CN101408621A - Ultra-resolution method based on polarization synthetic aperture radar image - Google Patents
Ultra-resolution method based on polarization synthetic aperture radar image Download PDFInfo
- Publication number
- CN101408621A CN101408621A CNA2008102095725A CN200810209572A CN101408621A CN 101408621 A CN101408621 A CN 101408621A CN A2008102095725 A CNA2008102095725 A CN A2008102095725A CN 200810209572 A CN200810209572 A CN 200810209572A CN 101408621 A CN101408621 A CN 101408621A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- synthetic aperture
- aperture radar
- resolution image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention provides a super-resolution method based on polarimetric synthetic aperture radar images and relates to the radar image processing field. The method is provided for solving the disadvantage that the traditional super-resolution algorithms of the polarimetric synthetic aperture radar images can not keep phase information and polarimetric scattering property of a scatterer. The method is realized by the following steps: (1) reading in radar image data; (2) obtaining different scattering components by preprocessing; (3) forming an initial high-resolution image; (4) obtaining super-resolution images of various scattering components from a n time; (5) computing a root-mean-square error; and (6) judging whether the iterative termination condition is satisfied, if not, returning to the step 4, otherwise obtaining the final image. The method can improve the image resolution, and can fully keep the polarimetric scattering property of images and the phase information.
Description
Technical field
The present invention relates to the method that radar image is handled, be specifically related to the method that a kind of polarization synthetic aperture radar image is handled.
Background technology
Polarimetric synthetic aperture radar is the New Type Radar that is based upon on traditional synthetic-aperture radar system, it utilizes different POLARIZATION CHANNEL that the polarization characteristic of Same Scene is provided, can distinguish the parameters such as finer structures, target directing and material composition of object, these information all have the effect that can't estimate in the military and civilian field.
Polarization target decomposition is the fundamental method of polarimetric radar Flame Image Process, the fundamental purpose of Polarization target decomposition is that polarization scattering matrix is resolved into some sums representing different scattering types with coherence matrix or covariance matrix, and each corresponding certain physical meaning.The outstanding advantage of Polarization target decomposition theory is exactly that they mostly have clear and definite physical interpretation.
But in the practical application, be subjected to the restriction of signal bandwidth and antenna size, the resolution of polarization synthetic aperture radar image can not be compared with remote sensing image.Usually in the middle of a width of cloth polarization diagrams picture, each resolution element has comprised several different scattering mechanisms.How to distribute in a resolution element if know these scattering mechanisms, the detailed information of image just can be enhanced so, and this also is improved with regard to the resolution that means image.The super-resolution processing method of traditional polarization synthetic aperture radar image can improve the resolution of image.Yet in processing procedure, the phase information and the polarization information that are comprised in the middle of the original image can be lost.
Summary of the invention
The present invention is in order to solve the deficiency that traditional polarization synthetic aperture radar image super-resolution algorithm can not keep the complete polarization scattering properties of phase information and scatterer, and the ultra-resolution method based on polarization synthetic aperture radar image that proposes.
Based on the ultra-resolution method of polarization synthetic aperture radar image, described method is realized by following steps;
Step 1: read in the polarimetric synthetic aperture radar view data according to data layout;
Step 2: the polarimetric synthetic aperture radar view data that reads is carried out pre-service, and utilize the Polarization target decomposition method to obtain different scattering compositions;
Step 3: each pixel of the original low-resolution image of each scattering composition is divided into 2 * 2 sub-pixel, constitutes initial full resolution pricture;
Step 4: in full resolution pricture, determine to handle window, utilize the polarized spatial correlativity to obtain the proportion of each sub-pixel in handling window, and then obtain the super resolution image of the n time acquisition of each scattering composition;
Step 5: when described n=1, then calculate the root-mean-square error of this super resolution image and initial full resolution pricture; When described n>1, then calculate the root-mean-square error of this super resolution image and the last super resolution image that obtains;
Step 6: judge whether n reaches maximum iteration time n
MaxOr whether root-mean-square error less than the precision ε that sets, and when satisfying aforementioned any one condition, then this super resolution image is final super resolution image; Otherwise, on described super resolution image basis, return execution in step four, and make n=n+1;
Described n is a natural number, and initial value is 1.
