CN103995258B - Radar target self-adaptation fusion detection method under complex clutter peripheral surroundings - Google Patents

Abstract

The invention discloses radar target self-adaptation fusion detection method under a kind of complex clutter peripheral surroundings, belong to radar signal processing field.Under complex clutter environment before and after clutter edge may appear at simultaneously in sliding window, the deficiencies such as existing CFAR detection method false alarm control capability is poor in order to overcome, detection perform declines, operand is excessive, the present invention proposes radar target self-adaptation fusion detection method under a kind of complex clutter peripheral surroundings; The method carries out homogeneity judgement based on neighboring reference data difference, the change of clutter edge number and position can be adaptive to, judge clutter edge and range unit relative position to be detected, filter out the homogeneous reference data that can represent range unit clutter background to be detected, and merge formation detection threshold; While maintenance constant false alarm rate characteristic, effectively improve the false alarm control capability of radar in complex clutter peripheral surroundings and detection perform, avoid circulation or iterative operation, calculated amount is little, is convenient to Project Realization.

Description

Radar target self-adaptation fusion detection method under complex clutter peripheral surroundings

Technical field

The present invention is under the jurisdiction of radar signal processing field, is specifically related to radar target self-adaptation fusion detection method under a kind of complex clutter peripheral surroundings.

Background technology

Radar is the important tool utilizing radio-wave sounding target in military and civilian field.False-alarm problem during radar detects automatically is one of unavoidable major issue in radar system design.Constant false alarm rate (CFAR) technology is the important means controlling false alarm rate in radar automatic checkout system, plays an important role in the automatic testing process of radar.The object of CFAR design is to provide and is adapted to background and mixes and make an uproar and the detection threshold of interference variations, makes Automatic Targets process have constant false-alarm probability.

What clutter edge described is the interregional zone of transition situation of detection background different qualities, and the exemplary of this situation is the edge of rainfall area, land, ocean intersection etc.If detecting unit is in weak clutter district, and be in strong clutter district with reference to some other reference unit in sliding window, even if so signal to noise ratio (S/N ratio) also can produce coating effect to target detection very greatly, detection probability and false-alarm probability all can decline.If detecting unit is in strong clutter district, and some other reference unit is in weak clutter district, so false-alarm probability can sharply rise.This problem is the major issue should considered in Radar Design, and classical cell-average (CA)-CFAR detecting device detection perform under clutter edge environment sharply declines and false alarm rate too rises.As the amendment scheme of CA-CFAR detecting device, large (the GO)-CFAR detection method of choosing is specially for the solution of clutter edge situation, GO-CFAR method slides the greater in window two partial estimation before and after getting, as the clutter power horizontal estimated that detecting device is total, the situation that clutter edge appears at single sliding window can be successfully managed; But when clutter local heterogeneity increases, before and after occurring most probably all there are the complex situations of clutter edge in sliding window, and now the false alarm control capability of GO-CFAR method declines, and cannot meet CFAR testing requirement.In addition, Ordered Statistic (OS) class CFAR detection method is based on the ordered sequence of reference sample, unknown clutter power intensity is estimated by selecting representational reference data, effectively improve the multi-sources distinguishing ability of detection method, but the false alarm control capability of these class methods in clutter edge is still difficult to meet CFAR testing requirement.In addition, current most of self-adaptation CFAR detection method selects representative reference data by loop iteration deleting technique, the complex loops wherein related to or iterative operation result in huge calculated amount, constrain the effective application of this kind of self-adaptation CFAR detection method under complex clutter edge background.

In practical application, before and after clutter edge may appear at simultaneously in sliding window.For the average power level obtaining range unit clutter background to be detected is estimated, the relative position of clutter edge and range unit to be detected need being determined, selecting the reference unit that can represent range unit clutter background to be detected, for calculating detection threshold.Because clutter edge may appear in the sliding window in front and back simultaneously, and its locus is also unknown, therefore, accurate judgement clutter edge position, available information in effective integration reference data, simplicity of design efficient adaptive CFAR detection method, makes it to be adaptive to clutter edge change, has important practical significance.

