CN105738883A - Method for detecting smooth generalized likelihood ratio in part uniform sea clutter background - Google Patents

Abstract

The invention discloses a method for detecting a smooth generalized likelihood ratio in the partially uniform sea clutter background. The method is a target detection method based on the smooth generalized likelihood ratio in the partially uniform sea clutter background. Firstly, a mathematical expression of a GLRT detector is given. Then, the normalized sample covariance matrix algorithm or the approximated maximum likelihood estimation algorithm is used as the estimation algorithm of a covariance M, and the median algorithm is used as the estimation algorithm of a smooth factor theta h. Finally, a scale parameter beta in the GLRT mathematical expression is replaced by beta theta h, and the expression of the smooth GLRT detector is obtained.

Description

A kind of smooth generalized likelihood test method under the uniform sea clutter background of part

Technical field

The present invention relates to the smooth generalized likelihood test method under the uniform sea clutter background of a kind of part, belong to Radar Targets'Detection technical field.

Background technology

In sea-surface target detects, adopting the adaptive targets detection algorithm being matched with sea clutter statistics and correlation properties is a kind of commonly used technological means.Thus, the characteristic of unit clutter to be detected and the design of adaptive detector, detection performance are closely related.Along with the further raising of distance by radar resolution, radar receives echo and is become non-homogeneous clutter environment from previous uniform clutter.The uniform sea clutter of part has the advantages that instantaneous power fluctuation is bigger, and this bigger frequency fluctuation can directly affect the performance of detector.At present, adaptive detector under sea clutter background designs often in order to simplify calculating, reduces the complexity of detector, and assuming that handled radar receives echo is uniform noise performance, again or in order to tackle part uniformly sea clutter, and introduce complex signal processing algorithm.Such as, the patent of invention of Beijing environmental characteristics institute application: ocean clutter cancellation and the method and system (number of patent application: CN201310556638.9, publication number: CN103645467A) of target detection in sea clutter background.This patent application extracts actual measurement speed term parameter from actual measurement time-space dispersive relation, it is determined that go out the intrinsic time-space dispersive relation of sea clutter based on described intrinsic speed item parameter, and then reconstruct obtains the course figure of the sea clutter one-dimensional range profile estimated.Eventually through surveying the gained course figure course figure with the estimation view data subtracting each other the course figure of the one-dimensional range profile obtaining suppression sea clutter, thus reaching to eliminate the purpose of Doppler frequency shift.This patent application is disadvantageous in that: during the fact that consider the Doppler frequency shift that elimination radar movable causes, have ignored part uniformly this another objective fact of sea clutter.So, it finally gives the interference characteristic suppressing the view data of sea clutter to introduce non-homogeneous sea clutter.Again such as, the patent of invention of Xian Electronics Science and Technology University's application: sub-band adaptive GLRT-LTD detection method (number of patent application: CN201510030360.0 under sea clutter background, publication number: CN104569948A), this patent application tackles part uniformly sea clutter by the mode of constructor band filter, thus realizing detecting judgement accurately, improve detection performance.But the main deficiency of this patent is: the filial generation wave filter of introducing is excessively complicated, and amount of calculation is bigger.This will certainly affect the conversion speed of detector.And the present invention can solve problem above well.

Summary of the invention

Present invention aim at solving above-mentioned the deficiencies in the prior art, it is proposed to the smooth generalized likelihood test method under the uniform sea clutter background of a kind of part, the method, under the premise not increasing GLRT detector computation complexity, improves the performance of detector.

This invention address that its technical problem is adopted the technical scheme that: the smooth generalized likelihood test method under the uniform sea clutter background of a kind of part, the method, under the premise not increasing computation complexity, can obtain in Observed sea clutter is tested and better detect performance.

Method flow:

Step 1: adopt GLRT detector as the mathematics prototype of S-GLRT detector；Described GLRT detector mathematic(al) representation is:

$\frac{|z^H{M}^{-1}p{|}^2}{\left(\mathrm{\&beta;}+z^H{M}^{-1}z\right)\left(p^H{M}^{-1}p\right)}\underset{H_0}{\overset{H_1}{\begin{array}{c}>\\ <\end{array}}}\mathrm{\&xi;}$

Wherein M represents the covariance matrix of clutter, and p is Doppler's steering vector, and z represents that radar receives the echo of unit to be detected, and H represents conjugate transpose, and β is scale parameter, and ξ is decision threshold.

Step 2: with normalization sample covariance matrix (normalizedsamplecovariancematrix, NSCM) estimation or progressive maximum likelihood (approximatedmaximumlikelihood, AML) estimate as the algorithm for estimating of clutter covariance matrix M, to take mediant estimation as smoothing factorAlgorithm for estimating；Normalization sample covariance matrix (normalizedsamplecovariancematrix, the NSCM) estimated form of described clutter covariance matrix M is:

${\hat{M}}_{NSCM}=\frac{N}{K}\underset{k=1}{\overset{K}{\&Sigma;}}\frac{z_k{z}_k^H}{z_k^H{z}_k}$

Progressive maximum likelihood (approximatedmaximumlikelihood, AML) estimated form is:

$\begin{array}{c}{\hat{M}}_{AML}(i+1)=\frac{N}{K}\underset{k=1}{\overset{K}{\&Sigma;}}\frac{z_k{z}_k^H}{z_k^H{\left({\hat{M}}_{AML}\right(i\left)\right)}^{-1}{z}_k}\\ i=0,1,\mathrm{...},3\end{array}$

Accordingly, the smoothing factor of mediant estimation is takenExpression formula is:

${\mathrm{\&theta;}}_{Med-NSCM}=median\left\{z_k^H{z}_k\right|k=1,2,...,K\}$

${\mathrm{\&theta;}}_{Med-AML}=median\left\{z_k^H{{\hat{M}}_{AML}}^{-1}{z}_k\right|k=1,2,...,K\}$

Step 3, replaces with the scale parameter β in GLRT detector mathematic(al) representationObtain the correction form of GLRT detector, i.e. the expression formula of smooth GLRT (smoothGLRT, S-GLRT) detector；The expression formula of described smooth GLRT (smoothGLRT, S-GLRT) detector is:

Beneficial effect:

The present invention has the advantage that compared with the prior art

(1) smooth GLRT (smoothGLRT, the S-GLRT) detector that the present invention proposes is compared with GLRT detector, under the premise not increasing computation complexity, can obtain and better detect performance in Observed sea clutter is tested.

(2) the S-GLRT detector that the present invention proposes, the smoothing factor of its introducing is primarily to the impact on detector performance of the uniform sea clutter in weakened part.But not losing versatility, for the target detection under uniform sea clutter background, S-GLRT detector still has the detection performance close with GLRT detector.Meet the clutter environment requirement of reality.

(3) scale parameter is had CFAR characteristic by the S-GLRT that the present invention proposes.

(4) smoothing factor in S-GLRT detectorEmploying takes mediant estimation algorithm, has good performance in actual environment.

Accompanying drawing explanation

Fig. 1 is the method flow diagram of the present invention.

Fig. 2 is S-GLRT and the GLRT that proposes of present invention Performance comparision figure in actual measurement clutter situation.

Detailed description of the invention

Below in conjunction with Figure of description, the invention is described in further detail.

The present invention under part uniformly sea clutter background, the method improving GLRT detector performance, technical problem underlying therein includes:

(1) smoothing factorThe selection of algorithm for estimating.

(2) derivation of S-GLRT detector mathematic(al) representation.

In the uniform sea clutter of part of the present invention, the smooth adaptive detection algorithm of radar target includes techniques below measure: first, provides the mathematical model of GLRT detector.Then, clutter covariance matrix M is respectively adopted NSCM and AML algorithm for estimating, smoothing factorEmploying takes mediant estimation algorithm.Finally, the scale parameter β in GLRT detector mathematic(al) representation is replaced withObtain the mathematical model of smooth GLRT (smoothGLRT, S-GLRT) detector.

As it is shown in figure 1, the smooth generalized likelihood test method that the invention provides under the uniform sea clutter background of a kind of part, the method includes:

Step 1: initially with the mathematic(al) representation of GLRT detector as mathematics prototype:

In formula (1), M represents the covariance matrix of clutter, and p is Doppler's steering vector, and z represents that radar receives the echo of unit to be detected, and H represents conjugate transpose, and β is scale parameter, and ξ is decision threshold.

Step 2: when estimating as the algorithm for estimating of clutter covariance matrix M with normalization sample covariance matrix (normalizedsamplecovariancematrix, NSCM), using take mediant estimation algorithm asValue algorithm time,Expression formula be

What the k in formula (2) represented is the number of samples of radar return.Taking the mediant estimation algorithm median concrete meaning represented is, to all numeric ratios relatively size in brace, takes size that numerical value placed in the middle as value.

When estimating as the algorithm for estimating of clutter covariance matrix M with progressive maximum likelihood (approximatedmaximumlikelihood, AML), using take mediant estimation algorithm asValue algorithm time,Expression formula be

Step 3: for formula (1), scale parameter β is replaced withCorresponding GLRT detector is revised as:

Formula (4) is smooth GLRT (smoothGLRT, the S-GLRT) detector that the present invention proposes.

The following experiment of can passing through of smooth GLRT (smoothGLRT, the S-GLRT) detector that the present invention proposes is verified further.Experiment uses the sea clutter data of IPIX radar collection to analyze the detection performance of S-GLRT, there is provided the network address of data: http://soma.mcmaster.ca/ipix.php, data are called: 19980223-170435 (range resolution ratio is 15m), HH polarizes, these data contain 60000 time pulses altogether, 34 distance unit.Considering that partial distance cell data is likely to contaminated, so inventor have chosen the data of 26 pure sea clutter unit, target is added on the 15th distance unit.

Fig. 2 is S-GLRT and the tradition GLRT detection Performance comparision under different covariance matrix that the present invention proposes.Obviously, whether NSCM estimator or AML estimator, in actual measurement clutter, the detection performance of S-GLRT is substantially better than the detection performance of GLRT.

Claims (4)

1. the smooth generalized likelihood test method under the uniform sea clutter background of part, it is characterised in that described method comprises the steps:

Step 1: adopt GLRT detector mathematic(al) representation as the mathematics prototype of S-GLRT detector；

Step 2: estimate using normalization sample covariance matrix or progressive maximal possibility estimation is as the algorithm for estimating of clutter covariance matrix M, to take mediant estimation as smoothing factor θ_hAlgorithm for estimating；

Step 3, replaces with β θ by the scale parameter β in GLRT detector mathematic(al) representation_h, obtain the correction form of GLRT detector, i.e. the expression formula of smooth GLRT detector.

2. the smooth generalized likelihood test method under the uniform sea clutter background of part according to claim 1, wherein the GLRT detector mathematic(al) representation described in step 1 is:

Wherein M represents the covariance matrix of clutter, and p is Doppler's steering vector, and z represents that radar receives the echo of unit to be detected, and H represents conjugate transpose, and β is scale parameter, and ξ is decision threshold.

3. the smooth generalized likelihood test method under the uniform sea clutter background of part according to claim 1, wherein the normalization sample covariance matrix estimated form of the clutter covariance matrix M described in step 2 is:

${\hat{M}}_{NSCM}=\frac{N}{K}\underset{k=1}{\overset{K}{\&Sigma;}}\frac{z_k{z}_k^H}{z_k^H{z}_k}$

Progressive maximum likelihood, namely the estimated form of AML is:

${\hat{M}}_{AML}(i+1)=\frac{N}{K}\underset{k=1}{\overset{K}{\&Sigma;}}\frac{z_k{z}_k^H}{z_k^H{\left({\hat{M}}_{AML}\right(i\left)\right)}^{-1}{z}_k}$

I=0,1 ..., 3

Accordingly, the smoothing factor θ of mediant estimation is taken_hExpression formula is:

${\mathrm{\&theta;}}_{Med-NSCM}=median\left\{z_k^H{z}_k\right|k=1,2,...,K\}$

${\mathrm{\&theta;}}_{Med-AML}=median\left\{z_k^H{{\hat{M}}_{AML}}^{-1}{z}_k\right|k=1,2,...,K\}.$

4. the smooth generalized likelihood test method under the uniform sea clutter background of part according to claim 1, wherein the expression formula of the smooth GLRT detector described in step 3 is: