CN103852759A - Scanning radar super-resolution imaging method - Google Patents

Abstract

The invention discloses a scanning radar super-resolution imaging method which specifically comprises the following steps: transmitting large-time broadband width product linear frequency modulation signal, and realizing distance direction high resolution through a pulse compression technology; estimating a global peak signal to clutter/noise ratio on the data subjected to range walk correction; if the peak signal to clutter/noise ratio is higher than an initial set value, performing target area detection on the signal subjected to range walk correction, performing high-iteration super-resolution imaging processing on the detected target area, and performing low iteration on scene data; and finally, recombining the target area and the scene data, and finishing imaging processing in the whole imaging area; and if the global peak signal to clutter/noise ratio is lower than the initial set value, selecting reasonable iterations to perform unified processing on the data in the whole imaging area. According to the method, a strong target area and the scene in the imaging area are divided, super-resolution imaging is respectively performed, a high super-resolution factor in the strong target area is obtained, and the contour features of the scene information are guaranteed.

Description

Scanning radar super-resolution imaging method

Technical field

The invention belongs to Radar Signal Processing Technology field, particularly the formation method in scanning radar (Scanning Radar) technology.

Background technology

Microwave Imaging Technique has the feature of round-the-clock, all weather operations as a kind of Aeronautics and Astronautics remote sensing of active, have a wide range of applications in the field such as geological mapping, disaster monitoring, become at present high resolving power earth observation and global resources and managed one of most important means.Real Beam radar forms distance to high-resolution by the large bandwidth signal of transmitting, but azimuthal resolution is subject to the restriction of antenna beamwidth, generally not high.Improving the direct method of real Beam radar azimuthal resolution is to increase antenna physical pore size, but is subject to the restriction of antenna weight, size and some other physical factor, is difficult to obtain orientation high-resolution.

Because scanning radar orientation can be regarded as the convolution of antenna radiation pattern and target scattering coefficient to signal, therefore can pass through the method reconstructed object information of deconvolution, thereby break through real Beam radar azimuthal resolution restriction.Because the existence at noise and antenna radiation pattern zero point, having caused deconvolution is an intrinsic ill-conditioning problem, and the impact of deconvolution method performance trusted miscellaneous noise ratio is larger.Document " improves the Generalized Inverse Filtering of real aperture radar angular resolution " (electronic letters, vol, 1993, No.9, Vol.21, and " a kind of monopulse radar multi-channel deconvolution forward sight formation method " (signal processing, 2007, No.5 pp:15-19), Vol.23, pp:699-703), adopt respectively generalized inverse filtering, design and poor passage deconvolver, realize orientation to super-resolution, but the situation that is only applicable to high letter miscellaneous noise ratio, super-resolution performance is limited, document " Radar Angular Super-resolution Algorithm Based on Bayesian Approach " (ICSP2010Proceedings, Beijing, China, and " Improving angular resolution based on maximum a posteriori criterion for scanning radar " (IEEE Radar Conference October2010), USA, pp.0451-0454, May2012) thinking that adopts respectively maximum likelihood and maximum a posteriori criterion to carry out iteration realizes super-resolution processing, although can tolerate certain low letter miscellaneous noise ratio, but under high iterations condition, atural object profile is often destroyed, cannot obtain the scene profile that higher super-resolution multiple is become reconciled simultaneously.

Summary of the invention

The object of the invention is the defect existing for background technology, propose a kind of scanning radar super-resolution imaging method, to overcome the contradictory problems of super-resolution performance and scene information integrality in existing processing procedure.

Technical scheme of the present invention is: a kind of scanning radar super-resolution imaging method, specifically comprises the following steps:

Step 1: echo obtains,

If transmit as linear FM signal

wherein, τ be distance to time variable, K _rfor chirp rate, T _rwide during for pulse, f ₀for carrier frequency;

From echo expression formula after down coversion of point target P (x, y) reflection be:

$S(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\mathrm{rect}\left[\frac{\mathrm{\&tau;}-{\mathrm{\&tau;}}_d}{T_r}\right]\&times;\exp \left\{\mathrm{j\&pi;}{K}_r{[\mathrm{\&tau;}-{\mathrm{\&tau;}}_d]}^2\right\}\&times;\exp \{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000487731780000022.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}{\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">f</figure-callout>}}_0{\mathrm{\&tau;}}_d\}$