The present invention is the ultra-resolution method that is used for polarization synthetic aperture radar image, it has considered the polarization scattering characteristics and the polarized spatial correlativity of atural object comprehensively, not only can improve the resolution of image, and can keep the polarization scattering characteristics and the phase information of image fully, for follow-up image classification and image interpretation and classification provide information more accurately.
Description of drawings
Fig. 1 is the process flow diagram of the inventive method; Fig. 2 is the Polarization target decomposition image of polarimetric synthetic aperture radar image; Fig. 3 is the synoptic diagram of original low-resolution image; Fig. 4 is the synoptic diagram of high-definition picture; The super-resolution Polarization target decomposition image that Fig. 5 obtains for the inventive method.
Embodiment
Embodiment one: in conjunction with Fig. 1 present embodiment is described, the described method of present embodiment is realized by following steps;
Step 1: read in the polarimetric synthetic aperture radar view data according to data layout;
Step 2: the polarimetric synthetic aperture radar view data that reads is carried out pre-service, and utilize the Polarization target decomposition method to obtain different scattering compositions;
Step 3: each pixel of the original low-resolution image of each scattering composition is divided into 2 * 2 sub-pixel, constitutes initial full resolution pricture;
Step 4: in full resolution pricture, determine to handle window, utilize the polarized spatial correlativity to obtain the proportion of each sub-pixel in handling window, and then obtain the super resolution image of the n time acquisition of each scattering composition;
Step 5: when described n=1, then calculate the root-mean-square error of this super resolution image and initial full resolution pricture; When described n>1, then calculate the root-mean-square error of this super resolution image and the last super resolution image that obtains;
Step 6: judge whether n reaches maximum iteration time n
MaxOr whether root-mean-square error less than the precision ε that sets, and when satisfying aforementioned any one condition, then this super resolution image is final super resolution image; Otherwise, on described super resolution image basis, return execution in step four, and make n=n+1;
Described n is a natural number, and initial value is 1.
Embodiment two: present embodiment and embodiment one difference are that the Polarization target decomposition method adopts the Pauli decomposition method in the step 2; It is the most classical relevant target decomposition method that Pauli decomposes, under the situation of reciprocity, scattering matrix is decomposed into odd scattering, even scattering and with horizontal direction the even scatterings at 45 degree inclination angles is arranged,
Parameter wherein
Parameter
Parameter
Other step is identical with embodiment one.
Embodiment three: present embodiment is described in conjunction with Fig. 2~Fig. 5, present embodiment and embodiment one difference are that step 4 utilizes the very strong polarized spatial correlativity of neighbor, obtain the proportion of each sub-pixel in handling window, and obtain super resolution image.
Fig. 3 is original low-resolution image, handles through super-resolution, and each low-resolution pixel is (as A
5) be divided into four high resolving power sub-pixels (as A
51, A
52, A
53, A
54), can obtain high-definition picture shown in Figure 4.
Get sub-pixel A
51With its 3 * 3 neighborhoods, shown in the heavy line square frame among Fig. 4, define sub-pixel A
51Space correlation coefficient be:
R
51=|α
51-α
14|
2+|α
51-α
23|
2+|α
51-α
24|
2+|α
51-α
42|
2
(2)
+|α
51-α
52|
2+|α
51-α
44|
2+|α
51-α
53|
2+|α
51-α
54|
2
In like manner can define sub-pixel A
52, A
53And A
54Space correlation coefficient:
R
52=|α
52-α
23|
2+|α
52-α
24|
2+|α
52-α
33|
2+|α
52-α
51|
2
(3)
+|α
52-α
61|
2+|α
52-α
53|
2+|α
52-α
54|
2+|α
52-α
63|
2
R
53=|α
53-α
42|
2+|α
53-α
51|
2+|α
53-α
52|
2+|α
53-α
44|
2
(4)
+|α
53-α
54|
2+|α
53-α
72|
2+|α
53-α
81|
2+|α
53-α
82|
2
R
54=|α
54-α
51|
2+|α
54-α
52|
2+|α
54-α
61|
2+|α
54-α
53|
2
(5)
+|α
54-α
63|
2+|α
54-α
81|
2+|α
54-α
82|
2+|α
54-α
91|
2
R
5i, i=1,2,3,4 have reflected sub-pixel A
5iWith the correlativity of sub-pixel in its 3 * 3 neighborhoods, R
5iValue more little, sub-pixel A is described
5iApproaching more with the sub-pixel around it, correlativity is strong more.Order:
R
5=R
51+R
52+R
53+R
54 (6)
It has reflected as low-resolution pixel A
5Be divided into four high resolving power sub-pixel A
51, A
52, A
53, A
54The time, the degree of closeness between four sub-pixels and its sub-pixel on every side, R
5Value more little, the decomposition method that this sub pixel is described makes sub-pixel and the sub-pixel around its approaching more, correlativity is strong more, and is just reasonable more.And four sub-pixels will finding the solution have following relation of plane:
α
51+α
52+α
53+α
54=α
5 (7)
So in fact the super-resolution of carrying out is handled is exactly to find the solution under the restrictive condition that satisfies formula (7), makes R
5α when reaching minimal value
5i, i=1,2,3,4 value.Can utilize lagrange's method of multipliers to find the solution such conditional extremum.Order:
R
5+λ(α
51+α
52+α
53+α
54-α
5)=0 (8)
In theory, following formula is asked about α
5i, i=1,2,3,4 partial derivative is added the restrictive condition of formula (7), can solve α
5i, I=1,2,3,4 and the value of λ.And in fact, function of a complex variable
f(z)=|z|
2 (9)
Remove outside (0,0) point can be led, all can not lead at other some places, so the conditional extremum of formula (9) can not directly be found the solution out.We notice:
f(z)=|z|
2=|x+yj|
2=x
2+y
2=f(x,y) (10)
Be about x, the binary function of y, and can lead everywhere.And the restrictive condition of formula (7) can be write as:
R in the following formula and I represent to get real part and imaginary part respectively.So the constrained extremal problem of finding the solution can change under the restrictive condition that satisfies formula (11), asks R
5Minimal value.R wherein
5Be with α
5iRAnd α
5iI, i=1,2,3,4 is the real function of independent variable.Utilize lagrange's method of multipliers, order:
R
5+λ
1(α
51R+α
52R+α
53R+α
54R-α
5R)+λ
2(α
51I+α
52I+α
53I+α
54I-α
5I)=0 (12)
Following formula is asked about α
5iRAnd α
5iI, i=1,2,3,4 partial derivative is added the restrictive condition of formula (11), can solve λ
1, λ
2, α
5iRAnd α
5iI, i=1,2,3,4 value is respectively:
C in the following formula
i, i=1,2,3,4 is α
5i3 * 3 neighborhoods in remove α
51, α
52, α
53, α
54The time, all the other 5 sub-pixel sums.Order again
α
5i=α
5iR+α
5iIj,i=1,2,3,4 (14)
It promptly is the value of each sub-pixel.Fig. 2 and Fig. 5 be not for adopting the effect comparison diagram of this method and employing this method.Other step is identical with embodiment one.
Embodiment four: present embodiment and embodiment one difference are that step 4, step 5 and step 6 are interative computation.
The analysis and the derivation of equation in the step 4 are all carried out on high-definition picture, the result who obtains is the high-definition picture that a width of cloth polarized spatial correlativity improves, step 5 and step 6 are judged, be that identical processing is carried out on the basis with the result who obtains again, so just formed the process of iteration.
If I
N-1Be that initial full resolution pricture or last iteration obtains high-definition picture, and I
nBe the result that obtains of iteration this time, then the root-mean-square error of two width of cloth images is:
M in the formula (15) and N are respectively the line number and the columns of image.As e during less than predefined a certain constant ε, iteration finishes.If iteration convergence, but can not reach we predefined precision ε the time, iterative process will not be sustained and can finish, and in order to address this problem, maximum iteration time n will be set
Max, when iterations reaches n
MaxThe time, iterative process finishes.
Along with the increase of iterations, root-mean-square error descends on the whole, illustrates that whole iterative process is a convergent, and the polarized spatial correlativity super-resolution algorithm that this paper proposes is effective; Just tend towards stability substantially after dropping to a certain degree, this just illustrates that this algorithm is limited improving aspect the polarized spatial coherence ability, tells us reasonably to choose ε and n simultaneously
MaxCan effectively reduce program runtime, improve program run efficient.Other step is identical with embodiment one.
Claims (2)
1,, it is characterized in that described method is realized by following steps based on the ultra-resolution method of polarization synthetic aperture radar image;
Step 1: read in the polarimetric synthetic aperture radar view data according to data layout;
Step 2: the polarimetric synthetic aperture radar view data that reads is carried out pre-service, and utilize the Polarization target decomposition method to obtain different scattering compositions;
Step 3: each pixel of the original low-resolution image of each scattering composition is divided into 2 * 2 sub-pixel, constitutes initial full resolution pricture;
Step 4: in full resolution pricture, determine to handle window, utilize the polarized spatial correlativity to obtain the proportion of each sub-pixel in handling window, and then obtain the super resolution image of the n time acquisition of each scattering composition;
Step 5: when described n=1, then calculate the root-mean-square error of this super resolution image and initial full resolution pricture; When described n>1, then calculate the root-mean-square error of this super resolution image and the last super resolution image that obtains;
Step 6: judge whether n reaches maximum iteration time n
MaxOr whether root-mean-square error less than the precision ε that sets, and when satisfying aforementioned any one condition, then this super resolution image is final super resolution image; Otherwise, on described super resolution image basis, return execution in step four, and make n=n+1;
Described n is a natural number, and initial value is 1.
2, the ultra-resolution method based on polarization synthetic aperture radar image according to claim 1 is characterized in that step 4, step 5 and step 6 are interative computation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008102095725A CN101408621B (en) | 2008-11-28 | 2008-11-28 | Ultra-resolution method based on polarization synthetic aperture radar image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008102095725A CN101408621B (en) | 2008-11-28 | 2008-11-28 | Ultra-resolution method based on polarization synthetic aperture radar image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101408621A true CN101408621A (en) | 2009-04-15 |
CN101408621B CN101408621B (en) | 2011-12-28 |
Family
ID=40571714
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008102095725A Expired - Fee Related CN101408621B (en) | 2008-11-28 | 2008-11-28 | Ultra-resolution method based on polarization synthetic aperture radar image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101408621B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102053248A (en) * | 2010-11-12 | 2011-05-11 | 哈尔滨工业大学 | Polarimetric synthetic aperture radar image target detection method based on quotient space granular computing |
CN102207548A (en) * | 2010-03-31 | 2011-10-05 | 中国科学院电子学研究所 | MIMO SAR imaging method by employing minimum mean square error estimation |
CN102645651A (en) * | 2012-04-23 | 2012-08-22 | 电子科技大学 | SAR (synthetic aperture radar) tomography super-resolution imaging method |
CN102967858A (en) * | 2012-11-14 | 2013-03-13 | 电子科技大学 | Radar foresight super-resolution imaging method |
CN104122554A (en) * | 2014-07-31 | 2014-10-29 | 西安电子科技大学 | Method for extracting characteristics of attribute scattering center of high-resolution SAR (synthetic aperture radar) image targets |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100510772C (en) * | 2006-01-23 | 2009-07-08 | 武汉大学 | Small target super resolution reconstruction method for remote sensing image |
CN100510773C (en) * | 2006-04-14 | 2009-07-08 | 武汉大学 | Single satellite remote sensing image small target super resolution ratio reconstruction method |
-
2008
- 2008-11-28 CN CN2008102095725A patent/CN101408621B/en not_active Expired - Fee Related
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102207548A (en) * | 2010-03-31 | 2011-10-05 | 中国科学院电子学研究所 | MIMO SAR imaging method by employing minimum mean square error estimation |
CN102207548B (en) * | 2010-03-31 | 2013-09-25 | 中国科学院电子学研究所 | MIMO SAR imaging method by employing minimum mean square error estimation |
CN102053248A (en) * | 2010-11-12 | 2011-05-11 | 哈尔滨工业大学 | Polarimetric synthetic aperture radar image target detection method based on quotient space granular computing |
CN102053248B (en) * | 2010-11-12 | 2012-12-19 | 哈尔滨工业大学 | Polarimetric synthetic aperture radar image target detection method based on quotient space granular computing |
CN102645651A (en) * | 2012-04-23 | 2012-08-22 | 电子科技大学 | SAR (synthetic aperture radar) tomography super-resolution imaging method |
CN102645651B (en) * | 2012-04-23 | 2013-12-11 | 电子科技大学 | SAR (synthetic aperture radar) tomography super-resolution imaging method |
CN102967858A (en) * | 2012-11-14 | 2013-03-13 | 电子科技大学 | Radar foresight super-resolution imaging method |
CN102967858B (en) * | 2012-11-14 | 2014-03-05 | 电子科技大学 | Radar foresight super-resolution imaging method |
CN104122554A (en) * | 2014-07-31 | 2014-10-29 | 西安电子科技大学 | Method for extracting characteristics of attribute scattering center of high-resolution SAR (synthetic aperture radar) image targets |
CN104122554B (en) * | 2014-07-31 | 2016-08-17 | 西安电子科技大学 | The attribute scattering center feature extracting method of High Resolution SAR Images target |
Also Published As
Publication number | Publication date |
---|---|
CN101408621B (en) | 2011-12-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Camps-Valls et al. | Physics-aware Gaussian processes in remote sensing | |
CN102194099B (en) | For senior background estimation technique and the circuit of EO-1 hyperion object detection method | |
CN102938072B (en) | A kind of high-spectrum image dimensionality reduction and sorting technique based on the tensor analysis of piecemeal low-rank | |
Wu et al. | An error-bound-regularized sparse coding for spatiotemporal reflectance fusion | |
CN101408621B (en) | Ultra-resolution method based on polarization synthetic aperture radar image | |
CN102842124B (en) | Multispectral image and full-color image fusion method based on matrix low rank decomposition | |
CN104751181B (en) | A kind of high spectrum image solution mixing method based on relative abundance | |
CN105069482A (en) | Hyperspectral remote sensing image classification method based on spatial regularization manifold learning algorithm | |
CN104978573A (en) | Non-negative matrix factorization method applied to hyperspectral image processing | |
CN108154094A (en) | Hyperspectral image unsupervised waveband selection method based on subinterval division | |
CN103425995B (en) | Hyperspectral image classification method based on region similarity low rank expression dimension reduction | |
CN102073039B (en) | Thermal infrared hyperspectral emissivity simulation method and system | |
Girard et al. | Sparse representations and convex optimization as tools for LOFAR radio interferometric imaging | |
CN105913092A (en) | Figure regular hyperspectral image band selection method based on subspace learning | |
Qiao et al. | Estimating maize LAI by exploring deep features of vegetation index map from UAV multispectral images | |
Zhang et al. | Improving model parsimony and accuracy by modified greedy feature selection in digital soil mapping | |
CN107274360A (en) | A kind of high spectrum image denoising method based on Fisher dictionary learnings, low-rank representation | |
CN106097252A (en) | High spectrum image superpixel segmentation method based on figure Graph model | |
CN102789641B (en) | Based on high spectrum image and the infrared image fusion method of figure Laplce | |
Meng et al. | Improving the spatial resolution of hyperspectral image using panchromatic and multispectral images: An integrated method | |
Shetty et al. | Performance evaluation of dimensionality reduction techniques on hyperspectral data for mineral exploration | |
Li et al. | A multispectral remote sensing data spectral unmixing algorithm based on variational Bayesian ICA | |
Du et al. | A novel unsupervised bands selection algorithm for hyperspectral image | |
Song et al. | Physical-aware radar image synthesis with projective network | |
Hamzeh et al. | Retrieval of Sugarcane Leaf Area Index From Prisma Hyperspectral Data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20111228 Termination date: 20121128 |