CN103698749B - A kind of method utilizing small data set SAR image sequential extraction procedures Permanent scatterers - Google Patents

Abstract

The invention provides a kind of method utilizing small data set SAR image sequential extraction procedures Permanent scatterers.Technical scheme utilizes the sequence comprising 8 to 15 width SAR image, first carries out the pre-service of SAR image sequence, then carries out spatial domain signal to noise ratio and estimate and the estimation of time domain signal to noise ratio, judge, obtain PS and extract result finally by PS.The invention has the beneficial effects as follows: realization is simple, accuracy is high, applied widely.

Description

A kind of method utilizing small data set SAR image sequential extraction procedures Permanent scatterers

Technical field

The present invention relates to the interleaving techniques field of microwave remote sensing and signal transacting, particularly one utilizes small data set SAR(Synthetic Aperture Radar, synthetic-aperture radar) method of image zooming-out Permanent scatterers.

Background technology

PS(Permanent Scatterers, Permanent scatterers) technology is a kind of typical long-term sequence differential interferometry technology, by process Spaceborne SAR System not in the same time to the same area obtain some width SAR image, millimetre-sized Ground Deformation information can be obtained, therefore be widely used in Natural calamity monitoring, Ground Deformation extraction, city prevention and cure of settlement etc., there is important using value.

PS technical finesse flow process comprises PS extraction, PS network struction, PS network parameter solve and optimize, the step such as Surface Deformation Field extraction, wherein, the processing accuracy of the follow-up each step of accuracy impact that PS extracts and sane degree, finally affect the extraction accuracy of Surface Deformation Field.

According to Italian scholar A.Ferretti to the definition of PS, PS refers to that, in longer time range, scattering phase place keeps stable point target.Existing PS extracting method has amplitude departure method and coefficient of coherence method etc.The basic thought of amplitude departure method is, due under given conditions, scattering amplitude and the scattering phase place of pixel have close degree of stability, and the amplitude stability by pixel extracts PS.This method needs the SAR image of more than process tens width to obtain effectively to extract result usually.Coefficient of coherence method utilizes the average coherence coefficient of pixel on whole image sequence as the evaluation index of its scattering stability, but the defect of this algorithm is the block clustering phenomena of PS easily occurred.By the restriction of SAR system data retrieval capabilities, in real world applications, often need the SAR image sequential extraction procedures regional deformation field utilizing small data set (picture number is 8 to 15 width).Be difficult to obtain gratifying extraction result because existing PS extracting method is applied to small data set SAR image sequence, at this moment we will adopt new, effective PS extracting method.

Summary of the invention

The technical problem to be solved in the present invention is, provides a kind of that be applicable to small data set, that effective PS extracts method.This method make use of the time-space domain information of SAR image data, thus decreases the requirement to SAR image number, extracts PS realize breaking through utilizing small data set SAR image.

The basic ideas of technical solution of the present invention are, because pixel signal to noise ratio and scattering phase stability exist positive correlation, first utilize the time-space domain information of SAR image data accurately to estimate the signal to noise ratio of each pixel in SAR image, then judge each pixel whether as PS according to the size of signal to noise ratio.

Technical scheme of the present invention is, based on Spaceborne SAR System not in the same time to M (8≤M≤15) the width SAR image sequence S that the same area obtains _i(1≤i≤M), carry out following process:

Step one: SAR image sequence pre-service.

The pre-service of SAR image sequence comprises SAR image sequence registration, resampling and radiant correction three steps, concrete grammar is with reference to Ph.D. Dissertation's " research based on Permanent scatterers radar differential interferometry search coverage Ground Deformation " (Southwest Jiaotong University, Chen Qiang, 2006.8,40th page to the 66th page) process.

Step 2: signal to noise ratio is estimated.This step comprises spatial domain signal to noise ratio and estimates and the estimation of time domain signal to noise ratio.

Wherein, spatial domain signal to noise ratio is estimated to comprise following three steps:

(1) step, the filtering of SAR image sequence time domain average.

To pretreated M width SAR image sequence S _i, utilize following formula to carry out time domain average filtering, obtain average amplitude figure

$\stackrel{\‾}{S}=\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{S}_i$

(2) step, estimates spatial domain signal to noise ratio.To average amplitude figure following formula is utilized to estimate the noise intensity of a jth pixel j represents average amplitude figure pixel number.

