CN105869146B - SAR image change detection based on conspicuousness fusion - Google Patents
SAR image change detection based on conspicuousness fusion Download PDFInfo
- Publication number
- CN105869146B CN105869146B CN201610164208.6A CN201610164208A CN105869146B CN 105869146 B CN105869146 B CN 105869146B CN 201610164208 A CN201610164208 A CN 201610164208A CN 105869146 B CN105869146 B CN 105869146B
- Authority
- CN
- China
- Prior art keywords
- contrast
- pixel
- notable
- indicate
- sar 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Abstract
The invention discloses a kind of SAR image change detections based on saliency fusion, mainly solve influence of the SAR image variation detection vulnerable to speckle noise in existing method, and the problem that detection accuracy is not high.The specific steps of the present invention are as follows: (1) inputting SAR image;(2) it filters;(3) the log ratio value of pixel is calculated;(4) global notable figure is constructed;(5) the contrast saliency value of different scale is calculated;(6) local notable figure is constructed;(7) local notable figure and global notable figure are merged;(8) fuzzy clustering;(9) output variation testing result.The present invention has the advantages that the influence of noise robustness for changing detection to SAR image is good and detection accuracy is high.
Description
Technical field
The invention belongs to technical field of image processing, further relate in the change detection techniques field of remote sensing images
A kind of SAR image change detection based on conspicuousness fusion.The present invention realizes the synthesis hole obtained to two width difference phases
The detection in the image change region diameter radar SAR (Synthetic Aperture Radar) can cover and utilization, nature in atural object
Disaster monitoring and assessment, urban planning, the fields such as map rejuvenation are widely applied.
Background technique
SAR image variation detection refers to using two width SAR images of different phase areals come detection and analysis ground
Situation of change.Since SAR technology has the characteristics that round-the-clock, round-the-clock work compared with ordinary optical remote sensing technology, so that
SAR image variation detection has a wide range of applications in national economy and Military Application field.In recent years, it is carried out using SAR image
Variation detection is highly valued in international remote sensing fields, has become one of main direction of studying of image procossing.
" the Two-Stage Change Detection for Synthetic that Miriam Cha et al. is delivered at it
Aperture Radar”(IEEE Transactions on Geoscience&Remote Sensing,2015,53(12):
A kind of SAR image change detection of two-phase method 6547-6560) is proposed in paper.The first stage of this method is first
The pixel pair as caused by the target of large scale with unequal variance is detected, initial variation detection figure is obtained;There to be phase again
Etc. variances or close variance pixel to the input as second stage.Second stage is detected using Berger correlation estimation
There is the region of minor change in SAR image, so that it is determined that final change-detection images.Since this method only considers two images
The variance of respective pixel, although higher pixel contrast can be obtained in initial detecting figure, this method is still had
Shortcoming be that this method does not account for the spatial information of SAR image, so detection accuracy is not high asks there are still variation
Topic.
Xian Electronics Science and Technology University is in patent " method for detecting change of remote sensing image based on image co-registration " (Shen of its application
Please the patent No.: 201210414782.4, publication number: 102968790A) in propose a kind of remote sensing images based on image co-registration
Change detecting method.This method extracts the difference disparity map and ratio difference figure of variation front and back two images first, and extracts difference
It is worth the variance matrix of image;With KI thresholding method, optimal threshold T is obtained;According to optimal threshold T and variance matrix by differential chart
It is merged with ratio figure, obtains fused final disparity map D;Carrying out classification number to the gray value of resulting final disparity map D is 2
Fuzzy C-means clustering segmentation, by cluster centre value it is biggish it is a kind of as variation class, another kind of is non-changing class, obtain
Change testing result.Although this method has preferable detection for the disparity map with conspicuous object, this method is still deposited
Shortcoming be that for the disparity map detection effect with unconspicuous variation targets and bad, and cannot effectively inhibit
The influence of speckle noise.
Patent " SAR image change detection based on low-rank matrix decomposed " of the Xian Electronics Science and Technology University in its application
It proposes and a kind of is decomposed based on low-rank matrix in (number of applying for a patent: 201210193347.3, publication number: 102722892A)
SAR image change detection.This method carries out drop spot pretreatment to two width SAR image to be measured first, obtains more smooth
SAR image;Then the log ratio of the two images after construction drop spot, then carries out low-rank sparse decomposition for log ratio,
Obtain the low-rank part and sparse part of log ratio;Column are pressed again by sparse Partial Transformation into sparse matrix;Finally calculated with K mean value
Method clusters obtained sparse matrix, obtains final variation testing result.Although this method can be detected accurately
To region of variation, but the shortcoming that this method still has is, due to not accounting for detailed information and space neighborhood information,
So the problem unsatisfactory there are still detection accuracy.
Summary of the invention
It is an object of the invention to overcome above-mentioned the deficiencies in the prior art, a kind of SAR based on conspicuousness fusion is proposed
Image change detection method.The present invention compared with other SAR image change detection techniques, effectively can inhibit spot to make an uproar in existing
Sound, and improve the detection accuracy of SAR image variation detection.
