CN117970331B - InSAR earth surface deformation monitoring method and system - Google Patents
InSAR earth surface deformation monitoring method and system Download PDFInfo
- Publication number
- CN117970331B CN117970331B CN202410395059.9A CN202410395059A CN117970331B CN 117970331 B CN117970331 B CN 117970331B CN 202410395059 A CN202410395059 A CN 202410395059A CN 117970331 B CN117970331 B CN 117970331B
- Authority
- CN
- China
- Prior art keywords
- layer
- representing
- sar
- sar image
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000012216 screening Methods 0.000 claims abstract description 11
- 238000006243 chemical reaction Methods 0.000 claims description 56
- 230000006870 function Effects 0.000 claims description 30
- 230000005489 elastic deformation Effects 0.000 claims description 28
- 230000008859 change Effects 0.000 claims description 23
- 238000011176 pooling Methods 0.000 claims description 22
- 238000013136 deep learning model Methods 0.000 claims description 18
- 230000004913 activation Effects 0.000 claims description 17
- 238000006073 displacement reaction Methods 0.000 claims description 16
- 230000003746 surface roughness Effects 0.000 claims description 16
- 238000010586 diagram Methods 0.000 claims description 14
- 239000000463 material Substances 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 11
- 230000004927 fusion Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 8
- 239000003673 groundwater Substances 0.000 claims description 5
- 230000002093 peripheral effect Effects 0.000 claims description 4
- 238000005303 weighing Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000013527 convolutional neural network Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 6
- 230000002829 reductive effect Effects 0.000 description 5
- 230000001174 ascending effect Effects 0.000 description 4
- 230000007423 decrease Effects 0.000 description 4
- 238000004590 computer program Methods 0.000 description 3
- 238000005286 illumination Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 206010061274 Malocclusion Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005305 interferometry Methods 0.000 description 2
- 238000012502 risk assessment Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 210000003127 knee Anatomy 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000012732 spatial analysis Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to the technical field of satellite geographic information, in particular to an InSAR earth surface deformation monitoring method and system, comprising the following steps: acquiring a first SAR image shot on a target surface area when an SAR satellite ascends and a second SAR image shot on the target surface area when the SAR satellite descends; screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain a credible pixel grid in the SAR image; fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area; and inputting the fused SAR image into a ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of the target ground surface area. The invention effectively improves the deformation monitoring accuracy and the data reliability of the InSAR technology under the complex terrain condition.
Description
Technical Field
The invention relates to the technical field of satellite geographic information, in particular to an InSAR earth surface deformation monitoring method and system.
Background
The synthetic aperture radar interferometry (InterferometricSyntheticApertureRadar, inSAR) technology, which is an efficient remote sensing monitoring technology, can provide high-resolution and large-range surface deformation data, and is important for understanding the fields of topographic surface process, disaster monitoring, environment assessment and the like.
Although InSAR technology is excellent in many aspects, there are some limitations in processing complex terrains (e.g., mountainous areas, canyons, etc.), and in particular, the change in the angle of incidence of the radar signal has a significant impact on the accuracy of InSAR deformation monitoring results. The change in incident angle causes a change in surface reflectivity, thereby affecting the quality and reliability of the deformation signal. Furthermore, due to the topography relief and the presence of different earth surface coverage types, the difference in angle of incidence can lead to phase errors in the deformation monitoring, affecting the accuracy of the final deformation interpretation. The less suitable choice of incidence angle in the prior art results in lower accuracy of the InSAR deformation monitoring result.
Disclosure of Invention
In order to solve the technical problem that the accuracy of the InSAR deformation monitoring result is low due to the fact that the incidence angle is selected improperly in the prior art, the invention aims to provide an InSAR earth surface deformation monitoring method and system, and the adopted technical scheme is as follows:
In a first aspect, the invention provides an InSAR earth surface deformation monitoring method, which comprises the following steps:
Collecting SAR images of a target surface area in a preset time range, wherein the SAR images comprise a first SAR image shot by the target surface area when an SAR satellite ascends and a second SAR image shot by the target surface area when the SAR satellite descends;
Screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain a credible pixel grid in the SAR image;
Fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area;
And inputting the fused SAR image into a ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of the target ground surface area.
Preferably, the surface deformation monitoring model comprises a deep learning model module, a feature conversion module and an elastic half-space model module which are connected in cascade;
The deep learning model module is used for extracting an image feature set from the fused SAR image, wherein the image feature set comprises surface texture information, surface phase information and image coherence loss information;
The feature conversion module is used for determining an estimated value of the surface roughness of the target surface area according to the surface texture information; determining an estimated value of the earth surface elevation of the target earth surface area according to the earth surface phase information; determining an SAR signal quality estimated value according to the image coherence loss information; determining a displacement variation corresponding to the target surface area according to the surface roughness estimated value, the surface elevation and depression estimated value and the SAR signal quality estimated value;
And the elastic half-space model module is used for determining the earth surface elastic deformation index of the displacement variation and judging whether the earth surface deformation risk exists in the target earth surface area according to the earth surface elastic deformation index.
Preferably, the deep learning model module adopts a multi-task CNN, wherein the multi-task CNN comprises a cascade input layer, a convolution layer, a pooling layer, a full connection layer and a multi-task output layer;
the input layer is used for receiving the fused SAR image;
The structure of the convolution layer is as follows:
;
wherein, Represents the/>Output feature map of layer convolution layer,/>And/>Respectively represent the/>Weights and offsets of layer convolution layers,/>Represents the/>Output of layer convolution layer or fusion SAR image,/>Representing convolution operations, reLU represents an activation function;
The structure of the pooling layer is as follows: ; wherein/> Representing a characteristic diagram after pooling operation, maxPool representing maximum pooling operation; flattening the feature map after pooling operation;
the structure of the full-connection layer is as follows: ; wherein/> Representing the output of the fully connected layer,/>And/>Respectively represent the weight and bias of the fully connected layer,/>To activate the function,/>Representing the flattened feature map;
the structure of the multi-task output layer is as follows:
;
;
;
wherein, For output surface texture information,/>For output earth surface phase information,/>Loss of information for image coherence; /(I)、/>And/>Respectively represent the activation functions of the corresponding tasks,/>、/>AndModel layer weights respectively representing corresponding tasks,/>、/>And/>Respectively representing model layer biases for corresponding tasks.
Preferably, the feature conversion module comprises a plurality of initial feature conversion layers and a comprehensive feature conversion layer; the input end of each initial feature conversion layer is connected to the output end of the multiplexing CNN, and the output end of each initial feature conversion layer is connected to the input end of the comprehensive feature conversion layer.
