CN106932766A - Range extension target self-adapting detecting method based on variable element generalized structure - Google Patents

Abstract

The invention discloses a kind of range extension target self-adapting detecting method based on variable element generalized structure, belong to radar signal processing field.The problem of transition clutter environment is difficult in adapt to for existing covariance matrix estimation method and self adaptation extended distance detector, rationally utilize clutter non-gaussian characteristic information, construct unified covariance matrix fusion and estimate framework, common feature based on existing optimal distance extension target detection statistic under specific clutter environment, range extension target adaptive detector is constructed using product form, range extension target detection device under transition clutter environment is realized by single parameter and designs Synchronization Control with corresponding clutter covariance matrix method of estimation, detector arrangement and corresponding clutter covariance matrix method of estimation is set to match, parameter setting is succinct, detection performance of the wideband radar under complex clutter environment can effectively be lifted, with application value.

Description

Range extension target self-adapting detecting method based on variable element generalized structure

Technical field

The present invention is under the jurisdiction of radar signal processing field, and in particular to a kind of extended distance based on variable element generalized structure Objective self-adapting detection method.

Background technology

Compared with traditional low resolution Narrow-band Radar, wideband radar has the range resolution ratio of bigger bandwidth and Geng Gao, There is obvious advantage at the aspect such as anti-interference, counterreconnaissance, accurately detecting and imaging, high precision tracking, target identification, in the modern times Military and civilian field obtains extensive attention and applies, and has become an important directions of modern radar development.Tradition Narrow-band Radar is low due to range resolution ratio, and its Range resolution unit is much larger than the physical dimension of frequent goal, target echo signal A Range resolution unit is only occupied, realistic objective is often processed as " point target ".With " point target " of Narrow-band Radar Difference, the echo-signal of common broadband radar target not only only occupies a Range resolution unit, and is distributed across different In radial distance resolution cell, it is rendered as " one-dimensional range profile ", is formed " range extension target ".With extensively should for wideband radar With, range extension target detection problem is just received more and more attention, the focus as Radar Signal Processing circle in recent years and One of difficulties.

On the one hand, because the target echo that wideband radar is observed is distributed on multiple radial distance units, if still using The point target detecting method of conventional narrow-band radar, target detection is carried out to echo-signal for single range cell, and using adjacent Near range cell sampling carries out background clutter statistical property estimation, then the energy of range extension target strong scattering point can be leaked into " signal contamination " phenomenon is formed in adjacency unit, and further treats test point target configuration shadowing effect, led to not Point target is detected, due to failing to make full use of the whole energy for being distributed in target echo in multiple range cells, this is not only not The advantage of wideband radar can be embodied, the detectability of extension target of adjusting the distance can be reduced on the contrary.On the other hand, in practical application often The pure clutter assistance data neighbouring with detected unit is often utilized, unknown clutter covariance matrix is estimated, and clutter Statistical property plunders that ditch is closely related with radar, and its non-Gaussian feature strengthens with the reduction for plunderring ditch.It is larger ditch is plunderred In the case of, clutter statistical characteristicses Gaussian distributed, now the Gaussian Clutter covariance matrix based on assistance data is maximum seemingly So it is estimated as the sample covariance matrix of classics, based on two-step method generalized likelihood-ratio test criterion, corresponding optimal distance Extension objective self-adapting detector has the structure of broad sense adaptive matched filter.The reduction in ditch, background are plunderred with radar Clutter shows stronger non-homogeneous, non-Gaussian feature, can be now distributed with complex Gaussian and is modeled, i.e., clutter is represented by The product of the multiple Gauss speckle component that the non-negative texture component that space-time becomes slowly becomes soon with space-time.Wherein, texture component is used to describe The fluctuating of different distance unit clutter power level, and speckle component is used to describe the correlation inside clutter multidimensional echo-signal Property.Different from Gaussian Background, the Compound-Gaussian Clutter covariance matrix structure maximal possibility estimation based on assistance data is not closed Type expression formula, and it is related to the solution of transcendental equation, existing normalization sample covariance matrix, near-maximum-likelihood estimated matrix Etc. suboptimal estimation method is, based on two-step method generalized likelihood-ratio test criterion, corresponding optimal distance extension target is certainly Adapting to detector has the form of broad sense self adaptation integration detection device.

Although for the range extension target detection problem under gaussian sum Compound-Gaussian Clutter background, there is optimal or suboptimum Clutter covariance matrix method of estimation and corresponding adaptive detector structure, but the non-Gaussian feature of actual clutter often exists Time and spatially gradual change, the optimal or suboptimum clutter covariance square under above-mentioned specific clutter background with the change of environment Battle array method of estimation and corresponding detector are difficult in adapt to the quick change of clutter environment, cause to detect performance and CFAR accordingly Rate (CFAR) characteristic deteriorates.For the space-time gradually changeable of clutter non-gaussian degree in actual environment, in covariance matrix and In adaptive detector structured design process, extreme gaussian sum Compound-Gaussian Clutter environment should be considered, also to take into account between Gaussian sum complex Gaussian transition clutter environment therebetween.Suitable clutter covariance matrix method of estimation how is designed, is led to The characteristics of taking into full account transition clutter environment is crossed, adaptive detector structure design and corresponding clutter covariance matrix is estimated Method matches, and is to improve the key that wideband radar detects performance under complex clutter environment.

The content of the invention

Current sample covariance matrix method of estimation only considered Gaussian Clutter background, and normalize sample covariance matrix Or near-maximum-likelihood method of estimation also only considered complex Gaussian background, for the range extension target under transition clutter environment Self-adapting detecting problem, these covariance matrix estimation methods only considered the single particular case of Gauss or complex Gaussian, And sample covariance matrix method of estimation only matches with the broad sense adaptive matched filter under Gaussian Background, and normalization is adopted Sample covariance matrix or near-maximum-likelihood method of estimation also only with complex Gaussian background under broad sense self adaptation integration detection device Match, i.e., existing covariance matrix estimation method and self adaptation extended distance detector are difficult in adapt to transition clutter environment Feature.For the space-time gradually changeable of clutter non-gaussian degree in actual environment, how in range extension target adaptive detector The characteristics of synchronously considering transition clutter environment in structure design and corresponding clutter covariance matrix method of estimation, makes detector knot Structure and corresponding clutter covariance matrix method of estimation match, and need simultaneously take into account the particular surroundings such as gaussian sum complex Gaussian Under broadband radar target detection demand.

Range extension target self-adapting detecting method based on variable element generalized structure of the present invention includes following technology Measure：

Step 1 obtains data to be tested vector from L range cell to be detected, from the R nothing neighbouring with detected unit Target range unit obtains pure clutter assistance data, and is estimated using the fusion that helper data vectors calculate unknown covariance matrix ValueExisting sample covariance matrix, normalization sample covariance matrix, near-maximum-likelihood are covered by adjusting parameter α Optimal or suboptimal estimation method under the specific clutter background such as estimated matrix so that covariance matrix result is adaptive to transition The change of the non-gaussian degree of clutter environment；Specific steps include：

L data to be tested vector z is constituted by the L multiple amplitude of range cell echo to be detected_t, t=1,2 ..., L, its In, L is the natural number more than 1, centered on range cell to be detected, continuously takes a number of not including respectively before and after it The multiple amplitude of the range cell echo of target, constitutes R pure clutter helper data vectors x_m, m=1,2 ..., R, wherein, z_tAnd x_m It is the vector of N × 1 dimension, N represents that radar receives the product of array number and Coherent processing umber of pulse；

The fusion estimate of unknown covariance matrixWhereinRealized by following iterative process

In above formula, the conjugate transposition computing of subscript H representing matrixs, the inversion operation of the representing matrix of subscript -1, k=0, 1,...,K-1；K represents final iterations, is configured according to covariance matrix required precision and operand requirement, If covariance matrix required precision is higher, K takes higher value, if it is desired to operand is smaller, then K takes smaller value；Table Show kth time iterative estimate matrix during given parameters α；To strengthen the universality of estimation procedure, the initialization matrix of formula (1) Unit matrix is used, and final iteration result is the fusion estimate of unknown covariance matrixParameter alpha satisfaction 0≤ α≤1, and set according to the non-gaussian degree of actual clutter, enhance the generalization ability of subsequent detectors, the wherein not high of clutter This degree is bigger, then the value of α is bigger；When α=0It is equivalent to sample covariance matrix；When α=1It is equivalent near Like maximal possibility estimation matrix；When α=1 and K=1It is equivalent to normalize sample covariance matrix；

Step 2 is based on the fusion estimate of unknown covariance matrixBroad sense Adaptive matching is merged by product form The general character part of wave filter and broad sense self adaptation integration detection device in detection statistic, builds based on variable element generalized structure Range extension target self-adapting detecting statistic λ_α, by single parameter α to self adaptation range extension target detection device structure and Corresponding clutter covariance matrix is estimated to realize synchronous adjustment so that range extension target detection device structure adaptive is in transition clutter The change of the non-gaussian degree of environment；

Wherein, using the fusion estimate of unknown covariance matrixWith L data to be tested vector z_t, t=1, 2 ..., L, the range extension target self-adapting detecting statistic λ based on variable element generalized structure_αIt is expressed as

In above formula, p represents known space-time steering vector, is a unit vector for N × 1 dimension, according to radar system work Make parameter determination；Formula (2) realizes synchronous adjustment with parameter alpha in formula (1), detection statistic is adaptive to transition clutter environment Non-gaussian degree, can cover the optimal or sub-optimum detectors under existing gaussian sum Compound-Gaussian Clutter background；When α=0, formula (2) detection statistic λ_αThe broad sense adaptive matched filter based on sample covariance matrix is equivalent to, Gauss can be obtained miscellaneous Optimal detection performance under ripple background；When α=1, the detection statistic λ of formula (2)_αIt is equivalent to estimate based on near-maximum-likelihood The broad sense self adaptation integration detection device of matrix, range extension target is adaptive under being suitable for the stronger complex Gaussian background of non-Gaussian system Should detect, and can be according to covariance matrix required precision and the final iterations K of operand requirement adjustment；

Step 3 is the constant false alarm rate characteristic of holding detection method, and detection threshold T is set according to default false-alarm probability；Will Data to be tested vector z_t, t=1,2 ..., the self-adapting detecting statistic λ of L_αIt is compared with thresholding T, if λ_α>=T, then judge There is range extension target, echo vector z in L range cell to be detected_t, t=1,2 ..., L is not as other range cells Assistance data；If otherwise λ_α<T, then judge that L range cell to be detected does not exist range extension target, echo vector z_t, t=1, 2 ..., L as follow-up other range cells assistance data.

Compared with background technology, the beneficial effects of the invention are as follows：1) merged by unified covariance matrix and estimate framework, So that covariance matrix is adaptive to the non-gaussian degree of transition clutter environment, existing sampling covariance square can be covered Optimal or suboptimal estimation side under the specific clutter backgrounds such as battle array, normalization sample covariance matrix, near-maximum-likelihood estimated matrix Method, and can make to reach optimization balance between covariance matrix precision and operand requirement by adjusting final iterations； 2) for the space-time gradually changeable of clutter non-gaussian degree in actual environment, clutter non-gaussian characteristic information is rationally utilized, based on spy Determine the common feature of existing optimal detection statistic under clutter environment, extended distance mesh is constructed using succinct product form Mark adaptive detector, range extension target detection device designs miscellaneous with corresponding under realizing transition clutter environment by single parameter The Synchronization Control of ripple covariance matrix estimation method, parameter setting is succinctly effective；3) the range extension target detection device for being proposed Structure can be compatible and covers optimal or suboptimum clutter covariance matrix estimation side under the specific clutter background such as gaussian sum complex Gaussian Method and corresponding range extension target detection device, and the transition clutter environment in gaussian sum complex Gaussian therebetween is adapted to, The adaptive performance to actual clutter non-gaussian degree space-time gradually changeable is embodied, with very strong generalization ability, can effectively be carried Rise detection performance of the wideband radar under complex clutter environment.

Brief description of the drawings

Fig. 1 is the work(of the range extension target self-adapting detecting method based on variable element generalized structure proposed by the invention Can module map.In Fig. 1,1. covariance matrix merges estimation module, and 2. range extension target fusion detection device builds module, 3. examines Survey judging module.

Specific embodiment

The invention will be further described below in conjunction with the accompanying drawings.The embodiment of the present invention is used for illustrating the present invention, without It is to limit the invention, in the protection domain of spirit and claims of the present invention, what the present invention was made any repaiies Change and change, both fall within protection scope of the present invention.

With reference to Figure of description 1, specific embodiment of the invention is divided into following steps：

Step 1 carry out radar illumination without target zone around detection zone by treating, and obtains adjacent with unit to be detected The near pure clutter assistance data not comprising target, constitutes R helper data vectors x_m, m=1,2 ..., R, wherein, x_m, m=1, 2 ..., R are the vector of N × 1 dimension, and N represents that radar receives the product of array number and Coherent processing umber of pulse；And by reference data to Amount delivers to covariance matrix and merges estimation module (1), and the fusion estimate of unknown covariance matrix is calculated according to formula (1)And WillDeliver to range extension target fusion detection device and build module (2)；

It is worth noting that, work as α=0, the fusion estimate of the unknown covariance matrix calculated according to formula (1)Represent For

From formula (3), estimate is merged when α=0Exactly sample covariance matrix, illustrates covariance of sampling Matrix is fusion estimateSpecial case at α=0, it is shown that the fusion estimation procedure shown in formula (1) has very strong pervasive Property；

And when α=1, the iterative process that formula (1) is represented is reduced to

From formula (4), estimate is merged when α=1Iterative process and near-maximum-likelihood estimated matrix repeatedly It is completely the same for process, illustrate that near-maximum-likelihood estimated matrix is fusion estimateSpecial case at α=1, further shows Show that the fusion estimation procedure shown in formula (1) has very strong universality；

Especially when α=1 and K=1, the fusion estimate of the unknown covariance matrix calculated according to formula (1)It is expressed as

From formula (5), estimate is merged when α=1 and K=1Sample covariance matrix is exactly normalized, is said Bright normalization sample covariance matrix is fusion estimateSpecial case at α=1 and K=1, it is shown that melting shown in formula (1) Closing estimation procedure has very strong universality；

Final iterations K in formula (1) can be adjusted according to covariance matrix required precision and operand requirement Whole, if covariance matrix required precision is higher, K takes higher value, if it is desired to operand is smaller, then K takes smaller value；For Non-Gaussian clutter environment, when K >=3, merges estimatePreferable estimated accuracy can be obtained, algorithm operation quantity is considered Requirement with detector design to covariance matrix precision, takes iteration result the melting as unknown covariance matrix of K=3 Close estimate；

In general, the fusion method of estimation of covariance matrix can be estimated by unified covariance matrix fusion in the present invention Meter framework, covers the spies such as existing sample covariance matrix, normalization sample covariance matrix, near-maximum-likelihood estimated matrix Determine optimal or suboptimal estimation method under clutter background, and can by adjust final iterations make covariance matrix precision with Optimization balance is reached between operand requirement；

Step 2 constitutes L data to be tested vector z by the L multiple amplitude of range cell echo to be detected_t, t=1,2 ..., L, and data to be tested vector is delivered into range extension target fusion detection device structure module (2)；In range extension target fusion Detector is built in module (2), and calculating the range extension target self-adapting detecting based on variable element generalized structure according to formula (2) unites Metering λ_α, and according to the non-gaussian degree setting ratio factor-alpha of actual clutter, the non-gaussian degree of clutter is bigger, then the value of α It is bigger；By λ_αDeliver to detection judging module (3)；

It is worth noting that, when background clutter distribution it is Gaussian stronger when, the value of α is smaller, is carried on the back in extreme Gauss Under scape, α=0 is taken, the fusion estimate for now being calculated by formula (1)It is sample covariance matrix, and the inspection calculated by formula (2) Survey statistic λ_αThe broad sense adaptive matched filter based on sample covariance matrix is equivalent to, can be obtained under Gaussian Clutter background Optimal detection performance, illustrate that the broad sense adaptive matched filter based on sample covariance matrix is inspection proposed by the invention Special case of the survey method at α=0；And when the non-Gaussian system of background clutter distribution is stronger, the value of α is larger, is answered in extreme Close under the non-gaussian background of Gaussian Profile, take α=1, the fusion estimate for now being calculated by formula (1)It is near-maximum-likelihood Estimated matrix, and the detection statistic λ calculated by formula (2)_αIt is equivalent to the broad sense based on near-maximum-likelihood estimated matrix adaptive Integration detection device is answered, illustrates that broad sense self adaptation integration detection device is the special case of detection method proposed by the invention at α=1；

In sum, detection method proposed by the invention can be for the space-time of clutter non-gaussian degree in actual environment gradually Denaturation, rationally utilizes clutter non-gaussian characteristic information, and the general character based on existing optimal detection statistic under specific clutter environment is special Levy, range extension target adaptive detector is constructed using succinct product form, transition is realized by single parameter Range extension target detection device designs the Synchronization Control with corresponding clutter covariance matrix method of estimation under clutter environment, and can root Factually the non-gaussian degree self-adaptative adjustment parameter of border clutter distribution, makes range extension target detection device structure and clutter covariance Matrix estimation method is adapted to the change of actual clutter environment simultaneously, with very strong generalization ability；In addition, the distance for being proposed Extension target detection device structure can be compatible and covers optimal or suboptimum clutter association under the specific clutter background such as gaussian sum complex Gaussian Variance matrix method of estimation and corresponding range extension target detection device, and adapt in gaussian sum complex Gaussian therebetween Transition clutter environment, embodies the adaptive performance to actual clutter non-gaussian degree space-time gradually changeable, can effectively lift broadband Detection performance of the radar under complex clutter environment；

Step 3 carries out range extension target detection and adjudicates and export testing result in detection judging module (3), to keep The constant false alarm rate characteristic of detection method, detection threshold T is set according to default false-alarm probability；By data to be tested vector z_t, t= 1,2 ..., the self-adapting detecting statistic λ of L_αIt is compared with thresholding T, if λ_α>=T, then judge that L range cell to be detected is deposited In range extension target, echo vector z_t, t=1,2 ..., assistance datas of the L not as other range cells；If otherwise λ_α<T, Then judge that L range cell to be detected does not exist range extension target, echo vector z_t, t=1,2 ..., L as it is follow-up other The assistance data of range cell.

Claims (3)

1. the range extension target self-adapting detecting method of variable element generalized structure is based on, it is characterised in that comprised the following steps：

Step 1 obtains data to be tested vector from L range cell to be detected, from the R neighbouring with detected unit without target Range cell obtains pure clutter assistance data, and the fusion estimate of unknown covariance matrix is calculated using helper data vectorsCover existing sample covariance matrix, normalization sample covariance matrix, near-maximum-likelihood by adjusting parameter α to estimate Optimal or suboptimal estimation method under the specific clutter backgrounds such as meter matrix so that it is miscellaneous that covariance matrix result is adaptive to transition The change of the non-gaussian degree of ripple environment；

Step 2 is based on the fusion estimate of unknown covariance matrixBroad sense adaptive matched filter is merged by product form The general character part of device and broad sense self adaptation integration detection device in detection statistic, builds the distance based on variable element generalized structure Extension objective self-adapting detection statistic λ_α, by single parameter α to self adaptation range extension target detection device structure and accordingly Clutter covariance matrix is estimated to realize synchronous adjustment so that range extension target detection device structure adaptive is in transition clutter environment Non-gaussian degree change；

Step 3 is the constant false alarm rate characteristic of holding detection method, and detection threshold T is set according to default false-alarm probability；Will be to be checked Survey data vector z_t, t=1,2 ..., the self-adapting detecting statistic λ of L_αIt is compared with thresholding T, if λ_α>=T, then judge L There is range extension target, echo vector z in range cell to be detected_t, t=1,2 ..., auxiliary of the L not as other range cells Data；If otherwise λ_α<T, then judge that L range cell to be detected does not exist range extension target, echo vector z_t, t=1, 2 ..., L as follow-up other range cells assistance data.

2. the range extension target self-adapting detecting method based on variable element generalized structure according to claim 1, it is special Levy and be, the step 1 is specially：

L data to be tested vector z is constituted by the L multiple amplitude of range cell echo to be detected_t, t=1,2 ..., L, wherein, L is Natural number more than 1, centered on range cell to be detected, continuously takes before and after it a number of not comprising target respectively The multiple amplitude of range cell echo, constitutes R pure clutter helper data vectors x_m, m=1,2 ..., R, wherein, z_tAnd x_mIt is N × 1 The vector of dimension, N represents that radar receives the product of array number and Coherent processing umber of pulse；

The fusion estimate of unknown covariance matrixWhereinRealized by following iterative process

${\hat{M}}_{\mathrm{\&alpha;}}^{(k+1)}=\frac{1}{R}\underset{m=1}{\overset{R}{\&Sigma;}}\frac{x_m{x}_m^H}{{\&lsqb;x_m^H{(N\&CenterDot;{\hat{M}}_{\mathrm{\&alpha;}}^{\left(k\right)})}^{-1}{x}_m\&rsqb;}^{\mathrm{\&alpha;}}}---\left(1\right)$

In above formula, the conjugate transposition computing of subscript H representing matrixs, the inversion operation of the representing matrix of subscript -1, k=0,1 ..., K- 1；K represents final iterations, is configured according to covariance matrix required precision and operand requirement, if covariance Matrix Estimation required precision is higher, then K takes higher value, if it is desired to operand is smaller, then K takes smaller value；Represent given ginseng Kth time iterative estimate matrix during number α；To strengthen the universality of estimation procedure, the initialization matrix of formula (1)Using unit Battle array, and final iteration result is the fusion estimate of unknown covariance matrixParameter alpha meets 0≤α≤1, and Non-gaussian degree setting according to actual clutter, enhances the generalization ability of subsequent detectors, wherein the non-gaussian degree of clutter Bigger, then the value of α is bigger；When α=0It is equivalent to sample covariance matrix；When α=1It is equivalent to approximate maximum Possibility predication matrix；When α=1 and K=1It is equivalent to normalize sample covariance matrix.

3. the range extension target self-adapting detecting method based on variable element generalized structure according to claim 1, it is special Levy and be, the step 2 is specially：

Using the fusion estimate of unknown covariance matrixWith L data to be tested vector z_t, t=1,2 ..., L, based on change The range extension target self-adapting detecting statistic λ of Parameter Generalized structure_αIt is expressed as

${\mathrm{\&lambda;}}_{\mathrm{\&alpha;}}=-\underset{t=1}{\overset{L}{\&Sigma;}}ln\&lsqb;1-\frac{|p^H{\hat{M}}_{\mathrm{\&alpha;}}^{-1}{z}_t{|}^2}{\left(p^H{\hat{M}}_{\mathrm{\&alpha;}}^{-1}p\right){\left(z_t^H{\hat{M}}_{\mathrm{\&alpha;}}^{-1}{z}_t\right)}^{\mathrm{\&alpha;}}}\&rsqb;---\left(2\right)$

In above formula, p represents known space-time steering vector, is a unit vector for N × 1 dimension, is worked according to radar system and joined Number determines；Formula (2) realizes synchronous adjustment with parameter alpha in formula (1), detection statistic is adaptive to the not high of transition clutter environment This degree, can cover the optimal or sub-optimum detectors under existing gaussian sum Compound-Gaussian Clutter background；When α=0, formula (2) Detection statistic λ_αThe broad sense adaptive matched filter based on sample covariance matrix is equivalent to, Gaussian Clutter background can be obtained Under optimal detection performance；When α=1, the detection statistic λ of formula (2)_αIt is equivalent to based on near-maximum-likelihood estimated matrix Broad sense self adaptation integration detection device, is suitable for range extension target self adaptation inspection under the stronger complex Gaussian background of non-Gaussian system Survey, and can be according to covariance matrix required precision and the final iterations K of operand requirement adjustment.