${\hat{I}}_j=\frac{1}{K}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{a_{\mathrm{jk}}}^2,a_{\mathrm{jk}}\∈{\mathrm{\Ω}}_j$

Wherein, Ω _jfor the clutter estimation region of a jth pixel, use spatial domain windowing method selected (concrete grammar can with reference to " synthetic aperture radar target etection theory, algorithm and application " the 122 to 125 page).A _jkfor average amplitude figure a jth pixel clutter estimation region Ω _jthe amplitude of an interior kth pixel, K is clutter estimation region Ω _jpixel number.

Suppose that a jth pixel is at average amplitude figure amplitude be P _j, then its spatial domain signal to noise ratio for:

${S\hat{C}R}_{1_j}=\left\{\begin{array}{cc}10\log \left(\frac{{P_j}^2-{\hat{I}}_j}{{\hat{I}}_j}\right)& {P_j}^2>{\hat{I}}_j\\ 0& {P_j}^2\≤{\hat{I}}_j\end{array}\right.;$

In addition, time domain signal to noise ratio is estimated to comprise following two steps:

(1) step, estimates the amplitude deviation value of pixel.

Based on the SAR image sequence obtained after pre-service, following formula is utilized to estimate the amplitude deviation value of a jth pixel, j=1,2 ..., J:

${\hat{D}}_j=\frac{\sqrt{\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a_{\mathrm{ij}}}^2-{\left(\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a}_{\mathrm{ij}}\right)}^2}}{\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a}_{\mathrm{ij}}}$

Wherein, a _ijfor a jth pixel is at the SAR image amplitude in i moment.

(2) step, utilizes estimate the time domain signal to noise ratio of a jth pixel be shown below:

${S\hat{C}R}_{2_j}=20\log \left(\frac{{\hat{D}}_j^2}{2}\right)$

Step 3: PS judges.

Setting signal to noise ratio threshold value is 8db(decibel, decibel), according to the spatial domain signal to noise ratio of a jth pixel with time domain signal to noise ratio judge it whether as PS:

As with all be greater than threshold value, then judge that a jth pixel is as PS; Otherwise, judge that this pixel is as non-PS.After so all pixels being judged, final PS can be obtained and extract result.

The invention has the beneficial effects as follows: realization is simple, accuracy is high, applied widely.Owing to make use of spatial domain and the time-domain information of SAR image data, the present invention on extraction accuracy higher than existing extracting method.The present invention relaxes the requirement to SAR image number, even if utilize the SAR image sequence of small data set, also can obtain and extract result more accurately, and therefore realization is simple, usable range is wide, and this is that existing method is difficult to realize.

Accompanying drawing explanation

Fig. 1 is principle process schematic diagram of the present invention;

Fig. 2 utilizes the embodiment of the present invention to carry out the result figure tested;

Fig. 3 is for carrying out the conclusion of PS Accuracy evaluation to experimental result shown in Fig. 2.

Embodiment

Fig. 1 is principle process schematic diagram of the present invention.Technical scheme comprises carries out the pre-service of SAR image sequence to based on satellite-borne SAR image sequence, then carries out spatial domain signal to noise ratio and estimates and the estimation of time domain signal to noise ratio, judged, obtain PS and extract result by PS.

Fig. 2 utilizes the embodiment of the present invention to carry out the result figure tested.Wherein, SAR image corresponding region is the Anaheim city of the U.S., horizontal ordinate be orientation to (km), ordinate be distance to (km), the PS of extraction indicates with white round dot.Wherein, the SAR image sequence utilized comprises 12 width images, i.e. M=12, hollow windowing method is used to select clutter estimation region, the total size of rectangular window is 12 pixel × 12 pixels, window hollow area size 5 pixel × 5 pixel, i.e. pixel-5 pixel × 5, pixel number K=119=12 pixel × 12 pixel of clutter estimation region.

Fig. 3 is for carrying out the conclusion of PS Accuracy evaluation to experimental result shown in Fig. 2.

The PS accuracy index adopted is that (circular is with reference to international conference paper " Testing And Evaluation Of Permanent Scatterers Candidates Selection Methods " for PS network optimization rate, the National University of Defense Technology, Chen Shao strength, 2013.05).