Preferably, the first initial feature conversion layer for the surface texture information has a structure of:
; wherein/> Representing an estimated surface roughness value; /(I)And/>Model layer weights and biases respectively representing a first initial feature transformation layer,/>Is an activation function;
the structure of the second initial feature conversion layer aiming at the earth surface phase information is as follows:
; wherein/> Representing the estimated value of the earth surface rise and fall; /(I)And/>Respectively representing model layer weights and biases of the second initial feature conversion layer;
The third initial feature conversion layer for the image coherence loss information has a structure as follows:
; wherein/> Representing SAR signal quality estimates; /(I)And/>Model layer weights and biases respectively representing a second initial feature transformation layer,/>Is an activation function;
the structure of the comprehensive characteristic conversion layer is as follows:
;
In the method, in the process of the invention, Representing the displacement variation; /(I)、/>And/>Respectively representing the corresponding relation weighing weights of the surface roughness, the surface elevation and the SAR signal qualityRepresenting preset adjustment parameters for SAR signal quality.
Preferably, the elastic half-space model module has the structure that:
;
wherein, Representing the elastic deformation index of the earth surface; /(I)Poisson's ratio for surface material; /(I)Young's modulus of the surface material,/>Representing a preset distance; /(I)Integrating information corresponding to the target surface area; /(I)Representing displacement variation caused by underground point source,/>Representing the change quantity of the water level of the reservoir/>Indicating the variation of the groundwater pressure; /(I)Representing a first subsurface stress coefficient of variation; /(I)And representing a second subsurface stress change coefficient, wherein the target surface area is a reservoir peripheral area.
Preferably, the filtering the pixel grid according to the included angle between the satellite line of sight of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain the trusted pixel grid in the SAR image includes:
For any SAR image, taking an included angle between the satellite line of sight of a ground unit of a pixel grid in the SAR image and the ground normal direction as a local incident angle corresponding to the pixel grid; and taking the pixel grid corresponding to the local incidence angle smaller than the preset angle threshold value as a trusted pixel grid in the SAR image.
Preferably, the determining whether the target surface area has a surface deformation risk according to the surface elastic deformation index includes:
Determining an average value of image acquisition moments corresponding to the first SAR image and the second SAR image as a target image acquisition moment; acquiring air humidity information and barometric pressure information corresponding to the target surface area at the target image acquisition moment;
calibrating the surface elastic deformation index based on the air humidity information and the atmospheric pressure information, and determining a target surface deformation score; and when the target surface deformation score is larger than a preset score threshold, the surface deformation risk exists in the target surface area.
Preferably, said calibrating said surface elastic deformation index based on said air humidity information and said barometric pressure information, determining a target surface deformation score, comprises:
Inputting the air humidity information and the atmospheric pressure information into a preset calibration model to obtain corresponding calibration parameters; the calibration model adopts a multi-layer perceptron; and correcting the surface elastic deformation index according to the calibration parameters to obtain a target surface deformation score.
In a second aspect, the present invention provides an InSAR earth surface deformation monitoring system, comprising:
The acquisition unit is used for acquiring SAR images of the target surface area in a preset time range, and comprises a first SAR image shot on the target surface area when the SAR satellite ascends and a second SAR image shot on the target surface area when the SAR satellite descends;
The trusted pixel grid screening unit is used for screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain the trusted pixel grid in the SAR image;
The image fusion unit is used for fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area;
the ground surface deformation risk identification unit is used for inputting the fused SAR image into the ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of the target ground surface area.
The embodiment of the invention has at least the following beneficial effects:
In a complex terrain area, as the local incidence angle is continuously increased, the SAR image coherence coefficient is continuously reduced, so that the SAR deformation signal quality is severely interfered. According to the invention, the surface deformation risk analysis is carried out by selecting the credible pixel grid area with the local incidence angle smaller than the angle threshold value, so that the negative influence of the local incidence angle change on the deformation signal quality can be effectively reduced, and the high-precision deformation interpretation result is realized. Because the proportion and the range of the trusted pixel grid area in the single SAR image relative to the whole target surface area are limited, the invention can effectively improve the range of the trusted pixel grid area by fusing the trusted pixel grid areas in the SAR image shot by the track lifting and the track lowering, and finally improve the accuracy of the deformation interpretation result. The reliable pixel grids in the lifting rail SAR data are screened and fused, so that the deformation monitoring accuracy and the data reliability of the InSAR technology under the complex terrain condition are effectively improved, potential geological disaster risks such as landslide, ground subsidence and the like can be identified early by timely and accurately monitoring the surface deformation, and the method has important application value in the fields of geological disaster prevention and environmental assessment.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of steps of an InSAR earth surface deformation monitoring method provided by an embodiment of the invention;
FIG. 2 is a schematic three-dimensional schematic of a localized angle of incidence provided by an embodiment of the present invention;
FIG. 3 is a schematic diagram of a principle of solving local incidence angles provided by an embodiment of the present invention;
FIG. 4 is a plot of the relationship between the local incidence angle of the derailment and the rate of change of the backscatter coefficient provided by an embodiment of the present invention;
FIG. 5 is a plot of the relationship between derailment CC change rate and LIA provided by an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a surface deformation monitoring model according to an embodiment of the present invention;
Fig. 7 is a schematic structural diagram of a feature conversion module according to an embodiment of the present invention;
FIG. 8 is a schematic flow chart of identifying a surface deformation risk of a target surface area according to a surface elastic deformation index according to an embodiment of the present invention;
FIG. 9 is a block diagram of an InSAR earth surface deformation monitoring system provided by an embodiment of the present invention;
Fig. 10 is a schematic diagram of a hardware structure of an electronic device for executing the method for monitoring the surface deformation of the InSAR according to the embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description is given below of a specific implementation, structure, characteristics and effects of the InSAR earth surface deformation monitoring method and system according to the invention in combination with the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The invention provides a specific scheme of an InSAR earth surface deformation monitoring method and system by combining a drawing.
An embodiment of an InSAR earth surface deformation monitoring method:
referring to fig. 1, a flowchart illustrating steps of an InSAR earth surface deformation monitoring method according to an embodiment of the present invention is shown, where the method includes the following steps:
Step S110, collecting SAR images of a target surface area in a preset time range, wherein the SAR images comprise a first SAR image shot on the target surface area when an SAR satellite ascends and a second SAR image shot on the target surface area when the SAR satellite descends.
Specifically, the first SAR image is a SAR image taken for a target surface area when the SAR satellite is in orbit, and the second SAR image is a SAR image taken for the target surface area when the SAR satellite is in orbit. The time difference between the first image acquisition time corresponding to the first SAR image and the second image acquisition time corresponding to the second SAR image is smaller than a preset time threshold, for example, the first SAR image and the second SAR image are acquired by corresponding orbit ascending satellites and orbit descending satellites respectively at the same time so as to ensure the timeliness matching of the orbit ascending SAR image and the orbit descending SAR image, and the surface deformation condition of the target surface area can be reflected more timely and objectively. Meanwhile, both the first SAR image and the second SAR image may be referred to as SAR images of the target surface area.
The track lifting shooting means that a radar satellite shoots on a track which is gradually raised on the track, a radar beam irradiates the ground at a certain angle, and the incidence angle of the radar beam and the ground is gradually increased along with the rise of the satellite. The down-orbit shooting is to shoot a radar satellite on an orbit in a gradually decreasing track, the radar beam irradiates the ground at a certain angle, and the incidence angle of the radar beam with the ground gradually decreases along with the decrease of the satellite.
