论文研究 - 自适应MTI滤波器的优化.pdf-资料库

Int. J. Communications, Network and System Sciences, 2017, 10, 206-217 http://www.scirp.org/journal/ijcns ISSN Online: 1913-3723 ISSN Print: 1913-3715 Optimization of Adaptive MTI Filter Wenxu Zhang, Shudi Ma, Qiuying Du College of Information and Communication Engineering, Harbin Engineering University, Harbin, China How to cite this paper: Zhang, W.X., Ma, S.D. and Du, Q.Y. (2017) Optimization of Adaptive MTI Filter. Int. J. Communica- tions, Network and System Sciences, 10, 206-217. https://doi.org/10.4236/ijcns.2017.108B022 Received: May 31, 2017 Accepted: August 11, 2017 Published: August 14, 2017 Abstract Moving target indication (MTI) is an effective means for radar to find moving targets in clutter environment. This paper introduces the basic principles of MTI, how to avoid the blind speed problem and the optimization of MTI fil- ter. Implementing the multi-notch adaptive moving target indication (AMTI) filter that designed by using the stagger code in varied cases, which is based on a feature vector method optimization. Keywords Adaptive Moving Target Indication (AMTI), Stagger Code, Feature Vector Method, Multi-Notch 1. Introduction MTI band-stop filter as a “single channel”, followed by detection is relatively simple. When the target speed is large and the repetition frequency is low, make sure that there is no distance blur, through the “variable week” variable repeat cycle or repeat and “time varying” [1]. Can overcome the blind speed problem, the drawback is no improvement in noise. In general, the mess is not very strong, the radar can handle a limited number of pulses, suitable for the use of repetitive and time-varying weighted system. The adaptive has a variety of ways to achieve, in which the performance is better “first order” and “second order”. The first- order basic method is to use the interval-based velocity measurement and the zero-point distribution method to determine the weighting parameters of the clutter cancellation filter to obtain the filter whose notch is aligned with the cen- ter of the clutter spectrum [2]. Its advantages are simpler, the disadvantage is that it cannot be adaptive with the clutter spectrum, so sometimes the perfor- mance is worse. The second-order basic method is to estimate the clutter cova- riance matrix, and then use matrix inversion or feature decomposition feature vector method to determine the filter weight coefficient. This paper first analyzes the moving target indication (MTI), on this basis, the DOI: 10.4236/ijcns.2017.108B022 August 14, 2017

W. X. Zhang et al. MTI is optimized, and the appropriate filter coefficients are designed by the fea- ture vector method, which can effectively suppress the clutter. And the use of stagger code design MTI filter to eliminate the impact of blind speed. For mo- tion clutter, the spectral center is not at zero frequency, and is time-varying. In order to suppress such clutter, this paper adopts adaptive motion clutter sup- pression technique AMTI, and designs multi-notch AMTI filter [3]. 2. Research on Adaptive Clutter Suppression Algorithm The earliest MTI filter is a delay line canceller, is currently one of the most commonly used MTI filter. According to the different number of cancellation, but also divided into single delay line canceller, double delay line canceller and multi-delay line canceller [4]. Single delay line canceller as shown in Figure 1, the impulse response of the is equal to the ( ) x t ( ) y t and the input ( )h t ( )h t , and output single delay line canceller is expressed as convolution between the impulse response The impulse response of the counter is: ( − δ δ ( ) h t ( ) t = [5]. )r t T − (1) The power gain of the single delay line canceller is: H ( ) ω 2 2  4 sin =      rT ω 2       (2) Double delay line canceller as shown in Figure 2. The response of the double delay line canceller is ( ) h t = δ ( ) t − ( 2 δ t T − r ) + ( δ t − T 2 r ) (3) x(t) h(t) + DelayTr Σ - y(t) x(t) Figure 1. Single delay line canceller. h(t) + Σ - + Σ - DelayTr DelayTr Figure 2. Double delay line canceller. y(t) 207

W. X. Zhang et al. The double delay line canceller impulse response is: H ( ) ω 2 = H 1 ( ) ω 2 H 1 ( ) ω 2 = 16 sin       rT ω 2 4       (4) The adaptive moving target indication (AMTI) filter is usually composed of a FIR filter with a horizontal structure. The output of the MTI filter is: ( ) ( ) Y n W X n = T = 1 − N ∑ i = 0 w x n i − i ( ) (5) where W is the weight vector and quency response of this filter is: ) ( H f = N 1 − ∑ i = 0 X n is the input signal vector. The fre- ( ) w i exp ( − j fTπ 2 i ) (6) In the radar system, in order to avoid the occurrence of blind effects, usually the use of “variable T” approach, that is, by regularly changing the radar launch pulse period so that the frequency of blindness is greater than the target possible Doppler frequency. Adaptive clutter suppression is compatible with parametric techniques, meaning that the clutter suppression filter must be time-varying. For the determined N value, the frequency characteristic of the MTI filter is de- termined only by the weight vector, so the calculation of the weight vector is the core of the MTI process, according to different design methods, the optimal weight vector is generally different. In engineering practice, the improvement factor is often used to measure the performance of MTI system. The improve- . Obviously, the ment factor of the MTI filter is defined as greater the I , the better the effect of the system on clutter suppression. It has been proved that the optimal weight vector of the MTI filter should be the ei- genvector corresponding to the minimum eigenvalue of the covariance matrix of the input clutter, in order to maximize the average improvement factor of the MTI. At this point the improvement factor is max λ= min S C o 0 S C i i ) ( / 1/ = ( ) I I / / [6]. 2.1. Optimal Design of Filter The so-called optimization design requires a set of optimal filter coefficients, to maximize the improvement factor, a lot of design methods. In the case of the va- riable T, the better methods are feature vector method, matching algorithm, ze- ro-point allocation method and linear prediction method [7]. The feature vector method is the solution that minimizes the clutter output power when the target gain is constant. The zero-point assignment method is to set the frequency re- sponse zero at the notch when designing the band-stop filter. The matching al- gorithm and the linear prediction method are the solutions that minimize the clutter output power when one of the elements of the weight vector is constant. So the feature vector method has better performance [8]. The feature vector method is a clutter suppression method based on the maximum improvement factor. It is usually assumed that the clutter has a Gaussian power spectrum, the fσ , and the spectral density function 0f , the spectral width is spectral center is 208

W. X. Zhang et al. is: ( C f ) = 1 2 πσ f exp )2 ( −     f f − 2 σ f 0 2     (7) According to the Wiener filter theory, if the clutter is a stationary stochastic process, its power spectrum and autocorrelation function are Fourier transform ( cr m n is the Fourier pairs. Therefore, the clutter autocorrelation function transform of its power spectrum ( )C f ( ) C f e ( r m n , c . ( t t − m n df f 2 π = +∞ ) ) , ) j ∫ ∫ = −∞ +∞ −∞ 1 2 πσ f exp ( −     f f − 2 2 σ 0 f 2 ) e     j f 2 π ( t t − m n ) df (8) t m τ = mn then − is the relevant time. If the center of the clutter spectrum is zero, t n ( cr m n , ) 2 2 2 −= e πσ τ 2 f mn (9) We obtain the clutter autocorrelation matrix A of N pulses N N  1, ) ( 0,0 ( ) 1,0  −     ( 0, ( 1,    =     ( r N c ( r N c ) 1,1 r c r c r c r c r c r c 1,0 R c − ) ) 1 − ) 1 − N − ) 1        (10) ) ( 0,1 ( ) 1,1  − ) ( r N c ( S f B f , the Doppler spectrum r pressed as of the target echo signal can be ex- ( S f )  −=  1 ，    B 2 0 other ， f B 2 (11) The target autocorrelation function is ( r m n , s ) = = B /2 B − /2 1 ∫ B sin 2 π τ f mn j e df = ( B π τ mn B π τ mn )  =   1, 0, j e   1 j B 2 π τ mn m n = m n ≠ 2 π τ B mn /2 − j − e 2 π τ B mn /2   (12) Assume that the clutter data and the target data of the N pulse MTI input are respectively C  =  ( c t 1 ) , ( c t 2 S  =  ( s t 1 ) , ( s t 2 ) ) ,  , ( c t N ,  , ( s t N ) ) T   T   (13) (14) Then the MTI output of the clutter power and signal power are C 0 = S 0 = E w C H H E w S       2 2       = = H C w R w i c H S w R w i s (15) (16) where iC and iS represent the clutter power and the signal power at the MTI 209

W. X. Zhang et al. filter input, respectively, w is the weight vector of the FIR filter. According to the definition of the improvement factor of the MTI filter o I = / / S S i C C i ( sr m n know, ) , o By = S C o i S C i o × = H S w R w i s S i × H C i H w R w C w R w w R w i = H c c s (17) sR for the unit array, therefore, I = H w w H w R w c (18) The characteristic equation of λ= n R w c n cR is w n , n =  (19) 0,1, N , where nw is the eigenvector corresponding to the eigenvalue nλ . Among them λ λ 0 1    λ n In the eigenvalues of cR , the subspace of the eigenvector corresponding to the large eigenvalue is the subspace of the signal, and the main points of the clutter are located in this subspace. The subspace of the eigenvector corres- ponding to the small eigenvalue is the noise subspace. Since the noise subspace is orthogonal to the signal subspace, the eigenvector B corresponding to the min- 0w of the MTI filter, this imum eigenvalue can suppress the clutter component to the greatest extent, which is the biggest improvement factor [9]. 0λ is taken as the weight vector 2.2. Stagger Repetition Frequency In general, it is not possible to obtain a PRF that can meet the required ambi- guous distance and Doppler coverage. Therefore, a method of stagger repetition frequency is proposed. Stagger repetition frequency is a measure that can be used to prevent blind influence [10]. If the radar uses N repetition frequencies, their repetition periods can be ex- pressed as K T = ∆ 1 K T ∆ = 2 (20) 1 2 T r T r       T  rN = = 1/ 1/ r 1 f f r 2  f K T ∆ T∆ is the maximum convention period for [ 1/ = = rN N T T , r r 1 , T , rN 2 ] , then the odds ratio is: T T : r r 1 2 : = T rN : K K 1 : 2 :  : K N (21) ] [ 2 N : : : K K K 1 is the stagger code, the ratio of the largest K value to the minimum K value in the parametric code is called the maximum ratio r of the azimuth cycle. r (22) max ] [ / K K 1 K K 1   K K = [ ] : : : : : : N N 2 2 If iK is mutually different and satisfies Equation (22), then the first true blind velocity corresponds to the Doppler frequency bnf . 210

W. X. Zhang et al. bnf = 1 T ∆ (23) The average repetition period of the radar is T r = 1 N ∑ N = i 1 T ri = K T ∆ av (24) avK is the mean of the difference. Therefore K av = T r T ∆ f bn = T f r bn = f bn f r (25) = K f av r (26) Because f r = 1/ T r is the average radar repetition frequency, it is also called avK for the blind expansion factor. The coefficient of the MTI filter between the pulses is different for each pulse of the three pulse canceller, so it is a time-varying filter. If the radar uses three 3T at one time, three sets of MTI filters work in repetition frequencies turn. The depth of the stagger MTI filter speed response notch is independent of the form of the canceller and is independent of the pulse received in the radar antenna beam and is related to the maximum ratio of the azimuth cycle. The larger the maximum change ratio, the shallower the corresponding notch depth. 2T , 1T , df 2.3. Optimization of Adaptive MTI Filter of the motion clutter in the input In the clutter region, the spectral center signal is estimated to obtain the Doppler frequency estimate of the center of the clutter spectrum. And then estimate the spectral width B to obtain the es- ∧ timated value B of the spectral width. Then we obtain the weight coefficient of the multi-notch filter by using the obtained estimator into the feature vector method, and design the MTI filter with multi-notch. As shown in Figure 3. ∧ and B df df ∧ First estimate the motion of the clutter spectrum center. The radar suffers from narrowband clutter and noise that can be expressed as (27) ( ) A t e ( ) u t ( ) n t ( ω ϕ+ 0 = + ) j t d Input u(t) Output AMTI filter The weight vector is calculated by the feature vector method Spectral width estimation of motion clutter spectrum Motion clutter spectrum center estimation Figure 3. Optimization design of adaptive MTI filter. 211

is the amplitude, ( )n t dω is the Doppler frequency of the clutter, 0ϕ is is the additive noise. Noise is not related to clutter, W. X. Zhang et al. ( )A t the initial phase, and and noise between different PRI is uncorrelated. Delay the signal after a PRI ) ( ) A t T e = ( )u t ( The correlation function of ( ) E u t u t T r and )  −  ( u t T r ( R T r − − =   ) * r ( ω d j ( t T ϕ− r 0 + ) ) + ( n t T r − ) (28) ( )r u t T− is ( ) ( E A t A t T =   r − )   T d r j e ω (29) Therefore, the center frequency estimate of the clutter spectrum is obtained ∧ f d = 1 arctan T 2 π r Im Re ∧ ( R T r ∧ ( R T r       ) )       (30) After obtaining the center frequency of the clutter spectrum, the spectral width estimation is performed by the integral method. Combined with the Gauss spectrum, there are Gaussian power spectra ( C f ) = P c 1 2 πσ f exp ( f −     f − d 2 2 σ f )2     (31) fσ is the frequency variance of the Gaussian power spectrum, is the cP is the corresponding power spectrum at center of the power spectrum, and zero Doppler frequency. According to the definition of half power points f ∆ dB 3 According to the nature of Gaussian distribution, there are 2.355 σ f ≅ df . { P µ σ { P − µ σ { P − µ σ − < ≤ + 2 < ≤ + 3 < ≤ + } ) 0.6826 1 = Φ − Φ − = µ σ } ) ( 0.9544 2 2 − Φ − = Φ = µ σ ) ( } 0.9974 3 3 − Φ − = = Φ µ σ ( ) 1 ( ) 2 ( ) 3 x x x (      (32) Prior to the estimated spectrum as the center to both sides of the center df of the accumulated clutter power spectrum (corresponding to integration), to 3dBf∆ 95.44% for the energy threshold, and then using the relationship between of the spectral estimate Gauss. After obtain- and ing the estimated spectral center and estimating the spectrum width, the weight coefficient of the filter is obtained by using the feature vector method. fσ to the spectral width B ∧ ∧ It is found that the power spectrum is the sum of their respective power spec- tra for the stagger clutter of multiple Gaussian spectra. The autocorrelation function should also have the sum of the corresponding multi-clutter compo- nents. Thus, we can derive the weight coefficients of two or more notch filters to design a multi-notch AMTI filter. 3. Simulation and Performance Analysis In Figure 4, obviously, the frequency response of the single delay line canceller and rf . The peak the double delay line canceller changes cyclically, and the period is 0n ≥ . As nf= appears at , r , and the zero value appears at ) ( 1 f 2 r = + n 2 ( ) f f 212

20 0 -20 -40 -60 -80 B d - e s n o p s e r e d u t i l p m A -100 0 0.5 1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 s t l o V - e s n o p s e r e d u t i l p m A 0 0 0.5 1 W. X. Zhang et al. single c double anceler anceler c 3.5 4 4.5 5 single c double anceler anceler c 3.5 4 4.5 5 1.5 2 3 Normalized frequency - f/fr 2.5 (a) 1.5 2 3 Normalized frequency - f/fr 2.5 (b) Figure 4. Normalized frequency response of single delay line canceller and double delay line supporter. (a) dB. (b) Volt. can be seen from the figure, the double delay line canceller has a deeper notch and a more flat passband response than a single delay line canceller. In Figure 5, the frequency response is still cyclical when the T is equal. It can be clearly seen from the figure that the notch depth is significantly enhanced compared to the delay line canceller, the passband response is also more flat, and the frequency of the notches can be set at the same time. In Figure 6, it can be seen that the use of staggered repetition frequency can greatly improve the first blind speed. The larger stagger ratio, the lighter the 213

资料库

论文研究 - 自适应MTI滤波器的优化.pdf

相关推荐

开发技术

热门标签

最新资料