Wherein, σ ₀for target scattering coefficient, ω _a(t) be antenna radiation pattern modulating function, ω _a(t)=A (θ ₀-θ), A (θ) is antenna radiation pattern function, θ ₀for azimuth of target, rect[] be that distance is to time window function, τ _d=2R (t)/c is round trip echo delay,

for any point target P in scene is to Texas tower instantaneous distance, R ₀during for target P zero, be carved into Texas tower distance, t is orientation time variable, and V is platform movement velocity, and c is the light velocity;

Step 2: distance to pulse compression,

Structure distance is to pulse compression frequency matching function

echoed signal, along distance to FFT, in apart from frequency domain-orientation time domain, multiply each other with adaptation function, then the signal expression that contravariant is changed in two-dimensional time-domain is:

S _rc(τ,t)=σ ₀w _a(t)sinc{B[τ-τ _d]}×exp{-j2πf ₀τ _d}

Wherein, sinc{} is apart from pulse pressure response function, and B is transmitted signal bandwidth;

Step 3: range walk judgement and correction,

According to the instantaneous distance expression formula in step 1, can obtain range walk amount is Δ R=VT _scos θ ₀, wherein,

for beam scanning residence time, θ _betafor antenna beamwidth, ω is sweep velocity.Judge whether it crosses over range unit

wherein, f _rfor distance is to sampling rate;

If meet Δ R< Δ r, directly carry out step 4; If Δ R> Δ r, is being multiplied by the distance frequency domain-orientation time domain data in step 2 after pulse compression frequency matching function, then is being multiplied by Range Walk Correction function

and then carry out distance to IFFT, obtain Range Walk Correction after signal expression be:

$S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\sin c\left\{B\right[\mathrm{\&tau;}-\frac{2R_0}{c}\left]\right\}\&times;\exp \{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000487731780000022.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}{\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">f</figure-callout>}}_0{\mathrm{\&tau;}}_d\}$

Step 4: calculate peak value letter miscellaneous noise ratio,

After calculating Range Walk Correction, data peaks is believed miscellaneous noise ratio:

$\mathrm{SCNR}=20*{\mathrm{<figure-callout\; id="10"\; label="log"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">log</figure-callout>}}_{10}\frac{\max \left(S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)\right)}{\mathrm{mean}\left(S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)\right)}$

Wherein, max (S _rcmc(τ, t)) be data S after Range Walk Correction _rcmc(τ, t) maximal value, mean (S _rcmc(τ, t)) be mean value; By peak value letter miscellaneous noise ratio SCNR and the initial set value SCNR0 comparison of calculating, if SCNR>=SCNR0 carries out step 5; If SCNR<SCNR0, according to Lucy-Richardson iterative solution Convolution Formula, after the normal moveout correction of adjusting the distance away, the rational iterations of data selection carries out super-resolution imaging processing;

Step 5: calculate global threshold,

According to DP-CFAR rate (CFAR) detection method based on rayleigh distributed, after the normal moveout correction of adjusting the distance away, data are carried out CFAR detection, and detection criteria is:

$S_{\mathrm{CFAR}}(i,j)=\left\{\begin{array}{c}{S}_{\mathrm{rcmc}}(i,j),\frac{(S_{\mathrm{rcmc}}(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}>\mathrm{<figure-callout\; id="0"\; label="Th"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">Th</figure-callout>}\\ 0,\frac{(S_{\mathrm{rcmc}}(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}<\mathrm{Th}\end{array}\right.$

Wherein, S _cFAR(i, j) for CFAR detect after data, μ _cdata mean value in reference window while detection for CFAR, σ _cfor data standard deviation in reference window, Th is the thresholding factor, according to rayleigh distributed probability density function, calculates Th and false-alarm probability P _fabetween pass be:

$\mathrm{Th}=\frac{(2\sqrt{-\ln P_{\mathrm{fa}}}-\sqrt{\mathrm{\&pi;}})}{\sqrt{4-\mathrm{\&pi;}}}$

According to the global threshold β that maximizes the rear data of inter-class variance method calculating CFAR detection;

After CFAR detects, data letter miscellaneous noise ratio can improve, and has reduced the impact on global threshold β of noise and clutter, has improved global threshold and has detected rear target detection probability;

Taking β as threshold value, data are carried out to Preliminary division, tentatively determine that target area is:

S _target=S _cFAR(τ, t), wherein, | S _cFAR(τ, t) |>=β.

Step 6: target area is detected.

Destroyed through the object edge information in the preliminary definite target area of step 5, in the time carrying out orientation to deconvolution, can affect deconvolution result.For can better making target information retain, along S _targetorientation is to carrying out the detection of object edge information.