And step S120, screening the pixel grids according to the included angle between the satellite line of sight of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain the credible pixel grid in the SAR image.
It should be noted that, the pixel grid of the SAR image refers to a ground area unit represented by each pixel in the SAR image, and the corresponding pixel grid amount can be measured by meters/pixel. Due to the variation of the angle of incidence, the scattering properties of the radar beam to the ground will also vary, resulting in different scattering patterns in the radar image.
Fig. 2 shows a schematic three-dimensional schematic of an example of a local angle of incidence, in which,Representing the local angle of incidence. The incidence angle of electromagnetic signals, the slope of the earth's surface and the slope direction directly affect the magnitude of backscattering at the incident surface, the local incidence angle/>Is a function of the target surface slope and the target surface direction relative to the radar illumination angle, which is the angle between the incidence direction of the radar electromagnetic wave and the local ground normal. A slope of a given material type facing radar illumination (small local angle of incidence) produces a stronger backscatter signal than if the same slope were offset from radar illumination (large local angle of incidence).
And for any SAR image, taking the included angle between the satellite line of sight of the ground unit of the pixel grid in the SAR image and the ground normal direction as the local incident angle corresponding to the pixel grid. And taking the pixel grid corresponding to the local incidence angle smaller than the preset angle threshold value as a trusted pixel grid in the SAR image.
Specifically, a first local incidence angle corresponding to each pixel grid in the first SAR image is calculated, a second local incidence angle corresponding to each pixel grid in the second SAR image is calculated, the pixel grid corresponding to the first local incidence angle smaller than a preset angle threshold value is determined to be a first credible pixel grid, and the pixel grid corresponding to the second local incidence angle smaller than the angle threshold value is determined to be a second credible pixel grid.
Fig. 3 is a schematic diagram for solving the principle of the local incident angle, in fig. 3, vs represents the satellite sight line, vn represents the surface normal, va represents the combined vector of the satellite sight line and the surface normal,Representing the surface gradient angle,/>Representing the angle between the normal lines of the horizontal plane of the target surface area.
The direction of incidence of electromagnetic waves is expressed by two parameters: an incident angle and an azimuth angle, wherein the incident angle represents an included angle between a satellite sight line and a horizontal plane normal line of a target area, and the azimuth angle represents a satellite orbit direction and is a rotation angle value in the horizontal plane in a clockwise direction with the north-positive direction being zero degrees.
The horizontal angle of the satellite sight line direction represents the shooting azimuth angle of the satellite, and is a rotation angle value in the horizontal plane in the clockwise direction with the north of the north being zero degrees. The side view mode of the satellite needs to be added for solving, namely, the right view is as the case may be: ; left view condition: /(I) ; Wherein/>Representing the horizontal angle of the satellite line of sight direction,/>Representing azimuth angle.
Taking the right case as an example, the three components of the satellite right view line direction are expressed as:
wherein, 、/>And/>Three components representing the direction of the right view line of the satellite,/>Representing the vector distance between the satellite line of sight and the plane of the target surface area,/>Is the horizontal angle of the satellite sight line direction.
The surface normal is also generally described by two parameters: the earth surface gradient angle and the earth surface azimuth angle, wherein the earth surface gradient angle represents an included angle between the ground and the horizontal plane, the earth surface azimuth angle represents projection of the earth surface slope direction on the horizontal plane, and the earth surface gradient angle is an angle value rotating clockwise with the north of north as zero degrees. The three components of the surface normal are expressed as:
wherein, 、/>And/>Three components of the surface normal representing the surface normal,/>Representing the vector distance between the surface normal and the target surface area plane,/>Is the surface gradient angle,/>Is the azimuth of the earth's surface.
Three-component representation of the resultant vector of the satellite line of sight and the surface normal:,,/> Wherein/> 、/>And/>Representing the resultant vector/>Is included in the three components of (a).
Then from satellite gaze direction vectorAnd surface normal vector/>And a resultant vector/>A triangle is formed, and the calculation formula of the local incidence angle can be obtained by the cosine law:
wherein, The ground slope angle is the ground slope angle and represents the included angle between the ground unit corresponding to the pixel grid and the horizontal plane. /(I)The azimuth angle of the earth surface is the projection of the slope of the earth surface on the horizontal plane. /(I)Representing the angle between the satellite line of sight direction and the normal to the horizontal plane.The horizontal angle of the satellite's line of sight represents the azimuth angle of the satellite.
In some embodiments, the SAR data used is Radarsat-2 derated mode programming data. The parameters are incident angle 39.83 degrees, azimuth angle 189.48 degrees, and the gradient and slope direction of the ground surface can be obtained by DEM (Digital Elevation Model ) of the target ground surface area through spatial analysis of raster data. And substituting the slope and slope data and the Radarsat-2 derating mode shooting parameters into a calculation formula of the local incidence angle, so that the local incidence angle distribution map can be solved.
Since the local incident angle has a fractional angle, the data distribution is very scattered and is not suitable for rule statistics, and thus, the local incident angle distribution map is reclassified according to a group of every 10 degrees in the present embodiment, and 10 angle classes can be classified in the region.
More specifically, the gradient is calculated by the DEMAnd slope/>The formula of (2) is:
In the method, in the process of the invention, Is an elevation value. Furthermore,/>And/>The spatial derivatives of elevation in the x-axis and y-axis, respectively.
The research finds that in the complex terrain area, the SAR image coherence coefficient is continuously reduced along with the continuous increase of the local incidence angle, so that the SAR deformation signal quality is severely interfered.
Specifically, as shown in fig. 4, a graph of a relationship between a local incidence angle and a backscattering coefficient change rate of a down-track is shown, wherein an abscissa of the graph is an angle of the local incidence angle, an ordinate of the graph is the backscattering coefficient change rate, a difference value between each classification unit of the backscattering coefficient in the down-track shooting mode is a Y-axis of a statistical graph, and a corresponding local incidence angle is an X-axis, so as to obtain the graph of the local incidence angle and the backscattering coefficient change rate. From the graph, the backscattering coefficient of the SAR image is closely related to the local incidence angle, and the data change rate shows an inflection point at about 40-50 degrees.
Referring to fig. 5, a graph of a relationship between the derailment CC (correlation coefficient) change rate and LIA (local incidence angle) is shown, in which the abscissa is the angle of the local incidence angle and the ordinate is the derailment correlation coefficient change rate, according to an embodiment of the present application. And drawing a relation diagram between the statistic value of the coherence coefficient of each group and the local incidence angle classification by taking the statistic value of the coherence coefficient value between each pair as a Y axis and the classification number of the local incidence angle classification as an X axis. As can be seen from the graph group, the regularity of the influence of the local incidence angle on the coherence coefficient between the SAR images is also obvious, and the coherence coefficient is consistent with the change of the local incidence angle in the six sets of SAR data interference image pairs shot in the derailment mode. With the continuous increase of the local incidence angle, the coherence coefficient is continuously reduced. And small fluctuations occur in the fourth group classification. From the graph, it can be seen that the backscattering coefficient of the SAR image is closely related to the local incidence angle. The data change rate shows a knee point trend at about 40-50 degrees.