Summary of the invention

1. the technical matters that will solve

Under complex clutter peripheral surroundings, poor in order to overcome existing CFAR detection method false alarm control capability, detection perform declines, the problems such as operand is excessive, the present invention is based on neighboring reference data difference and carry out homogeneity judgement, filter out and can represent range unit to be detected and to mix the homogeneous reference data of background of making an uproar, and merge formation detection threshold, radar target self-adaptation fusion detection method under proposition complex clutter peripheral surroundings, accurately judging on the basis of clutter edge and range unit relative position to be detected, improve the false alarm control capability of radar in complex clutter peripheral surroundings and detection perform.

2. technical scheme

Under complex clutter peripheral surroundings of the present invention, radar target self-adaptation fusion detection method comprises following technical measures:

Step 1 obtains reference data, carries out variance evaluation, calculates neighboring reference data difference

Centered by range unit observed reading x to be detected, before and after it, get the radar return video observing measured value of M range unit respectively, form 2M reference data (x _m, m=1,2 ..., 2M);

Calculate the estimate of variance of all 2M reference data namely

${\widehat{\mathrm{\σ}}}^2=\frac{1}{2M}\underset{m=1}{\overset{2M}{\Σ}}{(x_m-\frac{1}{2M}\underset{k=1}{\overset{2M}{\Σ}}{x}_k)}^2---\left(1\right)$

Calculate the difference (y of neighbor distance elements reference data _m, m=2,3 ..., 2M), namely

y _m＝x _m-x _m-1,m＝2,3,…,2M(2)

Step 2 carries out the judgement of reference data homogeneity according to neighboring reference data difference, determines clutter edge position

Find out the maximal value of the absolute value of all 2M-1 difference, be labeled as p by under the difference of correspondence, namely

$p=\arg \underset{m}{max}\left(\right|y_m\left|\right),m=2,3,...,2M---\left(3\right)$

In above formula, || represent and take absolute value, the value of m when representing that in bracket, parameter value is maximum;

(1) if maximum absolute difference y _pmeet

$\left|y_p\right|<2\mathrm{\&beta;}\widehat{\mathrm{\&sigma;}}---\left(4\right)$

Then show to there is not clutter edge in 2M reference data, perform step 3; In formula (4), β is regulation coefficient, can set according to actual conditions, generally need meet β >=1;

If y _pdo not meet formula (4), then with subscript p for boundary, be divided into two groups of (x with reference to data _m, m=1 ..., p-1 and x _k, k=p ..., 2M), and respectively data consistency judgement is carried out to two groups of reference datas according to following step (2);

(2) data consistency judges

For the data sequence (z that length is L _l, l=1 ..., L), consistency check statistic V _lcan be expressed as

$V_L=L\underset{l=1}{\overset{L}{\&Sigma;}}{z}_l^2/{\left(\underset{l=1}{\overset{L}{\&Sigma;}}{z}_l\right)}^2---\left(5\right)$

In general, V _lless, represent that the consistance of data sequence is better;

If V _lmeet

V _l<H _l(6) then decision data sequence has consistance; Otherwise decision data sequence is inconsistent; In formula (6), H _lrepresent consistance decision threshold, can set according to data sequence length L;

(3) if two groups of reference data (x _m, m=1 ..., p-1 and x _k, k=p ..., 2M) all there is consistance, then show that 2M reference data can be divided into uniform two groups, perform step 3, otherwise perform step (4);

(4) to all difference data (y _m, m=2,3 ..., 2M) divide into groups by numerical value is positive and negative, will with y _pthe difference data set that symbol is contrary is designated as S _p, namely

S _p＝{y _m|y _my _p<0,m＝2,3,…,2M}(7)

Find out data acquisition S _pthe difference of middle maximum absolute value, is labeled as q by under the difference of correspondence, namely

$q=\arg \underset{m}{max}\left(\right|y_m\left|\right),y_m\&Element;S_p---\left(8\right)$