SAR image sequence in utilizing Fig. 2 to test, adopt existing amplitude departure method and coefficient of coherence method (with reference to National University of Defense Technology grandson Chiron Ph.D. Dissertation " SAR chromatography and difference chromatography imaging technique are studied " the 30th page to the 40th page) and this method to extract PS respectively, then calculate the PS network optimization rate index of each method.Result of calculation shows, when process 12 scape SAR image, the PS accuracy index of this method is far above existing amplitude departure method and coefficient of coherence method, and this has absolutely proved the superiority that this method is utilizing small data set SAR image to extract PS.

Claims (2)

1. utilize a method for small data set SAR image sequential extraction procedures Permanent scatterers, based on Spaceborne SAR System not in the same time to the M width SAR image sequence S that the same area obtains _i, wherein SAR represents synthetic-aperture radar, 8≤M≤15, and 1≤i≤M, is characterized in that, comprises the steps:

Step one: SAR image sequence pre-service:

The pre-service of SAR image sequence comprises SAR image sequence registration, resampling and radiant correction three steps;

Step 2: signal to noise ratio is estimated:

This step comprises spatial domain signal to noise ratio and estimates and the estimation of time domain signal to noise ratio;

Wherein, spatial domain signal to noise ratio is estimated to comprise following two steps:

(1) step, the filtering of SAR image sequence time domain average:

To pretreated M width SAR image sequence S _i, utilize following formula to carry out time domain average filtering, obtain average amplitude figure

$\stackrel{\‾}{S}=\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{S}_i$

(2) step, estimate spatial domain signal to noise ratio:

To average amplitude figure following formula is utilized to estimate the noise intensity of a jth pixel j represents average amplitude figure pixel number;

${\hat{I}}_j=\frac{1}{K}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{a_{\mathrm{jk}}}^2,a_{\mathrm{jk}}\∈{\mathrm{\Ω}}_j$

Wherein, Ω _jfor the clutter estimation region of a jth pixel, windowing method in spatial domain is used to select; a _jkfor average amplitude figure a jth pixel clutter estimation region Ω _jthe amplitude of an interior kth pixel, K is clutter estimation region Ω _jpixel number;

Suppose that a jth pixel is at average amplitude figure amplitude be P _j, then its spatial domain signal to noise ratio for:

${\widehat{\mathrm{SCR}}}_{1_j}=\left\{\begin{array}{cc}10\log \left(\frac{{P_j}^2-{\hat{I}}_j}{{\hat{I}}_j}\right)& {P_j}^2>{\hat{I}}_j\\ 0& {P_j}^2\≤{\hat{I}}_j\end{array}\right.;$

Wherein, time domain signal to noise ratio is estimated to comprise following two steps:

(1) step, estimate the amplitude deviation value of pixel:

Based on the SAR image sequence obtained after pre-service, following formula is utilized to estimate the amplitude deviation value of a jth pixel, j=1,2 ..., J:

${\hat{D}}_j=\frac{\sqrt{\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a_{\mathrm{ij}}}^2-{\left(\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a}_{\mathrm{ij}}\right)}^2}}{\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{a}_{\mathrm{ij}}}$

Wherein, a _ijfor a jth pixel is at the SAR image amplitude in i moment;

(2) step, utilizes estimate the time domain signal to noise ratio of a jth pixel be shown below:

${\widehat{\mathrm{SCR}}}_{2_j}=20\log \left(\frac{{\hat{D}}_j^2}{2}\right);$

Step 3: PS judges;

Setting signal to noise ratio threshold value is 8 decibels, according to the spatial domain signal to noise ratio of a jth pixel with time domain signal to noise ratio judge it whether as PS:

As with all be greater than threshold value, then judge that a jth pixel is as PS; Otherwise, judge that this pixel is as non-PS; After so all pixels being judged, final PS can be obtained and extract result; Wherein PS represents Permanent scatterers.

2. the method utilizing small data set SAR image sequential extraction procedures Permanent scatterers according to claim 1, it is characterized in that, when using hollow windowing method to select clutter estimation region, the total size of rectangular window is 12 pixel × 12 pixels, window hollow area size 5 pixel × 5 pixel.