Note that, since the case of the up-track SAR data is identical to the case of the down-track SAR data, the description thereof will not be repeated here.
From this, it can be seen that the local angle of incidence can directly affect the quality of SAR data. Along with the gradual increase of the local incidence angle, the quality of SAR data is gradually reduced, and the position of the inflection point of the curve is 40-50 degrees. A region with a local angle of incidence less than 50 deg. is referred to as a trusted region, if it is indicated that the data quality is more reliable.
The range of the angle threshold value is [40 degrees, 50 degrees ], so that the classification of the credible pixel grids and the unreliable pixel grids in the SAR image is realized. For example, when the angle threshold is 50 °, and when the local incident angle is smaller than 50 °, the reliability is higher, and the corresponding pixel grid is a reliable pixel grid. When the local incidence angle is larger than 50 degrees, the data quality is poor, the reliability of the result is low, and the partial area is called an untrusted area.
Reclassifying the local incident angle profile of the derailment mode with a threshold of 50 deg., and treating the local incident angle region less than 50 deg. as a trusted region.
According to statistics of the grid quantity, the grid quantity of the reliable area with better data quality is 125846, and the grid quantity of the unreliable area with poorer data quality is 72985, so that the reliable area range occupies 63.3% of the area of the research area. Therefore, the reliable region with strong backscattering coefficient and coherence coefficient in SAR data shot in the derailment mode does not cover the whole region.
And step S130, fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area.
By fusing the reliable pixel grids in the ascending SAR data and the descending SAR data, the reliable pixel grid region distribution with the largest range in the fused SAR image can be realized.
Because of the problem of reliability caused by data quality, the monitoring of the whole area is difficult to complete by only using one shooting mode (such as a track-down mode), so that in the embodiment, the image supplement analysis is further fused on the reliable area of SAR data in the track-up mode.
Under the condition that the track lifting mode and the track descending mode are shot simultaneously, the grid quantity of the trusted area is 194965, and the trusted area accounts for 98.1% of the total area of the research area. Therefore, the combined monitoring of the track lifting mode and the track descending mode can effectively increase the effective monitoring range of the monitoring area, and the trusted area is increased by 98.1 percent, and the untrusted areas are distributed in tiny and sporadic manner or are in long and thin strips, so that the trusted areas are not consistent with the characteristics of landslide, and the possibility of landslide distribution in the untrusted areas can be eliminated. Thereby further enabling the certainty of monitoring to reach full area coverage.
By the embodiment of the application, the backscattering coefficient of the SAR image and the coherence coefficient between the image pairs are closely related to the included angle between the sight direction of the SAR satellite and the surface normal, namely the local incident angle. As the included angle increases, the values of the backscatter coefficient and the coherence coefficient between the pair decrease, indicating that the quality of the SAR data also decreases. And obtaining the critical angle of the local incident angle of the trusted region to be 40-50 degrees according to the relation graph of the backscattering coefficient and the coherence coefficient between the image pairs and beta. Where the SAR data would be unreliable in areas where the local angle of incidence is greater than this angle, SAR monitoring with a single shooting mode for the same area would not cover the full area. Furthermore, the local incidence angle distribution of the Radarsat-2 track lifting condition is manufactured by the same method, and the near-full-coverage wide-area ground surface deformation monitoring, such as landslide monitoring, can be achieved by adopting a track lifting and track lowering combined monitoring method.
And step S140, inputting the fused SAR image into a ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of the target ground surface area.
In some embodiments, the surface deformation monitoring model may employ various known or potential artificial intelligence models, such as those of the prior art that directly multiplex to predict surface deformation risk. Preferably, a mixed model of a fused deep learning model module and a physical model module is adopted, so that a higher-precision ground surface deformation monitoring result can be realized.
Specifically, as shown in fig. 6, the surface deformation monitoring model 1000 includes a deep learning model module 1010, a feature conversion module 1020 and an elastic half-space model module 1030, which are cascaded. Therefore, the surface deformation monitoring model 1000 adopts a mixed model design, so that not only can the complexity of InSAR data be effectively processed, but also the interpretability and generalization capability of the model can be improved through a physical model.
The deep learning model module 1010 is configured to extract an image feature set from the fused SAR image, where the image feature set includes surface texture information, surface phase information, and image coherence loss information.
The deep learning model module 1010 may employ a deep learning model of various structural types suitable for processing classification tasks of image input data, such as CNN, transfer learning model, and the like. In some embodiments, for different types of image features, corresponding models may be employed for processing, respectively.
In this embodiment, the deep learning model module 1010 employs a multi-tasking CNN that includes cascaded input layers, convolution layers, pooling layers, full connection layers, and multi-tasking output layers.
Specifically, the input layer is configured to receive the fused SAR image;
The structure of the convolution layer is as follows:
wherein, Represents the/>Output feature map of layer convolution layer,/>And/>Respectively represent the/>Weights and offsets of layer convolution layers,/>Represents the/>Output of layer convolution layer or fusion SAR image,/>Representing a convolution operation, reLU represents an activation function.
The structure of the pooling layer is as follows:
wherein, Representing a characteristic diagram after pooling operation, maxPool representing maximum pooling operation; flattening the feature map after pooling operation: /(I); Wherein/>And representing the flattened characteristic diagram.
Through the above operation, although the shape of the feature map is changed, the element values and their relative order in the feature map are preserved, which ensures that important features obtained through pooling do not lose information when transferred to the fully connected layer. Furthermore, the characteristics extracted by the convolution layer and the pooling layer can be used for the traditional fully-connected neural network to perform classification, regression or other types of tasks through flattening operation.
The structure of the full-connection layer is as follows:
wherein, Representing the output of the fully connected layer,/>And/>Respectively represent the weight and bias of the fully connected layer,/>To activate the function,/>And representing the flattened characteristic diagram.
The structure of the multi-task output layer is as follows:
wherein, For output surface texture information,/>For output earth surface phase information,/>Loss of information for image coherence; /(I)、/>And/>Respectively represent the activation functions of the corresponding tasks,/>、/>AndModel layer weights respectively representing corresponding tasks,/>、/>And/>Respectively representing model layer biases for corresponding tasks.
Based on a multi-task learning framework in the CNN model, the model can learn and predict various types of surface information simultaneously. Then, through different output layers, the model can respectively and effectively predict and analyze the surface deformation, texture characteristics and coherence loss. Therefore, the utilization efficiency and comprehensive prediction capability of the model on InSAR data are improved.
A feature conversion module 1020 for determining an estimated surface roughness value of the target surface area according to the surface texture information; determining an estimated value of the earth surface elevation of the target earth surface area according to the earth surface phase information; determining an SAR signal quality estimated value according to the image coherence loss information; and determining the displacement variation corresponding to the target surface area according to the surface roughness estimated value, the surface elevation and depression estimated value and the SAR signal quality estimated value.
In some embodiments, for the input parameters and output parameters of the feature transformation module 1020, corresponding mapping functions may be specifically designed to transform the output feature map of the deep learning model module 1010 into the input parameters required by the physical model, so as to ensure compatibility between the deep learning model and the physical model. In particular, these mapping functions may be optimized through additional training processes to ensure that the transformed parameters maximally preserve the feature information extracted by the deep learning model module and can be effectively used by the physical model.
As shown in fig. 7, the feature conversion module 1100 includes a plurality of initial feature conversion layers including a first initial feature conversion layer 1101, a second initial feature conversion layer 1102, and a third initial feature conversion layer 1103, and a comprehensive feature conversion layer 1120. The input of each initial feature conversion layer is connected to the output of the multitasking CNN and the output of each initial feature conversion layer is connected to the input of the comprehensive feature conversion layer 1120.
The structure of the first initial feature conversion layer for the surface texture information is as follows:
wherein, Representing an estimated surface roughness value; /(I)And/>Model layer weights and biases respectively representing a first initial feature transformation layer,/>Is an activation function; thus, the texture features extracted by the CNN are adjusted through weighting and biasing, and then the roughness estimated value in a reasonable range is output through the tanh activation function, so that the output value can be limited in a certain range, and the method can be preferably applied to the limited amount of roughness.
The structure of the second initial feature conversion layer aiming at the earth surface phase information is as follows:
wherein, Representing the estimated value of the earth surface rise and fall; /(I)And/>Respectively representing model layer weights and biases of the second initial feature conversion layer; it should be noted that, because the earth surface elevation may cover a larger range, the earth surface elevation can be estimated by directly performing linear transformation on the phase information extracted by the CNN by adopting a more suitable linear transformation function.
The third initial feature conversion layer for the image coherence loss information has a structure as follows:
wherein, Representing SAR signal quality estimates; /(I)And/>Model layer weights and biases respectively representing a second initial feature transformation layer,/>To activate the function, the activation function is used to limit the output between 0 and 1, and the coherence loss information extracted by the CNN is converted into a probability value to intuitively reflect the quality condition of the SAR signal.
The structure of the comprehensive characteristic conversion layer is as follows:
In the method, in the process of the invention, Representing the displacement variation; /(I)、/>And/>Respectively representing the corresponding relation weighing weights of the surface roughness, the surface elevation and the SAR signal qualityRepresenting preset adjustment parameters for SAR signal quality. Is a hyperbolic tangent function for introducing nonlinearities and mapping signal quality estimates into a bounded range.
The elastic half-space model module 1030 is configured to determine an elastic deformation index of the earth surface of the displacement variation, and determine whether the earth surface deformation risk exists in the target earth surface area according to the elastic deformation index of the earth surface.
It should be noted that the elastic half-space model is a mathematical physical model for simulating deformation of the earth's surface, and it is assumed that the earth's surface is a uniform, isotropic, linear elastic half-space, and that the earth's surface is elastically deformed when subjected to external forces.
In one example, the elastic half-space model module may employ various general elastic half-space models. In another example of an embodiment of the present application, an elastic half-space model module is enhanced in combination with an actual reservoir geological monitoring scenario.
Specifically, the target surface area is a reservoir peripheral area, and the elastic half-space model module has the structure that:
wherein, Representing the elastic deformation index of the earth surface; /(I)Defining the proportional relation of transverse strain and longitudinal strain of the surface material for the Poisson's ratio of the surface material; /(I)The Young's modulus of the surface material is quantized, and the deformation resistance of the material is quantized; /(I)Representing a predetermined distance, such as a distance between a reflection point of a satellite radar beam and a predetermined subsurface point source (e.g., a calibrated seismic fault); Integrating information corresponding to the target surface area; /(I) Representing displacement variation caused by underground point source,/>Representing the change quantity of the water level of the reservoir/>Indicating the variation of the groundwater pressure; /(I)Representing a first subsurface stress coefficient of variation, the first subsurface stress coefficient of variation being determined from reservoir level variation information; /(I)And representing a second subsurface stress coefficient of variation, the second subsurface stress coefficient of variation being determined from the groundwater pressure change information.
In the embodiment of the application, the contribution of the reservoir water level change and the underground water pressure change to the surface deformation is integrated respectively, so that the geological and hydrological conditions of the surrounding area of the reservoir are more comprehensively considered, the surface deformation is predicted more accurately and reliably, and the potential geological disaster risks such as landslide or ground subsidence can be found in time.
In one example, if the surface elastic deformation indexToo large, it may be determined that there is a corresponding potential risk of geological disasters. In another example of an embodiment of the present application, the elastic semi-space model module is utilized to output results (i.e. the elastic deformation index/>) Meanwhile, other factors can be comprehensively considered, so that accuracy of recognition results aiming at geological disaster risks is further improved.
The atmospheric pressure refers to the pressure of air to the ground, the atmospheric pressure affects the refractive index of the radar beam, the radar beam is affected by the atmospheric pressure in the propagation process to be refracted, and the higher the atmospheric pressure is, the higher the refractive index of the radar beam is, and even the propagation path of the radar beam is bent. In addition, air humidity refers to the content of water vapor in the air, and water vapor is a strong absorption medium that absorbs energy in the radar beam, resulting in a weakening of the radar signal and occurrence of phase errors that affect the geometric accuracy of the SAR image and the accuracy of the interferometry results.
Fig. 8 is a schematic flow chart of identifying the surface deformation risk of the target surface area according to the surface elastic deformation index.
In step S1210, an average value of the image acquisition moments corresponding to the first SAR image and the second SAR image is determined as the target image acquisition moment.
Step S1220, acquiring air humidity information and barometric pressure information corresponding to the target surface area at the target image acquisition time.
Specifically, the acquisition log of the meteorological sensor is queried to determine air humidity information and barometric pressure information corresponding to the target surface area at the target image acquisition time.
Step S1230, calibrating the surface elastic deformation index based on the air humidity information and the atmospheric pressure information, and determining a target surface deformation score.
And inputting the air humidity information and the atmospheric pressure information into a preset calibration model to determine corresponding calibration parameters by the calibration model, wherein the calibration model adopts a multi-layer perceptron. And correcting the surface elastic deformation index according to the calibration parameters to determine a target surface deformation score.
Specifically, the calibration model includes an input layer, a feature processing layer, an intermediate layer, and an output layer. The input variables corresponding to the input layer include air humidityAnd atmospheric pressure/>Index of elastic deformation of the surface ]. Normalized pretreatment of humidity and pressure data based on feature processing layer,/>,/>Wherein/>Representing a normalization function,/>And/>Is normalized humidity and pressure.
The intermediate layer then uses one or more hidden layers, containing neurons, for capturing complex nonlinear relationships between humidity and pressure data and deformation index:
wherein, Is/>Output of the hidden layer,/>And/>Representing intermediate layer weights and biases, initial layer/>By/>The composition is formed.
Further, the output layer outputs a calibration parameterThis parameter will be used to adjust the original deformation index:
wherein, Representing calibration parameters,/>And/>Is the weight and bias of the output layer,/>May be a sigmoid function.
Finally, the elastic deformation index of the earth surface is measured by using the calibration parametersAnd (3) performing calibration:
wherein, Representing the calibration target surface deformation score.
Step S1240, when the target surface deformation score is greater than a preset score threshold, the target surface area has a surface deformation risk.
According to the embodiment of the invention, the atmospheric condition data of the target surface area to be monitored are comprehensively considered and are incorporated into deformation analysis, a multi-layer perceptron is used for capturing complex relations among humidity, pressure and surface deformation, then the deformation index is dynamically adjusted according to real-time environment data, self-adaptive adjustment of synchronous atmospheric conditions is realized, and the reliability and accuracy of the surface deformation risk assessment result can be further improved.
It should be noted that, for simplicity of description, the foregoing method embodiments are all illustrated as a series of acts combined, but it should be understood and appreciated by those skilled in the art that the present application is not limited by the order of acts, as some steps may be performed in other orders or concurrently in accordance with the present application. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily required for the present application. In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to related descriptions of other embodiments.
The InSAR earth surface deformation monitoring system provided by the application is described below, and the InSAR earth surface deformation monitoring system described below and the InSAR earth surface deformation monitoring method described above can be correspondingly referred to each other.
Fig. 9 shows a block diagram of an InSAR earth surface deformation monitoring system. As shown in fig. 9, the InSAR surface deformation monitoring system 1300 includes an acquisition unit 1310, a trusted pixel grid screening unit 1320, an image fusion unit 1330, and a surface deformation risk identification unit 1340.
The acquiring unit 1310 is configured to acquire, in a preset time range, a SAR image of a target surface area, where the SAR image includes a first SAR image captured by the target surface area during an ascending of the SAR satellite and a second SAR image captured by the target surface area during a descending of the SAR satellite.
The trusted pixel grid screening unit 1320 is configured to screen the pixel grid according to an included angle between a satellite line of sight of a ground unit of the pixel grid in the SAR image and a ground normal direction, so as to obtain the trusted pixel grid in the SAR image.
And the image fusion unit 1330 is configured to fuse the trusted pixel grids in the first SAR image and the second SAR image, so as to obtain a fused SAR image of the target surface area.
The surface deformation risk identification unit 1340 is used for inputting the fused SAR image into a surface deformation monitoring model to obtain a surface deformation monitoring result of the target surface area.
In some embodiments, embodiments of the present application provide a non-transitory computer readable storage medium having stored therein one or more programs including execution instructions that can be read and executed by an electronic device (including, but not limited to, a computer, a server, or a network device, etc.) for performing the above-described InSAR earth surface deformation monitoring method of the present application.
In some embodiments, embodiments of the present application also provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, cause the computer to perform the above-described InSAR earth surface deformation monitoring method.
In some embodiments, the present application further provides an electronic device, including: the system comprises at least one processor and a memory communicatively connected with the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform an InSAR earth surface deformation monitoring method.
Fig. 10 is a schematic hardware structure of an electronic device for performing an InSAR surface deformation monitoring method according to another embodiment of the present application, as shown in fig. 10, where the device includes:
One or more processors 1410, and memory 1420, one processor 1410 being illustrated in fig. 10.
The apparatus for performing the InSAR surface deformation monitoring method may further include: an input device 1430 and an output device 1440.
Processor 1410, memory 1420, input device 1430, and output device 1440 may be connected by a bus or other means, for example in fig. 10.
The memory 1420 is used as a non-volatile computer readable storage medium for storing non-volatile software programs, non-volatile computer executable programs, and modules, such as program instructions/modules corresponding to the InSAR surface deformation monitoring method in the embodiments of the present application. The processor 1410 executes various functional applications of the server and data processing, i.e., implements the InSAR earth surface deformation monitoring method of the above-described method embodiments, by running non-volatile software programs, instructions and modules stored in the memory 1420.
Memory 1420 may include a storage program area that may store an operating system, at least one application program required for functionality, and a storage data area; the storage data area may store data created according to the use of the electronic device, etc. In addition, memory 1420 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, memory 1420 may optionally include memory located remotely from processor 1410, which may be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 1430 may receive input numeric or character information and generate signals related to user settings and function control of the electronic device. The output device 1440 may include a display device such as a display screen.
The one or more modules are stored in the memory 1420, which when executed by the one or more processors 1410, perform the InSAR earth surface deformation monitoring method in any of the method embodiments described above.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the application and are intended to be included within the scope of the application.
Claims (4)
1. The InSAR earth surface deformation monitoring method is characterized by comprising the following steps of:
Collecting SAR images of a target surface area in a preset time range, wherein the SAR images comprise a first SAR image shot by the target surface area when an SAR satellite ascends and a second SAR image shot by the target surface area when the SAR satellite descends;
Screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain a credible pixel grid in the SAR image;
Fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area;
Inputting the fused SAR image into a ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of a target ground surface area;
The earth surface deformation monitoring model comprises a deep learning model module, a feature conversion module and an elastic half-space model module which are connected in cascade;
The deep learning model module is used for extracting an image feature set from the fused SAR image, wherein the image feature set comprises surface texture information, surface phase information and image coherence loss information;
The feature conversion module is used for determining an estimated value of the surface roughness of the target surface area according to the surface texture information; determining an estimated value of the earth surface elevation of the target earth surface area according to the earth surface phase information; determining an SAR signal quality estimated value according to the image coherence loss information; determining a displacement variation corresponding to the target surface area according to the surface roughness estimated value, the surface elevation and depression estimated value and the SAR signal quality estimated value;
The elastic half-space model module is used for determining the earth surface elastic deformation index of the displacement variation and judging whether the earth surface deformation risk exists in the target earth surface area according to the earth surface elastic deformation index;
the deep learning model module adopts a multi-task CNN, wherein the multi-task CNN comprises a cascade input layer, a convolution layer, a pooling layer, a full connection layer and a multi-task output layer;
the input layer is used for receiving the fused SAR image;
The structure of the convolution layer is as follows:
;
wherein, Represents the/>Output feature map of layer convolution layer,/>And/>Respectively represent the/>Weights and offsets of layer convolution layers,/>Represents the/>Output of layer convolution layer or fusion SAR image,/>Representing convolution operations, reLU represents an activation function;
The structure of the pooling layer is as follows: ; wherein/> Representing a characteristic diagram after pooling operation, maxPool representing maximum pooling operation; flattening the feature map after pooling operation;
the structure of the full-connection layer is as follows: wherein/> Representing the output of the fully connected layer,/>And/>Respectively represent the weight and bias of the fully connected layer,/>To activate the function,/>Representing the flattened feature map;
the structure of the multi-task output layer is as follows:
;
;
;
wherein, For output surface texture information,/>For output earth surface phase information,/>Loss of information for image coherence; /(I)、/>And/>Respectively represent the activation functions of the corresponding tasks,/>、/>And/>Model layer weights respectively representing corresponding tasks,/>、/>And/>Respectively representing model layer bias of corresponding tasks;
the feature conversion module comprises a plurality of initial feature conversion layers and a comprehensive feature conversion layer; the input end of each initial feature conversion layer is connected to the output end of the multi-task CNN, and the output end of each initial feature conversion layer is connected to the input end of the comprehensive feature conversion layer;
the structure of the first initial feature conversion layer for the surface texture information is as follows:
; wherein/> Representing an estimated surface roughness value; /(I)And/>Model layer weights and biases respectively representing a first initial feature transformation layer,/>Is an activation function;
the structure of the second initial feature conversion layer aiming at the earth surface phase information is as follows:
; wherein/> Representing the estimated value of the earth surface rise and fall; /(I)And/>Respectively representing model layer weights and biases of the second initial feature conversion layer;
The third initial feature conversion layer for the image coherence loss information has a structure as follows:
; wherein/> Representing SAR signal quality estimates; /(I)And/>Model layer weights and biases respectively representing a second initial feature transformation layer,/>Is an activation function;
the structure of the comprehensive characteristic conversion layer is as follows:
;
In the method, in the process of the invention, Representing the displacement variation; /(I)、/>And/>Respectively representing the corresponding relation weighing weights of the surface roughness, the surface elevation and the SAR signal qualityRepresenting preset adjustment parameters for SAR signal quality;
the elastic half-space model module has the structure that:
;
wherein, Representing the elastic deformation index of the earth surface; /(I)Poisson's ratio for surface material; /(I)Young's modulus of the surface material,/>Representing a preset distance; /(I)Integrating information corresponding to the target surface area; /(I)Representing displacement variation caused by underground point source,/>Representing the change quantity of the water level of the reservoir/>Indicating the variation of the groundwater pressure; /(I)Representing a first subsurface stress coefficient of variation; /(I)Representing a second subsurface stress variation coefficient, wherein the target surface area is a reservoir peripheral area;
the method for screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain the credible pixel grid in the SAR image comprises the following steps:
For any SAR image, taking an included angle between the satellite line of sight of a ground unit of a pixel grid in the SAR image and the ground normal direction as a local incident angle corresponding to the pixel grid; and taking the pixel grid corresponding to the local incidence angle smaller than the preset angle threshold value as a trusted pixel grid in the SAR image.
2. The method for monitoring the surface deformation of the InSAR according to claim 1, wherein the determining whether the target surface area has the surface deformation risk according to the surface elastic deformation index comprises:
Determining an average value of image acquisition moments corresponding to the first SAR image and the second SAR image as a target image acquisition moment; acquiring air humidity information and barometric pressure information corresponding to the target surface area at the target image acquisition moment;
calibrating the surface elastic deformation index based on the air humidity information and the atmospheric pressure information, and determining a target surface deformation score; and when the target surface deformation score is larger than a preset score threshold, the surface deformation risk exists in the target surface area.
3. The InSAR ground deformation monitoring method of claim 2, wherein the calibrating the ground elastic deformation index based on the air humidity information and the barometric pressure information, determining a target ground deformation score, comprises:
Inputting the air humidity information and the atmospheric pressure information into a preset calibration model to obtain corresponding calibration parameters; the calibration model adopts a multi-layer perceptron; and correcting the surface elastic deformation index according to the calibration parameters to obtain a target surface deformation score.
4. An InSAR earth's surface deformation monitoring system, comprising:
The acquisition unit is used for acquiring SAR images of the target surface area in a preset time range, and comprises a first SAR image shot on the target surface area when the SAR satellite ascends and a second SAR image shot on the target surface area when the SAR satellite descends;
the trusted pixel grid screening unit is used for screening the pixel grids according to the included angle between the satellite sight line of the ground unit of the pixel grid in the SAR image and the ground normal direction to obtain the trusted pixel grid in the SAR image, and comprises the following steps:
for any SAR image, taking an included angle between the satellite line of sight of a ground unit of a pixel grid in the SAR image and the ground normal direction as a local incident angle corresponding to the pixel grid; taking a pixel grid corresponding to a local incidence angle smaller than a preset angle threshold as a trusted pixel grid in the SAR image;
The image fusion unit is used for fusing the trusted pixel grids in the first SAR image and the second SAR image to obtain a fused SAR image of the target surface area;
the ground surface deformation risk identification unit is used for inputting the fused SAR image into a ground surface deformation monitoring model to obtain a ground surface deformation monitoring result of the target ground surface area;
The earth surface deformation monitoring model comprises a deep learning model module, a feature conversion module and an elastic half-space model module which are connected in cascade;
The deep learning model module is used for extracting an image feature set from the fused SAR image, wherein the image feature set comprises surface texture information, surface phase information and image coherence loss information;
The feature conversion module is used for determining an estimated value of the surface roughness of the target surface area according to the surface texture information; determining an estimated value of the earth surface elevation of the target earth surface area according to the earth surface phase information; determining an SAR signal quality estimated value according to the image coherence loss information; determining a displacement variation corresponding to the target surface area according to the surface roughness estimated value, the surface elevation and depression estimated value and the SAR signal quality estimated value;
The elastic half-space model module is used for determining the earth surface elastic deformation index of the displacement variation and judging whether the earth surface deformation risk exists in the target earth surface area according to the earth surface elastic deformation index;
the deep learning model module adopts a multi-task CNN, wherein the multi-task CNN comprises a cascade input layer, a convolution layer, a pooling layer, a full connection layer and a multi-task output layer;
the input layer is used for receiving the fused SAR image;
The structure of the convolution layer is as follows:
;
wherein, Represents the/>Output feature map of layer convolution layer,/>And/>Respectively represent the/>Weights and offsets of layer convolution layers,/>Represents the/>Output of layer convolution layer or fusion SAR image,/>Representing convolution operations, reLU represents an activation function;
The structure of the pooling layer is as follows: ; wherein/> Representing a characteristic diagram after pooling operation, maxPool representing maximum pooling operation; flattening the feature map after pooling operation;
the structure of the full-connection layer is as follows: wherein/> Representing the output of the fully connected layer,/>And/>Respectively represent the weight and bias of the fully connected layer,/>To activate the function,/>Representing the flattened feature map;
the structure of the multi-task output layer is as follows:
;
;
;
wherein, For output surface texture information,/>For output earth surface phase information,/>Loss of information for image coherence; /(I)、/>And/>Respectively represent the activation functions of the corresponding tasks,/>、/>And/>Model layer weights respectively representing corresponding tasks,/>、/>And/>Respectively representing model layer bias of corresponding tasks;
the feature conversion module comprises a plurality of initial feature conversion layers and a comprehensive feature conversion layer; the input end of each initial feature conversion layer is connected to the output end of the multi-task CNN, and the output end of each initial feature conversion layer is connected to the input end of the comprehensive feature conversion layer;
the structure of the first initial feature conversion layer for the surface texture information is as follows:
; wherein/> Representing an estimated surface roughness value; /(I)And/>Model layer weights and biases respectively representing a first initial feature transformation layer,/>Is an activation function;
the structure of the second initial feature conversion layer aiming at the earth surface phase information is as follows:
; wherein/> Representing the estimated value of the earth surface rise and fall; /(I)And/>Respectively representing model layer weights and biases of the second initial feature conversion layer;
The third initial feature conversion layer for the image coherence loss information has a structure as follows:
; wherein/> Representing SAR signal quality estimates; /(I)And/>Model layer weights and biases respectively representing a second initial feature transformation layer,/>Is an activation function;
the structure of the comprehensive characteristic conversion layer is as follows:
;
In the method, in the process of the invention, Representing the displacement variation; /(I)、/>And/>Respectively representing the corresponding relation weighing weights of the surface roughness, the surface elevation and the SAR signal qualityRepresenting preset adjustment parameters for SAR signal quality;
the elastic half-space model module has the structure that:
;
wherein, Representing the elastic deformation index of the earth surface; /(I)Poisson's ratio for surface material; /(I)Young's modulus of the surface material,/>Representing a preset distance; /(I)Integrating information corresponding to the target surface area; /(I)Representing displacement variation caused by underground point source,/>Representing the change quantity of the water level of the reservoir/>Indicating the variation of the groundwater pressure; /(I)Representing a first subsurface stress coefficient of variation; /(I)And representing a second subsurface stress change coefficient, wherein the target surface area is a reservoir peripheral area.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410395059.9A CN117970331B (en) | 2024-04-02 | 2024-04-02 | InSAR earth surface deformation monitoring method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410395059.9A CN117970331B (en) | 2024-04-02 | 2024-04-02 | InSAR earth surface deformation monitoring method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117970331A CN117970331A (en) | 2024-05-03 |
CN117970331B true CN117970331B (en) | 2024-05-31 |
Family
ID=90861470
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410395059.9A Active CN117970331B (en) | 2024-04-02 | 2024-04-02 | InSAR earth surface deformation monitoring method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117970331B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106066478A (en) * | 2016-05-27 | 2016-11-02 | 中国矿业大学 | Merge pixel skew to follow the tracks of and the mining area surface deformation calculation method of short baseline set |
CN106526590A (en) * | 2016-11-04 | 2017-03-22 | 山东科技大学 | Method for monitoring and resolving three-dimensional ground surface deformation of industrial and mining area by means of multi-source SAR image |
CN113534154A (en) * | 2021-09-16 | 2021-10-22 | 成都理工大学 | SAR sight line direction deformation and slope direction sensitivity calculation method |
CN115951354A (en) * | 2023-02-14 | 2023-04-11 | 中国铁道科学研究院集团有限公司铁道建筑研究所 | D-InSAR deformation monitoring method fusing lifting rail |
CN116973917A (en) * | 2023-06-27 | 2023-10-31 | 中国四维测绘技术有限公司 | Slope hidden danger identification method based on multi-source information fusion |
CN117008128A (en) * | 2023-06-12 | 2023-11-07 | 东华理工大学 | Creep type landslide time sequence InSAR technology multidimensional deformation monitoring and predicting method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20230064454A1 (en) * | 2021-08-31 | 2023-03-02 | Satsure Analytics India Private Limited | System and method for generating soil moisture data from satellite imagery using deep learning model |
-
2024
- 2024-04-02 CN CN202410395059.9A patent/CN117970331B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106066478A (en) * | 2016-05-27 | 2016-11-02 | 中国矿业大学 | Merge pixel skew to follow the tracks of and the mining area surface deformation calculation method of short baseline set |
CN106526590A (en) * | 2016-11-04 | 2017-03-22 | 山东科技大学 | Method for monitoring and resolving three-dimensional ground surface deformation of industrial and mining area by means of multi-source SAR image |
CN113534154A (en) * | 2021-09-16 | 2021-10-22 | 成都理工大学 | SAR sight line direction deformation and slope direction sensitivity calculation method |
CN115951354A (en) * | 2023-02-14 | 2023-04-11 | 中国铁道科学研究院集团有限公司铁道建筑研究所 | D-InSAR deformation monitoring method fusing lifting rail |
CN117008128A (en) * | 2023-06-12 | 2023-11-07 | 东华理工大学 | Creep type landslide time sequence InSAR technology multidimensional deformation monitoring and predicting method |
CN116973917A (en) * | 2023-06-27 | 2023-10-31 | 中国四维测绘技术有限公司 | Slope hidden danger identification method based on multi-source information fusion |
Non-Patent Citations (3)
Title |
---|
Deep Learning-Based Landslide Recognition Incorporating Deformation Characteristics;Li, Zhihai et al.;《Remote Sensing》;20240312;全文 * |
基于SBAS-InSAR技术对昆明机场高边坡沉降监测研究;麻源源;左小清;麻卫峰;凌涛;张子文;;工程勘察;20180628(第07期);全文 * |
基于时序InSAR的辽河油田地表形变监测及油层参数多模型反演;于冰 等;《大地测量与地球动力学》;20240219;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN117970331A (en) | 2024-05-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103456018A (en) | Remote sensing image change detection method based on fusion and PCA kernel fuzzy clustering | |
CN105247461A (en) | Determining pitch and yaw for touchscreen interactions | |
CN112884342A (en) | Water color satellite atmospheric layer top radiation product quality evaluation and cross calibration method | |
CN115730247A (en) | Multi-dimensional space-time-frequency domain characteristic parameter fusion seabed sediment classification method | |
Lucchese et al. | Attribute selection using correlations and principal components for artificial neural networks employment for landslide susceptibility assessment | |
CN105205484A (en) | Synthetic aperture radar target detection method based on curvelet transformation and Wiener filtering | |
Ndehedehe et al. | Hydrological hotspots of climatic influence in Brazil: a two-step regularization approach | |
CN116295285A (en) | Shallow sea water depth remote sensing inversion method based on region self-adaption | |
Chew et al. | Flooding and inundation maps using interpolated CYGNSS reflectivity observations | |
Saponaro et al. | Evaluation of aerosol and cloud properties in three climate models using MODIS observations and its corresponding COSP simulator, as well as their application in aerosol–cloud interactions | |
Fayad et al. | Hy-TeC: a hybrid vision transformer model for high-resolution and large-scale mapping of canopy height | |
Brune et al. | The local wavelet‐based organization index–Quantification, localization and classification of convective organization from radar and satellite data | |
Xing et al. | A land cover change detection method combing spectral values and class probabilities | |
Marzban et al. | Verification with variograms | |
CN117970331B (en) | InSAR earth surface deformation monitoring method and system | |
Teegavarapu | Mean areal precipitation estimation: methods and issues | |
CN108629382A (en) | Ship seakeeping system in a kind of gunz optimization SAR Radar Seas | |
Ho et al. | Fuzzy-based spatial modeling approach to predict island karst distribution: a conceptual model | |
CN111950198B (en) | Ground clutter simulation method based on neural network | |
Hu et al. | Grain yield predict based on GRA-AdaBoost-SVR model | |
Song et al. | Instrument identification technology based on deep learning | |
Mahesh et al. | Identifying atmospheric rivers and their poleward latent heat transport with generalizable neural networks: ARCNNv1 | |
Liu et al. | A generic composite measure of similarity between geospatial variables | |
Yang et al. | Improving tropical cyclone precipitation forecasting with deep learning and satellite image sequencing | |
CN116821694B (en) | Soil humidity inversion method based on multi-branch neural network and segmented model |
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 |