If one party bit location S _target(i, j) non-zero is respectively expanded half beam angle data of reservation, that is: centered by Xiang Yigai unit, orientation to both sides

S _target(i, j1)=S _rcmc(i, j1), wherein, (j-N _beta/ 2)≤j1≤(j+N _beta/ 2).

N in formula _betafor antenna radiation pattern main lobe sampling number.Can make like this data orientation to the complete as far as possible reservation of modulation, to reach better deconvolution effect.

Step 7: orientation is to super-resolution processing.

Adopt Lucy-Richardson iterative solution Convolution Formula in step 4, to the target area data S obtaining in step 5 _targetwith whole imaging scene S _rcmc, choose respectively rational iterations and carry out super-resolution iterative solution process of convolution, obtain super-resolution result and be respectively P _targetand P _rcmc.

Step 8: data recombination.

Find target area S _targetin all non-zero corresponding pixel position I={ (i, j) | S _target(i, j) ≠ 0}, then by P _rcmcin Data Update corresponding to pixel in all I of belonging to be P _targetin the data of corresponding point, i.e. P _rcmc(i, j)=P _target(i, j), (i, j) ∈ I, has obtained final imaging results.

Beneficial effect of the present invention: method of the present invention is launched the long-pending linear FM signal of wide bandwidth when large, realizes apart to high-resolution by pulse compression technique; The data estimation global peak letter miscellaneous noise ratio of adjusting the distance away after normal moveout correction, if peak value letter miscellaneous noise ratio is higher than initial set value, the signal of adjusting the distance away after normal moveout correction carries out target area detection, and high order iterative processing is carried out in the target area detecting, and imaging scene is carried out low order iteration; Finally, by target area and contextual data restructuring, complete the imaging processing of whole imaging region; If peak value letter miscellaneous noise ratio lower than initial set value, selects rational iterations to process whole data are unified.Method of the present invention is divided strong target area and the scene of imaging region, has carried out respectively super-resolution processing, in obtaining the higher super-resolution multiple in strong target area, has ensured the contour feature of scene information.

Brief description of the drawings

Fig. 1 is the schematic flow sheet of scanning radar formation method of the present invention.

Fig. 2 is the system construction drawing of scanning radar formation method of the present invention.

Fig. 3 is the scanning radar system parameter table that the specific embodiment of the invention adopts.

Fig. 4 is the Area Objects scene graph adopting in the specific embodiment of the invention.

Fig. 5 be in specific embodiment, adjust the distance migration proofread and correct after data add 10dB white Gaussian noise result.

Fig. 6 is to adding 20 results of the direct super-resolution iteration of data after noise in specific embodiment.

Fig. 7 is to adding 50 results of the direct super-resolution iteration of data after noise in specific embodiment.

Fig. 8 is to adding the target area of data after global threshold detects after noise in specific embodiment.

Fig. 9 is 50 results of target area super-resolution iteration in specific embodiment.

Figure 10 is final imaging results after data recombination in specific embodiment.

Embodiment

The present invention mainly adopts the method for emulation experiment to verify, institute in steps, conclusion all on Matlab2010 checking correct.Below in conjunction with the drawings and specific embodiments, the inventive method is further elaborated.

As shown in Figure 1, detailed process is as follows for the schematic flow sheet of scanning radar formation method of the present invention:

Step 1: echo obtains.

Adopt radar system platform as shown in Figure 2, choose radar simulator system parameter as shown in Figure 3, establish and transmit as linear FM signal wherein, τ be distance to variable, K _r=5 × 10 ¹¹hz/s is chirp rate, T _rwide when=4 μ s are pulse, f ₀=30GHz is carrier frequency.Emulation original face target used scene graph as shown in Figure 4.

Echo expression formula after down coversion is:

$S(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\mathrm{rect}\left[\frac{\mathrm{\&tau;}-{\mathrm{\&tau;}}_d}{T_r}\right]\&times;\exp \left\{\mathrm{j\&pi;}{K}_r{[\mathrm{\&tau;}-{\mathrm{\&tau;}}_d]}^2\right\}\&times;\exp \{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000487731780000022.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}{\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">f</figure-callout>}}_0{\mathrm{\&tau;}}_d\}$

Wherein, σ ₀for target scattering coefficient, ω _a(t) be antenna radiation pattern modulating function, ω _a(t)=A (θ ₀-θ), A (θ) is antenna radiation pattern function, θ ₀for azimuth of target, variation range is-13 °～13 °.Rect[] be that distance is to time window function, τ _d=2R (t)/c is round trip echo delay, R (t) ≈ R ₀-Vtcos θ ₀for any point target P in scene is to Texas tower instantaneous distance, R ₀=30km is carved into Texas tower distance while being scene center point target zero, t is orientation time variable, and V=100m/s is platform movement velocity, c=3 × 10 ⁸m/s is the light velocity.

