CN116068511B - Deep learning-based InSAR large-scale system error correction method - Google Patents
Deep learning-based InSAR large-scale system error correction method Download PDFInfo
- Publication number
- CN116068511B CN116068511B CN202310220889.3A CN202310220889A CN116068511B CN 116068511 B CN116068511 B CN 116068511B CN 202310220889 A CN202310220889 A CN 202310220889A CN 116068511 B CN116068511 B CN 116068511B
- Authority
- CN
- China
- Prior art keywords
- data
- phase
- unwrapping
- longitude
- neural network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000012937 correction Methods 0.000 title claims abstract description 23
- 238000013135 deep learning Methods 0.000 title claims abstract description 13
- 238000003062 neural network model Methods 0.000 claims abstract description 34
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000006835 compression Effects 0.000 claims description 18
- 238000007906 compression Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000007670 refining Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 description 7
- 230000009286 beneficial effect Effects 0.000 description 6
- 238000012876 topography Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005305 interferometry Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 241001420622 Meris Species 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000005433 ionosphere Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/40—Means for monitoring or calibrating
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B15/00—Measuring arrangements characterised by the use of electromagnetic waves or particle radiation, e.g. by the use of microwaves, X-rays, gamma rays or electrons
- G01B15/06—Measuring arrangements characterised by the use of electromagnetic waves or particle radiation, e.g. by the use of microwaves, X-rays, gamma rays or electrons for measuring the deformation in a solid
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
- G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
- G01S13/9021—SAR image post-processing techniques
- G01S13/9023—SAR image post-processing techniques combined with interferometric techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks
Abstract
The invention discloses an InSAR large-scale system error correction method based on deep learning, which comprises the following steps: collecting data of a coverage target area and carrying out data preprocessing on the data to obtain unwrapped phase data after geocoding, DEM data after clipping and longitude and latitude data; building a deep neural network model according to the relation between the related data; inputting the data obtained through pretreatment into a deep neural network model to obtain a simulated large-scale system error unwrapping phase; obtaining a corrected unwrapping phase; inverting the deformation parameters to obtain corrected deformation results. The invention provides a model capable of correcting the atmosphere phase, the track residual phase and part of turbulence phase related to the terrain, and can realize the large-scale system error correction of InSAR in high and steep mountain areas, hilly areas and flat areas.
Description
Technical Field
The invention relates to the technical field of synthetic aperture radar interferometry, in particular to an InSAR large-scale system error correction method based on deep learning.
Background
Synthetic aperture radar interferometry is a satellite-borne sensor-based earth observation measurement technique that has evolved rapidly over the last thirty years. The strain monitoring device has the strain monitoring capability in a large range, all over the day and all weather. Along with the continuous improvement of the time resolution and the spatial resolution of SAR images and the continuous development and progress of multi-source multi-phase SAR image alternative and InSAR time sequence algorithms, the InSAR technology is widely applied to the early identification and monitoring fields of geological disaster hidden danger, such as: volcanic monitoring, ground subsidence monitoring, landslide monitoring, seismic monitoring, and glacial movement. The technique can effectively make up the defects of the traditional measurement method, however, due to the existence of large-scale phase errors, such as: ionosphere, topography dependent atmospheric phase, track residual phase. These errors can have a serious impact on the accuracy of the deformation measurement. Such as: when the atmospheric relative humidity changes in a space-time manner by 20%, errors of a few decimeters are caused to deformation results; the track phase is similar to the spatial characteristics of large-scale deformation, and can seriously affect the monitoring of the large-scale deformation, such as earthquake and volcanic motion. Therefore, the comprehensive removal of the large-scale phase error is of great significance.
In order to remove the large-scale InSAR system error, in the last few decades, many scholars have proposed a large number of methods for different large-scale system error components. Correction methods based on external data such as digital weather products (weather models), GNSS data, spectroscopic analysis (MODIS, MERIS), or combinations thereof are proposed for the topography-dependent atmospheric phase; the second is a correction model-based method, such as: a linear correction method and a power rate correction method. A satellite orbit track-based compensation model is provided for the orbit residual phase, and the baseline error in the InSAR imaging geometry is accurately estimated; secondly, based on the characteristics of the track phase in a space domain and a frequency domain; then assistance based on external data, such as GNSS; finally, a mathematical fitting method is mainly used for models such as linear fitting, polynomial fitting and the like.