Smaller in p and q and the greater are designated as a and b respectively, namely

a＝min(p,q)(9)

b＝max(p,q)(10)

In formula, min () and max () represents respectively and gets minimum value and maximal value; With subscript a and subscript b for boundary, 2M reference data can be divided into uniform three groups, perform step 3;

Step 3, according to the homogeneity court verdict of step 2, determines the homogeneous reference data of screening

The homogeneous reference data of all N number of screenings are designated as X _m, m=1 ..., N;

If judgement 2M reference data (x in step (1) _m, m=1,2 ..., 2M) in there is not clutter edge, then all 2M reference data is considered to uniform, setting

N＝2M(11)

X _m＝x _m(12)

Otherwise, if judgement 2M reference data is with subscript p for boundary is divided into uniform two groups in step (3), then determine the homogeneous reference data of screening according to the magnitude relationship of p and M;

If

M<p-1(13)

Then by front p-1 reference data (x _m, m=1 ..., p-1) and as the homogeneous reference data of screening, namely

N＝p-1(14)

X _m＝x _m(15)

Otherwise the result according to step (4) determines the homogeneous reference data of screening;

If

M<a-1(16)

Then by front a-1 reference data (x _m, m=1 ..., a-1) and as the homogeneous reference data of screening, namely

N＝a-1(17)

X _m＝x _m(18)

Otherwise, if

M≥b(19)

Then by rear 2M-b+1 reference data (x _m, m=b ..., 2M) and as the homogeneous reference data of screening, namely

N＝2M-b+1(20)

X _m＝x _m+b-1(21)

Otherwise, by the b-a of centre reference data (x _m, m=a ..., b-1) and as the homogeneous reference data of screening, namely

N＝b-a(22)

X _m＝x _m+a-1(23)

Step 4 forms detection threshold according to the homogeneous reference data fusion of screening, and carries out adjudicating with presence or absence of target

Utilize the homogeneous reference data (X of screening _m, m=1 ..., N) merge, estimate the assorted mean intensity Z that makes an uproar, namely

$Z=\underset{m=1}{\overset{N}{\&Sigma;}}{X}_m/N---\left(24\right)$

Detection threshold T is set, namely based on the assorted strength estimations Z that makes an uproar

In T=α Z (25) above formula, threshold factor α sets according to default false-alarm probability, guarantees the CFAR characteristic of detection method;

Finally, treat detecting distance unit observed reading x and carry out detection judgement, if x >=T, then judge that unit to be detected exists target; Otherwise, if x<T, then judge that unit to be detected does not exist target.

3. beneficial effect

Compared with background technology, the invention has the beneficial effects as follows: slide the complex clutter environment of window before and after clutter edge may appear at simultaneously under, the inventive method can be adaptive to the change of clutter edge number and position, filter out the homogeneous reference data that can represent range unit clutter background to be detected, improve the false alarm control capability of radar in complex clutter peripheral surroundings and target detection performance, avoid circulation or iterative operation, calculated amount is little, is convenient to Project Realization.

Accompanying drawing explanation

Accompanying drawing is the functional block diagram of detection method proposed by the invention.In figure, 1. variance and difference calculating module, 2. clutter edge position estimation module, 3. reference data screening module, 4. detection threshold computing module, 5. detects judging module.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described.The embodiment of the present invention is used for explaining and the present invention is described, instead of limits the invention, and in the protection domain of spirit of the present invention and claim, any amendment make the present invention and change, all fall into protection scope of the present invention.