Step 2: distance is to pulse compression.

According to pulse compression principle, structure distance is to pulse compression frequency matching function

echoed signal, along distance to FFT, in apart from frequency domain-orientation time domain, multiply each other with adaptation function, then the signal expression that contravariant is changed in two-dimensional time-domain is:

S _rc(τ,t)=σ ₀w _a(t)sinc{B[τ-τ _d]}×exp{-j2πf ₀τ _d}

Wherein, sinc{} is apart from pulse pressure response function, and B=2MHz is transmitted signal bandwidth.

Step 3: range walk is judged and proofreaied and correct.

According to simulation parameter in Fig. 3, the range walk amount that can calculate scene center point is range unit

wherein f _r=4MHz is that distance is to sampling rate.Because Δ R< Δ r, so in the present embodiment, do not need to carry out Range Walk Correction.If Δ R> Δ r in actual applications, is being multiplied by the distance frequency domain-orientation time domain data in step 2 after pulse compression frequency matching function, then is being multiplied by Range Walk Correction function

and then carry out distance to IFFT, obtain Range Walk Correction after signal expression be:

$S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\sin c\left\{B\right[\mathrm{\&tau;}-\frac{2R_0}{c}\left]\right\}\&times;\exp \{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000487731780000022.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}{\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">f</figure-callout>}}_0{\mathrm{\&tau;}}_d\}$

Step 4: calculate peak value letter miscellaneous noise ratio.

In this embodiment, because be emulated data, there is no noise, is the situation of more approaching reality, step 3 data after treatment is added to 10dB white Gaussian noise, as shown in Figure 5.

Set initial letter miscellaneous noise ratio SCNR0=15dB, after calculating adds noise, data peaks is believed miscellaneous noise ratio

wherein max (S _n(τ, t)) for adding data S after noise _n(τ, t) maximal value, mean (S _n(τ, t)) be mean value.Obviously SCNR>SCNR0, carries out step 5.

Fig. 6 directly uses 20 results of Lucy-Richardson iterative solution convolution algorithm iteration to data after adding noise, from Fig. 6, can scene obtain good treatment effect, but target in target area could not enough be separated because iterations is lower.Fig. 7 is to adding 50 results of the direct super-resolution iteration of data after noise, can find out that target can separate, but scene information is because high order iteration is destroyed.

Lucy-Richardson iterative solution Convolution Formula is:

${\mathrm{\&sigma;}}_{k+1}={\mathrm{\&sigma;}}_k\left[H^T\left(\frac{r}{H{\mathrm{\&sigma;}}_k}\right)\right]$

Wherein, σ _kbe the k time iteration result, primary iteration value can be chosen σ ₀=H ^tr, H is the convolution kernel matrix of antenna radiation pattern sequence structure, [] ^trepresent transposition computing, r adds data after noise along orientation to a certain range unit data.

Step 5: calculate global threshold.

First according to DP-CFAR rate (CFAR) detection method based on rayleigh distributed, data after adding noise are carried out to CFAR detection, detection criteria is:

$S_{\mathrm{CFAR}}(i,j)=\left\{\begin{array}{c}{S}_N(i,j),\frac{(S_N(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}>\mathrm{<figure-callout\; id="0"\; label="Th"\; filenames="HDA0000487731780000032.png,HDA0000487731780000033.png"\; state="\{\{state\}\}">Th</figure-callout>}\\ 0,\frac{(S_N(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}<\mathrm{Th}\end{array}\right.$

Wherein, S _cFAR(i, j) for CFAR detect after data, μ _cdata mean value in reference window while detection for CFAR, σ _cfor data standard deviation in reference window.Th is the thresholding factor, according to rayleigh distributed probability density function, can calculate Th and false-alarm probability P _fabetween pass be:

$\mathrm{Th}=\frac{(2\sqrt{-\ln P_{\mathrm{fa}}}-\sqrt{\mathrm{\&pi;}})}{\sqrt{4-\mathrm{\&pi;}}}$

False-alarm probability P in this embodiment _fa=1 × 10 ^-6.