The present invention realizes that the thinking of above-mentioned purpose is: first scheming to the SAR for the different phase areals of input being registrated
Log ratio operation is carried out as carrying out non-local mean filtering, then to filtered SAR image, then constructs global notable figure,
The contrast saliency value of different scale is calculated filtered SAR image simultaneously, then constructs local notable figure, it will be global significant
Figure and local notable figure merge to obtain conspicuousness fusion figure, finally carry out fuzzy C-means clustering to conspicuousness fusion figure and are become
Change testing result.
The specific steps that the present invention realizes include the following:
(1) SAR image is inputted:
The SAR image I that two width of input areal difference phase have been registrated, have corrected1And I2;
(2) it filters:
Using non-local mean filtering device to two width SAR image I of input1And I2It is filtered, obtains filtered respectively
SAR image X1And X2;
(3) the log ratio value of pixel is calculated:
To filtered SAR image X1And X2It takes the gray level of respective pixel to carry out log ratio operation, obtains log ratio
Disparity map DL;
(4) according to the following formula, the global saliency value of the pixel of log ratio disparity map is calculated:
Wherein, D (i, j) indicates the global saliency value of the pixel of the i-th row jth column in log ratio disparity map, I (i, j) table
Show that the gray value for the pixel that the i-th row jth arranges in log ratio disparity map, I (i, j)=0,1,2 ..., 255, ∑ indicate summation behaviour
Make,Indicate optional sign, ∈ expression belongs to symbol, and I (m, n) is indicated in log ratio disparity map except the m except I (i, j)
The gray value for the pixel that row n-th arranges, I (m, n)=0,1,2 ..., 255, DLIndicate log ratio disparity map, | | | | expression takes
Gray value distance operation;
(5) the contrast saliency value of different scale is calculated:
(5a) is respectively to filtered SAR image X1And X21 × 1 pixel is taken, using pixel contrast method, construction 1 × 1
The contrast notable figure of pixel dimension;
(5b) is with the sliding window of 3 × 3 pixels respectively to filtered image X1And X2Take block;
Pixel in all blocks is pulled into a column by column respectively by (5c), obtains three-dimensional matrice L1And L2;
(5d) uses vector contrast method, by three-dimensional matrice L1And L2The contrast for constructing 3 × 3 pixel dimensions is significant
Value;
(5e) is with the sliding window of 5 × 5 pixels respectively to filtered image X1And X2Take block;
Pixel in all blocks is pulled into a column by column respectively by (5c), obtains three-dimensional matrice L3And L4;
(5g) uses vector contrast method, by three-dimensional matrice L3And L4Construct the contrast saliency value of 5 × 5 pixel dimensions;
(6) weighted mean method for using following formula, merges the contrast notable figure of different scale, obtains local notable figure:
S'=α0·S0+α1·S1+α2·S2
Wherein, S' indicates local notable figure, S0Indicate the contrast notable figure of 1 × 1 pixel dimension, S1Indicate 3 × 3 pixels
The notable figure of scale, S2Indicate the contrast notable figure of 5 × 5 pixel dimensions, α0、α1、α2Indicate weighting coefficient, α0+α1+α2, 0≤
α0、α1、α2≤1;
(7) according to the following formula, amalgamation of global notable figure and local notable figure, obtain conspicuousness fusion figure:
S (i, j)=D (i, j) exp (- S'(i, j))
Wherein, S indicates that conspicuousness fusion figure, D indicate global notable figure, and S' indicates local notable figure, and exp indicates index behaviour
Make symbol, i and j respectively indicate conspicuousness fusion figure, in global notable figure and local notable figure the row of the pixel of same position and
Column coordinate value;
(8) fuzzy clustering:
(8a) uses fuzzy C-means clustering method, clusters to conspicuousness fusion figure, obtains every in conspicuousness fusion figure
Being under the jurisdiction of non-changing class and changing class for a pixel is subordinate to angle value;
(8b) is judged to the ownership of the pixel for being subordinate to angle value for being subordinate to angle value and being greater than variation class of non-changing class for non-changing class picture
Element;The pixel for being subordinate to angle value for being subordinate to angle value and being less than variation class of non-changing class is judged to the ownership of to change class pixel;Obtain SAR figure
The variation testing result of picture;
(9) output variation testing result.
Compared with the prior art, the present invention has the following advantages:
First, due to the present invention local notable figure is calculated using the contrast notable figure in conjunction with different scale method,
The image block of 1 × 1,3 × 3 and 5 × 5 three kinds of different scales is chosen respectively to calculate the contrast of center pixel, to combine
The neighborhood information of image, overcomes and does not account for neighborhood information in the prior art, and SAR image variation detection is caused to be made an uproar vulnerable to spot
The influence of sound so that the present invention is conducive to inhibit coherent speckle noise, and then reduces the false detection rate of variation detection.
Second, since the present invention is using the method for conspicuousness fusion, global notable figure and local notable figure are merged,
On the basis of global notable figure highlights variation targets, local characteristic is increased, overcomes and SAR image is become in the prior art
It is not high to change detection accuracy, to the problem of noise robustness difference, so that the present invention is to the detection essence for improving SAR image variation detection
Degree, the robustness of enhanced SAR Image Change Detection.
