CN114252879A - InSAR inversion and multi-influence factor based large-range landslide deformation prediction method - Google Patents
InSAR inversion and multi-influence factor based large-range landslide deformation prediction method Download PDFInfo
- Publication number
- CN114252879A CN114252879A CN202111567548.0A CN202111567548A CN114252879A CN 114252879 A CN114252879 A CN 114252879A CN 202111567548 A CN202111567548 A CN 202111567548A CN 114252879 A CN114252879 A CN 114252879A
- Authority
- CN
- China
- Prior art keywords
- deformation
- data
- sequence
- item
- category
- 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.)
- Pending
Links
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
- 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
- G01S13/904—SAR modes
- G01S13/9092—SAR modes combined with monopulse 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
- 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
- G01S13/885—Radar or analogous systems specially adapted for specific applications for ground probing
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a large-range landslide deformation prediction method based on InSAR inversion and multiple influence factors, which comprises the following steps of: performing InSAR inversion processing on SAR image data to obtain time sequence deformation data; clustering the time sequence deformation data to obtain a plurality of categories of time sequence deformation data; decomposing the time sequence deformation data of each category into a period item deformation sequence and a trend item deformation sequence; determining an influence factor which is obviously related to the deformation of the period item; respectively establishing an LSTM model of each category of time sequence deformation data to realize prediction of each category of deformation; and adding the cycle item deformation prediction value and the trend item deformation prediction value of each category to obtain deformation prediction results of each category, and combining the deformation prediction results of each category to obtain a large-range landslide deformation prediction result. The method can effectively predict the large-range landslide deformation, and solves the defects of small prediction range, high cost and the like in the prior art.
Description
Technical Field
The invention relates to the field of geological disaster prediction, in particular to a large-range landslide deformation prediction method based on InSAR inversion and multiple influence factors.
Background
Landslide is a complex geological evolution process triggered by the combined action of internal and external factors such as geological structure, rainfall and the like. Frequent and widely distributed landslides and chain disasters thereof seriously affect the development of regional water energy resources, novel urbanization construction, construction and operation of projects such as railway and highway traffic trunks and the like. Landslide deformation prediction is one of basic works for landslide disaster prevention and control as an effective means for realizing landslide disaster prediction, and related theories and methods are rapidly developed and applied in recent years.
The invention patent with the publication number of CN112270400A considers the influence of landslide deformation influence factors on landslide, establishes a model by using a deep learning LSTM (Long Short-Term Memory) network, considers the dynamic characteristic of the self evolution of landslide and improves the prediction precision. However, in the scheme, single-point monitoring data obtained by traditional landslide monitoring is used as a main data source, the method is limited by small traditional landslide monitoring range and high cost, and the single-point monitoring/predicting result cannot comprehensively reflect the overall deformation characteristic of the landslide.
The invention patent with publication number CN113251947A adopts the technical scheme that an InSAR technology is used for obtaining a surface deformation time sequence result, and then a deep learning LSTM network is combined to establish a model for predicting the surface deformation. On one hand, however, the method does not classify and process the earth surface deformation inversion result of InSAR, and can not consider different deformation trends of different areas, which can affect the prediction effect of deformation; on the other hand, the method does not consider the influence of landslide deformation influence factors, and is not suitable for landslide areas greatly influenced by dynamic influence factors.
Disclosure of Invention
In view of this, the present invention aims to overcome the defects in the prior art, and provides a method for predicting large-scale landslide deformation based on InSAR inversion and multiple influence factors, which can realize effective prediction of large-scale landslide deformation, and solve the problems of small prediction range, high cost, and the like in the prior art.
The invention discloses a large-range landslide deformation prediction method based on InSAR inversion and multiple influence factors, which comprises the following steps:
s1, collecting SAR image data of a target area, and performing InSAR inversion processing on the SAR image data to obtain time sequence deformation data;
s2, clustering the time sequence deformation data to obtain a plurality of categories of time sequence deformation data;
s3, decomposing the time sequence deformation data of each category into a periodic item deformation sequence and a trend item deformation sequence;
s4, extracting a plurality of influence factors, calculating the association degree between each influence factor and each type of periodic item deformation sequence, selecting the influence factors which are obviously related to the periodic item deformation, and taking the obviously related influence factors as target influence factors;
s5, respectively establishing an LSTM model for the time sequence deformation data of each category, which specifically comprises the following steps:
the method comprises the following steps of (1) deformation of trend items in the same category, taking the deformation of the trend items as variable input, and establishing a trend item univariate LSTM model;
periodic item deformation of the same category, taking the periodic item deformation and a target influence factor as variable inputs, and establishing a periodic item multivariable LSTM model;
s6, network model training is carried out on the univariate LSTM model and the multivariate LSTM model in the step S5 to obtain a trained LSTM model, deformation of each category is predicted by using the trained LSTM model to obtain a period item deformation predicted value and a trend item deformation predicted value of each category;
and S7, adding the cycle item deformation prediction value and the trend item deformation prediction value of each category to obtain deformation prediction results of each category, and combining the deformation prediction results of each category to obtain a large-range landslide deformation prediction result.
Further, performing InSAR inversion processing on the SAR image data specifically comprises the following steps:
s11, preprocessing SAR image data; the preprocessing comprises data import and data clipping;
s12, selecting a super main image from the SAR image, and registering all other images to the super main image;
s13, obtaining basic data for SBAS inversion estimation through baseline estimation, connection diagram generation, differential interference, interference diagram filtering and phase unwrapping processing;
s14, removing residual terrain errors and residual atmospheric phases from the basic data, and performing unit conversion on the deformation information to obtain a deformation result; the deformation result comprises a deformation rate and a deformation time sequence;
and S15, transferring the deformation result to a geographical coordinate system to obtain a plurality of high coherence points of the target area, wherein each high coherence point has time sequence deformation data of a plurality of time periods.
Further, clustering the time-series deformation data specifically includes:
s21, randomly determining N data in all time sequence deformation data to be used as clustering centers respectively;
s22, calculating cosine similarity values CS of each time sequence deformation data and the N clustering centers respectively, and distributing the data to the cluster with the highest CS value;
s23, calculating a CS value of each piece of data in each cluster and other data in the cluster respectively, and redefining the data with the highest average CS value with the other data as a cluster center;
s24, repeating the steps S22-S23 to iterate until the data in each cluster is not changed or iterated for K times;
s25, calculating a CS value of each clustering center data, if the CS value is larger than A, adding 1 to the number N of the categories, and clustering again; if the CS value is smaller than B, subtracting 1 from the number N of the categories, and clustering again; and respectively outputting the data of each category after the clustering is finished.
Further, the cosine similarity value CS is calculated according to the following formula:
wherein the content of the first and second substances, andeach representing two different time series data, x0,…,xN-1And y0,…,yN-1Respectively representing the deformation of N time sequences.
Further, in step S3, the decomposition is performed by using wavelet decomposition, where the decomposition function of the wavelet is Daubechies, and the decomposition layers of the wavelet are four layers.
Further, the plurality of influence factors include rainfall, reservoir level elevation and reservoir level variation.
Further, calculating the association degree between each influence factor and each category of periodic item deformation sequence according to the following steps:
s41, taking the periodic item deformation sequence asTaking n sequences of influencing factors asAnd deforming the sequence of the period itemAnd the sequence of influencing factorsCarrying out normalization processing;
s42, calculating absolute difference | x of corresponding elements of the period item deformation sequence and the influence factor sequence0(t)-xi(t) |; wherein x is0(t) is a sequence of deformation of the periodic termValue at time t, xi(t) is a sequence of influencing factorsThe value at time t;
s43, calculating a correlation coefficient xi0i(t):
Wherein ξ0i(t) is a sequence of deformation of the periodic termAnd the sequence of influencing factorsThe correlation coefficient at the time t is,representing the minimum absolute difference of the two sequences,representing the maximum absolute difference value of the two sequences, wherein rho is a resolution coefficient;
s44, calculating the relevance r0i:
Wherein n is the length of the time sequence.
The invention has the beneficial effects that: the invention discloses a large-scale landslide deformation prediction method based on InSAR inversion and multiple influence factors, which takes an InSAR inversion result as a data source and solves the problems of small range, high cost and the like caused by the utilization of traditional landslide monitoring data in the prior art; compared with the prior art, the method can obtain a large-range deformation prediction result and can better reflect the deformation characteristics of the whole landslide area. By clustering the surface deformation results inverted by the InSAR technology and respectively predicting the clustering results, the problem of low prediction precision caused by directly using the InSAR surface deformation results for prediction in the prior art is solved. By decomposing the clustered surface deformation result and adding landslide deformation dynamic influence factors such as rainfall, reservoir water level and the like, the InSAR inversion result is suitable for landslide prediction, so that the prediction is more accurate.
Drawings
The invention is further described below with reference to the following figures and examples:
FIG. 1 is a flow chart of a prediction method according to the present invention;
FIG. 2 is a diagram illustrating the time-series deformation decomposition result of a high coherence point in category 1 according to the present invention;
FIG. 3 is a schematic diagram of the structure of the LSTM cell unit of the present invention;
FIG. 4 is a schematic diagram illustrating a comparison between deformation amount prediction results of selected point locations in category 1 and actual deformation amounts according to the present invention;
FIG. 5 is a diagram illustrating a prediction result of a large-scale landslide deformation according to the present invention.
Detailed Description
The invention is further described with reference to the accompanying drawings, in which:
the invention discloses a large-range landslide deformation prediction method based on InSAR inversion and multiple influence factors, which comprises the following steps:
s1, collecting SAR image data of a target area, and performing InSAR inversion processing on the SAR image data to obtain time sequence deformation data;
s2, clustering the time sequence deformation data to obtain a plurality of categories of time sequence deformation data;
s3, decomposing the time sequence deformation data of each category into a periodic item deformation sequence and a trend item deformation sequence;
s4, extracting a plurality of influence factors, calculating the association degree between each influence factor and each type of periodic item deformation sequence, selecting the influence factors which are obviously related to the periodic item deformation, and taking the obviously related influence factors as target influence factors;
s5, respectively establishing an LSTM model for the time sequence deformation data of each category, which specifically comprises the following steps:
the method comprises the following steps of (1) deformation of trend items in the same category, taking the deformation of the trend items as variable input, and establishing a trend item univariate LSTM model;
periodic item deformation of the same category, taking the periodic item deformation and a target influence factor as variable inputs, and establishing a periodic item multivariable LSTM model;
s6, network model training is carried out on the univariate LSTM model and the multivariate LSTM model in the step S5 to obtain a trained LSTM model, deformation of each category is predicted by using the trained LSTM model to obtain a period item deformation predicted value and a trend item deformation predicted value of each category;
and S7, adding the cycle item deformation prediction value and the trend item deformation prediction value of each category to obtain deformation prediction results of each category, and combining the deformation prediction results of each category to obtain a large-range landslide deformation prediction result.
In this embodiment, with the development of the Synthetic Aperture interferometry (interferometric Synthetic Aperture radar), the method has been widely applied to surface deformation monitoring. The inversion result of InSAR can be used as a reliable data source for predicting landslide deformation.
Selecting a certain research area as a target area, using Envi software, and processing 30-scene time sequence Sentinel-1SAR images of the research area by using an SBAS-InSAR technology, wherein the time sequence interval of each period of the images is 12 days, and the process is as follows:
s11, preprocessing SAR image data; the preprocessing comprises data import and data clipping; the data cutting is specifically to cut the range of the research area;
s12, selecting a scene SAR image with better quality as a super main image, and completely registering all other images to the super main image so as to realize the registration of the images;
s13, obtaining basic data for SBAS inversion estimation through baseline estimation, connection diagram generation, differential interference, interference diagram filtering and phase unwrapping processing;
s14, removing residual terrain errors and residual atmospheric phases from the basic data through SBAS inversion estimation, and performing unit conversion on deformation information to obtain a deformation result; the deformation result comprises a deformation rate and a deformation time sequence;
and S15, transferring the deformation result to a geographical coordinate system to obtain 2765 high coherence points of the target area, wherein each high coherence point has time sequence deformation data of 30 periods.
