rnggrid,fasttime,fc,fs,prf,speed,crossrangeresolution,c)

this function demonstrates the time-domain back projection algorithm for imaging the side-looking synthetic aperture radar. the pulsed compressed synthetic aperture data is taken as input in this algorithm. initialize the output matrix.

data = zeros(size(sigdata));
azimuthdist = -200:speed/prf:200;%azimuth distance

limit the range and cross-range pixels being processed to reduce processing time.

rangelims = [700 1400];
crossrangelims = [-10 10];

index the range grid in accordance with the range limits.

rangeidx =  [find(rnggrid>rangelims(1), 1) find(rnggrid'last')];

index the azimuth distance in accordance with the cross-range limits.

crossrangeidxstart = find(azimuthdist>crossrangelims(1),1);
crossrangeidxstop = find(azimuthdist'last');
for i= rangeidx(1):rangeidx(2)  % iterate over the range indices
    
    % using desired cross-range resolution, compute the synthetic aperture
    % length
    lsynth= (c/fc)* (c*fasttime(i)/2)/(2*crossrangeresolution);
    lsar = round(lsynth*length(azimuthdist)/azimuthdist(end)) ;
    % ensure lsar is an odd number
    lsar = lsar   mod(lsar,2);    
    
    % construct hanning window for cross-range processing, to suppress the
    % azimuthal side lobes
    hn= hanning(lsar).';
    % iterate over the cross-range indices
    for j= crossrangeidxstart:crossrangeidxstop 
        % azimuth distance in x direction over cross-range indices
        posx= azimuthdist(j);
        % range in y-direction over range indices
        posy= c*fasttime(i)/2;
        % initializing count to zero
        count= 0;
        % iterate over the synthetic aperture
        for k= j-lsar/2  1:j  lsar/2 
            % time delay for each of range and cross-range indices
            td= sqrt((azimuthdist(k)- posx)^2   posy^2)*2/c;
            cell= round(td*fs)  1 ;
            signal = sigdata(cell,k);
            count= count   hn(k -(j-lsar/2))*signal *exp(1j*2*pi*fc*(td));
        end
        % processed data at each of range and cross-range indices
        data(i,j)= count;
    end
    
end
end