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

Step 1 obtains reference data, carries out variance evaluation, calculates neighboring reference data difference

The signal that radar antenna receives through amplifying and after mixing, carrying out energy accumulation, then obtain radar return vision signal through wave detector by matched filter, and carries out sampling by Range resolution unit and obtains radar return video observing measured value; Centered by range unit observed reading x to be detected, before and after it, get the radar return video observing measured value of M range unit respectively, form 2M reference data (x _m, m=1,2 ..., 2M), and send into variance and difference calculating module (1), the variance of all 2M reference data is estimated according to formula (1) and the difference (y of neighbor distance elements reference data is calculated according to formula (2) _m, m=2,3 ..., 2M);

Step 2 is by the variance of step 1 gained with difference data (y _m, m=2,3,, 2M) and send into clutter edge position estimation module (2), carry out the judgement of reference data homogeneity according to neighboring reference data difference, determine clutter edge position, specifically adopt following steps (1) to step (4);

(1) the maximal value y of the absolute value of difference data is obtained according to formula (3) _p, according to the regulation coefficient β in actual conditions setting formula (4), if y _pmeet the condition of formula (4), then perform step 3;

(2) if y _pdo not meet the condition of formula (4), then with subscript p for boundary, be divided into two groups of (x with reference to data _m, m=1 ..., p-1 and x _k, k=p ..., 2M), and respectively data consistency judgement is carried out to two groups of reference datas according to formula (5) and formula (6);

(3) if two groups of reference data (x _m, m=1 ..., p-1 and x _k, k=p ..., 2M) all there is consistance, then perform step 3, otherwise perform step (4);

(4) obtain and y according to formula (7) _pthe difference data S set that symbol is contrary _p, find out S set according to formula (8) _pthe subscript q that the difference of middle maximum absolute value is corresponding, according to formula (9) and formula (10) acquisition subscript a and subscript b, and performs step 3;

The homogeneity court verdict of step 2 is sent into reference data screening module (3) by step 3, filters out N number of homogeneous reference data (X according to formula (11) to formula (23) _m, m=1 ..., N);

Homogeneous reference data (the X that step 3 is screened by step 4 _m, m=1 ..., N) and send into detection threshold computing module (4), merge according to formula (24) and form the assorted mean intensity estimation Z that makes an uproar, and according to formula (25) setting detection threshold T; Finally detection threshold T is sent into and detect judging module (5), detection judgement is carried out and output detections result, if x>=T to data to be tested x, then judge that unit to be detected exists target, otherwise, if x<T, then judge that unit to be detected does not exist target.

Claims (1)

1. radar target self-adaptation fusion detection method under complex clutter peripheral surroundings, is characterized in that comprising the following steps:

Step 1 obtains reference data, carries out variance evaluation, calculates neighboring reference data difference

Centered by range unit observed reading x to be detected, before and after it, get the radar return video observing measured value of M range unit respectively, form 2M reference data x _m, wherein m=1,2 ..., 2M;

Calculate the estimate of variance of all 2M reference data namely

${\widehat{\mathrm{\&sigma;}}}^2=\frac{1}{2M}\underset{m=1}{\overset{2M}{\&Sigma;}}{(x_m-\frac{1}{2M}\underset{k=1}{\overset{2M}{\&Sigma;}}{x}_k)}^2---\left(1\right)$

Calculate the difference y of neighbor distance elements reference data _m, wherein m=2,3 ..., 2M, namely

y _m＝x _m-x _m-1,m＝2,3,…,2M(2)

Step 2 carries out the judgement of reference data homogeneity according to neighboring reference data difference, determines clutter edge position

Find out the maximal value of the absolute value of all 2M-1 difference, be labeled as p by under the difference of correspondence, namely

$p=\arg \underset{m}{max}\left(\right|y_m\left|\right),m=2,3,...,2M---\left(3\right)$

In above formula, || represent and take absolute value, the value of m when representing that in bracket, parameter value is maximum;

(1) if maximum absolute difference y _pmeet

$\left|y_p\right|<2\mathrm{\&beta;}\widehat{\mathrm{\&sigma;}}---\left(4\right)$

Then show to there is not clutter edge in 2M reference data, perform step 3; In formula (4), β is regulation coefficient, can set according to actual conditions, need meet β >=1;

If y _pdo not meet formula (4), then with subscript p for boundary, be divided into x with reference to data _mand x _ktwo groups, wherein m=1 ..., p-1 and k=p ..., 2M, and respectively data consistency judgement is carried out to two groups of reference datas according to following step (2);

(2) data consistency judges

Be the data sequence z of L for length _l, wherein l=1 ..., L, consistency check statistic V _lcan be expressed as

$V_L=L\underset{l=1}{\overset{L}{\&Sigma;}}{z}_l^2/{\left(\underset{l=1}{\overset{L}{\&Sigma;}}{z}_l\right)}^2---\left(5\right)$

Wherein, V _lless, represent that the consistance of data sequence is better;

If V _lmeet

V _L<H _L(6)

Then decision data sequence has consistance; Otherwise decision data sequence is inconsistent; In formula (6), H _lrepresent consistance decision threshold, can set according to data sequence length L;

(3) if two groups of reference data x _mand x _kall there is consistance, wherein m=1 ..., p-1 and k=p ..., 2M, then show that 2M reference data can be divided into uniform two groups, performs step 3, otherwise perform step (4);

(4) to all difference data y _mdivide into groups by numerical value is positive and negative, wherein m=2,3 ..., 2M, and will with y _pthe difference data set that symbol is contrary is designated as S _p, namely

S _p＝{y _m|y _my _p<0,m＝2,3,…,2M}(7)

Find out data acquisition S _pthe difference of middle maximum absolute value, is labeled as q by under the difference of correspondence, namely

$q=\arg \underset{m}{max}\left(y_m\right|),y_m\&Element;S_p---\left(8\right)$

Smaller in p and q and the greater are designated as a and b respectively, namely

a＝min(p,q)(9)

b＝max(p,q)(10)

In formula, min () and max () represents respectively and gets minimum value and maximal value; With subscript a and subscript b for boundary, 2M reference data can be divided into uniform three groups, perform step 3;

Step 3, according to the homogeneity court verdict of step 2, determines the homogeneous reference data of screening

The homogeneous reference data of all N number of screenings are designated as X _m, m=1 ..., N;

If judgement 2M reference data x in step (1) _min there is not clutter edge, wherein m=1,2 ..., 2M, then all 2M reference data is considered to uniform, setting

N＝2M(11)

X _m＝x _m(12)

Otherwise, if judgement 2M reference data is with subscript p for boundary is divided into uniform two groups in step (3), then determine the homogeneous reference data of screening according to the magnitude relationship of p and M;

If

M<p-1(13)

Then by a front p-1 reference data x _mas the homogeneous reference data of screening, wherein m=1 ..., p-1, namely

N＝p-1(14)

X _m＝x _m(15)

Otherwise the result according to step (4) determines the homogeneous reference data of screening;

If

M<a-1(16)

Then by a front a-1 reference data x _mas the homogeneous reference data of screening, wherein m=1 ..., a-1, namely

N＝a-1(17)

X _m＝x _m(18)

Otherwise, if

M≥b(19)

Then by a rear 2M-b+1 reference data x _mas the homogeneous reference data of screening, wherein m=b ..., 2M, namely

N＝2M-b+1(20)

X _m＝x _m+b-1(21)

Otherwise, by the b-a of a centre reference data x _mas the homogeneous reference data of screening, wherein m=a ..., b-1, namely

N＝b-a(22)

X _m＝x _m+a-1(23)

Step 4 forms detection threshold according to the homogeneous reference data fusion of screening, and carries out adjudicating with presence or absence of target

Utilize the homogeneous reference data X of screening _mmerge, wherein m=1 ..., N, estimates the assorted strength estimations Z that makes an uproar, namely

$Z=\underset{m=1}{\overset{N}{\&Sigma;}}{X}_m/N---\left(24\right)$

Detection threshold T is set, namely based on the assorted strength estimations Z that makes an uproar

T＝αZ(25)

In above formula, threshold factor α sets according to default false-alarm probability, guarantees the CFAR characteristic of detection method;

Finally, treat detecting distance unit observed reading x and carry out detection judgement, if x >=T, then judge that unit to be detected exists target; Otherwise, if x<T, then judge that unit to be detected does not exist target.