In this embodiment, the cosine similarity is used as a similarity measure, and the time sequence deformation result data obtained by the InSAR technology is clustered to define different deformation regions. The method comprises the following specific steps:
s21, setting the number of the types of clustering as N, and randomly determining N data in all time sequence deformation data to be used as clustering centers respectively;
s22, calculating cosine similarity CS of each time sequence deformation data and N clustering centers respectively, and distributing the data to a cluster with the highest CS value; all data is finally classified into N classes;
s23, calculating a CS value of each piece of data in each cluster and other data in the cluster respectively, and redefining the data with the highest average CS value with the other data as a cluster center;
s24, repeating the steps S22-S23 to iterate until the data in each cluster is not changed or iterated for K times; the value of K is 100;
s25, calculating CS values among the data of all clustering centers, if the CS values are larger than A, considering that all the categories are similar, and needing to be further classified, adding 1 to the number N of the categories, and clustering again; if the CS value is smaller than B, the difference of each category is considered to be too large, the classification is properly reduced, the number N of the categories is reduced by 1, and the clustering is repeated; and respectively outputting the data of each category after the clustering is finished. Wherein, the value of A is 0.8, and the value of B is 0.6.
In this embodiment, the cosine similarity CS is calculated according to the following formula:
wherein the content of the first and second substances, andeach representing two different time series data, x0,…,xN-1And y0,…,yN-1Represents N time-series deformations;
the similarity between the two data is calculated by the formula, and only represents the correlation degree between the two data; regarding the determination of the cluster center, except that the data is initially determined randomly in S21, then in the loop of S23, the similarity of each piece of data with other data in the cluster is calculated, and the piece of data with the highest average similarity is determined as the cluster center again;
according to different deformation trends, after 2765 high-coherence points obtained by InSAR earth surface deformation inversion are clustered, and three different types of deformation are finally obtained through iterative calculation, wherein 846 data are available in the type 1, 1355 data are available in the type 2, and 564 data are available in the type 3.
In this embodiment, in step S3, the time-series deformation of each piece of data in each category is decomposed into a period item deformation sequence and a trend item deformation sequence:
S(t)=X(t)+Y(t);
wherein, S (t) is a time sequence deformation sequence, X (t) is a period deformation sequence, and Y (t) is a trend deformation sequence.
And decomposing the time sequence deformation of each category into period term time sequence deformation and trend term time sequence deformation. The Matlab software is used for wavelet decomposition calculation, a proper wavelet function and decomposition layer number need to be selected, and multiple experiments are carried out, in the example, the Daubechies wavelet and four-layer decomposition are adopted, so that a good decomposition effect can be obtained on the time sequence deformation of the research area, and fig. 2 shows the time sequence deformation decomposition result of a certain high coherence point in the category 1.
In the present embodiment, in step S4, influence factors such as rainfall and reservoir level in multiple dimensions are extracted from the relevant monitoring data, and five influence factors such as 12-day rainfall, 24-day rainfall, reservoir level elevation, 12-day reservoir level variation, and 24-day reservoir level variation in 30 periods are selected in this example. After the data processing is finished, performing grey correlation analysis on each influence factor and the periodic item deformation of each category of clustering centers, taking category 1 as an example, wherein the correlation analysis result is shown in table 1, and the result shows that the selected influence factors are closely related to the periodic item deformation.
TABLE 1
In this embodiment, the association degree between each influence factor and each category of the periodic item deformation sequence is calculated according to the following steps:
s41, taking the periodic item deformation sequence asTaking n sequences of influencing factors asAnd deforming the sequence of the period itemAnd the sequence of influencing factorsCarrying out normalization processing;
s42, calculating the deformation of the period itemAbsolute difference | x between corresponding elements of the sequence and the sequence of influence factors0(t)-xi(t) |; wherein x is0(t) is a sequence of deformation of the periodic termValue at time t, xi(t) is a sequence of influencing factorsThe value at time t;
s43, calculating a correlation coefficient xi0i(t):
Wherein ξ0i(t) is a sequence of deformation of the periodic termAnd the sequence of influencing factorsThe correlation coefficient at the time t is,representing the minimum absolute difference of the two sequences,representing the maximum absolute difference value of the two sequences, wherein rho is a resolution coefficient, and the value interval of rho is (0, 1), usually 0.5;
s44, calculating the relevance r0i:
Wherein n is the length of a time sequence;
comparing the calculated association degree with a set threshold, if the association degree of the influence factor and the period item deformation sequence is greater than the set threshold, the influence factor is obviously related to the period item deformation, otherwise, the influence factor is not obviously related to the period item deformation; wherein the set threshold is 0.6.
