CN102495408A - Method for automatically searching point of synthetic aperture radar dot-matrix target image data - Google Patents

Abstract

The invention discloses a method for automatically searching a point of synthetic aperture radar dot-matrix target image data. The method comprises the following steps of: 1, correcting an antenna directional diagram weighing factor; 2, correcting radar receiving power; 3, searching a bidimensional maximum value; 4, positioning a pixel position of a point target; 5, checking the position; and 6, performing bidimensional frequency domain interpolation, and positioning a precise position of the point target. Automatic point search for the synthetic aperture radar dot-matrix target image data is realized, so that the working quantity of manual point search operation is reduced; the non-uniformity of the amplitude of the point target caused by reasons, such as antenna weighing, is overcome; the point search accuracy is high; the precise position of a dot-matrix target in the synthetic aperture radar image data can be obtained, and is used for precisely positioning a calibrator; the positioning precision of the target position is adjusted by controlling the interpolation magnification; automatic estimation for the dot-matrix target in the synthetic aperture radar image data is realized; an average performance index of the image data is obtained; and an actual performance index of a system can be reflected.

Description

A kind of synthetic-aperture radar dot matrix destination image data seek point methods automatically

Technical field

The invention belongs to the signal processing technology field, what be specifically related to a kind of synthetic-aperture radar dot matrix destination image data seeks point methods automatically.

Background technology

(Synthetic Aperture Radar SAR) is a kind of round-the-clock, round-the-clock active Information Acquisition System to synthetic-aperture radar.The SAR system transmits and receives mode through gating pulse, obtains SAR original echo data.These data along the carrier heading be the orientation to, perpendicular to the orientation to direction be distance to.Through SAR original echo data are carried out data processing, obtain visual radar image.

Be different from optical imagery, typical point target is not simple impulse function in the SAR view data.Point target shows as the main lobe with certain broadening, and around it, is accompanied by a large amount of secondary lobes.The patent " self-adaption two-dimensional interpolation method that is used for synthetic aperture radar point target imaging quality assessment " of applications such as the article " double-base SAR point target image quality measure algorithm " that the patent of applications such as Sun Bing " a kind of point target resolution appraisal procedure of Squint SAR ", Ceng Tao etc. deliver, Yu Ze utilizes this characteristics of point target to measure just.

Because the influence of antenna gain, radar received power; The a single point target is assessed the actual performance that the image property index of gained can not the complete reaction system; Need in a scene, arrange the dot matrix target; And ask for the performance index of this each point target of scene, obtain the average image performance index.Existing method selected element target adopts the pattern of man-machine interaction, and only the dot matrix of 16 * 16 sizes just needs manually to select 256 times complex operation.And existing method only can provide the Pixel-level position of point target in the SAR image.

But there are many difficult points in the automatic searching technology of point target.Secondary lobe has and point target main lobe character much at one, and when point target quantity increases, the shape of secondary lobe will be more complicated.Antenna radiation pattern and radar received power make can make hard intensity point target side-lobe energy surpass weak intensity point target main lobe by SAR view data mid point target strength skewness, further causes the difficulty of differentiating point target.

Summary of the invention

What the present invention proposed a kind of synthetic-aperture radar dot matrix destination image data seeks point methods automatically; This method is to dot matrix target synthetic-aperture radar complex pattern data; The weighting influence of compensation antenna radiation pattern and radar received power; Automatically seek the point target Pixel-level position in these data, utilize interpolation to obtain the exact position of point target at last.The precision that this method is sought a little is higher than pixel precision, helps the accurate geographic location of point target, can realize the automated quality assessment of synthetic-aperture radar dot matrix target complex pattern data, obtains the average behavior index of image.

A kind of synthetic-aperture radar dot matrix destination image data seek point methods automatically, comprise following step:

Step 1: antenna radiation pattern weighted correction;

Initialization is also read SAR image complex data, calculates the pairing antenna radiation pattern weighting factor of each range gate, obtains realizing the antenna radiation pattern weighted correction to SAR image complex data along the antenna compensation vector of distance to distribution.

Step 2: the radar received power is proofreaied and correct;

Obtain the corresponding radar power equation of each range gate, to through image complex data S ' edge of antenna radiation pattern weighted correction apart to compensating, obtain the uniform SAR image of dot matrix target amplitude complex data S ".

Step 3: seek two-dimentional maximum point;

Ask for the uniform SAR image of dot matrix target amplitude complex data S " in two-dimentional maximum point, make the point target position sought all be contained among the set K.Remove then set K middle distance to or the incomplete extreme point of azimuth accumulation.

Step 4: point target location of pixels location;

Utilize sensitive factor to judge, among two-dimentional maximum value set K, extract the position of dot matrix target, form new dot matrix goal set E, the Pixel-level position of point target has been preserved in this set.

Step 5: location verification;

Whether the point target among the checking dot matrix goal set E satisfies the typical characteristics of SAR system imaging target, if do not satisfy then turn back to step 4, carries out location, point target exact position if meet then get into step 6.

Step 6: two-dimensional frequency interpolation, location, point target exact position.

Pixel-level position with point target among the dot matrix goal set E is the center, in original SAR image complex data, gets on every side the evaluating matrix L that 32 * 32 matrix element constitutes this point target.Evaluating matrix L is carried out the two-dimensional frequency interpolation, select the interpolation multiple, obtain the exact position of point target to positioning accuracy request.

The invention has the advantages that:

(1) realizes seeking automatically a little of synthetic-aperture radar dot matrix destination image data, reduced artificial workload of seeking point operation;

(2) overcome the non-homogeneous characteristic of point target amplitude that causes owing to reasons such as antenna weightings, sought an accuracy rate height;

(3) can obtain the exact position of dot matrix target in the diameter radar image data, can be used for scaler and accurately locate, through control interpolation multiple, the bearing accuracy of scalable target location;

(4) can be used for realizing that diameter radar image data dot matrix target assesses automatically, obtain the average behavior index of view data, more can reflect the actual performance index of system.

Description of drawings

Fig. 1 is the independent positioning method process flow diagram of seeking of the present invention;

Fig. 2 is bearing accuracy of the present invention and interpolation multiple graph of a relation.

Embodiment

To combine accompanying drawing and embodiment that the present invention is done further detailed description below.

The present invention be a kind of synthetic-aperture radar dot matrix destination image data seek point methods automatically, processing to as if synthetic-aperture radar dot matrix target complex pattern data, the result who obtains is the exact position of each point target in data.Synthetic-aperture radar dot matrix target complex pattern data are that a two-dimentional complex matrix

size is M * N.Wherein M representes the orientation to pulse number, and N representes that distance is to sampling number.The present invention has at first compensated non-homogeneousization of point target amplitude that the complex pattern data cause apart from the antenna radiation pattern weighted sum radar received power that makes progress; Constitute two-dimentional maximum point object set again; Utilize iteration adjustment sensitive factor to remove secondary lobe point and form Pixel-level point target collection in two-dimentional maximum point target tightening; At last two-dimensional interpolation is carried out in Pixel-level point target position, obtain the exact position of point target in the diameter radar image data.

The flow process of method is as shown in Figure 1, comprises following step:

Step 1: antenna radiation pattern weighted correction;

Initialization is also read SAR image complex data, calculates the pairing visual angle of each range gate, obtains this visual angle corresponding antenna compensating factor, forms constructing antennas directional diagram correction matrix, realizes the antenna radiation pattern weighted correction to SAR image complex data.

Be specially:

Obtain each range gate j (j=1,2,3...N) pairing oblique distance distance R _j:

$R_j=R_1+\frac{(j-1)\·C}{2f_s}$

Wherein: R ₁The corresponding oblique distance distance of this point during for first sampled point (being j=1), C represents the light velocity, f _sExpression distance to sampling rate.

Obtain this oblique distance distance R _jCorresponding view angle theta _j:

θ _j＝arccos(R _ref/H)-arccos(R _j/H)

Wherein: R _RefThe corresponding oblique distance of representative antennas sight line center irradiation, H is an aircraft altitude.

Obtain this view angle theta _jPairing antenna compensating factor a _jFor:

$a_j=\frac{\mathrm{\π}{D}_r\sin \left({\mathrm{\θ}}_j\right)/\mathrm{\λ}}{\sin (\mathrm{\π}{D}_r\sin \left({\mathrm{\θ}}_j\right)/\mathrm{\λ})}$

Wherein: D _rFor the distance to antenna length, λ is the wavelength that transmits.

Repeat above step, obtain the compensating factor a of each distance to sampled point corresponding antenna directional diagram _j(j=1,2,3...N).

Constructing antennas directional diagram correction matrix A, this matrix are the diagonal angle real number matrix of N * N, and this matrix diagonal element is antenna radiation pattern compensating factor a _j, off-diagonal element is zero, promptly

Complex pattern data matrix S and antenna radiation pattern correction matrix A carry out matrix multiple and obtain the diameter radar image complex data S ' after antenna radiation pattern is proofreaied and correct, and realize the antenna radiation pattern weighted correction to SAR image complex data.

Step 2: the radar received power is proofreaied and correct;

Obtain the corresponding radar received power compensating factor of each range gate, to through image complex data S ' edge of antenna radiation pattern weighted correction apart to compensating, obtain the uniform SAR image of dot matrix target amplitude complex data S ".

Be specially:

Obtain j (j=1,2,3...N) the pairing oblique distance of individual range gate is apart from radar received power compensating factor:

$p_j=\sqrt{4\mathrm{\π}{R_j}^4{\mathrm{\λ}}^2}$

Repeat above step, obtain each distance to sampled point corresponding antenna directional diagram compensating factor p _j(j=1,2,3...N).Structure radar received power correction matrix P, this matrix is the diagonal angle real number matrix of N * N, this matrix diagonal element is antenna radiation pattern compensating factor p _j, off-diagonal element is zero, promptly

Make data matrix S ' carry out matrix multiple with radar received power correction matrix P, to through image complex data S ' edge the distance of antenna radiation pattern weighted correction to compensating, obtain the diameter radar image complex data S after the radar received power compensates ".

Step 3: seek two-dimentional maximum point;

Ask for the uniform SAR image of dot matrix target amplitude complex data S " in two-dimentional maximum point; make the point target position sought all be contained among the two-dimentional maximum value set K, remove then two-dimentional maximum value set K middle distance to or the incomplete extreme point of azimuth accumulation.

Be specially:

The complex data matrix S " is asked mould to handle, obtained new real number matrix

Get real number matrix R _SMiddle arbitrary element r _Ij, judge whether this element satisfies r _{I, j}＞r _{I+1, j}, whether satisfy r _{I, j}＞r _{I, j+1}, whether satisfy r _{I, j}＞r _{I-1, j}, whether satisfy r _{I, j}＞r _{I, j-1}, whether satisfy r _{I, j}＞r _{I+1, j+1}, whether satisfy r _{I, j}＞r _{I-1, j-1}, whether satisfy r _{I, j}＞r _{I+1, j-1}, whether satisfy r _{I, j}＞r _{I-1, j+1}If above-mentioned condition all satisfies, then with this element at real number matrix R _SIn (i j) is recorded among the two-dimentional maximum value set K in the position.

Repeat above step, to real number matrix R _SMiddle all elements r _Ij(i=2,3,4...M-1, j=2,3,4...N-1) judge, form two-dimentional maximum value set K.

For the element among the two-dimentional maximum value set K (p, q), r _{P, q}Be R _SIn element, p, q are that this element is at R _SIn the position.

(p q), judges whether p satisfies T to get among the two-dimentional maximum value set K arbitrary element _pf _s＜p＜N-T _pf _s, judge whether q satisfies T _aPrf＜q＜M-T _aPrf, wherein T _pBe pulse width, f _sBe sampling rate, T _aBe the synthetic aperture time, prf is a pulse repetition rate.If above-mentioned condition has one not satisfy, then with this element (p q) removes in two-dimentional maximum value set K, then remove two-dimentional maximum value set K middle distance to or the incomplete extreme point of azimuth accumulation.

If two-dimentional maximum value set K is an empty set, then judge not have the dot matrix target in the complex pattern data matrix S, withdraw from institute in steps.If only there is an element in two-dimentional maximum value set K, then directly get into step 6.Two dimension maximum value set K interior element gets into step 4 more than or equal to two.

Step 4: point target Pixel-level position;

Utilize sensitive factor to judge, among two-dimentional maximum value set K, extract the position of dot matrix target, form new dot matrix goal set E, the Pixel-level position of point target has been preserved in this set.

Be specially:

Ask for real number matrix R _SThe maximum element r of middle numerical value _{X, y}=max (R _S).

Count representes number of iterations, and for the first time during iteration, count is 0, and iteration and later count will be provided by step 5 for the second time.

The definition sensitive factor does

$g_e=r_{x,y}\&CenterDot;{10}^{\frac{-13+(\mathrm{count}-1)\&CenterDot;\mathrm{<figure-callout\; id="20"\; label="d"\; filenames="HDA0000116858370000021.png"\; state="\{\{state\}\}">d</figure-callout>}}{20}}$

In the formula, d has represented the fine degree that sensitive factor changes in each iteration.The more a little bigger target localization of d value is accurate more, and locating speed is slow more.

For element among the two-dimentional maximum value set K (p, q), if r _{P, q}＞g _e, then this element position is recorded among the dot matrix object set E.

Repeat above step, handle all elements among the two-dimentional maximum value set K, obtain dot matrix object set E.For the arbitrary element among the dot matrix object set E (x, y), r _{X, y}Be R _SIn element, x, y are e _{X, y}At R _SIn the position.

If dot matrix object set E is an empty set, judge then not have the dot matrix target in the complex pattern data matrix S that institute in steps after withdrawing from.If only there is an element in dot matrix object set E, then get into step 6.If element number is more than or equal to two among the dot matrix object set E, get into step 5.

Step 5: location verification;

Whether the point target among the checking dot matrix goal set E satisfies the typical characteristics of SAR system imaging target, if do not satisfy then turn back to step 4, carries out location, point target exact position if meet then get into step 6.

Be specially:

For the arbitrary element among the dot matrix object set E (x, y):

In dot matrix object set E, seek to make | t-x| is minimum, and distance to the position be y element (t, y).If 20log10 is (r _{T, y}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6.

In dot matrix object set E, seek to make | f-y| is minimum, and the orientation to the position be x element (x, f).If 20log10 is (r _{X, f}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6.

Repeat above step, after each element among the dot matrix object set E is verified, accomplish the checking of dot matrix target location.

Element among the dot matrix object set E has write down the location of pixels of point target in the complex pattern data matrix S.

Step 6: two-dimensional frequency interpolation, location, point target exact position.

Pixel-level position with point target among the dot matrix goal set E is the center, in original SAR image complex data, gets on every side the evaluating matrix L that 32 * 32 matrix element constitutes this point target.Evaluating matrix L is carried out the two-dimensional frequency interpolation, select the interpolation multiple, obtain the exact position of point target to positioning accuracy request.

Be specially:

(x, y), the element of its correspondence in the complex pattern data matrix S is s for the arbitrary element among the dot matrix object set E _{X, y}, be the center with this element position, get this some evaluating matrix L of this point target of 32 * 32 bit matrix elements formation on every side.Promptly in the complex pattern data matrix S, get top-left position for (x-15, y-15), the position, bottom right is that (assignment is given evaluating matrix L for x+16, square formation y+16).

For the evaluating matrix L of this point target, at first the edge distance is to carrying out Fast Fourier Transform (FFT), promptly

Then along the orientation to carrying out Fast Fourier Transform (FFT), promptly

New two-dimensional matrix is carried out the edge distance carry out the zero padding operation to frequency spectrum shift and in the centre, the zero padding figure place is U=32 * (2 ^M-1) individual, 2 ^MFor distance to the interpolation multiple, wherein M is a positive integer, concrete mode does

$({l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}{l_{i3}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{i17}}^{\&prime;\&prime;}{l_{i18}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{1\&times;U}{l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}...{l_{i16}}^{\&prime;\&prime;})$

New two-dimensional matrix is carried out carrying out the zero padding operation along the orientation to frequency spectrum shift and in the centre, and the zero padding figure place is V=32 * (2 ^N-1) individual, 2 ^NFor the orientation to the interpolation multiple, wherein N is a positive integer, concrete mode does

$({l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}{l_{3j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{17j}}^{\&prime;\&prime;}{l_{18j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{V\&times;1}{l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}...{l_{16j}}^{\&prime;\&prime;})$

Carry out two-dimentional inverse Fourier transform again, obtain the evaluating matrix L ' after the interpolation.

L ' asks mould to evaluating matrix, obtains new real number matrix R '.Ask for the middle maximal value of real number matrix R ' position (r _x', r _y').Then this point target pairing exact position in the complex pattern data matrix S does

$(\frac{{r_x}^{\&prime;}+2^M-1}{2^M}+x-15,\frac{{r_y}^{\&prime;}+2^N-1}{2^N}+y-15).$

Repeat above step, each element among the dot matrix object set E is carried out two-dimensional interpolation, obtain the pairing exact position of each point target.

In the SAR system, the position of target by distance to the orientation to sampling rate decision, the position of target possibly dropped in the sampled point in the SAR image hardly just usually.Utilize the method for this patent, can obtain the exact position of target, the levels of precision of target location by the distance to interpolation multiple 2 ^MDecision, the relation of target exact position and difference multiple is as shown in Figure 2.In the calibration process of SAR system, the accurate location of target is significant to system calibration.

Embodiment:

Table 1 radar parameter

According to radar parameter that table 1 is given, original SAR image complex data size is 8192 * 16384 complex matrixs, and this matrix orientation is to totally 8192 sampled points, and distance remembers that to totally 16384 sampled points this complex matrix does

Step 1: antenna radiation pattern weighted correction;

Obtain each range gate j (j=1,2,3...16384) pairing oblique distance distance R _j:

$R_j=7071.068+\frac{(j-1)\&CenterDot;3\&times;{10}^8}{2\&times;400\&times;{10}^6}$

Obtain this oblique distance distance R _jCorresponding view angle theta _j:

θ _j＝arccos(7071.068/10000)-arccos(7071.068/R _j)

Obtain this view angle theta _jPairing compensating factor a _jFor:

$a_j=\frac{\mathrm{\&pi;}\&times;1\&times;\sin \left({\mathrm{\&theta;}}_j\right)/0.05}{\sin (\mathrm{\&pi;}\&times;1\&times;\sin \left({\mathrm{\&theta;}}_j\right)/0.05)}$

Repeat above step, obtain each distance to sampled point corresponding antenna directional diagram compensating factor a _j(j=1,2,3...16384).

Constructing antennas directional diagram correction matrix A, this matrix are 16384 * 16384 diagonal angle real number matrix, then

Diameter radar image complex data after then proofreading and correct through antenna radiation pattern is 8192 * 16384 complex matrix, i.e. S ' _{8192 * 16384}=S _{8192 * 16384}* A _{16384 * 16384}

Step 2: the radar received power is proofreaied and correct;

Obtain j (j=1,2,3...16384) the pairing oblique distance of individual range gate is apart from radar received power compensating factor:

$p_j=\sqrt{4\mathrm{\&pi;}\&times;{R_j}^4\&times;{0.05}^2}$

Repeat above step, obtain each range gate corresponding antenna directional diagram compensating factor p _j(j=1,2,3...16384).

Structure radar received power correction matrix P, this matrix is the diagonal angle real number matrix of N * N, this matrix diagonal element is antenna radiation pattern compensating factor p _j, off-diagonal element is zero, promptly

Make data matrix S ' carry out the diameter radar image complex data S after matrix multiple obtains radar received power compensation " with radar received power correction matrix P.

Diameter radar image complex data after then compensating through the radar received power is 8192 * 16384 complex matrix, i.e. S " _{8192 * 16384}=S ' _{8192 * 16384}* P _{16384 * 16384}

Step 3: seek two-dimentional maximum point;

The complex data matrix S " is asked mould to handle, obtained new real number matrix

Get real number matrix R _SMiddle arbitrary element r _Ij, judge whether this element satisfies r _{I, j}＞r _{I+1, j}, whether satisfy r _{I, j}＞r _{I, j+1}, whether satisfy r _{I, j}＞r _{I-1, j}, whether satisfy r _{I, j}＞r _{I, j-1}, whether satisfy r _{I, j}＞r _{I+1, j+1}, whether satisfy r _{I, j}＞r _{I-1, j-1}, whether satisfy r _{I, j}＞r _{I+1, j-1}, whether satisfy r _{I, j}＞r _{I-1, j+1}If above-mentioned condition all satisfies, then with this element at real number matrix R _SIn the position (i, j) be recorded in the set K in.

Repeat above step, to real number matrix R _SMiddle all elements r _Ij(i=2,3,4...8191, j=2,3,4...16383) judge, form two-dimentional maximum value set K.

(p q), judges whether p satisfies 0.00001 * 400000000＜p＜16384-0.00001 * 400000000, judges whether q satisfies 3.333 * 1000＜q＜8192-3.333 * 1000 to get arbitrary element among the two-dimentional maximum value rendezvous value set K.If above-mentioned condition has one not satisfy, then (p q) removes in two-dimentional maximum value set K with this element.

If two-dimentional maximum value set K is an empty set, then judge not have the dot matrix target in the complex pattern data matrix S, withdraw from institute in steps.If only there is an element in two-dimentional maximum value set K, then directly get into step 6.Two dimension maximum value set K interior element gets into step 4 more than or equal to two.

Step 4: point target Pixel-level position;

Ask for real number matrix R _SThe maximum element r of middle numerical value _{X, y}=max (R _S).

Count representes number of iterations, and for the first time during iteration, count is 0, and iteration and later count will be provided by step 5 for the second time.

Sensitive factor is:

$g_e=r_{x,y}\&CenterDot;{10}^{\frac{-13+(\mathrm{count}-1)\&CenterDot;\mathrm{<figure-callout\; id="20"\; label="d"\; filenames="HDA0000116858370000021.png"\; state="\{\{state\}\}">d</figure-callout>}}{20}}$

(p q), judges that this element is at matrix R to element among the two-dimentional maximum value set K _SIn whether satisfy r _{P, q}＞g _e, then this element position is recorded among the dot matrix object set E as satisfying.

Repeat above step, handle all elements among the two-dimentional maximum value set K, obtain dot matrix object set E.

If dot matrix object set E is an empty set, judge then not have the dot matrix target in the complex pattern data matrix S that institute in steps after withdrawing from.If only there is an element in dot matrix object set E, then get into step 6.If element number is more than or equal to two among the dot matrix object set E, get into step 5.

Step 5: location verification;

To the arbitrary element among the dot matrix object set E (x, y) carry out location verification:

In dot matrix object set E, seek to make | t-x| is minimum, and distance to the position be y element (t, y).If 20log10 is (r _{T, y}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6.。

In dot matrix object set E, seek to make | f-y| is minimum, and the orientation to the position be x element (x, f).If 20log10 is (r _{X, f}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6.。

Repeat above step, after each element among the dot matrix object set E is verified, accomplish the checking of dot matrix target location.

Element among the dot matrix object set E has write down the location of pixels of point target in the complex pattern data matrix S.

Step 6: two-dimensional frequency interpolation, location, point target exact position.

(x, y), the element of its correspondence in the complex pattern data matrix S is s for the arbitrary element among the dot matrix object set E _{X, y}, be the center with this element position, get this some evaluating matrix L of this point target of 32 * 32 bit matrix elements formation on every side.Promptly in the complex pattern data matrix S, get top-left position for (x-15, y-15), the position, bottom right is that (assignment is given evaluating matrix L for x+16, square formation y+16).

For the evaluating matrix L of this point target, at first the edge distance is to carrying out Fast Fourier Transform (FFT), promptly

Then along the orientation to carrying out Fast Fourier Transform (FFT), promptly

New two-dimensional matrix is carried out the edge distance carry out the zero padding operation to frequency spectrum shift and in the centre, instance adopts 64 times of distances to sample interpolation, and then zero padding figure place is U=32 * (2 ⁶-1)=2016.Concrete mode does

$({l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}{l_{i3}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{i17}}^{\&prime;\&prime;}{l_{i18}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{1\&times;U}{l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}...{l_{i16}}^{\&prime;\&prime;})$

New two-dimensional matrix is carried out carrying out the zero padding operation along the orientation to frequency spectrum shift and in the centre, and instance adopts 64 times of orientation to sample interpolation, and the zero padding figure place is V=32 * (2 ⁶-1)=2016.Concrete mode does

$({l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}{l_{3j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{17j}}^{\&prime;\&prime;}{l_{18j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{V\&times;1}{l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}...{l_{16j}}^{\&prime;\&prime;})$

Carry out two-dimentional inverse Fourier transform again, obtain the evaluating matrix L ' after the interpolation.

L ' asks mould to evaluating matrix, obtains new real number matrix R '.Ask for the middle maximal value of real number matrix R ' position (r _x', r _y').Then the Pixel-level position is that (x, the pairing exact position in the complex pattern data matrix S of point target y) does

$(\frac{{r_x}^{\&prime;}+63}{64}+x-15,\frac{{r_y}^{\&prime;}+63}{64}+y-15).$

Repeat above step, each element among the dot matrix object set E is carried out two-dimensional interpolation, obtain the pairing exact position of each point target.

Table 2 has provided and has utilized the present invention to seek point methods in different SAR image complex datas, to seek a result.Can be found out that by table 2 it is good to the adaptability of object scene that the present invention seeks point methods, the extensive matrix of small-scale dot matrix to 128 * 128 from 2 * 2 can be searched for the position of all point targets automatically.

Table 2 utilizes a result that seeks of this patent method

	Point target number in the SAR image	The present invention seeks a number
			2 * 2 dot matrix SAR images	4	4
4 * 4 dot matrix SAR images	16	16
			8 * 8 dot matrix SAR images	64	64
16 * 16 dot matrix SAR images	256	256
			32 * 32 dot matrix SAR images	1024	1024
64 * 64 dot matrix SAR images	4096	4096
			128 * 128 dot matrix SAR images	16384	16384

Claims (7)

1. A synthetic-aperture radar dot matrix destination image data seek point methods automatically, it is characterized in that, processing to as if synthetic-aperture radar dot matrix target complex pattern data, the result who obtains is the exact position of each point target in data; Size that synthetic-aperture radar dot matrix target complex pattern data are

   

   is M * N; Wherein M representes the orientation to pulse number, and N representes that distance to sampling number, comprises following step:

   Step 1: antenna radiation pattern weighted correction;

   Initialization is also read SAR image complex data, calculates the pairing antenna radiation pattern weighting factor of each range gate, obtains realizing the antenna radiation pattern weighted correction to SAR image complex data along the antenna compensation vector of distance to distribution;

   Step 2: the radar received power is proofreaied and correct;

   Obtain the corresponding radar power equation of each range gate, to through image complex data S ' edge of antenna radiation pattern weighted correction apart to compensating, obtain the uniform SAR image of dot matrix target amplitude complex data S ";

   Step 3: seek two-dimentional maximum point;

   Ask for the uniform SAR image of dot matrix target amplitude complex data S " in two-dimentional maximum point, make the point target position sought all be contained among the set K.Remove then set K middle distance to or the incomplete extreme point of azimuth accumulation;

   Step 4: point target location of pixels location;

   Utilize sensitive factor to judge, among two-dimentional maximum value set K, extract the position of dot matrix target, form new dot matrix goal set E, the Pixel-level position of point target has been preserved in this set;

   Step 5: location verification;

   Whether the point target among the checking dot matrix goal set E satisfies the typical characteristics of SAR system imaging target, if do not satisfy then turn back to step 4, carries out location, point target exact position if meet then get into step 6;

   Step 6: two-dimensional frequency interpolation, location, point target exact position.

   Pixel-level position with point target among the dot matrix goal set E is the center; In original SAR image complex data, get on every side the evaluating matrix L that 32 * 32 matrix element constitutes this point target; Evaluating matrix L is carried out the two-dimensional frequency interpolation; Select the interpolation multiple to positioning accuracy request, obtain the exact position of point target.
2. 2. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 1 is specially:

   Obtain the pairing oblique distance distance R of each range gate j _j, j=1,2,3...N:

   $R_j=R_1+\frac{(j-1)\&CenterDot;C}{2f_s}$

   Wherein: R ₁The corresponding oblique distance distance of this point during for first sampled point, C represents the light velocity, f _sExpression distance to sampling rate; Obtain this oblique distance distance R _jCorresponding view angle theta _j:

   θ _j＝arccos(R _ref/H)-arccos(R _j/H)

   Wherein: R _RefThe corresponding oblique distance of representative antennas sight line center irradiation, H is an aircraft altitude;

   Obtain this view angle theta _jPairing antenna compensating factor a _jFor:

   $a_j=\frac{\mathrm{\&pi;}{D}_r\sin \left({\mathrm{\&theta;}}_j\right)/\mathrm{\&lambda;}}{\sin (\mathrm{\&pi;}{D}_r\sin \left({\mathrm{\&theta;}}_j\right)/\mathrm{\&lambda;})}$

   Wherein: D _rFor the distance to antenna length, λ is the wavelength that transmits;

   Repeat above step, obtain the compensating factor a of each distance to sampled point corresponding antenna directional diagram _j

   Constructing antennas directional diagram correction matrix A, this matrix are the diagonal angle real number matrix of N * N, and this matrix diagonal element is antenna radiation pattern compensating factor a _j, off-diagonal element is zero, promptly

   

   Complex pattern data matrix S and antenna radiation pattern correction matrix A carry out matrix multiple and obtain the diameter radar image complex data S ' after antenna radiation pattern is proofreaied and correct, and realize the antenna radiation pattern weighted correction to SAR image complex data.
3. 3. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 2 is specially:

   Obtain j the pairing oblique distance of range gate apart from radar received power compensating factor:

   $p_j=\sqrt{4\mathrm{\&pi;}{R_j}^4{\mathrm{\&lambda;}}^2}$

   Repeat above step, obtain each distance to sampled point corresponding antenna directional diagram compensating factor p _j

   Structure radar received power correction matrix P, this matrix is the diagonal angle real number matrix of N * N, this matrix diagonal element is antenna radiation pattern compensating factor p _j, off-diagonal element is zero, promptly

   

   Make data matrix S ' carry out matrix multiple with radar received power correction matrix P, to through image complex data S ' edge the distance of antenna radiation pattern weighted correction to compensating, obtain the diameter radar image complex data S after the radar received power compensates ".
4. 4. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 3 is specially:

   The complex data matrix S " is asked mould to handle, obtained new real number matrix

   Get real number matrix R _SMiddle arbitrary element r _Ij, judge whether this element satisfies r _{I, j}＞r _{I+1, j}, whether satisfy r _{I, j}＞r _{I, j+1}, whether satisfy r _{I, j}＞r _{I-1, j}, whether satisfy r _{I, j}＞r _{I, j-1}, whether satisfy r _{I, j}＞r _{I+1, j+1}, whether satisfy r _{I, j}＞r _{I-1, j-1}, whether satisfy r _{I, j}＞r _{I+1, j-1}, whether satisfy r _{I, j}＞r _{I-1, j+1}If above-mentioned condition all satisfies, then with this element at real number matrix R _SIn (i j) is recorded among the two-dimentional maximum value set K, i=2 wherein, 3,4...M-1, j=2,3,4...N-1 in the position;

   Repeat above step, to real number matrix R _SMiddle all elements r _IjJudge, form two-dimentional maximum value set K;

   For the element among the two-dimentional maximum value set K (p, q), r _{P, q}Be R _SIn element, p, q are that this element is at R _SIn the position;

   (p q), judges whether p satisfies T to get among the two-dimentional maximum value set K arbitrary element _pf _s＜p＜N-T _pf _s, judge whether q satisfies T _aPrf＜q＜M-T _aPrf, wherein T _pBe pulse width, f _sBe sampling rate, T _aBe the synthetic aperture time, prf is a pulse repetition rate; If above-mentioned condition has one not satisfy, then with this element (p q) removes in two-dimentional maximum value set K, then remove two-dimentional maximum value set K middle distance to or the incomplete extreme point of azimuth accumulation;

   If two-dimentional maximum value set K is an empty set, then judge not have the dot matrix target in the complex pattern data matrix S, withdraw from institute in steps; If only there is an element in two-dimentional maximum value set K, then directly get into step 6; Two dimension maximum value set K interior element gets into step 4 more than or equal to two.
5. 5. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 4 is specially:

   Ask for real number matrix R _SThe maximum element r of middle numerical value _{X, y}=max (R _S);

   Count representes number of iterations, and for the first time during iteration, count is 0, and iteration and later count will be provided by step 5 for the second time;

   The definition sensitive factor does

   $g_e=r_{x,y}\&CenterDot;{10}^{\frac{-13+(\mathrm{count}-1)\&CenterDot;d}{20}}$

   In the formula, d has represented the fine degree that sensitive factor changes in each iteration; The more a little bigger target localization of d value is accurate more, and locating speed is slow more;

   For element among the two-dimentional maximum value set K (p, q), if r _{P, q}＞g _e, then this element position is recorded among the dot matrix object set E;

   Repeat above step, handle all elements among the two-dimentional maximum value set K, obtain dot matrix object set E; For the arbitrary element among the dot matrix object set E (x, y), r _{X, y}Be R _SIn element, x, y are e _{X, y}At R _SIn the position;

   If dot matrix object set E is an empty set, judge then not have the dot matrix target in the complex pattern data matrix S that institute in steps after withdrawing from; If only there is an element in dot matrix object set E, then get into step 6; If element number is more than or equal to two among the dot matrix object set E, get into step 5.
6. 6. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 5 is specially:

   For the arbitrary element among the dot matrix object set E (x, y):

   In dot matrix object set E, seek to make | t-x| is minimum, and distance to the position be y element (t, y); If 20log10 is (r _{T, y}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6;

   In dot matrix object set E, seek to make | f-y| is minimum, and the orientation to the position be x element (x, f); If 20log10 is (r _{X, f}/ r _{X, y})>=-10 then returns step 4, and count=count+1, otherwise, get into step 6;

   Repeat above step, after each element among the dot matrix object set E is verified, accomplish the checking of dot matrix target location;

   Element among the dot matrix object set E has write down the location of pixels of point target in the complex pattern data matrix S.
7. 7. a kind of synthetic-aperture radar dot matrix destination image data according to claim 1 seek point methods automatically, it is characterized in that described step 6 is specially:

   (x, y), the element of its correspondence in the complex pattern data matrix S is s for the arbitrary element among the dot matrix object set E _{X, y}, be the center with this element position, get this some evaluating matrix L of this point target of 32 * 32 bit matrix elements formation on every side; Promptly in the complex pattern data matrix S, get top-left position for (x-15, y-15), the position, bottom right be that (assignment is to evaluating matrix L for x+16, square formation y+16);

   For the evaluating matrix L of this point target, at first the edge distance is to carrying out Fast Fourier Transform (FFT), promptly

   

   Then along the orientation to carrying out Fast Fourier Transform (FFT), promptly

   

   New two-dimensional matrix is carried out the edge distance carry out the zero padding operation to frequency spectrum shift and in the centre, the zero padding figure place is U=32 * (2 ^M-1) individual, 2 ^MFor distance to the interpolation multiple, wherein M is a positive integer, concrete mode does

   $({l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}{l_{i3}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{i17}}^{\&prime;\&prime;}{l_{i18}}^{\&prime;\&prime;}...{l_{i32}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{1\&times;U}{l_{i1}}^{\&prime;\&prime;}{l_{i2}}^{\&prime;\&prime;}...{l_{i16}}^{\&prime;\&prime;})$

   New two-dimensional matrix is carried out carrying out the zero padding operation along the orientation to frequency spectrum shift and in the centre, and the zero padding figure place is V=32 * (2 ^N-1) individual, 2 ^NFor the orientation to the interpolation multiple, wherein N is a positive integer, concrete mode does

   $({l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}{l_{3j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;})\&DoubleRightArrow;({l_{17j}}^{\&prime;\&prime;}{l_{18j}}^{\&prime;\&prime;}...{l_{32j}}^{\&prime;\&prime;}{(\mathrm{0,0}...0)}_{V\&times;1}{l_{1j}}^{\&prime;\&prime;}{l_{2j}}^{\&prime;\&prime;}...{l_{16j}}^{\&prime;\&prime;})$

   Carry out two-dimentional inverse Fourier transform again, obtain the evaluating matrix L ' after the interpolation;

   L ' asks mould to evaluating matrix, obtains new real number matrix R '; Ask for the middle maximal value of real number matrix R ' position (r _x', r _y'); Then this point target pairing exact position in the complex pattern data matrix S does

   $(\frac{{r_x}^{\&prime;}+2^M-1}{2^M}+x-15,\frac{{r_y}^{\&prime;}+2^N-1}{2^N}+y-15).$

   Repeat above step, each element among the dot matrix object set E is carried out two-dimensional interpolation, obtain the pairing exact position of each point target.

Images