The proposed correction methods described above have proven to be somewhat successful in removing the terrain-related atmospheric phase and the track residual phase, but these methods are limited by their inherent limitations. Conventional methods based on linear models, polynomials, etc. cannot fully describe each interference pair and orbit characteristics and can be affected by spatial characteristics and orbit phase and topography-dependent atmospheric phase deformations.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides an InSAR large-scale system error correction method based on deep learning, provides a model capable of correcting the atmosphere phase, the track residual phase and part of turbulence phase related to the terrain at the same time, and solves the problem of limitation of the traditional method.
In order to achieve the aim of the invention, the invention adopts the following technical scheme: an InSAR large-scale system error correction method based on deep learning, which is characterized by comprising the following steps:
s1: collecting SAR data, DEM data and longitude and latitude data covering a target area, preprocessing the data, removing the area with poor coherence, and obtaining unwrapped phase data after geocoding, and DEM data and longitude and latitude data after clipping;
s2: according to the relation between the related data, a deep neural network model is built, wherein the related data comprises DEM data, longitude and latitude data, a terrain related atmosphere phase, a track residual phase and a turbulence phase;
s3: inputting the pre-processed geocoded unwrapping phase data, the cut DEM data and the latitude and longitude data into a deep neural network model to obtain a simulated large-scale system error unwrapping phase;
s4: subtracting the original unwrapping phase after geocoding from the unwrapping phase of the large-scale system error after simulation of the deep neural network model to obtain a corrected unwrapping phase;
s5: and (3) reversely geocoding the corrected unwrapping phase to an SAR coordinate system, and inverting deformation parameters to obtain corrected deformation results.
The beneficial effect of above-mentioned scheme is: according to the technical scheme, the track phase and the terrain-related atmosphere phase are comprehensively considered, the functional relation between the track phase and the unwrapping phase is obtained, the spatial characteristics of the turbulence phase are considered while the terrain-related atmosphere phase and the track phase are removed, part of the turbulence phase can be removed, and meanwhile, the terrain-related atmosphere phase, the track residual phase and part of the turbulence phase are corrected, so that the problem that the traditional method is limited is solved, and deformation can be accurately identified.
Further, in S1, the step of performing the InSAR data preprocessing on the SAR data includes the following steps:
s1-1: acquiring two SLC images of the same research area in time sequence, and selecting one of the images as a main image;
s1-2: reducing the influence of speckle noise by setting a multi-view ratio of 10:2 on the distance and the azimuth of the image, setting corresponding filtering parameters according to the size of the noise, and filtering to obtain filtered coherence data;
s1-3: setting an unwrapping threshold value through the data distribution of the coherence data after filtering, ensuring the sufficiency of training data and preventing the influence of unwrapping error data;
s1-4: selecting a minimum cost flow method to perform phase unwrapping on the wrapped phase obtained by calculating the selected image interferogram according to the unwrapping threshold;
s1-5: and transferring the unwrapping phase data in the SAR coordinate system to the same geographic coordinates as the cut DEM data and longitude and latitude data through geographic coding, and obtaining the unwrapping phase data after geographic coding.
The beneficial effects of the above-mentioned further scheme are: through the technical scheme, the collected data is preprocessed, so that the data required by training the deep neural network is obtained.