In this embodiment, in step S5, different types of deformations are obtained by clustering, and models are respectively built to realize refined prediction.
Using a Keras framework, taking TensorFlow as a rear end, and using a python coding language to decompose the periodic item and the trend item deformation obtained by the same category, and constructing a double-LSTM model for prediction, namely, only using the trend item deformation as variable input, establishing a trend item univariate LSTM model, and predicting the trend item deformation; and (3) using the periodic item deformation and the target influence factor as variable inputs, establishing a periodic item multivariable LSTM model, and predicting the periodic item deformation.
The structure of the LSTM cell unit is shown in fig. 3, and the corresponding algorithm for the main structure is:
it=σ(Wxixt+Whiht-1+Wcict-1+bi)
ft=σ(Wxfxt+Whfht-1+Wcfct-1+bf)
ct=ftct-1+ittanh(Wxcxt+Whcht-1+bc)
Ot=σ(Wxoxt+Whoht-1+Wcoct+bo)
ht=Ottanh(ct)
in the formula, i, f, c and o respectively represent an input gate, a forgetting gate, a cell state and an output gate; x is the number oftIndicates the input at time t, htAn implicit state output representing the corresponding cell unit; w and b are respectively corresponding weight coefficient matrix and bias item; σ and tanh are sigmoid and hyperbolic tangent activation functions, respectively.
In this embodiment, network model training is performed on a univariate LSTM model and a multivariate LSTM model, an input length and an output length are determined, a training set, a validation set, and a test set are divided, the training set is used for model training, the validation set is used for hyper-parameter fine tuning in a model training process, the test set is used for model testing, and finally, a cycle item and a trend item predicted value of each category are obtained.
The method specifically comprises the following steps: the input length is selected to be 24, the output length is 1, namely the deformation of the previous 24 phases, and the future 1 phase is subjected to single-step prediction. Taking category 1 as an example, the training set is 846 high coherence points, deformation time sequences of 1-25 periods, the periods 1-24 are taken as samples, the period 25 is taken as a label, seventy percent is taken as the training set, and thirty percent is taken as the verification set; the test set was for period 2-26 deformations of 846 high coherence points. And (3) performing parameter optimization according to a grid search method, wherein the number of LSTM layers in the trend term univariate LSTM model is set to be 2, the number of hidden neurons is 32, the number of LSTM layers in the periodic term multivariate LSTM model is set to be 3, and the number of hidden neurons is 64.
And directly predicting the test set by using the trained model, predicting the cycle item deformation and the trend item deformation in the 26 th stage, and adding the cycle item deformation and the trend item deformation predicted values of all the categories to obtain the deformation quantity prediction result of each high coherence point in all the categories. The pair of the deformation amount prediction result and the true deformation amount of the partial point location selected from the category 1 is shown in fig. 4, in which the average absolute error of each point in the figure is 1.871mm, and the maximum point location error is 3.755 mm. Wherein, the average absolute error:in the formula, xiAndthe real value and the predicted value are respectively, and N is the number of samples.
By combining the deformation prediction quantities of various categories and visually outputting, a large-range landslide deformation prediction result as shown in fig. 5 can be obtained.
Finally, the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered in the claims of the present invention.
