CN103970932A - High-resolution permanent scatterer modeling method for separation of building and background - Google Patents

Abstract

The invention discloses a high-resolution permanent scatterer modeling method for separation of a building and a background. The method comprises the steps that firstly, PS points on the building are separated from PS points on a background ground object; secondly, networks are constructed independently for the PS points on the building and the PS points on the background ground object; thirdly, elevation and deformation parameters of the PS points on the building are estimated, and then a final deformation graph is obtained. By means of the scheme, the elevation error is small, the deformation rate value is small, the influence of a high-phase gradient on solution of the elevation and deformation of the building is effectively lowered, and the method is suitable for telemetering modeling in a high-resolution urban environment.

Description

The permanent scatterer modeling method of a kind of high-resolution buildings and background separation

Technical field

The present invention relates to microwave radar remotely sensed image field, especially relate to permanent scatterer modeling and method for parameter estimation that under a kind of high resolving power condition, in urban environment, buildings separates with background atural object.

Background technology

It is the microwave radar remote sensing technology that development in recent years is got up that interfering synthetic aperture radar is measured (InSAR) technology.D-InSAR is as the extension of InSAR technology, and difference interfering synthetic aperture radar is measured, and is that the radar image that utilizes satellite to obtain through the same area for twice carries out differential interferometry, to extract land subsidence information.The advantage of D-InSAR technology is can round-the-clock collection radar image, and the precision of D-InSAR technology Ground Subsidence Monitoring can reach a millimeter rank in theory.The main limitation of D-InSAR technology is that time dephasing is dry, space dephasing is dry and atmosphere delay phase place.

For overcome that the atmospheric phase that exists in D-InSAR technology postpones and dephasing dry etc., main method is D-InSAR Time Series Analysis Method.The D-InSAR Time Series Analysis Method first SAR(based on a large amount of is generally greater than 25 scapes), close serious pixel and stable pixel collection is carried out to time series analysis by abandoning dephasing, then space-time decoherence and atmosphere delay impact be can farthest weaken by relevant statistical theory, precision and the reliability of Ground Deformation monitoring improved.Mainly contain at present following several D-InSAR Time Series Analysis Method: permanent scatterer interfere measurement technique, short baseline set interfere measurement technique (Short BaselinesInterferometry, SBAS), StaMPS (Stanford Method for Persistent Scatterer), interference point target analysis (Interferometric point target analysis) (Werner et al., 2003), Squeeze SAR (SqueeSAR).

D-InSAR Time Series Analysis Method is all with PSI(Persistent scattererinterferometric, and permanent scatterer is interfered) technology is theoretical foundation.The core concept of PSI technology is to PS(Persistent scatterer, permanent scatterer) interferometric phase of point carries out time series analysis, according to the space-time characteristic of each phase component, estimation atmospheric wave, DEM error and noise etc., it is separated one by one from differential interferometry phase place, finally obtain linearity and non-linear rate of deformation, atmosphere delay amount (Atmosphere Phase Screen, APS) and DEM error that each PS is ordered.PSI technology, compared with D-InSAR technology, has following features: first need a large amount of SAR images, it has been generally acknowledged that and need to be greater than 25 scapes, Hooper etc. think and SAR image number can be down to 12 scapes; Its research object is no longer whole scape image, but therefrom filters out the PS point with stable scattering properties, forms discrete point observation grid (higher than conventional deformation monitoring reticular density); Secondly, can not to PS point carry out apart to orientation to spectral filtering; Secondly, due to using PS point as object of observation, reduced the impact of Space Baseline on coherence, even under the condition of Critical baseline, still can be by analyzing the variation inverting deformation data of PS differential interferometry phase place; In addition, count clear and definite elaboration in the solution of DEM error, atmospheric phase estimation and non-linear deformation etc., can utilize low Accuracy Figure elevation model to estimate the DEM corrected value at PS point place, can also obtain the atmosphere delay phase place of main and auxiliary image; Through the processing of PSI method, deformation data is estimated to reach to the precision of submillimeter level.

The flow chart of data processing of PSI technology comprises: first, by covering the M width SAR image makeup time sequence of areal, choose 1 width image as public main image, all the other all images are all registrated on main image, generates M-1 interferogram; Then extract the PS point target that keeps high coherence on SAR image; Choose suitable region DEM, all DEM are carried out to coordinate conversion, be modeled to the phase diagram under radar fix system, all interference, to carrying out phase difference processing with the DEM of simulation one by one, obtain M-1 differential interferometry figure; Utilize the right spatial autocorrelation characteristic modeling of the adjacent PS point of short distance, the difference of the differential interferometry phase place that contiguous PS is ordered can be expressed as the functional form of deformation phase place, DEM error, atmosphere delay impact and noise etc.; Selected one of them high-quality PS point, as phase unwrapping reference point, utilizes regretional analysis to try to achieve Linear deformation rate and vertical error that each PS is ordered, and the residual phase that PS is ordered separates, and comprises non-linear deformation and atmosphere delay phase place; Finally deformation result is carried out to geocoding, obtain rate of deformation and the deformation sequence of areal coverage.

Conventional PSI method needs PS point to unify network forming in processing procedure, is connected to form network by segmental arc, then resolves its elevation and deformation parameter.But under high resolving power condition, in urban environment, buildings is under long base line condition, phase difference between PS point on PS point and background atural object on its interferometric phase is amplified, do not having under the condition of accurate urban surface model (DSM), make the segmental arc formation of connecting building thing and background atural object be similar to discrete abrupt slope phase place, therefore this segmental arc cannot meet corresponding threshold condition.The disposal route of PSI is mainly to reject this PS point that cannot meet in threshold condition segmental arc.Meanwhile, consider that buildings elevation phase place and deformation phase place easily lump together, in the situation that can not accurately resolving buildings elevation, will directly affect the extraction accuracy of buildings deformation data.Therefore, how to solve under high resolving power condition, the high phase gradient problem forming under long base line condition between City Building and background atural object, thus the estimated accuracy of raising building elevation deformation data is the main bugbear that high resolving power PSI technology faces.

Summary of the invention

Conventional PSI method completes after difference processing (comprising level land phase place removal and elevation phase compensation), in processing procedure, need PS point to unify network forming, form the network being connected by segmental arc, and then utilize the sequence of the difference of the differential phase of method to consecutive point that two-dimension periodic estimates to solve, and then the segmental arc integration that meets threshold condition resolves the deformation values of whole net.Under high resolution SAR condition, City Building projects to and on two dimensional surface, has formed together with multiple PS points are aliasing in the PS point on background atural object after oblique distance imaging, unify at PSI under the mode of network forming, PS point on buildings on PS point and background atural object inevitably links together, for the permanent scatterer of long baseline, between them, the difference of interferometric phase is amplified, and makes to form high phase gradient between consecutive point, easily makes follow-up permanent scatterer parameter estimation produce and resolves mistake.

The present invention is mainly devoted to study permanent scatterer modeling and the parameter estimation algorithm that under new high resolving power condition, in urban environment, buildings separates with background atural object, by separating the PS point on buildings and background atural object, given solution twines path and resolves detached building elevation and deformation data.The method can effectively be avoided producing segmental arc between the PS point on buildings and background atural object and connect, and ensures, in the time that the adjacent tie point of each segmental arc is set up to phase difference model, to produce phase place discontinuous on the phase increment in segmental arc, avoids the problem of high phase gradient.

The present invention is directed to above-mentioned technical matters is mainly solved by following technical proposals: the permanent scatterer modeling method of a kind of high-resolution buildings and background separation, comprises the following steps:

S01, buildings is separated with the PS point on background atural object;

S02, the PS point on buildings and background atural object is carried out to independent network forming;

S03, elevation and deformation parameter that the PS on buildings is ordered are estimated, obtain final deformation map.

As preferably, in described step S01, buildings is separated and is specially with the PS point on background atural object: first, according to closing on neighborhood pixels gray scale, texture etc. on SAR image, image is cut apart; Secondly, utilize Full Lambda-Schedule algorithm, on the basis in conjunction with gray scale and spatial information, iteration merges contiguous little patch; Then, calculate the attribute of buildings classification; Then, adopt the contiguous method of K SAR image to be classified in the Euclidean distance of N dimension space according to data to be sorted and training area element, obtain the mask file of different buildingss; Finally, according to the mask file of different buildingss, PS point is classified, PS point is referred on different buildingss, obtain final classification results.

As preferably, in described step S02, PS point on buildings and background atural object is carried out to independent network forming to be specially: first the PS point on buildings is formed to an independently Delaunay triangulation network, then the PS point on background atural object is formed to an independent Delaunay triangulation network, and between the triangulation network of every solitary building and the triangulation network of background atural object, set up a tie point, by tie point by two independently network couple together.

As preferably, in described step S03, elevation and deformation parameter that PS on buildings is ordered are estimated, obtaining final deformation map is specially: adopt multi-frame interferometry to the average weighted method of elevation phase result obtaining to weaken noise effect, improve the reliability of elevation phase estimation, concrete formula is:

$\widehat{\mathrm{\φ}}=\frac{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j\·{\mathrm{\φ}}_j}{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j}$

In formula, φ _jinterfere right solution for single width and twine phase place, for the optimum evaluation of elevation phase place, t is interferogram quantity, p _jfor power corresponding to every width interferogram; The combination elevation phase place of then PS on buildings being ordered is removed building elevation phase place and is contributed to extract buildings deformation phase information from interferogram, the deformation phase place of then ordering taking PS on buildings is handling object, utilize two-dimensional linear to return phase model, progressively revise height value by several times interative computation, resolve rate of deformation, finally realize the extraction to deformation sequence; Finally, by the deformation result of the deformation result of detached building and background atural object, be connected and form deformation map in a big way by common tie point.

The substantial effect that the present invention brings is, first the PS point that conventional method obtains, the PS point of top of building is crossed mostly disallowablely due to landform residual phase, causes PS point rare, on the method buildings that the present invention proposes, PS point quantity is many, substantially covers every solitary building; Secondly, the PS point rate of deformation that conventional method obtains is obviously bigger than normal, indivedual annual deformation of PS point even reaches centimetre-sized, this result is obviously with residual buildings elevation phase place, along with the variation of depth of building, affect the estimation of rate of deformation on buildings, made the rate of deformation result of buildings bigger than normal.Vertical error remaining in the rate of deformation result that the improved PSI calculation method that the present invention proposes obtains is obviously less, rate of deformation numerical value is less, effectively reduce the impact that high phase gradient resolves buildings elevation and deformation, the superiority of visible this programme in buildings deformation monitoring.

Brief description of the drawings

Fig. 1 is a kind of schematic flow sheet of the present invention.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment: the permanent scatterer modeling method of a kind of high-resolution buildings of the present embodiment and background separation, as shown in Figure 1, comprises the following steps:

S01, buildings is separated with the PS point on background atural object;

S02, the PS point on buildings and background atural object is carried out to independent network forming;

S03, elevation and deformation parameter that the PS on buildings is ordered are estimated, obtain final deformation map.

1, buildings and separating that PS on background atural object is ordered

The present invention has adopted OO sorting technique to realize buildings and separating that PS on background atural object is ordered, whole sorting technique as follows: first, according to closing on neighborhood pixels gray scale, texture etc. on SAR image, image is cut apart, use a kind of partitioning algorithm based on edge, this algorithm calculates very fast, and as long as input parameter just can produce multi-scale division result.By the difference control of different scale coboundary, thereby produce from thin to thick multi-scale division; Secondly, when Image Segmentation, because threshold value is too low, some features can be by wrong point, and a feature is also likely divided into a lot of parts.We have utilized FullLambda-Schedule algorithm, merge contiguous little patch in conjunction with iteration on the basis of gray scale and spatial information; Then, calculating the attribute of buildings classification, is mainly the gray feature (the gradation of image value of buildings is higher, and the gradation of image value of background atural object is lower) according to buildings, other half-tone information of compute classes; Then, adopt the contiguous method of K SAR image to be classified in the Euclidean distance of N dimension space according to data to be sorted and training area element, obtain the mask file of different buildingss; Finally, according to the mask file of different buildingss, PS point is classified, PS point is referred on different buildingss, obtain final classification results.

2, the independent network forming that on buildings and background atural object, PS is ordered

Obtain after the PS point on detached building, can carry out the network forming of PS point.Delaunay TIN is generally selected in conventional PS network forming, couples together and forms non-overlapping triangular net by all discrete PS points.But this network forming mode is the unified some connected mode under global conditions, the PS point on buildings and background atural object is inevitably linked together.Do not having accurate DSM to carry out under elevation phase compensation condition, be difficult to apply rate of deformation and vertical error correction between the correct estimation point of two-dimension periodic figure because the phase place between point is discontinuous.If the PS point on the buildings identifying and background atural object is carried out independent network forming by we, can effectively avoid the connection of segmental arc between PS point on buildings and background atural object.Therefore, we adopt buildings and background atural object independent network forming mode respectively.First the PS point on buildings is formed to an independently Delaunay triangulation network, then the PS point on background atural object is formed to an independent Delaunay triangulation network, and between the triangulation network of every solitary building and the triangulation network of background atural object, set up a tie point, by tie point we can by two independently network couple together.

3, on buildings, PS point height and deformation parameter are estimated

We adopt the mode of many interferograms combination to carry out the elevation phase estimation of detached building, eliminate atmosphere delay error and phase noise error is interfered the impact on the elevation phase result obtaining to single width with this.Concrete grammar be adopt multi-frame interferometry to the average weighted method of elevation phase result obtaining to weaken noise effect, improve the reliability of elevation phase estimation:

$\widehat{\mathrm{\φ}}=\frac{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j\·{\mathrm{\φ}}_j}{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j}---\left(1\right)$

In formula, φ _jinterfere right solution for single width and twine phase place, for the optimum evaluation of elevation phase place, t is interferogram quantity, p _jfor power corresponding to every width interferogram.Within the specific limits, interfere the vertical component of baseline larger, elevation result is more reliable.Therefore,, at fixed temporary necessary consideration vertical parallax, need in addition the factor of considering to comprise phase noise intensity and atmosphere delay impact.The coherence who considers buildings better and scope is less (is less than 1km ²) affected by atmosphere delay less, the present invention adopts the power method of determining of only considering baseline, and has relatively large vertical parallax B _⊥interference more reliable to the elevation result obtaining.General right to choose repeated factor and B _⊥ ²be directly proportional, i.e. p=B _⊥ ².Weighting formula can be rewritten into:

$\widehat{\mathrm{\φ}}=\frac{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{B_{\⊥j}}^2\·{\mathrm{\φ}}_j}{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{B_{\⊥j}}^2}---\left(2\right)$

The combination elevation phase place that we order the PS on buildings is removed building elevation phase place and is contributed to extract buildings deformation phase information from interferogram, the deformation phase place of then ordering taking PS on buildings is handling object, utilize two-dimensional linear to return phase model, progressively revise height value by interative computation repeatedly, resolve rate of deformation, finally realize the extraction to deformation sequence.Finally, by the deformation result of the deformation result of detached building and background atural object, be connected and form deformation map in a big way by common tie point.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendments or supplement or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.

Although more used the terms such as independent network forming, multi-frame interferometry herein, do not got rid of the possibility that uses other term.Use these terms to be only used to describe more easily and explain essence of the present invention; They are construed to any additional restriction is all contrary with spirit of the present invention.

Claims (4)

1. a permanent scatterer modeling method for high-resolution buildings and background separation, is characterized in that, comprises the following steps:

S01, buildings is separated with the PS point on background atural object;

S02, the PS point on buildings and background atural object is carried out to independent network forming;

S03, elevation and deformation parameter that the PS on buildings is ordered are estimated, obtain final deformation map.

2. the permanent scatterer modeling method of a kind of high-resolution buildings according to claim 1 and background separation, it is characterized in that, in described step S01, buildings is separated and is specially with the PS point on background atural object: first, according to closing on neighborhood pixels gray scale, texture etc. on SAR image, image is cut apart; Secondly, utilize Full Lambda-Schedule algorithm, on the basis in conjunction with gray scale and spatial information, iteration merges contiguous little patch; Then, calculate the attribute of buildings classification; Then, adopt the contiguous method of K SAR image to be classified in the Euclidean distance of N dimension space according to data to be sorted and training area element, obtain the mask file of different buildingss; Finally, according to the mask file of different buildingss, PS point is classified, PS point is referred on different buildingss, obtain final classification results.

3. the permanent scatterer modeling method of a kind of high-resolution buildings according to claim 1 and 2 and background separation, it is characterized in that, in described step S02, PS point on buildings and background atural object is carried out to independent network forming to be specially: first the PS point on buildings is formed to an independently Delaunay triangulation network, then the PS point on background atural object is formed to an independent Delaunay triangulation network, and between the triangulation network of every solitary building and the triangulation network of background atural object, set up a tie point, by tie point by two independently network couple together.

4. the permanent scatterer modeling method of a kind of high-resolution buildings according to claim 3 and background separation, it is characterized in that, in described step S03, elevation and deformation parameter that PS on buildings is ordered are estimated, obtaining final deformation map is specially: adopt multi-frame interferometry to the average weighted method of elevation phase result obtaining to weaken noise effect, improve the reliability of elevation phase estimation, concrete formula is:

$\widehat{\mathrm{\φ}}=\frac{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j\·{\mathrm{\φ}}_j}{\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{p}_j}$

In formula, φ _jinterfere right solution for single width and twine phase place, for the optimum evaluation of elevation phase place, t is interferogram quantity, p _jfor power corresponding to every width interferogram; The combination elevation phase place of then PS on buildings being ordered is removed building elevation phase place and is contributed to extract buildings deformation phase information from interferogram, the deformation phase place of then ordering taking PS on buildings is handling object, utilize two-dimensional linear to return phase model, progressively revise height value by several times interative computation, resolve rate of deformation, finally realize the extraction to deformation sequence; Finally, by the deformation result of the deformation result of detached building and background atural object, be connected and form deformation map in a big way by common tie point.