CN101545969A - Method for evaluating point target resolution of side-glance SAR - Google Patents

Abstract

The invention discloses a method for evaluating point target resolution of a side-glance SAR. The evaluation method is finished through three steps of two-dimensional interpolation, cross-sectional data extraction and resolution index calculation, and comprises the following steps: firstly, performing the two-dimensional interpolation on local point target data; secondly, drawing a power value contour line projection drawing of the two-dimensional interpolation result; and thirdly, determining inclination angles in a distance direction and an orientation direction through an interactive mode, respectively extracting distance direction cross-sectional data and orientation direction cross-sectional data in a two-dimensional interpolation result complex data matrix along the direction of the inclination angle, and calculating the resolution in the distance direction and the orientation direction according to formulas (6) and (7). The method mainly solves the problem of resolution measurement errors caused by inaccurate extraction of an inclined cross section. The method has the following main characteristics: (1) the method can still obtain an accurate cross section curve under the condition that the distance direction cross section is not vertical to the orientation direction cross section; and (2) the method obtains a more accurate cross section curve and then obtains a more accurate resolution through calculation according to the formulas (6) and (7), and can reflect the actual resolution indexes of the SAR point targets more.

Description

A kind of point target resolution appraisal procedure of Squint SAR

Technical field

The present invention relates to the Radar Technology field, specifically, be meant a kind of stravismus synthetic-aperture radar point target resolution appraisal procedure of (being called for short SAR).

Background technology

Whether reasonably the index evaluation of diameter radar image is quantitative evaluation SAR systematic parameter design important evidence, the particularly assessment of point target resolution.Because the SAR image is two dimensional image, comprise distance to the orientation to, for point target, equally also comprise distance to the orientation to the resolution index.The distance of point target to the orientation to resolution assessment divided for two steps greatly: the distance that one extracts point target to the orientation to cross-sectional data; It two is adjusted the distance respectively to carrying out interpolation calculation with the orientation to cross-sectional data, determine separately half-power width respectively as distance to resolution and orientation to resolution.Usually for positive side-looking and the very little stravismus (angle of squint is less than 3 degree) of angle, on two-dimensional SAR image, distance is to vertical to thinking with the orientation, if and the identical range gate data of each row representative of imaging results data two-dimensional matrix, on behalf of the mode of a pulse data, each row arrange (as shown in Figure 2), then column direction be distance to, line direction be the orientation to.

But for general Squint SAR, stravismus imaging meeting cause distance to the orientation to out of plumb, in data array shown in Figure 2, distance is to run-off the straight (shown in Fig. 3 (b)), angle of inclination and angle of squint relating to parameters.Select data will produce if still use this moment, thereby the point target assessment result is brought adverse effect than mistake as the distance profile data according to column direction.

Summary of the invention

The present invention when solving the assessment of classical SAR point target resolution because simple column direction and the line direction selected produces problem than mistake as cross-sectional data respectively, according to when stravismus distance to section and orientation to section off plumb characteristics, utilize interactive operation determine distance to the orientation to the section pitch angle, according to the pitch angle extract successively the distance to the orientation to cross-sectional data, on this basis further the calculation level target range to the orientation to the resolution index, realizes a kind of appraisal procedure of point target resolution that can the big Squint SAR of adaptation.

The point target resolution appraisal procedure of Squint SAR of the present invention realizes by following steps:

(1) two-dimensional interpolation: the two-dimentional complex data that will contain point target to be assessed is asked the power peak position, and be the center with the peak, get the regional area data and carry out two-dimentional Fourier's interpolation and obtain two-dimensional interpolation complex data as a result, and the power maximal value coordinate of image after definite interpolation.

(2) cross-sectional data is extracted: draw the two-dimensional interpolation performance number level line perspective view of complex data as a result, by interactive mode determine distance to the angle of inclination of orientation to section, and along the angle of inclination direction two-dimensional interpolation extract respectively in the complex data matrix as a result distance to cross-sectional data and orientation to cross-sectional data.

(3) the resolution index is calculated: adjusting the distance respectively obtains one dimension interpolation complex data to carrying out one dimension Fourier interpolation with the orientation to cross-sectional data, ask the maximum value position and the half-power position, both sides of one dimension interpolation complex data respectively, and according to corresponding formula distance to the orientation to resolution.

Above-mentioned distance is [45 °, 135 °] to the range of tilt angles of section, and the orientation is [45 °, 45 °] to the range of tilt angles of section.

The advantage of the appraisal procedure of the point target resolution of big Squint SAR provided by the invention: (1) method provided by the invention can be in distance to still obtaining section curve more accurately with the orientation under section off plumb condition; (2) resolution that calculates after obtaining section curve more accurately and according to correlation formula of method provided by the invention is more accurate, more can reflect the true resolution index of SAR point target.

Description of drawings

Fig. 1 is the flow chart of steps of the point target resolution appraisal procedure of Squint SAR of the present invention;

The data storage sequential schematic that Fig. 2 adopts for the inventive method;

The synoptic diagram of representative point target two dimensional cross-section when Fig. 3 (a) is positive side-looking;

The synoptic diagram of representative point target two dimensional cross-section when Fig. 3 (b) is big the stravismus;

Fig. 4 be the distance to the orientation to cross-sectional data extracting method synoptic diagram;

Fig. 5 is a cross-sectional data normalization logarithm half-power width synoptic diagram;

Fig. 6 is the power data contour map of case study on implementation complex data C2;

Fig. 7 selects synoptic diagram to section and orientation to section for the case study on implementation distance.

Embodiment

Be elaborated below in conjunction with the point target resolution appraisal procedure of accompanying drawing to Squint SAR of the present invention.

The point target resolution appraisal procedure of Squint SAR provided by the invention realizes by following steps, as Fig. 1

Shown in:

Step 1, two-dimensional interpolation: the two-dimentional complex data that will contain point target to be assessed is asked the power peak position, and be the center with the peak, get the regional area data and carry out two-dimentional Fourier's interpolation and obtain two-dimensional interpolation complex data as a result, and the power maximal value coordinate of image after definite interpolation.

(a) search contains the power maximal value of the two-dimentional complex data C0 of point target to be assessed, writes down the volume coordinate (i of this position correspondence ₀, j ₀);

(b) the target analysis window of setting up an office is N _Win* N _Win, then with among the two-dimentional complex data C0, with (i ₀, j ₀) be the center, window size is N _Win* N _WinComplex data matrix assignment to give size be N _Win* N _WinComplex matrix C1, promptly

C1[0：N _win-1，0：N _win-1]＝C0[i ₀-0.5N _win：i ₀+0.5N _win-1，j ₀-0.5N _win：j ₀+0.5N _win-1] (1)

For conveniently doing Fast Fourier Transform (FFT), common N _WinGet and be not less than 16 2 integral number power;

(c) complex matrix C1 is done two-dimentional Fourier's interpolation processing, the two-dimensional interpolation multiple is made as N ₁, obtaining dimension is N _WinN ₁* N _WinN ₁Complex matrix C2 after the interpolation;

(d) the power maximal value of two-dimentional complex data C2 behind the search value interposition writes down the volume coordinate (i of this position correspondence _p, j _p).

Step 2, cross-sectional data are extracted: draw the two-dimensional interpolation performance number level line perspective view of complex data as a result, by interactive mode select to determine distance to the orientation to the angle of inclination, and along the angle of inclination direction two-dimensional interpolation extract respectively in the complex data matrix as a result distance to cross-sectional data and orientation to cross-sectional data.

(a) the performance number data of complex data C2 are drawn out contour map;

(b) looked like (i at contour map _p, j _p) point and (i _r, j _r) point, make the two line L1 pass the peak of distance to secondary lobe, can calculate the inclination alpha of distance to section curve this moment _r, promptly

${\mathrm{\α}}_r=a\tan \frac{j_r-j_p}{i_r-i_p}---\left(2\right)$

(c) looked like (i at contour map _p, j _p) point and (i _a, j _a) point, make the two line L2 pass the peak of orientation to secondary lobe, can calculate the inclination alpha of orientation to section curve this moment _a, promptly

${\mathrm{\α}}_a=a\tan \frac{j_a-j_p}{i_a-i_p}---\left(3\right)$

(d) along straight line L1, in complex data C2 line by line successively the nearest point of chosen distance straight line L1 as the current line value of distance, promptly to cross-sectional data Dr

Dr[j]＝C2[round(i _p+(j-j _p)/tanα _r)，j] (4)

Round (i wherein _p+ (j-j _p)/tan α _r) represent i _p+ (j-j _p)/tan α _rRound.

(e), in complex data C2, pursue the nearest point of leu time chosen distance straight line L2 as the current line value of orientation, promptly to cross-sectional data Da along straight line L2

Da[j]＝C2[i，round(j _p+(i-i _p)·tan?α _a)] (5)

Round (j wherein _p+ (i-i _p) tan α _a) represent j _p+ (i-i _p) tan α _aRound.

Step 3, resolution index are calculated: adjusting the distance respectively obtains one dimension interpolation complex data to carrying out one dimension Fourier interpolation with the orientation to cross-sectional data, ask the maximum value position and the half-power position, both sides of one dimension interpolation complex data respectively, and according to corresponding formula distance to the orientation to resolution.

(a) adjust the distance and carry out one dimension Fourier interpolation processing to cross-sectional data Dr, one dimension interpolation multiple is made as N _r, obtaining dimension is N _WinN ₁N _rInterpolation after the distance to the section complex data, ask power to obtain power data Sr to it;

(b) one dimension Fourier interpolation processing is carried out to cross-sectional data Da in the orientation, one dimension interpolation multiple is made as N _a, obtaining dimension is N _WinN ₁N _aInterpolation after the distance to the section complex data, ask power to obtain power data Sa to it;

(c) the power maximal value Pr of point by point search power data Sr, the corresponding following Rp that is labeled as searches for left from Rp, and record is designated as R1 under the point of performance number less than 0.5*Pr for the first time, Rp searches for to the right again, and record is designated as Rr under the point of performance number less than 0.5*Pr for the first time;

(d) the power maximal value Pa of point by point search power data Sa, the corresponding following Ap that is labeled as searches for left from Ap, and record is designated as A1 under the point of performance number less than 0.5*Pa for the first time, Ap searches for to the right again, and record is designated as Ar under the point of performance number less than 0.5*Pa for the first time;

(e) establish distance to being respectively Δ to pixel separation with the orientation _rAnd Δ _a, distinguish computed range to resolution δ according to formula (6) and (7) _rWith the orientation to resolution δ _a:

${\mathrm{\&delta;}}_r=\sqrt{{\left(\frac{\mathrm{Rr}-\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="A200910081387E00131.png"\; state="\{\{state\}\}">R</figure-callout>}1+1}{N_1{N}_r}{\mathrm{\&Delta;}}_r\right)}^2+{\left(\frac{\mathrm{Rr}-\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="A200910081387E00131.png"\; state="\{\{state\}\}">R</figure-callout>}1+1}{N_1{N}_r\tan {\mathrm{\&alpha;}}_r}{\mathrm{\&Delta;}}_a\right)}^2}---\left(6\right)$

${\mathrm{\&delta;}}_a=\sqrt{{\left(\frac{\mathrm{Ar}-\mathrm{<figure-callout\; id="1"\; label="A"\; filenames="A200910081387E00131.png"\; state="\{\{state\}\}">A</figure-callout>}1+1}{N_1{N}_a}{\mathrm{\&Delta;}}_a\right)}^2+{\left(\frac{\mathrm{Ar}-\mathrm{<figure-callout\; id="1"\; label="A"\; filenames="A200910081387E00131.png"\; state="\{\{state\}\}">A</figure-callout>}1+1}{N_1{N}_r}\tan {{\mathrm{\&alpha;}}_a\mathrm{\&Delta;}}_r\right)}^2}---\left(7\right)$

Case study on implementation 1:At a two-dimentional impulse response function is that (i, ideal stravismus point target j) is carried out the resolution assessment to F.

N=64 wherein, M=50, ${\mathrm{\&delta;}}_a^{\&prime;}=2.5,$ ${\mathrm{\&delta;}}_r^{\&prime;}=3.0,$ The data matrix subscript is all since 0 when analyzing in this example.

Concrete appraisal procedure is specifically passed through following steps:

Step 1, two-dimensional interpolation

(a) dimension of establishing two-dimentional complex data C0 is 50 * 64, C0 (i, j) each element real part respectively corresponding F (i, j), imaginary part is 0, the power maximal value of search C0 equals 1, writes down the volume coordinate (i of this position correspondence ₀, j ₀)=(24,31);

(b) get N _Win=32, promptly the point target analysis window is 32 * 32, then with among the two-dimentional complex data C0, with (i ₀, j ₀)=(24,31) be the center, window size is 32 * 32 complex data matrix assignment to size is 32 * 32 complex matrix C1, promptly

C1[0:31，0:31]＝C0[8:39，15:46] (9)

(c) complex matrix C1 is done two-dimentional Fourier's interpolation processing, the two-dimensional interpolation multiple is got N ₁=32, obtaining dimension is N _WinN ₁* N _WinN ₁Complex matrix C2 after=1024 * 1024 interpolation;

(d) the power maximal value of two-dimentional complex data C2 behind the search value interposition writes down the volume coordinate (i of this position correspondence _p, j _p)=(512,512).

Step 2, cross-sectional data are extracted

(a) the performance number data of complex data C2 are drawn out contour map, as shown in Figure 6;

(b) get (i _r, j _r)=(752,994), looked like (i at contour map _p, j _p)=(512,512) with (i _r, j _r)=(752,994) point makes the two line L1 pass the peak (as shown in Figure 7) of distance to secondary lobe, can calculate the inclination alpha of distance to section curve according to formula (2) _r=63.53 °;

(c) get (i _a, j _a)=(888,512), looked like (i at contour map _p, j _p)=(512,512) with (i _a, j _a)=(888,512) point makes the two line L2 pass the peak (as shown in Figure 7) of orientation to secondary lobe, can calculate the inclination alpha of orientation to section curve according to formula (3) _a=0;

(d) along straight line L1, in complex data C2 line by line successively the nearest point of chosen distance straight line L1 as the current line value of distance, promptly to cross-sectional data Dr

Dr[j]＝C2[round(512-(j-512)/2.0083)，j]j∈[0，1024) (10)

(e) along straight line L2, in complex data C2 line by line successively the nearest point of chosen distance straight line L2 as the current line value of orientation, promptly to cross-sectional data Da

Da[0:1023]＝C2[0:1023，512](11)

Step 3, resolution index are calculated

(a) adjust the distance and carry out one dimension Fourier interpolation processing to cross-sectional data Dr, one dimension interpolation multiple is got N _r=64, obtaining dimension is N _WinN ₁N _rDistance asks power to obtain power data Sr to the section complex data to it after=65536 the interpolation;

(b) one dimension Fourier interpolation processing is carried out to cross-sectional data Da in the orientation, one dimension interpolation multiple is made as N _a=64, obtaining dimension is N _WinN ₁N _aDistance asks power to obtain power data Sa to the section complex data to it after=65536 the interpolation;

(c) the power maximal value Pr of point by point search power data Sr, the corresponding following Rp=32768 that is labeled as, search for left from Rp, record is designated as R1=30208 under the point of performance number less than 0.5*Pr for the first time, Rp searches for to the right again, and record is designated as Rr=35328 under the point of performance number less than 0.5*Pr for the first time;

(d) the power maximal value Pa of point by point search power data Sa, the corresponding following Ap=32768 that is labeled as, search for left from Ap, record is designated as A1=29698 under the point of performance number less than 0.5*Pa for the first time, Ap searches for to the right again, and record is designated as Ar=35838 under the point of performance number less than 0.5*Pa for the first time;

(e) this routine middle distance is to being respectively Δ with the orientation to pixel separation _r=1 and Δ _a=1, distinguish computed range to resolution δ according to formula (6) and (7) _r=2.7933m and orientation are to resolution δ _a=2.9990m.Extract distance to cross-sectional data along distance to each time side lobe peak as can be seen, measure the distance that obtains and more can embody the distance of point target reality to resolution characteristic to resolution.

The present invention be primarily aimed at stravismus imaging meeting cause distance to the orientation to section off plumb characteristics, utilize interactive operation determine respectively distance to the orientation to the section pitch angle, according to the pitch angle extract successively the distance to the orientation to cross-sectional data, on this basis further the calculation level target range to the orientation to the resolution index, realize a kind of appraisal procedure that can adapt to the point target resolution of big Squint SAR, and, the implementation process of this method has been described in further detail by instance analysis.

Claims (4)

1, a kind of point target resolution appraisal procedure of Squint SAR is characterized in that may further comprise the steps:

(1) two-dimensional interpolation: the two-dimentional complex data that will contain point target to be assessed is asked the power peak position, and be the center with the peak, get the regional area data and carry out two-dimentional Fourier's interpolation and obtain two-dimensional interpolation complex data as a result, and the power maximal value coordinate of image after definite interpolation.

(2) cross-sectional data is extracted: draw the two-dimensional interpolation performance number level line perspective view of complex data as a result, by interactive mode determine distance to the angle of inclination of orientation to section, and along the angle of inclination direction two-dimensional interpolation extract respectively in the complex data matrix as a result distance to cross-sectional data and orientation to cross-sectional data.

(3) the resolution index is calculated: adjusting the distance respectively obtains one dimension interpolation complex data to carrying out one dimension Fourier interpolation with the orientation to cross-sectional data, ask the maximum value position and the half-power position, both sides of one dimension interpolation complex data respectively, and according to corresponding formula (6) and (7) computed range to the orientation to resolution.

2, the point target resolution appraisal procedure of Squint SAR according to claim 1 is characterized in that: the described cross-sectional data of step (2) is extracted and is comprised the steps: that (a) draws out contour map with the power data of complex data C2 behind the two-dimensional interpolation; (b) looked like (i at contour map _p, j _p) point and (i _r, j _r) point, make the two line L1 pass the peak of distance to secondary lobe, can calculate the inclination angle of distance to section curve this moment ${\mathrm{\&alpha;}}_r=a\tan \frac{j_r-j_p}{i_r-i_p};$ (c) looked like (i at contour map _p, j _p) point and (i _a, j _a) point, make the two line L2 pass the peak of orientation to secondary lobe, can calculate the inclination angle of orientation to section curve this moment ${\mathrm{\&alpha;}}_a=a\tan \frac{j_a-j_p}{i_a-i_p};$ (d) along straight line L1, in complex data C2 line by line successively the nearest point of chosen distance straight line L1 as the current line value of distance to cross-sectional data Dr, Dr[j]=C2[round (i _p+ (j-j _p)/tan α _r), j], round (i wherein _p+ (j-j _p)/tan α _r) represent i _p+ (j-j _p)/tan α _rRound; (e) along straight line L2, in complex data C2 by the nearest point of leu time chosen distance straight line L2 as the orientation to cross-sectional data Da when the prostatitis value, make Da[j]=C2[i, round (j _p+ (i-i _p) tan α _a)], round (j wherein _p+ (i-i _p) tan α _a) represent j _p+ (i-i _p) tan α _aRound.

3, the point target resolution appraisal procedure of Squint SAR according to claim 1, it is characterized in that: the described resolution index of step (3) is calculated and is determined to comprise the steps: that (a) adjusts the distance carries out one dimension Fourier interpolation processing to cross-sectional data Dr, and one dimension interpolation multiple is made as N _r, obtaining dimension is N _WinN ₁N _rInterpolation after the distance to the section complex data, ask power to obtain power data Sr to it; (b) one dimension Fourier interpolation processing is carried out to cross-sectional data Da in the orientation, one dimension interpolation multiple is made as N _a, obtaining dimension is N _WinN ₁N _aInterpolation after the distance to the section complex data, ask power to obtain power data Sa to it; (c) the power maximal value Pr of point by point search power data Sr, the corresponding following Rp that is labeled as searches for left from Rp, and record is designated as R1 under the point of performance number less than 0.5*Pr for the first time, Rp searches for to the right again, and record is designated as Rr under the point of performance number less than 0.5*Pr for the first time; (d) the power maximal value Pa of point by point search power data Sa, the corresponding following Ap that is labeled as searches for left from Ap, and record is designated as A1 under the point of performance number less than 0.5*Pa for the first time, Ap searches for to the right again, and record is designated as Ar under the point of performance number less than 0.5*Pa for the first time; (e) the difference computed range is to resolution δ _rWith the orientation to resolution δ _a, ${\mathrm{\&delta;}}_r=\sqrt{{\left(\frac{\mathrm{Rr}-\mathrm{Rl}+1}{N_1{N}_r}{\mathrm{\&Delta;}}_r\right)}^2+{\left(\frac{\mathrm{Rr}-\mathrm{Rl}+1}{N_1{N}_r\tan {\mathrm{\&alpha;}}_r}{\mathrm{\&Delta;}}_a\right)}^2},$ ${\mathrm{\&delta;}}_a=\sqrt{{\left(\frac{\mathrm{Ar}-\mathrm{Al}+1}{N_1{N}_a}{\mathrm{\&Delta;}}_a\right)}^2+{\left(\frac{\mathrm{Ar}-\mathrm{Al}+1}{N_1{N}_a}\tan {\mathrm{\&alpha;}}_a{\mathrm{\&Delta;}}_r\right)}^2},$ Δ wherein _rAnd Δ _aBe respectively the distance to the orientation to pixel separation.

4, the point target resolution appraisal procedure of Squint SAR according to claim 1 is characterized in that: Squint SAR point target distance is [45 °, 135 °] to the range of tilt angles of section, and the orientation is [45 °, 45 °] to the range of tilt angles of section.

Images