Claims (7)
1. A large-range landslide deformation prediction method based on InSAR inversion and multiple influence factors is characterized by comprising the following steps: the method comprises the following steps:
s1, collecting SAR image data of a target area, and performing InSAR inversion processing on the SAR image data to obtain time sequence deformation data;
s2, clustering the time sequence deformation data to obtain a plurality of categories of time sequence deformation data;
s3, decomposing the time sequence deformation data of each category into a periodic item deformation sequence and a trend item deformation sequence;
s4, extracting a plurality of influence factors, calculating the association degree between each influence factor and each type of periodic item deformation sequence, selecting the influence factors which are obviously related to the periodic item deformation, and taking the obviously related influence factors as target influence factors;
s5, respectively establishing an LSTM model for the time sequence deformation data of each category, which specifically comprises the following steps:
the method comprises the following steps of (1) deformation of trend items in the same category, taking the deformation of the trend items as variable input, and establishing a trend item univariate LSTM model;
periodic item deformation of the same category, taking the periodic item deformation and a target influence factor as variable inputs, and establishing a periodic item multivariable LSTM model;
s6, network model training is carried out on the univariate LSTM model and the multivariate LSTM model in the step S5 to obtain a trained LSTM model, deformation of each category is predicted by using the trained LSTM model to obtain a period item deformation predicted value and a trend item deformation predicted value of each category;
and S7, adding the cycle item deformation prediction value and the trend item deformation prediction value of each category to obtain deformation prediction results of each category, and combining the deformation prediction results of each category to obtain a large-range landslide deformation prediction result.
2. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 1, characterized in that: performing InSAR inversion processing on SAR image data, which specifically comprises the following steps:
s11, preprocessing SAR image data; the preprocessing comprises data import and data clipping;
s12, selecting a super main image from the SAR image, and registering all other images to the super main image;
s13, obtaining basic data for SBAS inversion estimation through baseline estimation, connection diagram generation, differential interference, interference diagram filtering and phase unwrapping processing;
s14, removing residual terrain errors and residual atmospheric phases from the basic data, and performing unit conversion on the deformation information to obtain a deformation result; the deformation result comprises a deformation rate and a deformation time sequence;
and S15, transferring the deformation result to a geographical coordinate system to obtain a plurality of high coherence points of the target area, wherein each high coherence point has time sequence deformation data of a plurality of time periods.
3. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 1, characterized in that: clustering the time sequence deformation data, which specifically comprises the following steps:
s21, randomly determining N data in all time sequence deformation data to be used as clustering centers respectively;
s22, calculating cosine similarity values CS of each time sequence deformation data and the N clustering centers respectively, and distributing the data to the cluster with the highest CS value;
s23, calculating a CS value of each piece of data in each cluster and other data in the cluster respectively, and redefining the data with the highest average CS value with the other data as a cluster center;
s24, repeating the steps S22-S23 to iterate until the data in each cluster is not changed or iterated for K times;
s25, calculating a CS value of each clustering center data, if the CS value is larger than A, adding 1 to the number N of the categories, and clustering again; if the CS value is smaller than B, subtracting 1 from the number N of the categories, and clustering again; and respectively outputting the data of each category after the clustering is finished.
4. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 3, characterized in that: the cosine similarity value CS is calculated according to the following formula:
5. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 1, characterized in that: in step S3, the decomposition is performed by wavelet decomposition, where the decomposition function of the wavelet is Daubechies, and the number of decomposition layers of the wavelet is four.
6. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 1, characterized in that: the plurality of influence factors comprise rainfall, reservoir water level elevation and reservoir water level variation.
