Summary of the invention
The present invention provides a kind of alpine terrain deformation extracting method based on PSInSAR technology, with solve the prior art without
Method accurately extracts the problem of alpine terrain deformation.
A kind of alpine terrain deformation extracting method based on PSInSAR technology provided by the invention, comprising:
SAR image is chosen, the SAR image includes master image and auxiliary image;
Using the master image and the auxiliary image registration and interference processing is carried out, obtains coherence factor figure and interferometric phase
Figure;
PS point is chosen on the coherence factor figure using amplitude deviation and the dual thresholds method of coherence factor, obtains PS point
The triangulation network;
The interferometric phase image is subjected to flat filtering processing, using dem data to treated the interferometric phase image
Differential interferometry processing is carried out, differential interferometry figure is obtained;
Phase unwrapping processing is carried out to the differential interferometry figure, obtains Deformation Field.
In one embodiment of the invention, SAR image is chosen to specifically include:
N width SAR image is chosen, selects the width in the SAR image as the master map using comprehensive coherence factor method
Picture, remaining n-1 width is as the auxiliary image.
In one embodiment of the invention, phase unwrapping processing is carried out to the differential interferometry figure to specifically include:
It is utilized respectively network adjustment method, puppet 3D phase unwrapping method and Lambda algorithm and phase is carried out to the differential interferometry figure
Solution twines processing.
In one embodiment of the invention, the network adjustment method specifically includes:
The differential mode pattern three times of PS point is established, the coherence factor K in the differential mode pattern three times is calculated1,K2,B⊥, institute
State differential mode pattern three times are as follows:
Plumb line method of loci is introduced, Δ δ ' is chosenHSearch range and step-length with Δ v', by between the adjacent PS point of control
The suitable Δ δ ' of multiple correlation coefficient γ search one by oneHWith Δ v', wherein
It regard linear deformation speed difference corresponding to the maximum value of the coherence factor and elevation amendment difference as parameter Estimation
Value, establishes adjustment function model, and the adjustment function model is,
Solve linear equation:
In one embodiment of the invention, the puppet 3D phase unwrapping method specifically includes:
Delaunay triangulation network connection is done to the PS point and establishes the triangulation network, time dimension is connected by time relationship;
Residue points calculating is carried out along the side of the triangulation network;
The optimization initial value that space 2D phase unwrapping is done using time dimension disentanglement fruit is carried out using minimum Lp norm method
The optimal solution of space dimension.
In one embodiment of the invention, the Lambda algorithm specifically includes:
In n-1 width differential interferometry figure, the phase type of differential interferometry three times of the adjacent PS point of the triangulation network is formed into equation group:
It is denoted as y1=A1·a+B1·b;
In y1=A1·a+B1Imitation observation equation y is added in b2=A2·a+B2B constitutes y=Aa+Bb, wherein A2For 2*m null matrix, B2For 2*2 unit matrix;
It is acquired according to the principle of least square:
C=(A B)
It obtainsComplete cycle float solution, search for obtain the fixed solution of phase complete cycle using fuzziness decorrelation method
By the fixed solutionIt brings solution into and twines phase expression formulaIn reuse minimum two
Multiplication obtains:
After completing phase unwrapping, the Deformation Field is obtained by mutually high conversion and geocoding.
A kind of alpine terrain deformation extracting method based on PSInSAR technology provided by the invention, is obtained by piggyback satellite
Several High Resolution SAR Images are taken, and choose n width SAR image, wherein a width makees master map using comprehensive coherence factor method selection
Picture, remaining makees auxiliary image, then allow master image respectively with auxiliary image registration and carry out interference processing, obtain coherence factor figure and interference
Phase diagram;According to coherence factor figure, PS point is chosen using corresponding amplitude deviation and coherence factor dual thresholds method and obtains PS
The point triangulation network;Flat filtering is carried out to interferometric phase image, and utilizes external dem data, differential interferometry processing is carried out to interference pattern,
Obtain differential interferometry figure;It is utilized respectively three kinds of network adjustment method, puppet 3D phase unwrapping method and Lambda algorithm irregular nets of three-dimensional
Network phase unwrapping method carries out phase unwrapping to differential interferometry figure, obtains Deformation Field to the end.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples.
It is the stream of the alpine terrain deformation extracting method provided in an embodiment of the present invention based on PSInSAR technology referring to Fig. 1
Cheng Tu, this method comprises:
S100: SAR image is chosen, the SAR image includes master image and auxiliary image.
SAR image is chosen to specifically include:
N width SAR image is chosen, selects the width in the SAR image as the master map using comprehensive coherence factor method
Picture, remaining n-1 width is as the auxiliary image.
S200: using the master image and the auxiliary image registration and carrying out interference processing, obtains coherence factor figure and does
Relate to phase diagram.
S300: PS point is chosen on the coherence factor figure using amplitude deviation and the dual thresholds method of coherence factor, is obtained
To the PS point triangulation network.
S400: carrying out flat filtering processing for the interferometric phase image, using dem data to treated the interference phase
Bitmap carries out differential interferometry processing, obtains differential interferometry figure.
Dem data is digital elevation model (Digital Elevation Model), is by limited landform altitude number
The factually existing digitized simulation (i.e. the digital expression of topographical surface form) to ground surface or terrain, is with one group of orderly array of values
A kind of actual ground model of form expression ground elevation.In the embodiment of the present invention, using external dem data, to going flat filtering
Treated, and interferometric phase image carries out differential interferometry processing, obtains differential interferometry figure.
S500: phase unwrapping processing is carried out to the differential interferometry figure, obtains Deformation Field.
It is utilized respectively network adjustment method, puppet 3D phase unwrapping method and three kinds of Lambda algorithm three-dimensional scramble network phase solutions
Quick pushing manipulation carries out phase unwrapping to differential interferometry figure, obtains Deformation Field to the end.
Wherein, network adjustment method specifically includes:
The differential mode pattern three times of PS point is established, the coherence factor K in the differential mode pattern three times is calculated1,K2,B⊥, institute
State differential mode pattern three times are as follows:
Plumb line method of loci is introduced, Δ δ ' is chosenHSearch range and step-length with Δ v', by between the adjacent PS point of control
The suitable Δ δ ' of multiple correlation coefficient γ search one by oneHWith Δ v', wherein
It regard linear deformation speed difference corresponding to the maximum value of the coherence factor and elevation amendment difference as parameter Estimation
Value, establishes adjustment function model, and the adjustment function model is,
Solve linear equation:
Pseudo- 3D phase unwrapping method specifically includes:
Delaunay triangulation network connection is done to the PS point and establishes the triangulation network, time dimension is connected by time relationship.
Residue points calculating is carried out along the side of the triangulation network.Residue points necessarily occur in some triangle, corresponding triangle
Certain point in the dual network of net.
The optimization initial value that space 2D phase unwrapping is done using time dimension disentanglement fruit is carried out using minimum Lp norm method
The optimal solution of space dimension.
Lambda algorithm is a kind of searching algorithm based on least-squares estimation adjustment, is currently widely applied be based on
GPS carrier phase is positioned, and the algorithm of appearance and orientation is surveyed.
Lambda algorithm specifically includes:
In n-1 width differential interferometry figure, the phase type of differential interferometry three times of the adjacent PS point of the triangulation network is formed into equation group:
It is denoted as y1=A1·a+B1·b。
Due to Matrix Properties, equation group intangibility, therefore, in y1=A1·a+B1Imitation observation equation y is added in b2=
A2·a+B2B constitutes y=Aa+Bb, whereinA2For 2*m null matrix, B2For
2*2 unit matrix.
It is acquired according to the principle of least square:
C=(A B)
It obtainsComplete cycle float solution, search for obtain the fixed solution of phase complete cycle using fuzziness decorrelation method
By the fixed solutionIt brings solution into and twines phase expression formulaIn reuse minimum two
Multiplication obtains:
After completing phase unwrapping, the Deformation Field is obtained by mutually high conversion and geocoding.
The embodiment of the present invention is described in detail below with reference to Fig. 2, Fig. 3, Fig. 4 and Fig. 5.
Survey region is Mao County, and Erlongshan Mountains is located in river blueness block handover region, under the influence of plate motion and rainfall, mudstone
The geological disasters such as stream take place frequently, and Chengdu surface configuration changes climate and geological conditions influence is smaller, are affected by culture,
The embodiment of the present invention uses the TerraSAR-X image of precision 1m, Mao County, Erlongshan Mountains and Chengdu image information and master map such as 1 institute of table
Show.
1 test block image information of table and master map
Place |
Initial time |
Terminal time |
Master map |
Quantity |
Mao County |
16/04/23 |
17/12/19 |
16/06/17 |
18 |
Erlongshan Mountains |
14/04/04 |
15/04/02 |
14/11/21 |
15 |
Chengdu |
14/05/20 |
16/09/02 |
15/01/17 |
14 |
Fig. 2 and Fig. 3 illustrates the PS point triangular network in Mao County trial zone and Mao County area, includes apparent 7 in region
Electric power pylon, PS point screening technique use amplitude deviation and coherence factor Integrated Selection, obvious striped in the differential interferometry figure of Mao County
Appear in steel tower, rock and ridge region.
Fig. 4 and Fig. 5 illustrates the differential interferometry figure of Erlongshan Mountains and In Chengdu, in the region of Erlongshan Mountains it is affected by noise compared with
Greatly, interference fringe is only present in steel tower part, and Chengdu region is typical Plain urban area, in left side culture's the commercial house
Area, the middle of the road line take out existing obvious interference fringe.
Network adjustment method, puppet 3D phase unwrapping method and Lambda are used to the differential interferometry figure in Mao County, Erlongshan Mountains and Chengdu
Algorithm carries out solution and twines processing, wherein Mao County region linear deformation amount and elevation correction value are as shown in tables 2 and 3.
2 Mao County PS point Linear deformation rate of table is distributed mm/y
|
v≤-2 |
-2<v≤-1 |
-1<v≤0 |
0<v≤1 |
1<v≤2 |
v>2 |
Network adjustment |
0.0313 |
0.247 |
0.2615 |
0.2206 |
0.2181 |
0.0306 |
Pseudo- 3D |
0.0218 |
0.2517 |
0.2664 |
0.2254 |
0.2224 |
0.0215 |
Lambda |
0 |
0.1889 |
0.3510 |
0.2891 |
0.1801 |
0 |
3 Mao County elevation correction value m of table
|
h≤-15 |
-15<h≤-8 |
-8<h≤0 |
0<h≤8 |
8<h≤15 |
h>15 |
Network adjustment |
0 |
0.0367 |
0.4145 |
0.5073 |
0.0415 |
0 |
Lambda |
0 |
0.1984 |
0.2653 |
0.3179 |
0.2184 |
0 |
As seen from the table, Mao County rate of deformation is uniformly distributed, and PS point Ground Deformation performance lifting or downward trend are unified, tower
Body region has obvious rise to sink to changing, and elevation correction value deviation trend is unified, and both of which can get trusted elevation knot
Fruit.
Erlongshan Mountains regional deformation rate concentrates in -5~+7mm/y range, and 80% or more is distributed within the scope of ± 5mm/y,
Ascendant trend is presented in PS point more than half, and the PS point in steel tower region is stablized in 0mm/y or so, has stability, wherein 3D phase
Position solution twines algorithm resolving rate and more concentrates relatively.Elevation correction value is distributed within the scope of ± 10m, and Lambda algorithm is more than 80% collection
In in ± 5m.In electric power pylon region, there is network adjustment smaller elevation to correct difference, more confidence level.Chengdu region
The PS dot variable of rate of deformation 90% concentrates within the scope of ± 2.5mm/y, algorithms of different have certain proportion across 2.5~
5mm/y has the point greater than 5mm/y deformation.Horizontal road region all show as it is unified it is stage rise or fall, elevation amendment
Value 85% concentrates within the scope of ± 10m, and network adjustment method calculates the sparse elevation correction value of about 40m or so, this elevation
Amendment may be built by commodity building in region to be formed.
Erlongshan Mountains elevation correction value is minimum, and it is related to be chronically at no developing zone with Erlongshan Mountains.There is artificial trace in Mao County area
Mark and crustal movement combined influence, elevation also has to be changed to a certain degree.The elevation of In Chengdu has discrete sharp point, is built by city
If work causes.
Table 4, table 5 and table 6 illustrate deformation of the earliest auxiliary image with respect to master image in three regions and are distributed.
4 Mao County deformation quantity of table counts mm
|
d≤-2.4 |
-2.4<d≤-1.2 |
-1.2<d≤0 |
0<d≤1.2 |
1.2<d≤2.4 |
d>2.4 |
Network adjustment |
0.098 |
0.275 |
0.164 |
0.147 |
0.214 |
0.099 |
Pseudo- 3D |
0.083 |
0.225 |
0.156 |
0.170 |
0.283 |
0.082 |
Lambda |
0.004 |
0.319 |
0.220 |
0.180 |
0.273 |
0.002 |
5 Erlongshan Mountains deformation quantity of table counts mm
|
d≤-4.5 |
-4.5<d≤-3 |
-3<d≤0 |
0<d≤3 |
3<d≤4.5 |
d>4.5 |
Network adjustment |
0 |
0 |
0.4845 |
0.515 |
0 |
0 |
Pseudo- 3D |
0 |
0 |
0.545 |
0.455 |
0 |
0 |
Lambda |
0.005 |
0.035 |
0.416 |
0.504 |
0.037 |
0.003 |
6 Chengdu deformation quantity of table counts cm
|
d≤-6 |
-6<d≤-3 |
-3<d≤0 |
0<d≤3 |
3<d≤6 |
d>6 |
Network adjustment |
0 |
0.0865 |
0.4930 |
0.3301 |
0.0904 |
0 |
Pseudo- 3D |
0 |
05230 |
0.5097 |
0.4553 |
0 |
0 |
Lambda |
0 |
0.001 |
0.4368 |
0.5581 |
0.0040 |
0 |
In each research area, PS point in SAR image maximum imaging time interval have total deformation distribution, trend with
Linear deformation rate is related to time interval.The deformation of maximum ± 30mm occurs for Mao County, and nearly 60% PS point deformation is in ± 12mm
In deformation range.- 30~24mm deformation occurs for Erlongshan Mountains, and 70% deformation concentrates between ± 12mm.Chengdu occurs maximum
The deformation of 120mm, most of PS dot changes are distributed within the scope of 30m.By Sichuan Electric Power Network artificial line walking record show Mao County and
An iron rake with three to six teeth periphery is transmitted electricity in image imaging time as occurrence of large-area geological disaster in Erlongshan Mountains, does not also have collapse of iron tower case, says
Bright deformation, which is extracted, has confidence level, and steel tower is not threatened by deformation safely.
The specific embodiment of invention described above is not intended to limit the scope of the present invention..