Then according to global threshold β=10.8 that maximize the rear data of inter-class variance method calculating CFAR detection.Taking β as threshold value, data are carried out to Preliminary division, tentatively determine that target area is:

S _target=S _cFAR(τ, t), wherein, | S _cFAR(τ, t) |>=β.

Step 6: target area is detected.

Destroyed through the object edge information in the preliminary definite target area of step 5, in the time carrying out orientation to deconvolution, can affect deconvolution result.For can better making target information retain, along S _targetorientation is to carrying out the detection of object edge information.If one party bit location S _target(i, j) non-zero is respectively expanded reservation half beam angle data, that is: S centered by Xiang Yigai unit, orientation to both sides _target(i, j1)=S _rcmc(i, j1), wherein, (j-N _beta/ 2)≤j1≤(j+N _beta/ 2); Wherein, N _betafor antenna radiation pattern main lobe sampling number.Can make like this data orientation to the complete as far as possible reservation of modulation, to reach better deconvolution effect.The target area that detection obtains as shown in Figure 8.

Step 7: orientation is to super-resolution processing.

Adopt Lucy-Richardson iterative solution Convolution Formula in step 4, to the target area data S obtaining in step 5 _targetcarry out super-resolution iterative solution process of convolution 50 times, obtaining super-resolution result is P _target, as shown in Figure 9.To whole imaging scene S _ncarry out super-resolution iterative solution process of convolution 20 times, obtaining super-resolution result is P _n, as shown in Figure 6.

Step 8: data recombination.

Find target area S _targetin all non-zero corresponding pixel position I={ (i, j) | S _target(i, j) ≠ 0}, then by P _nin Data Update corresponding to pixel in all I of belonging to be P _targetin the data of corresponding point, i.e. P _n(i, j)=P _target(i, j), (i, j) ∈ I.Obtained final imaging processing result, as shown in figure 10.

Innovative point of the present invention is that strong target area and scene in imaging region are divided, and in obtaining the higher super-resolution multiple in strong target area, has ensured the contour feature of scene information.

Can find out, the inventive method, in imaging scene domain, has realized scanning radar and has ensured in to strong target area high-resolution imaging the contour feature of scene information.

Claims (2)

1. a scanning radar super-resolution imaging method, specifically comprises the following steps:

Step 1: echo obtains,

If transmit as linear FM signal wherein, τ be distance to time variable, K _rfor chirp rate, T _rwide during for pulse, f ₀for carrier frequency;

From echo expression formula after down coversion of point target P (x, y) reflection be:

$S(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\mathrm{rect}\left[\frac{\mathrm{\&tau;}-{\mathrm{\&tau;}}_d}{T_r}\right]\&times;\exp \left\{\mathrm{j\&pi;}{K}_r{[\mathrm{\&tau;}-{\mathrm{\&tau;}}_d]}^2\right\}\&times;\exp \{-j2\mathrm{\&pi;}{f}_0{\mathrm{\&tau;}}_d\}$

Wherein, σ ₀for target scattering coefficient, ω _a(t) be antenna radiation pattern modulating function, ω _a(t)=A (θ ₀-θ), A (θ) is antenna radiation pattern function, θ ₀for azimuth of target, rect[] be that distance is to time window function, τ _d=2R (t)/c is round trip echo delay, for any point target P in scene is to Texas tower instantaneous distance, R ₀during for target P zero, be carved into Texas tower distance, t is orientation time variable, and V is platform movement velocity, and c is the light velocity;

Step 2: distance to pulse compression,

Structure distance is to pulse compression frequency matching function

echoed signal, along distance to FFT, in apart from frequency domain-orientation time domain, multiply each other with adaptation function, then the signal expression that contravariant is changed in two-dimensional time-domain is:

S _rc(τ,t)=σ ₀w _a(t)sinc{B[τ-τ _d]}×exp{-j2πf ₀τ _d}

Wherein, sinc{} is apart from pulse pressure response function, and B is transmitted signal bandwidth;

Step 3: range walk judgement and correction,

According to the instantaneous distance expression formula in step 1, obtaining range walk amount is Δ R=VT _scos θ ₀, wherein,

for beam scanning residence time, θ _betafor antenna beamwidth, ω is sweep velocity.Judge whether it crosses over range unit

wherein, f _rfor distance is to sampling rate;

If meet Δ R< Δ r, directly carry out step 4; If Δ R> Δ r, is being multiplied by the distance frequency domain-orientation time domain data in step 2 after pulse compression frequency matching function, then is being multiplied by Range Walk Correction function

and then carry out distance to IFFT, obtain Range Walk Correction after signal expression be:

$S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)={\mathrm{\&sigma;}}_0{w}_a\left(t\right)\sin c\left\{B\right[\mathrm{\&tau;}-\frac{2R_0}{c}\left]\right\}\&times;\exp \{-j2\mathrm{\&pi;}{f}_0{\mathrm{\&tau;}}_d\}$

Step 4: calculate peak value letter miscellaneous noise ratio;

After calculating Range Walk Correction, data peaks is believed miscellaneous noise ratio:

$\mathrm{SCNR}=20*{\log }_{10}\frac{\max \left(S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)\right)}{\mathrm{mean}\left(S_{\mathrm{rcmc}}(\mathrm{\&tau;},t)\right)}$

Wherein, max (S _rcmc(τ, t)) be data S after Range Walk Correction _rcmc(τ, t) maximal value, mean (S _rcmc(τ, t)) be mean value; By peak value letter miscellaneous noise ratio SCNR and the initial set value SCNR0 comparison of calculating, if SCNR>=SCNR0 carries out step 5; If SCNR<SCNR0, according to Lucy-Richardson iterative solution Convolution Formula, after the normal moveout correction of adjusting the distance away, the rational iterations of data selection carries out super-resolution imaging processing;

Step 5: calculate global threshold,

According to DP-CFAR rate (CFAR) detection method based on rayleigh distributed, after the normal moveout correction of adjusting the distance away, data are carried out CFAR detection, and detection criteria is:

$S_{\mathrm{CFAR}}(i,j)=\left\{\begin{array}{c}{S}_{\mathrm{rcmc}}(i,j),\frac{(S_{\mathrm{rcmc}}(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}>\mathrm{Th}\\ 0,\frac{(S_{\mathrm{rcmc}}(i,j)-{\mathrm{\&mu;}}_c)}{{\mathrm{\&sigma;}}_c}<\mathrm{Th}\end{array}\right.$

Wherein, S _cFAR(i, j) for CFAR detect after data, μ _cdata mean value in reference window while detection for CFAR, σ _cfor data standard deviation in reference window, Th is the thresholding factor, according to rayleigh distributed probability density function, calculates Th and false-alarm probability P _fabetween pass be:

$\mathrm{Th}=\frac{(2\sqrt{-\ln P_{\mathrm{fa}}}-\sqrt{\mathrm{\&pi;}})}{\sqrt{4-\mathrm{\&pi;}}}$

According to the global threshold β that maximizes the rear data of inter-class variance method calculating CFAR detection;

Taking β as threshold value, data are carried out to Preliminary division, tentatively determine that target area is:

S _target=S _cFAR(τ, t), wherein, | S _cFAR(τ, t) |>=β.

Step 6: target area is detected,

Along S _targetorientation is to carrying out the detection of object edge information;

If one party bit location S _target(i, j) non-zero is respectively expanded half beam angle data of reservation, that is: centered by Xiang Yigai unit, orientation to both sides

S _target(i, j1)=S _rcmc(i, j1), wherein, (j-N _beta/ 2)≤j1≤(j+N _beta/ 2);

N in formula _betafor antenna radiation pattern main lobe sampling number;

Step 7: orientation is to super-resolution processing,

Adopt Lucy-Richardson iterative solution Convolution Formula in step 4, to the target area data S obtaining in step 5 _targetwith whole imaging scene S _rcmc, choose respectively predefined iterations and carry out super-resolution iterative solution process of convolution, obtain super-resolution result and be respectively P _targetand P _rcmc;

Step 8: data recombination;

Find target area S _targetin the corresponding pixel of all non-zero position

then by P _rcmcin Data Update corresponding to pixel in all I of belonging to be P _targetin the data of corresponding point, i.e. P _rcmc(i, j)=P _target(i, j), (i, j) ∈ I, has obtained final imaging results.

2. scanning radar super-resolution imaging method according to claim 1, is characterized in that, the Lucy-Richardson iterative solution Convolution Formula described in step 4 is:

${\mathrm{\&sigma;}}_{k+1}={\mathrm{\&sigma;}}_k\left[H^T\left(\frac{r}{H{\mathrm{\&sigma;}}_k}\right)\right]$

Wherein, σ _kbe the k time iteration result, primary iteration value can be chosen σ ₀=H ^tr, H is the convolution kernel matrix of antenna radiation pattern sequence structure, [] ^trepresent transposition computing, r adds data after noise along orientation to a certain range unit data.

Images