7. The InSAR inversion and multi-influence factor-based large-range landslide deformation prediction method according to claim 1, characterized in that: calculating the association degree between each influence factor and each category of periodic item deformation sequence according to the following steps:
s41, taking the periodic item deformation sequence asTaking n sequences of influencing factors asAnd deforming the sequence of the period itemAnd the sequence of influencing factorsCarrying out normalization processing;
s42, calculating absolute difference | x of corresponding elements of the period item deformation sequence and the influence factor sequence0(t)-xi(t) |; wherein x is0(t) is a sequence of deformation of the periodic termValue at time t, xi(t) is a sequence of influencing factorsThe value at time t;
s43, calculating a correlation coefficient xi0i(t):
Wherein ξ0i(t) is a sequence of deformation of the periodic termAnd the sequence of influencing factorsThe correlation coefficient at the time t is,representing the minimum absolute difference of the two sequences,representing the maximum absolute difference value of the two sequences, wherein rho is a resolution coefficient;
s44, calculating the relevance r0i:
Wherein n is the length of the time sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111567548.0A CN114252879A (en) | 2021-12-20 | 2021-12-20 | InSAR inversion and multi-influence factor based large-range landslide deformation prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111567548.0A CN114252879A (en) | 2021-12-20 | 2021-12-20 | InSAR inversion and multi-influence factor based large-range landslide deformation prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114252879A true CN114252879A (en) | 2022-03-29 |
Family
ID=80793370
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111567548.0A Pending CN114252879A (en) | 2021-12-20 | 2021-12-20 | InSAR inversion and multi-influence factor based large-range landslide deformation prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114252879A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114966685A (en) * | 2022-05-24 | 2022-08-30 | 中国水利水电科学研究院 | Dam deformation monitoring and predicting method based on InSAR and deep learning |
CN114966692A (en) * | 2022-07-19 | 2022-08-30 | 之江实验室 | Transformer-based InSAR technology frozen soil area multivariable time sequence deformation prediction method and device |
CN117349602A (en) * | 2023-12-06 | 2024-01-05 | 江西省水投江河信息技术有限公司 | Water conservancy facility operation state prediction method, system and computer |
CN117710776A (en) * | 2024-02-06 | 2024-03-15 | 中国地震局地质研究所 | Landslide deformation space-time prediction method and system |
-
2021
- 2021-12-20 CN CN202111567548.0A patent/CN114252879A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114966685A (en) * | 2022-05-24 | 2022-08-30 | 中国水利水电科学研究院 | Dam deformation monitoring and predicting method based on InSAR and deep learning |
CN114966685B (en) * | 2022-05-24 | 2023-04-07 | 中国水利水电科学研究院 | Dam deformation monitoring and predicting method based on InSAR and deep learning |
CN114966692A (en) * | 2022-07-19 | 2022-08-30 | 之江实验室 | Transformer-based InSAR technology frozen soil area multivariable time sequence deformation prediction method and device |
CN114966692B (en) * | 2022-07-19 | 2022-11-08 | 之江实验室 | Transformer-based InSAR technology frozen soil area multivariable time sequence deformation prediction method and device |
CN117349602A (en) * | 2023-12-06 | 2024-01-05 | 江西省水投江河信息技术有限公司 | Water conservancy facility operation state prediction method, system and computer |
CN117710776A (en) * | 2024-02-06 | 2024-03-15 | 中国地震局地质研究所 | Landslide deformation space-time prediction method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114252879A (en) | InSAR inversion and multi-influence factor based large-range landslide deformation prediction method | |
CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
Sugiartawan et al. | Prediction by a hybrid of wavelet transform and long-short-term-memory neural network | |
CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
CN113240187B (en) | Prediction model generation method, system, device, storage medium and prediction method | |
CN111242377B (en) | Short-term wind speed prediction method integrating deep learning and data denoising | |
CN114220271A (en) | Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network | |
CN111540193A (en) | Traffic data restoration method for generating countermeasure network based on graph convolution time sequence | |
Gramacy et al. | Parameter space exploration with Gaussian process trees | |
CN110969304A (en) | Method, system and device for predicting production capacity of digital factory | |
CN113486303A (en) | Long-time sequence prediction method based on modification model integration | |
CN114417693A (en) | Ocean three-dimensional temperature field inversion method based on deep learning | |
CN114065996A (en) | Traffic flow prediction method based on variational self-coding learning | |
CN115587666A (en) | Load prediction method and system based on seasonal trend decomposition and hybrid neural network | |
CN114266982A (en) | SAR image change detection method based on network parameter and structure collaborative optimization | |
CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
CN115905818A (en) | Landslide early warning method based on data mining | |
CN112784479A (en) | Flood flow prediction method | |
CN114692963A (en) | Landslide displacement prediction method based on long-term and short-term memory neural network | |
CN116169670A (en) | Short-term non-resident load prediction method and system based on improved neural network | |
Li et al. | Deep reliability learning with latent adaptation for design optimization under uncertainty | |
Pandhiani et al. | Time series forecasting by using hybrid models for monthly streamflow data | |
CN117408167A (en) | Debris flow disaster vulnerability prediction method based on deep neural network | |
CN117094431A (en) | DWTfar meteorological data time sequence prediction method and equipment for multi-scale entropy gating | |
CN116739168A (en) | Runoff prediction method based on gray theory and codec |
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 |