Detailed description of the invention
Fig. 1 is flow chart of the invention;
Fig. 2 is three groups of true SAR image data that emulation experiment uses and corresponding variation detection with reference to figure;
Fig. 3 is the variation testing result figure to the emulation experiment of the area Ottawa SAR image;
Fig. 4 is the variation testing result figure to the emulation experiment of Sardinia area SAR image;
Fig. 5 is the variation testing result figure to the emulation experiment of the Yellow River estuary SAR image.
Specific embodiment
The present invention will be further described with reference to the accompanying drawing.
Referring to Fig.1, the present invention realizes that specific step is as follows:
Step 1, SAR image is inputted.
The SAR image I that two width of input areal difference phase have been registrated, have corrected1And I2。
Step 2, it filters.
Using non-local mean filtering device to two width SAR image I of input1And I2It is filtered, obtains filtered respectively
SAR image X1And X2。
Step 3, the log ratio value of pixel is calculated.
To filtered SAR image X1And X2It takes the gray level of respective pixel to carry out log ratio operation, obtains log ratio
Disparity map DL。
Step 4, according to the following formula, the global saliency value of the pixel of log ratio disparity map is calculated:
Wherein, D (i, j) indicates the global saliency value of the pixel of the i-th row jth column in log ratio disparity map, I (i, j) table
Show the gray value of the i-th row jth column pixel in log ratio disparity map, I (i, j)=0,1,2 ..., 255, ∑ indicates sum operation,Indicate optional sign, ∈ expression belongs to symbol, and I (m, n) is indicated in log ratio disparity map except the m row except I (i, j) the
The gray value of the pixel of n column, I (m, n)=0,1,2 ..., 255, DLIndicate log ratio disparity map, | | | | expression takes gray scale
The distance of value operates.
Step 5, the contrast notable figure of different scale is calculated.
Respectively to filtered SAR image X1And X21 × 1 pixel is taken, using pixel contrast method, constructs 1 × 1 pixel
The contrast notable figure of scale.
Pixel contrast method specific steps are as follows:
The first step calculates the contrast saliency value of 1 × 1 block of pixels scale of contrast notable figure according to the following formula:
Wherein, S0(x1,x2) indicate to compare the contrast saliency value of 1 × 1 block of pixels scale of notable figure, x1And x2Table respectively
Show the index of current contrast saliency value coordinate position, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure of 1 × 1 pixel dimension according to the following formula:
Wherein, S0Indicate the contrast notable figure of 1 × 1 pixel dimension, S0(x1,x2) indicate the 1 × 1 of contrast notable figure
The contrast saliency value of block of pixels scale, x1And x2The position coordinates index of current contrast saliency value is respectively indicated, M and N are indicated
The line number and columns of contrast notable figure.
With the sliding window of 3 × 3 pixels respectively to filtered image X1And X2Take block.
The pixel in all blocks is pulled into a column by column respectively, obtains three-dimensional matrice L1And L2。
Using vector contrast method, by three-dimensional matrice L1And L2Construct the contrast saliency value of 3 × 3 pixel dimensions.
Vector contrast method specific steps are as follows:
The first step calculates the contrast saliency value of the contrast notable figure of 3 × 3 pixel dimensions according to the following formula:
Wherein, S1(l1,l2) indicate 3 × 3 pixel dimensions contrast notable figure contrast saliency value, i and j distinguish table
Show the index of the contrast saliency value position coordinates ranks of the contrast notable figure of 3 × 3 pixel dimensions, l1And l2Respectively indicate L1
And L2In column vector, T indicate transposition operation, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure of 3 × 3 pixel dimensions according to the following formula:
Wherein, S1Indicate the contrast notable figure under 3 × 3 pixel dimensions, S1(l1,l2) indicate 3 × 3 pixel dimensions pair
Than the contrast saliency value of degree notable figure, i and j indicate the index of the contrast notable figure position coordinates of 3 × 3 pixel dimensions, M and
N respectively indicates the line number and columns of the contrast notable figure of 3 × 3 pixel dimensions.
With the sliding window of 5 × 5 pixels respectively to filtered image X1And X2Take block.
The pixel in all blocks is pulled into a column by column respectively, obtains three-dimensional matrice L3And L4。
Using vector contrast method, by three-dimensional matrice L3And L4Construct the contrast saliency value of 5 × 5 pixel dimensions.
Vector contrast method specific steps are as follows:
The first step calculates the contrast saliency value of the contrast notable figure of 5 × 5 pixel dimensions according to the following formula:
Wherein, S2(l3,l4) indicate 5 × 5 pixel dimensions contrast notable figure contrast saliency value, i and j distinguish table
Show the index of the contrast saliency value position coordinates of the contrast notable figure of 5 × 5 pixel dimensions, l3And l4Respectively indicate L3And L4
In column vector, T indicate transposition operation, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure under 5 × 5 pixel dimensions according to the following formula:
Wherein, S2Indicate the scale contrast notable figure of 5 × 5 pixels, S2(l3,l4) indicate 5 × 5 pixel dimensions comparison
The contrast saliency value of notable figure is spent, i and j respectively indicate the index of the contrast notable figure pixel coordinate of 5 × 5 pixel dimensions,
l3And l4Respectively indicate L1And L2In column vector, M and N respectively indicate the contrast notable figure of 5 × 5 pixel dimensions line number and
Columns.