Further, the deep neural network model construction in S2 includes the following steps:
s2-1: the method comprises the steps that a full-connection layer network of a deep neural network model is used as a feature module for extracting and inputting DEM data and longitude and latitude data information, and a channel attention mechanism module of the deep neural network model is selected as a weighted feature extraction module of data by combining the relation among the DEM data, the longitude and latitude data, the terrain-related atmospheric phase, the track residual phase and the turbulence phase;
s2-2: performing feature compression operation through a weighted feature extraction module, performing feature compression on the space dimension, and generating corresponding weights for each related data feature dimension through restoration operation, wherein the method comprises the following formula:
the eigenvalue compression formula is as follows:
wherein ,
for input data, ++>
and />
Length and width of input data, respectively, +.>
For compression operation, +.>
For the compression of the original data +.>
Two-dimensional matrix->
and />
All are image pixel coordinates;
the weight calculation formula is as follows:
wherein ,
for the restoration operation, ++>
For RELU activation function, +.>
Dimension reduction layer for RELU activation function parameters, < ->
Up-level for RELU activation function parameters, < ->
For characteristic channel weight, ++>
For the result after the compression of the original data, +.>
Pooling for global averaging;
s2-3: the weight is applied to each original relevant data characteristic channel, and the formula is as follows:
The beneficial effects of the above-mentioned further scheme are: through the technical scheme, the depth neural network model is built in consideration of the used data, so that the simulated large-scale system error unwrapping phase can be obtained conveniently, and meanwhile, the weight is acted on each original related data characteristic channel, so that the importance of different data channels can be learned.
Further, the deep neural network model selects MSE as a loss function for parameter evaluation, and selects Adam optimization algorithm as a parameter optimization algorithm.
The beneficial effects of the above-mentioned further scheme are: by the technical scheme, the model can be fitted with a better global parameter by considering that the atmosphere phase related to the terrain and the terrain functional relationship are spatially changed, namely, the functional relationship between the terrain and the atmosphere phase in different areas is different.
Further, the step S3 comprises the following steps:
s3-1: the method comprises the steps of inputting pre-processed geocoded unwrapping phase data, cut DEM data and longitude and latitude data into a deep neural network model, extracting spatial features of the DEM data and the longitude and latitude data through a full-connection module of the deep neural network model, and obtaining weights of the DEM data, the longitude and latitude data and the unwrapping phase data through a channel attention mechanism module;
s3-2: obtaining a functional relation among the DEM data, the longitude and latitude data and the unwrapped phase data according to the weights of the DEM data, the longitude and latitude data and the unwrapped phase data, and obtaining a pre-training weight;
s3-3: and obtaining a final prediction result through a deep neural network model according to the pre-training weight to obtain a simulated large-scale system error unwrapping phase.
The beneficial effects of the above-mentioned further scheme are: through the technical scheme, after data are input into the deep neural network model for processing, the unwrapped phase after the large-scale system error is removed is obtained.
Further, S5 includes the following steps:
s5-1: recoding the corrected unwrapped phase to SAR coordinates, performing DINSAR inversion deformation operation, performing orbit phase refining and re-flattening through selection and refinement of stable ground control points, eliminating orbit error phase in differential interference, and inverting deformation parameters;
s5-2: and performing geocoding on the inversion deformation result to obtain a deformation result after the system error is corrected.
The beneficial effects of the above-mentioned further scheme are: and recoding the obtained corrected unwrapped phase to SAR coordinates, continuing the corresponding DINSAR inversion deformation operation to obtain a deformation result diagram after system error correction, and comparing the deformation result diagram to obviously see the advantages of the invention.
Drawings
FIG. 1 is a flow chart of an InSAR large scale system error correction method based on deep learning.
FIG. 2 is a technical flow chart of an InSAR large scale system error correction method based on deep learning.
FIG. 3 is a graph of the large scale error correction results of the InSAR of the FC-Net portion.
FIG. 4 is a graph comparing deformation results before and after FC-Net correction.
Detailed Description
The invention will be further described with reference to the drawings and specific examples.
As shown in fig. 1, a deep learning-based method for correcting an InSAR large-scale system error is characterized in that the method comprises the following steps:
s1: collecting SAR data, DEM data and longitude and latitude data covering a target area, preprocessing the data, removing the area with poor coherence, and obtaining unwrapped phase data after geocoding, and DEM data and longitude and latitude data after clipping;
s2: according to the relation between the related data, a deep neural network model is built, wherein the related data comprises DEM data, longitude and latitude data, a terrain related atmosphere phase, a track residual phase and a turbulence phase;
s3: inputting the pre-processed geocoded unwrapping phase data, the cut DEM data and the latitude and longitude data into a deep neural network model to obtain a simulated large-scale system error unwrapping phase;
s4: subtracting the original unwrapping phase after geocoding from the unwrapping phase of the large-scale system error after simulation of the deep neural network model to obtain a corrected unwrapping phase;
s5: and (3) reversely geocoding the corrected unwrapping phase to an SAR coordinate system, and inverting deformation parameters to obtain corrected deformation results.
In addition, in S1, the SAR data is subjected to InSAR data preprocessing, including the following steps:
s1-1: acquiring two SLC images of the same research area in time sequence, and selecting one of the images as a main image;
s1-2: reducing the influence of speckle noise by setting a multi-view ratio of 10:2 on the distance and the azimuth of the image, setting corresponding filtering parameters according to the size of the noise, and filtering to obtain filtered coherence data;
s1-3: setting an unwrapping threshold value through the data distribution of the coherence data after filtering, ensuring the sufficiency of training data and preventing the influence of unwrapping error data;
s1-4: selecting a minimum cost flow method to perform phase unwrapping on the wrapped phase obtained by calculating the selected image interferogram according to the unwrapping threshold;
s1-5: and transferring the unwrapping phase data in the SAR coordinate system to the same geographic coordinates as the cut DEM data and longitude and latitude data through geographic coding, and obtaining the unwrapping phase data after geographic coding.
The deep neural network model construction in the S2 comprises the following steps:
s2-1: the method comprises the steps that a full-connection layer network of a deep neural network model is used as a feature module for extracting and inputting DEM data and longitude and latitude data information, and a channel attention mechanism module of the deep neural network model is selected as a weighted feature extraction module of data by combining the relation among the DEM data, the longitude and latitude data, the terrain-related atmospheric phase, the track residual phase and the turbulence phase;
s2-2: performing feature compression operation through a weighted feature extraction module, performing feature compression on the space dimension, and generating corresponding weights for each related data feature dimension through restoration operation, wherein the method comprises the following formula:
the eigenvalue compression formula is as follows:
wherein ,
for input data, ++>
and />
Length and width of input data, respectively, +.>
For compression operation, +.>
For the compression of the original data +.>
Two-dimensional matrix->
and />
All are image pixel coordinates; />
The weight calculation formula is as follows:
wherein ,
for the restoration operation, ++>
For RELU activation function, +.>
The dimension reduction layer for the RELU activation function parameters,
up-level for RELU activation function parameters, < ->
For characteristic channel weight, ++>
For the result after the compression of the original data, +.>
Pooling for global averaging;
s2-3: the weight is applied to each original relevant data characteristic channel, and the formula is as follows:
The deep neural network model selects MSE as a loss function to perform parameter evaluation, and selects Adam optimization algorithm as a parameter optimization algorithm.
S3, the following steps are included:
s3-1: the method comprises the steps of inputting pre-processed geocoded unwrapping phase data, cut DEM data and longitude and latitude data into a deep neural network model, extracting spatial features of the DEM data and the longitude and latitude data through a full-connection module of the deep neural network model, and obtaining weights of the DEM data, the longitude and latitude data and the unwrapping phase data through a channel attention mechanism module;
s3-2: obtaining a functional relation among the DEM data, the longitude and latitude data and the unwrapped phase data according to the weights of the DEM data, the longitude and latitude data and the unwrapped phase data, and obtaining a pre-training weight;
s3-3: and obtaining a final prediction result through a deep neural network model according to the pre-training weight to obtain a simulated large-scale system error unwrapping phase.
S5, the following steps are included:
s5-1: recoding the corrected unwrapped phase to SAR coordinates, performing DINSAR inversion deformation operation, performing orbit phase refining and re-flattening through selection and refinement of stable ground control points, eliminating orbit error phase in differential interference, and inverting deformation parameters;
s5-2: and performing geocoding on the inversion deformation result to obtain a deformation result after the system error is corrected.
In one embodiment of the present invention, as shown in fig. 2, SAR image data and precision orbit data covering a target area are collected, SLC data is obtained through registration processing, an original interference pair is formed, meanwhile, a land leveling phase and a terrain phase are removed by using external elevation data, a differential interference pair is formed, a winding phase is unwrapped and geocoded through a minimum cost flow unwrapping algorithm, the unwrapped phase, elevation and longitude and latitude data after preprocessing are input into a deep neural network to obtain a simulated unwrapped phase, the simulated terrain-related atmospheric phase and the simulated orbit residual phase are included, then the unwrapped phase after geocoding is subtracted from the unwrapped phase simulated by a network model to obtain a corrected unwrapped phase, the corrected unwrapped phase is reversely geocoded to a SAR coordinate system, and deformation parameters are inverted, so as to obtain a corrected deformation result. In this embodiment, a fully-connected channel attention network is used as the deep neural network.
The invention provides a method for realizing large-scale system error correction of InSAR in high and steep mountain areas, hilly areas and flat areas, and provides a model capable of correcting the landform-related atmospheric phase, the track residual phase and part of turbulence phase at the same time, which has important significance for acquiring actual deformation rate and correct interpretation of InSAR. Fig. 3 provides error correction results, and gives a comparison graph of the interference pair after filtering, the original unwrapping phase, the simulated unwrapping phase and the corrected unwrapping phase results, and as can be seen from fig. 3 and fig. 4 (a and c are deformation results before correction, and b and d are deformation results after correction in fig. 4), the deformation can be identified from the area affected by the atmospheric influence and the track phase by greatly weakening the influence of the topography-related atmospheric phase, the track residual phase and part of the small-scale turbulence phase by using the method when the deformation is monitored by using the DInSAR.
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit of the invention, and such modifications and combinations are still within the scope of the invention.
Claims (4)
1. An InSAR large-scale system error correction method based on deep learning, which is characterized by comprising the following steps:
s1: collecting SAR data, DEM data and longitude and latitude data covering a target area, preprocessing the data, removing the area with poor coherence, and obtaining unwrapped phase data after geocoding, and DEM data and longitude and latitude data after clipping;
s2: according to the relation between the related data, a deep neural network model is built, wherein the related data comprises DEM data, longitude and latitude data, a terrain related atmosphere phase, a track residual phase and a turbulence phase;
s3: inputting the pre-processed geocoded unwrapping phase data, the cut DEM data and the latitude and longitude data into a deep neural network model to obtain a simulated large-scale system error unwrapping phase;
s4: subtracting the original unwrapping phase after geocoding from the unwrapping phase of the large-scale system error after simulation of the deep neural network model to obtain a corrected unwrapping phase;
s5: reversely geocoding the corrected unwrapping phase to an SAR coordinate system, and inverting deformation parameters to obtain corrected deformation results;
the deep neural network model construction in the S2 comprises the following steps:
s2-1: the method comprises the steps that a full-connection layer network of a deep neural network model is used as a feature module for extracting and inputting DEM data and longitude and latitude data information, and a channel attention mechanism module of the deep neural network model is selected as a weighted feature extraction module of data by combining the relation among the DEM data, the longitude and latitude data, the terrain-related atmospheric phase, the track residual phase and the turbulence phase;
s2-2: performing feature compression operation through a weighted feature extraction module, performing feature compression on the space dimension, and generating corresponding weights for each related data feature dimension through restoration operation, wherein the method comprises the following formula:
the eigenvalue compression formula is as follows:
wherein ,
for input data, ++>
and />
Length and width of input data, respectively, +.>
For compression operation, +.>
For the compression of the original data +.>
Two-dimensional matrix->
and />
All are image pixel coordinates;
the weight calculation formula is as follows:
wherein ,
for the restoration operation, ++>
For RELU activation function, +.>
Dimension reduction layer for RELU activation function parameters, < ->
Up-level for RELU activation function parameters, < ->
For characteristic channel weight, ++>
For the result after the compression of the original data, +.>
Pooling for global averaging;
s2-3: the weight is applied to each original relevant data characteristic channel, and the formula is as follows:
the step S3 comprises the following steps:
s3-1: the method comprises the steps of inputting pre-processed geocoded unwrapping phase data, cut DEM data and longitude and latitude data into a deep neural network model, extracting spatial features of the DEM data and the longitude and latitude data through a full-connection module of the deep neural network model, and obtaining weights of the DEM data, the longitude and latitude data and the unwrapping phase data through a channel attention mechanism module;
s3-2: obtaining a functional relation among the DEM data, the longitude and latitude data and the unwrapped phase data according to the weights of the DEM data, the longitude and latitude data and the unwrapped phase data, and obtaining a pre-training weight;
s3-3: and obtaining a final prediction result through a deep neural network model according to the pre-training weight to obtain a simulated large-scale system error unwrapping phase.
2. The method for correcting the error of the deep learning-based InSAR large scale system according to claim 1, wherein the step of preprocessing the InSAR data in S1 comprises the following steps:
s1-1: acquiring two SLC images of the same research area in time sequence, and selecting one of the images as a main image;
s1-2: reducing the influence of speckle noise by setting a multi-view ratio of 10:2 on the distance and the azimuth of the image, setting corresponding filtering parameters according to the size of the noise, and filtering to obtain filtered coherence data;
s1-3: setting an unwrapping threshold value through the data distribution of the coherence data after filtering, ensuring the sufficiency of training data and preventing the influence of unwrapping error data;
s1-4: selecting a minimum cost flow method to perform phase unwrapping on the wrapped phase obtained by calculating the selected image interferogram according to the unwrapping threshold;
s1-5: and transferring the unwrapping phase data in the SAR coordinate system to the same geographic coordinates as the cut DEM data and longitude and latitude data through geographic coding, and obtaining the unwrapping phase data after geographic coding.
3. The deep learning-based InSAR large scale system error correction method according to claim 1, wherein the deep neural network model selects MSE as a loss function for parameter evaluation, and selects Adam optimization algorithm as a parameter optimization algorithm.
4. The method for correcting the large-scale system error of the InSAR based on the deep learning according to claim 1, wherein the step of S5 comprises the following steps:
s5-1: recoding the corrected unwrapped phase to SAR coordinates, performing DINSAR inversion deformation operation, performing orbit phase refining and re-flattening through selection and refinement of stable ground control points, eliminating orbit error phase in differential interference, and inverting deformation parameters;
s5-2: and performing geocoding on the inversion deformation result to obtain a deformation result after the system error is corrected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310220889.3A CN116068511B (en) | 2023-03-09 | 2023-03-09 | Deep learning-based InSAR large-scale system error correction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310220889.3A CN116068511B (en) | 2023-03-09 | 2023-03-09 | Deep learning-based InSAR large-scale system error correction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116068511A CN116068511A (en) | 2023-05-05 |
| CN116068511B true CN116068511B (en) | 2023-06-13 |
Family
ID=86171652
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310220889.3A Active CN116068511B (en) | 2023-03-09 | 2023-03-09 | Deep learning-based InSAR large-scale system error correction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116068511B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112816983B (en) * | 2021-01-06 | 2023-09-19 | 中南大学 | Time sequence InSAR turbulence atmosphere delay correction method based on optimized interference atlas |
Family Cites Families (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109541596B (en) * | 2018-11-28 | 2022-05-20 | 中国电子科技集团公司电子科学研究院 | InSAR image processing method and device based on deep learning algorithm |
| CN110332990B (en) * | 2019-05-24 | 2021-08-24 | 深圳大学 | Atmospheric turbulence real-time compensation method and system based on deep learning |
| AU2020101276B4 (en) * | 2020-01-23 | 2021-04-01 | Yongchao Wang | Systems and methods for processing GNSS data streams for determination of hardware and atmosphere-delays |
| CN111915570B (en) * | 2020-07-10 | 2024-03-08 | 云南电网有限责任公司带电作业分公司 | Atmospheric delay estimation method based on back propagation neural network |
| CN111998766B (en) * | 2020-08-31 | 2021-10-15 | 同济大学 | Surface deformation inversion method based on time sequence InSAR technology |
| CN112748430B (en) * | 2020-12-02 | 2023-10-20 | 北京理工大学 | Multi-base atmosphere phase joint compensation method based on space division |
| CN113238227B (en) * | 2021-05-10 | 2022-09-30 | 电子科技大学 | Improved least square phase unwrapping method and system combined with deep learning |
| CN113311433B (en) * | 2021-05-28 | 2022-08-02 | 北京航空航天大学 | InSAR interferometric phase two-step unwrapping method combining quality map and minimum cost flow |
| CN114265062B (en) * | 2021-11-11 | 2023-11-10 | 电子科技大学 | InSAR phase unwrapping method based on phase gradient estimation network |
| CN114280608B (en) * | 2022-03-07 | 2022-06-17 | 成都理工大学 | Method and system for removing DInSAR elevation-related atmospheric effect |
| CN115293223A (en) * | 2022-05-17 | 2022-11-04 | 电子科技大学长三角研究院(湖州) | Method and system for optimizing detail loss of phase unwrapping in deep learning and storage medium |
| CN115270608A (en) * | 2022-07-09 | 2022-11-01 | 自然资源部第二海洋研究所 | Coastal zone ground settlement prediction method based on ARIMA and LSTM |
| CN115272683A (en) * | 2022-07-25 | 2022-11-01 | 广西科技大学 | Central differential information filtering phase unwrapping method based on deep learning |
| CN115423180A (en) * | 2022-08-31 | 2022-12-02 | 内蒙古大学 | Surface deformation monitoring and predicting method |
| CN115545100A (en) * | 2022-09-27 | 2022-12-30 | 重庆三峡学院 | GB-InSAR atmospheric phase compensation method based on LSTM |
| CN115629384B (en) * | 2022-12-08 | 2023-04-11 | 中南大学 | Correction method of time sequence InSAR error and related equipment |
-
2023
- 2023-03-09 CN CN202310220889.3A patent/CN116068511B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN116068511A (en) | 2023-05-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110738252B (en) | Space autocorrelation machine learning satellite precipitation data downscaling method and system | |
| López-Quiroz et al. | Time series analysis of Mexico City subsidence constrained by radar interferometry | |
| Bovenga et al. | Application of multi-temporal differential interferometry to slope instability detection in urban/peri-urban areas | |
| Kandregula et al. | Estimation of regional surface deformation post the 2001 Bhuj earthquake in the Kachchh region, Western India using RADAR interferometry | |
| Zhang et al. | Subsidence monitoring in coal area using time-series InSAR combining persistent scatterers and distributed scatterers | |
| CN103454636B (en) | Differential interferometric phase estimation method based on multi-pixel covariance matrixes | |
| CN113281742B (en) | SAR landslide early warning method based on landslide deformation information and meteorological data | |
| CN114966692B (en) | Transformer-based InSAR technology frozen soil area multivariable time sequence deformation prediction method and device | |
| Grohmann et al. | Influence of cell size on volume calculation using digital terrain models: A case of coastal dune fields | |
| CN116068511B (en) | Deep learning-based InSAR large-scale system error correction method | |
| Chen et al. | ARU-net: Reduction of atmospheric phase screen in SAR interferometry using attention-based deep residual U-net | |
| CN112051572A (en) | Method for monitoring three-dimensional surface deformation by fusing multi-source SAR data | |
| CN114858238A (en) | Underground water storage and transformation monitoring and early warning system | |
| Rigo et al. | Monitoring of Guadalentín valley (southern Spain) through a fast SAR Interferometry method | |
| Zhang et al. | Reconstruction of GF-1 soil moisture observation based on satellite and in situ sensor collaboration under full cloud contamination | |
| CN111650579B (en) | InSAR mining area three-dimensional deformation estimation method and device for rock migration parameter adaptive acquisition and medium | |
| CN113251914A (en) | Surface deformation prediction method combining InSAR technology and long-term memory neural network | |
| Tang et al. | Atmospheric correction in time-series SAR interferometry for land surface deformation mapping–A case study of Taiyuan, China | |
| CN115079172A (en) | MTInSAR landslide monitoring method, equipment and storage medium | |
| Liu et al. | A Strain‐Model Based InSAR Time Series Method and Its Application to The Geysers Geothermal Field, California | |
| CN115201825A (en) | Atmospheric delay correction method in InSAR (interferometric synthetic aperture radar) inter-seismic deformation monitoring | |
| Luo et al. | Dynamic analysis of urban ground subsidence in Beijing based on the permanent scattering InSAR technology | |
| Feng et al. | A hierarchical network densification approach for reconstruction of historical ice velocity fields in East Antarctica | |
| CN113091600B (en) | Monitoring method for monitoring deformation of soft soil foundation by utilizing time sequence InSAR technology | |
| CN112285808B (en) | Method for reducing scale of APHRODITE precipitation data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |