In this thesis, a radar clutter suppression algorithm based on the adaptive clutter prewhitening filter and Doppler filter bank is described which provides improved performance compared with conventional nonadaptive clutter suppression algorithms. The clutter prewhitening filter functions as an adaptive decorrelator to whiten the clutter spectrum in received radar signal from a particular rangeazimuth resolution cell. Then the Doppler filter bank is used to sum the clutter-suppressed signals from the output of the clutter prewhitening filter in the frequency domain, therefore enhanced target signals result at the output. The clutter prewhitening filter is implemented by using a prediction-error filter whose coefficients are periodically updated by the Burg's maximum entropy method in order to account for changes in nonstationary clutter and noise environments. The Doppler filter bank is implemented by using the discrete Fourier transform.
In simulated clutter environments, the performance of a proposed clutter suppression algorithm is obtained and compared with the conventional cascaded moving target indicator (MTI) and Doppler filter bank (DFB) algorithm. As a performance index, we use the improvement factor which is defined as "the signal-to-clutter plus noise ratio at the output of the clutter suppression filter divided by the signal-to-clutter plus noise ratio at the input of the filter". The simulated radar clutter signal is generated by sampling the power spectrum which represents typical clutter and noise spectra. The clutter signal is assumed stationary over the adaptation period of the clutter suppression filter and the shape of the clutter spectrum is assumed Gaussian. Computer simulation has been performed for single and bimodal clutter sources by varying the clutter power, mean Doppler frequency, and spectral width of the clutter spectrum.
In the simulated clutter environment, simulation results show that the performance of the proposed algorithm is about the same as that of the conventional cascaded MTI and DFB algorithm when the mean Doppler frequency of the clutter spectrum is around zero under the ground clutter assumption. As the mean Doppler frequency of clutter spectrum deviates from zero, the improvement factor of the conventional cascaded MTI and DFB algorithm is lowered, but the average improvement factor of the proposed algorithm is about the same as that in the ground clutter environment. However, when the main lobe of each filter of the Doppler filter bank is close to the spectrum of a strong clutter, the proposed clutter suppression filter yields the improvement factor lower than that can be obtained when the main lobe of the filter is far from the clutter spectrum. As clutter power increases, the average improvement factor increases and the filter response effectively nulls the clutter power spectrum.
본 논문에서는 adaptive clutter prewhitening filter와 Doppler filter bank를 이용한 radar clutter 제거 알고리즘의 성능에 관하여 연구하였다. Adaptive clutter prewhitening filter의 기능은 수신한 레이다 신호에 포함된 clutter를 decorrelation하는 whitening filter의 역할을 하며 Doppler filter bank (DFB) 는 prewhitening filter에 의해서 clutter가 제거된 신호를 integration하여 target 신호를 강화하는 역할을 한다. Transversal filter로 구성되는 prewhitening filter의 계수는 변화하는 clutter와 noise환경에 따라서 Burg의 maximum entropy 방법에 의해서 update되며 DFB는 discrete Fourier transform으로 구현된다.
본 논문에서 제안한 알고리즘의 성능 지표로는 improvement factor (IF) 를 사용하였으며 simulated clutter 상황에서 종래의 moving target indicator (MTI) 와 DFB를 직렬 연결한 알고리즘의 성능과 비교하였다. Simulated clutter는 가정한 clutter의 power spectrum sample로 부터 discrete Fourier transform으로 발생시켰으며 white noise는 Gaussian random number generator로 부터 얻었다. Clutter는 알고리즘이 adaptation하는 동안 stationary하고 스펙트럼의 형태는 Gaussian이라 가정하였다. Computer simulation은 clutter의 power, 평균 도플러 주파수, 그리고 스펙트럼의 폭으로 특징 지워지는 single clutter 또는 bimodal clutter에 대해서 수행하였다.
Simulation 결과는 다음과 같다. Ground clutter 상황과 같이 clutter의 평균 도플러 주파수가 영일 때 제안된 알고리즘의 성능은 종래의 MTI/DFT 알고리즘의 성능과 거의 같은 성능을 얻었다. 그러나 weather clutter 상황과 같이 스펙트럼의 평균 도플러 주파수가 증가하면 MTI/DFT 알고리즘의 성능은 매우 나빠지나 제안된 알고리즘의 평균 IF는 ground clutter 상황에서의 성능과 거의 같다. 그러나 DFB filter의 main lobe가 강한 clutter 스펙트럼에 가까우면 그 filter의 IF는 main lobe가 clutter 스펙트럼에서 떨어져 있는 filter의 IF보다 나쁘다. Clutter-to-noise ratio가 증가하면 알고리즘의 평균 IF는 증가하며 filter의 power transfer function이 변화하는 clutter 환경에 따라 적응되어 clutter를 제거한다.