Step 6, using the weighted mean method of following formula, the contrast notable figure of different scale is merged, local notable figure is obtained:
S'=α0·S0+α1·S1+α2·S2
Wherein, S' indicates local notable figure, S0Indicate the contrast notable figure of 1 × 1 pixel dimension, S1Indicate 3 × 3 pixels
The notable figure of scale, S2Indicate the contrast notable figure of 5 × 5 pixel dimensions, α0、α1、α2Indicate weighting coefficient, α0+α1+α2=1,
0≤α0、α1、α2≤1。
Step 7, according to the following formula, amalgamation of global notable figure and local notable figure, obtain conspicuousness fusion figure:
S (i, j)=D (i, j) exp (- S'(i, j))
Wherein, S indicates that conspicuousness fusion figure, D indicate global notable figure, and S' indicates local notable figure, and exp indicates index behaviour
Make symbol, i and j respectively indicate conspicuousness fusion figure, in global notable figure and local notable figure the row of the pixel of same position and
Column coordinate value.
Step 8, fuzzy clustering.
Using fuzzy C-means clustering method, conspicuousness fusion figure is clustered, obtains each picture in conspicuousness fusion figure
Being under the jurisdiction of non-changing class and changing class for element is subordinate to angle value.
The pixel for being subordinate to angle value for being subordinate to angle value and being greater than variation class of non-changing class is judged to the ownership of for non-changing class pixel;It will
The pixel for being subordinate to angle value for being subordinate to angle value and being less than variation class of non-changing class, is judged to the ownership of to change class pixel;Obtain the change of SAR image
Change testing result.
Step 9, output variation testing result.
Effect of the invention is described further below with reference to emulation experiment:
1, simulated conditions:
Emulation experiment of the invention is Inter (R) Core (TM) i5-3470CPU, the memory 4GB in dominant frequency 3.2GHz
It is carried out under hardware environment and the software environment of MATLAB R2015b.
Emulation experiment of the invention has used three groups of true SAR image data and corresponding variation detection with reference to figure, such as
Shown in Fig. 2.
The first group of true SAR image data and corresponding variation that emulation experiment of the present invention uses are detected with reference to figure
The SAR image in the area Ottawa, image size are 290 × 350.Wherein, Fig. 2 (a) is the area Ottawa in May, 1997
SAR image, Fig. 2 (b) are the SAR images in the area Ottawa of in August, 1997, and Fig. 2 (c) is the corresponding variation inspection in the area Ottawa
It surveys with reference to figure.
The second group of true SAR image data and corresponding variation that emulation experiment of the present invention uses are detected with reference to figure
The SAR image in the area Sardinia, image size are 300 × 412.Wherein, Fig. 2 (d) is the area Sardinia of nineteen ninety-five September
SAR image, Fig. 2 (e) is the SAR image in the area Sardinia in July, 1996, and Fig. 2 (f) is that the area Sardinia is corresponding
Variation detection is with reference to figure.
It is yellow that the true SAR image data of the third group that emulation experiment of the present invention uses and corresponding variation, which are detected with reference to figure,
The SAR image in river estuary area, image size are 291 × 306.Wherein, Fig. 2 (g) is the Yellow River estuary in June, 2008
The SAR image in area, Fig. 2 (h) are the SAR images in the Yellow River estuary area in June, 2009, and Fig. 2 (i) is the Yellow River estuary area
Corresponding variation detection is with reference to figure.
Simulation parameter used in emulation experiment of the present invention is as follows:
Missing inspection number: the number of pixels in the region that changes in statistical experiment result figure is calculated, and with reference to region of variation in figure
Number of pixels compare, with reference to changing in figure but be detected as unchanged number of pixels in experimental result picture, claim
For missing inspection number FN.
Erroneous detection number: the number of pixels in the region that do not change in statistical experiment result figure is calculated, and with reference to unchanged in figure
The number of pixels for changing region compares, with reference to the pixel for not changing but being detected as in experimental result picture variation in figure
Number, referred to as erroneous detection number FP.
Total error number/the total pixel number of accuracy PCC:PCC=1-.
Measure testing result figure and the Kappa coefficient with reference to figure consistency: Kappa=(PCC-PRE)/(1-PRE)
Wherein, accuracy PCC indicates actual concordance rate, the concordance rate of PRE representation theory.
2, emulation content and interpretation of result:
(LK method based on log ratio and K mean cluster is based on emulation experiment of the invention using four kinds of prior arts
The CDI-K method of disparity map fusion, the fuzzy clustering method based on Markov random field, the side SM based on significance measure
Method) with using the method for the present invention, the area Ottawa, the area Sardinia and the Yellow River estuary area SAR image are carried out respectively
The testing result of variation detection compares.
