Signal Processing Of FMCW Synthetic Aperture Radar Data

In the fields of airborne earth observations, there is a special interest in theSynthetic Aperture Radar (SAR) with near range, which is mounted on a smallunmanned aerial vehicle. It requires the radar system having small size, light weight,low cost and low power consuming. Convensional SAR works in pulse mode, which isusually a complex instrument, neither low cost nor compact. For the radar system withlow transmitted power, the couple between the transmitting and the receiving paths canbe resolved easily. Furthermore the Frequency Modulated Continuous Wave (FMCW)radar system is generally compact and relatively cheap. Therefore the combination ofFMCW technology and SAR techniques can lead to lightweight, cost effective, lowpower consuming imaging sensors of high resolution, which makes it feasible to mounton a small unmanned aerial vehicle. Meanwhile, it also has low probability ofinterception due to the low transmitted power.The dechirp-on-receive technology is adopted in the receiving path, i.e., the echo ismixed by the transmitted signal to yield a beat signal with narrow bandwidth. Thedemands for video receiving channels, A/D samples and processing rate are not severe.Conventional SAR works in pulse mode. Therefore, it requires high peak power underthe condition that the range is far. This results in high demands for the transmittingsystem, servo system, system weight, power consumption, cost and the platform. In thecontrast, FMCW SAR transmits signals in a long sweep equal to the pulse repetitioninterval, so the peak power is low, and the solid-state amplifier is sufficient.Furthermore, low cost makes it available in the civil market. Totally, FMCW SAR hasmany advantages, and the research on it has attracted increasing interests.Due to the fact that the FMCW SAR is continuously transmitting signals whileplatform moving, the"stop-and-go"approximation used for the derivation ofconventional SAR algorithms could not be anymore valid. Therefore it is necessary toanalyze specific algorithms for FMCW SAR. Here we start our work with FMCW SARecho analysis and signal processing, the contents including signal model derivation,signal processing algorithms in different modes, and moving target detection, parameterestimation and imaging, all proved through computer simulation.The thesis is organizedas follows.Chapter 1 provides a short overview of the pulse SAR and FMCW SAR. At the same time, as a demonstration of the increasing interest in FMCW SAR from thescientific and industrial community,the chapter reports some related works started atother institutes and introduces the main work of this thesis.Chapter 2 establishes the receiving signal model of FMCW SAR without thestop-and-go approximation, under the condition that the system adopts thedechirp-on-receive technology. The Doppler frequency shift effect, induced by theplatform's continuous motion while radar transmitting and receiving signals, isequivalent to azimuth Doppler frequency after formulation deduction. Also the effect ofDoppler frequency shift on SAR imaging is analyzed. To resolve the problem offrequency non-linearity, inherent to FMCW SAR, this chapter proposes a modifiedcorrection algorithm considering the deficiency of the existing methods.Chapter 3 researches monostatic FMCW SAR imaging, including squint FMCWSAR and spotlight FMCW SAR. For squint FMCW SAR imaging, Modified RD (MRD)algorithm and equivalent side-looking method are proposed. The former compensatesfor range walk in time domain, Doppler frequency shift and range curvature infrequency domain. The latter treats the squint SAR data as side-looking one afterpre-processing. To overcome the constraint of the azimuth focusing depth induced bythe pre-processing, the azimuth non-linear chirp scaling algorithm is utilized.For spotlight FMCW SAR imaging, the azimuthal pre-filtering processing isadopted to eliminate the azimuth spectral folding effect, which increases the samplingfrequency through interpolation. Since this procedure can realize bulk azimuth datacompression, the number of azimuth does not increase significantly. At last, based onthe Frequency Scaling Algorithm, residual data focusing is achieved.Chapter 4 studies bistatic FMCW SAR imaging. The tandem bistatic FMCW SARimaging with frequency non-linearity correction is considered first. The signalexpression in the two dimensional wave-number domains is deduced using the conceptof instantaneous frequency. Based on the signal analytical expression, a Modified RangeMigration Algorithm, aimed at tandem bistatic FMCW SAR, is presented with someapproximation. Furthermore, using the concept of instantaneous frequency and throughthe introduction of the parameters, i.e. the sum of the closest slant range, the differenceof the closest slant range and the half quasi bistatic angle, this chapter also deduces thereceiving signal expression of parallel bistatic spotlight FMCW SAR in the twodimensional wave-number domains, and subsequently, a Modified Frequency Scaling(MFS) algorithm, containing azimuth pre-filter processing and Doppler frequency shift compensation, is presented.Chapter 5 studies the downward-looking FMCW SAR imaging in threedimensions, and proposes a new method for its imaging. The bistatic configuration ofdownward-looking SAR is equivalent to monostatic one according to the equivalentphase center principle first, and then the data in range, along track and across track areprocessed, respectively. This method utilizes RD algorithm to process data in range andalong track dimensions. For the across track data processing, it can be considered as subaperture SAR, just with the small aperture length. The single point equivalent squintmode model is introduced, and we conclude that, after formulation deduction and theoryanalysis, although the squint mode model is different for the points at the same rangebin, it can be processed with the same reference function.Chapter 6 presents a new approach for forward-looking FMCW radar imaging,based on two dimensional Chirp-Z transform scaling. Inspired by Scan SAR processing,this algorithm needs zero padding in azimuth direction first, and then Chirp-Z transformscaling is used to remove range dependent RCM (Range Cell Migration) in rangeprocessing. For azimuth processing, Chirp-Z transform is adopted to adjust the outputinterval, combined with SPECAN algorithm to complete azimuth focusing. This chaptergives the complete deduction process and the expression of each compensation factor.The whole algorithm proposed only includes FFT and complex multiplication, withinterpolation free, and hence it is prone to implement in project.Chapter 7 presents the high speed ground moving target detection method usingtriangular modulation FMCW SAR, which can be applied to the case with Dopplerspectrum ambiguities. By analyzing the characteristic of the moving target echo, weillustrate that the moving target in the upslope and downslope FMCW images hasopposite shift in the range direction. According to this characteristic, we can detect themoving target by image subtraction after taking certain compensation procedure.Additionally, this chapter investigates the slow moving target parameter estimation andimaging after detection using FMCW.Chapter 8 summarizes the main results of the study which have led to this thesis;additionally, it draws conclusions and gives some recommendations for future work.