Fig. 3 is emulation experiment of the present invention to the area Ottawa, wherein Fig. 3 (a) is the variation of the area Ottawa SAR image
The reference picture of detection;Fig. 3 (b) is the simulation result diagram using the LK method based on log ratio and K mean cluster;Fig. 3 (c)
It is the simulation result diagram using the CDI-K method merged based on disparity map;Fig. 3 (d) is using based on Markov random field
The simulation result diagram of fuzzy clustering method MRFFCM;Fig. 3 (e) is the simulation result using the SM method based on significance measure
Figure;Fig. 3 (f) is the simulation result diagram using the method for the present invention.
It can be seen that using the noise spot in variation testing result figure of the invention from the visual effect of Fig. 3 than existing skill
Four kinds of methods of art are few, and change the edge clear of testing result figure.
Table 1 be emulation experiment of the invention using four kinds of prior arts and the method for the present invention to missing inspection number FP, erroneous detection number FN,
Accuracy PCC and Kappa coefficient is counted.In table " LK " indicate using based on log ratio and K mean cluster LK method,
" CDI-K " is indicated using CDI-K method merge based on disparity map in table, " MRFFCM " is indicated using being based on markov in table
" SM " indicates that SM method and the present invention based on significance measure use in the MRFFCM method and table of the fuzzy clustering of random field
Conspicuousness fusion method.From table 1 it follows that the variation of the method for the present invention detects accuracy PCC and Kappa coefficient, all
Higher than other four kinds of control methods, missing inspection number FP and erroneous detection number FN are relatively low, and detection effect is good.
1 area Ottawa SAR image of table changes testing result
Missing inspection number FP | Erroneous detection number FN | Accuracy PCC | Kappa coefficient | |
LK | 1219 | 2198 | 0.9663 | 0.8703 |
CDI-K | 2044 | 367 | 0.9762 | 0.9068 |
MRFFCM | 790 | 1711 | 0.9768 | 0.9151 |
SM | 5862 | 910 | 0.9333 | 0.7135 |
Conspicuousness fusion | 1006 | 634 | 0.9838 | 0.9399 |
Fig. 4 is emulation experiment of the present invention to the area Sardinia, and wherein Fig. 4 (a) is that the area Sardinia SAR image becomes
Change the reference picture of detection;Fig. 4 (b) is the simulation result diagram using the LK method based on log ratio and K mean cluster;Fig. 4
It (c) is the simulation result diagram for using the CDI-K method merged based on disparity map;Fig. 4 (d) is using based on Markov random field
Fuzzy clustering method MRFFCM simulation result diagram;Fig. 4 (e) is the simulation result using the SM method based on significance measure
Figure;Fig. 4 (f) is the simulation result diagram using the method for the present invention.
It can be seen that variation testing result figure of the invention from the visual effect of Fig. 4 and than four kinds of methods of the prior art
Miscellaneous point is less in variation testing result figure more effective compared to the detection of fine edge and of the invention.
Table 2 be emulation experiment of the invention using four kinds of prior arts and the method for the present invention to missing inspection number FP, erroneous detection number FN,
Accuracy PCC and Kappa coefficient is counted.In table " LK " indicate using based on log ratio and K mean cluster LK method,
" CDI-K " is indicated using CDI-K method merge based on disparity map in table, " MRFFCM " is indicated using being based on markov in table
" SM " indicates that SM method and the present invention based on significance measure use in the MRFFCM method and table of the fuzzy clustering of random field
Conspicuousness fusion method.As can be seen from Table 2, the missing inspection number FP of testing result of the present invention will than other four kinds of detection methods
It is low, and erroneous detection number FN, also in lower level, accuracy PCC and Kappa coefficient is above control methods, it is seen that the method for the present invention
The effect of SAR image variation detection is improved, and more preferable to the robustness of noise.
2 area Sardinia SAR image of table changes testing result
Missing inspection number FP | Erroneous detection number FN | Accuracy PCC | Kappa coefficient | |
LK | 836 | 1229 | 0.9833 | 0.8591 |
CDI-K | 913 | 1083 | 0.9839 | 0.8620 |
MRFFCM | 947 | 1523 | 0.9800 | 0.8333 |
SM | 1967 | 544 | 0.9797 | 0.8078 |
Conspicuousness fusion | 823 | 1126 | 0.9842 | 0.8663 |
Fig. 5 is emulation experiment of the present invention to the Yellow River estuary area, and wherein Fig. 5 (a) is the Yellow River estuary area SAR figure
As the reference picture of variation detection;Fig. 5 (b) is the simulation result diagram using the LK method based on log ratio and K mean cluster;
Fig. 5 (c) is the simulation result diagram using the CDI-K method merged based on disparity map;Fig. 5 (d) be using based on markov with
The simulation result diagram of the fuzzy clustering method MRFFCM on airport;Fig. 5 (e) is the emulation using the SM method based on significance measure
Result figure;Fig. 5 (f) is the simulation result diagram using the method for the present invention.
Can be seen that variation testing result figure of the invention from the visual effect of Fig. 5, there are four types of most connect compared with method with now
Closely with reference to figure, the miscellaneous point of variation testing result figure of the invention is minimum, and visual effect is best, and can effectively detect variation
Region.
3 the Yellow River estuary area SAR image of table changes testing result
Missing inspection number FP | Erroneous detection number FN | Accuracy PCC | Kappa coefficient | |
LK | 440 | 10170 | 0.8808 | 0.4263 |
CDI-K | 803 | 1816 | 0.9706 | 0.7577 |
MRFFCM | 472 | 1853 | 0.9738 | 0.7903 |
SM | 102 | 5209 | 0.9404 | 0.6317 |
Conspicuousness fusion | 712 | 350 | 0.9881 | 0.8893 |
Table 3 be emulation experiment of the invention using four kinds of prior arts and the method for the present invention to missing inspection number FP, erroneous detection number FN,
Accuracy PCC and Kappa coefficient is counted.In table " LK " indicate using based on log ratio and K mean cluster LK method,
" CDI-K " is indicated using CDI-K method merge based on disparity map in table, " MRFFCM " is indicated using being based on markov in table
" SM " indicates that SM method and the present invention based on significance measure use in the MRFFCM method and table of the fuzzy clustering of random field
Conspicuousness fusion method.As can be seen from Table 3, the erroneous detection number FN of testing result of the present invention is examined well below other four kinds of comparisons
Survey method, while the quantity of missing inspection number FP is controlled, and accuracy PCC and Kappa coefficient is even more significantly larger than other four kinds comparisons
Detection method.
Claims (4)
1. a kind of SAR image change detection based on conspicuousness fusion, includes the following steps:
(1) SAR image is inputted:
The SAR image I that two width of input areal difference phase have been registrated, have corrected1And I2;
(2) it filters:
Using non-local mean filtering device to two width SAR image I of input1And I2It is filtered respectively, obtains filtered SAR
Image X1And X2;
(3) the log ratio value of pixel is calculated:
To filtered SAR image X1And X2It takes the gray level of respective pixel to carry out log ratio operation, obtains log ratio difference
Scheme DL;
(4) according to the following formula, the global saliency value of the pixel of log ratio disparity map is calculated:
Wherein, D (i, j) indicates the global saliency value of the pixel of the i-th row jth column in log ratio disparity map, I (i, j) expression pair
The gray value of the pixel of the i-th row jth column, I (i, j)=0,1,2 ..., 255, ∑ indicate sum operation in number ratio difference figure,
Indicate optional sign, ∈ expression belongs to symbol, and I (m, n) is indicated in log ratio disparity map except the m row n-th except I (i, j)
The gray value of the pixel of column, I (m, n)=0,1,2 ..., 255, DLIndicate log ratio disparity map, | | | | expression takes gray scale
The distance of value operates;
(5) the contrast saliency value of different scale is calculated:
(5a) is respectively to filtered SAR image X1And X21 × 1 pixel is taken, using pixel contrast method, constructs 1 × 1 pixel
The contrast notable figure of scale;
(5b) is with the sliding window of 3 × 3 pixels respectively to filtered image X1And X2Take block;
Pixel in all blocks is pulled into a column by column respectively by (5c), obtains three-dimensional matrice L1And L2;
(5d) uses vector contrast method, by three-dimensional matrice L1And L2Construct the contrast saliency value of 3 × 3 pixel dimensions;
(5e) is with the sliding window of 5 × 5 pixels respectively to filtered image X1And X2Take block;
Pixel in all blocks is pulled into a column by column respectively by (5f), obtains three-dimensional matrice L3And L4;
(5g) uses vector contrast method, by three-dimensional matrice L3And L4Construct the contrast saliency value of 5 × 5 pixel dimensions;
(6) weighted mean method for using following formula, merges the contrast notable figure of different scale, obtains local notable figure:
S'=α0·S0+α1·S1+α2·S2
Wherein, S' indicates local notable figure, S0Indicate the contrast notable figure of 1 × 1 pixel dimension, S1Indicate 3 × 3 pixel dimensions
Notable figure, S2Indicate the contrast notable figure of 5 × 5 pixel dimensions, α0、α1、α2Indicate weighting coefficient, α0+α1+α2=1,0≤
α0、α1、α2≤1;
(7) according to the following formula, amalgamation of global notable figure and local notable figure, obtain conspicuousness fusion figure:
S (i, j)=D (i, j) exp (- S'(i, j))
Wherein, S indicates that conspicuousness fusion figure, D indicate global notable figure, and S' indicates local notable figure, and exp indicates index operation symbol
Number, i and j respectively indicate that conspicuousness fusion figure, the row and column of the pixel of same position is sat in global notable figure and local notable figure
Scale value;
(8) fuzzy clustering:
(8a) uses fuzzy C-means clustering method, clusters to conspicuousness fusion figure, obtains each picture in conspicuousness fusion figure
Being under the jurisdiction of non-changing class and changing class for element is subordinate to angle value;
(8b) is judged to the ownership of the pixel for being subordinate to angle value for being subordinate to angle value and being greater than variation class of non-changing class for non-changing class pixel;It will
The pixel for being subordinate to angle value for being subordinate to angle value and being less than variation class of non-changing class, is judged to the ownership of to change class pixel;Obtain the change of SAR image
Change testing result;
(9) output variation testing result.
2. the SAR image change detection according to claim 1 based on conspicuousness fusion, it is characterised in that: step
Specific step is as follows for pixel contrast method described in (5a):
The first step calculates the contrast saliency value of the contrast notable figure of 1 × 1 pixel dimension according to the following formula:
Wherein, S0(x1,x2) indicate 1 × 1 pixel dimension contrast notable figure contrast saliency value, x1And x2It respectively indicates and works as
The index of preceding contrast saliency value coordinate position, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure of 1 × 1 pixel dimension according to the following formula:
Wherein, S0Indicate the contrast notable figure of 1 × 1 pixel dimension, S0(x1,x2) indicate that the contrast of 1 × 1 pixel dimension is significant
The contrast saliency value of figure, x1And x2The position coordinates index of current contrast saliency value is respectively indicated, M and N respectively indicate comparison
Spend the line number and columns of notable figure.
3. the SAR image change detection according to claim 1 based on conspicuousness fusion, it is characterised in that: step
Specific step is as follows for vector contrast method described in (5d):
The first step calculates the contrast saliency value of the contrast notable figure of 3 × 3 pixel dimensions according to the following formula:
Wherein, S1(i, j) indicates the contrast saliency value of the contrast notable figure of 3 × 3 pixel dimensions, and i and j respectively indicate 3 × 3
The index of the contrast saliency value position coordinates ranks of the contrast notable figure of pixel dimension, l1And l2Respectively indicate L1And L2In
Column vector, T indicate transposition operation, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure of 3 × 3 pixel dimensions according to the following formula:
Wherein, S1Indicate the contrast notable figure under 3 × 3 pixel dimensions, S1(i, j) indicates that the contrast of 3 × 3 pixel dimensions is aobvious
The contrast saliency value of figure is write, i and j indicate the index of the contrast notable figure position coordinates of 3 × 3 pixel dimensions, M and N difference
Indicate the line number and columns of the contrast notable figure of 3 × 3 pixel dimensions.
4. the SAR image change detection according to claim 1 based on conspicuousness fusion, it is characterised in that: step
Specific step is as follows for vector contrast method described in (5g):
The first step calculates the contrast saliency value of the contrast notable figure of 5 × 5 pixel dimensions according to the following formula:
Wherein, S2(i, j) indicates the contrast saliency value of the contrast notable figure of 5 × 5 pixel dimensions, and i and j respectively indicate 5 × 5
The index of the contrast saliency value position coordinates of the contrast notable figure of pixel dimension, l3And l4Respectively indicate L3And L4In column
Vector, T indicate transposition operation, | | | | indicate Euclidean distance operation;
Second step constructs the contrast notable figure under 5 × 5 pixel dimensions according to the following formula:
Wherein, S2Indicate the contrast notable figure of 5 × 5 pixel dimensions, S2(i, j) indicates that the contrast of 5 × 5 pixel dimensions is significant
The contrast saliency value of figure, i and j respectively indicate the index of the contrast notable figure pixel coordinate of 5 × 5 pixel dimensions, and M and N divide
Not Biao Shi 5 × 5 pixel dimensions contrast notable figure line number and columns.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610164208.6A CN105869146B (en) | 2016-03-22 | 2016-03-22 | SAR image change detection based on conspicuousness fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610164208.6A CN105869146B (en) | 2016-03-22 | 2016-03-22 | SAR image change detection based on conspicuousness fusion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105869146A CN105869146A (en) | 2016-08-17 |
CN105869146B true CN105869146B (en) | 2019-03-01 |
Family
ID=56625558
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610164208.6A Active CN105869146B (en) | 2016-03-22 | 2016-03-22 | SAR image change detection based on conspicuousness fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105869146B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106898008A (en) * | 2017-03-01 | 2017-06-27 | 南京航空航天大学 | Rock detection method and device |
CN107274382B (en) * | 2017-05-03 | 2020-08-21 | 国网湖北省电力公司 | State identification method and device of hard pressing plate and electronic equipment |
CN107169498B (en) * | 2017-05-17 | 2019-10-15 | 河海大学 | A kind of fusion part and global sparse image significance detection method |
CN107341798B (en) * | 2017-07-06 | 2019-12-03 | 西安电子科技大学 | High Resolution SAR image change detection method based on the overall situation-part SPP Net |
CN107358261B (en) * | 2017-07-13 | 2020-05-01 | 西安电子科技大学 | High-resolution SAR image change detection method based on curvelet SAE |
CN107451992B (en) * | 2017-07-20 | 2020-08-11 | 广东工业大学 | Method and device for detecting SAR image change |
CN108257151B (en) * | 2017-12-22 | 2019-08-13 | 西安电子科技大学 | PCANet image change detection method based on significance analysis |
CN109242889B (en) * | 2018-08-27 | 2020-06-16 | 大连理工大学 | SAR image change detection method based on context significance detection and SAE |
CN110751135A (en) * | 2019-12-24 | 2020-02-04 | 广东博智林机器人有限公司 | Drawing checking method and device, electronic equipment and storage medium |
CN117830322A (en) * | 2024-03-06 | 2024-04-05 | 慧创科仪(北京)科技有限公司 | Method and device for performing significance difference analysis on near infrared data |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102722892A (en) * | 2012-06-13 | 2012-10-10 | 西安电子科技大学 | SAR (synthetic aperture radar) image change detection method based on low-rank matrix factorization |
CN103500453A (en) * | 2013-10-13 | 2014-01-08 | 西安电子科技大学 | SAR(synthetic aperture radar) image significance region detection method based on Gamma distribution and neighborhood information |
CN103955943A (en) * | 2014-05-21 | 2014-07-30 | 西安电子科技大学 | Non-supervision change detection method based on fuse change detection operators and dimension driving |
CN104794729A (en) * | 2015-05-05 | 2015-07-22 | 西安电子科技大学 | SAR image change detection method based on significance guidance |
US9239384B1 (en) * | 2014-10-21 | 2016-01-19 | Sandia Corporation | Terrain detection and classification using single polarization SAR |
-
2016
- 2016-03-22 CN CN201610164208.6A patent/CN105869146B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102722892A (en) * | 2012-06-13 | 2012-10-10 | 西安电子科技大学 | SAR (synthetic aperture radar) image change detection method based on low-rank matrix factorization |
CN103500453A (en) * | 2013-10-13 | 2014-01-08 | 西安电子科技大学 | SAR(synthetic aperture radar) image significance region detection method based on Gamma distribution and neighborhood information |
CN103955943A (en) * | 2014-05-21 | 2014-07-30 | 西安电子科技大学 | Non-supervision change detection method based on fuse change detection operators and dimension driving |
US9239384B1 (en) * | 2014-10-21 | 2016-01-19 | Sandia Corporation | Terrain detection and classification using single polarization SAR |
CN104794729A (en) * | 2015-05-05 | 2015-07-22 | 西安电子科技大学 | SAR image change detection method based on significance guidance |
Non-Patent Citations (3)
Title |
---|
SAR change detection based on intensity and texture changes;Maoguo Gong et al.;《ISPRS Journal of Photogrammetry and Remote Sensing》;20140731;第93卷;全文 |
基于空间相关性的SAR图像变化检测研究;张祎勃;《万方数据库》;20150415;全文 |
尺度自适应的SAR图像显著性检测方法;谢惠杰 等;《计算机工程与应用》;20151231;第51卷(第20期);全文 |
Also Published As
Publication number | Publication date |
---|---|
CN105869146A (en) | 2016-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105869146B (en) | SAR image change detection based on conspicuousness fusion | |
CN104574445B (en) | A kind of method for tracking target | |
Bujor et al. | Application of log-cumulants to the detection of spatiotemporal discontinuities in multitemporal SAR images | |
CN106228544B (en) | A kind of conspicuousness detection method propagated based on rarefaction representation and label | |
CN102096921B (en) | SAR (Synthetic Aperture Radar) image change detection method based on neighborhood logarithm specific value and anisotropic diffusion | |
CN103729854B (en) | A kind of method for detecting infrared puniness target based on tensor model | |
Liu et al. | A contrario comparison of local descriptors for change detection in very high spatial resolution satellite images of urban areas | |
CN103456020B (en) | Based on the method for detecting change of remote sensing image of treelet Fusion Features | |
Zhou et al. | CANet: An unsupervised deep convolutional neural network for efficient cluster-analysis-based multibaseline InSAR phase unwrapping | |
Chawan et al. | Automatic detection of flood using remote sensing images | |
CN108122008A (en) | SAR image recognition methods based on rarefaction representation and multiple features decision level fusion | |
CN104794729B (en) | SAR image change detection based on conspicuousness guiding | |
CN108401565B (en) | Remote sensing image registration method based on improved KAZE features and Pseudo-RANSAC algorithms | |
CN103955701A (en) | Multi-level-combined multi-look synthetic aperture radar image target recognition method | |
Qi et al. | FTC-Net: Fusion of transformer and CNN features for infrared small target detection | |
CN103065320A (en) | Synthetic aperture radar (SAR) image change detection method based on constant false alarm threshold value | |
CN104680536A (en) | Method for detecting SAR image change by utilizing improved non-local average algorithm | |
CN105205807B (en) | Method for detecting change of remote sensing image based on sparse automatic coding machine | |
CN104268574A (en) | SAR image change detecting method based on genetic kernel fuzzy clustering | |
Oga et al. | River state classification combining patch-based processing and CNN | |
Huang et al. | Superpixel-based change detection in high resolution sar images using region covariance features | |
Lin et al. | IR-TransDet: Infrared dim and small target detection with IR-transformer | |
Bostanci et al. | Feature coverage for better homography estimation: an application to image stitching | |
CN116912582A (en) | Strong robustness hyperspectral target detection method based on characterization model | |
CN103903258B (en) | Method for detecting change of remote sensing image based on order statistic spectral clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |