TAU-based GB-InSAR atmospheric disturbance compensation method, device and equipment and storage device
The TAU-based method is used to correct the atmospheric disturbance compensation of GB-InSAR data, and the TAU sub-model is used to predict the atmospheric disturbance phase and compensate, which solves the problem of atmospheric delay error in daytime continuous monitoring, and achieves high-precision deformation monitoring.
Patent Information
- Application Number
- CN202411193085.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2024-12-03
AI Technical Summary
The prior art cannot effectively reduce atmospheric delay errors during the daytime continuous monitoring process where temperature and water vapor change are faster, affecting the measurement accuracy of GB-InSAR.
Using a TAU-based method, a TAU submodel is randomly extracted from the differentially processed image as the TAU submodel training data, and a TAU submodel with corresponding structure is constructed, and the model is trained using k-fold cross-validation as the training criterion. The optimal number of iterations is obtained for weighted average, which is used to predict the atmospheric disturbance phase and to compensate for the original differential image.
During the continuous daytime monitoring process where temperature and water vapor change rapidly, the overall control accuracy error is within 0.7mm, effectively weakening the atmospheric disturbance error and having good adaptability to complex environments and climatic conditions.
Smart Images

Figure CN119066424A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of atmospheric disturbance compensation and correction, and in particular to a GB-InSAR atmospheric disturbance compensation method, device, equipment and storage device based on TAU. Background Art
[0002] In the field of GB-InSAR atmospheric disturbance compensation and correction, GB-InSAR is a mature high-precision deformation monitoring system. The GB-InSAR system monitors the deformation displacement of surface objects by transmitting and receiving electromagnetic wave signals. Since atmospheric conditions (temperature, humidity, pressure, etc.) vary with time and space, these changes will affect the propagation speed of radar waves through the troposphere, so the atmospheric effect causes transmission delay. Atmospheric disturbance is one of the main error sources affecting measurement accuracy. In order to achieve the sub-millimeter level of deformation monitoring accuracy, the wavelength of the band used by ground-based radar is short and very sensitive to changes in the ambient atmosphere. Therefore, a reliable atmospheric disturbance compensation correction method is urgently needed to compensate and correct the time-series differential interferometry phase, so as to obtain high-precision time-series deformation information. The atmospheric delay phase and the time-series deformation signal in the landslide creep process show different characteristics in the time-series differential interferometry phase. Specifically, the atmospheric delay phase is low-frequency in the time domain, that is, it appears in some time-differential interferograms, but suddenly disappears at other times. The temporal deformation signal in the landslide creep process is high-frequency in the temporal differential interferometry phase diagram, that is, the deformation phase is continuous. At present, the atmospheric correction methods are mainly divided into three categories: meteorological data correction method, permanent scatterer technology and distance function fitting method. The characteristics of these three methods are analyzed below:
[0003] Meteorological data correction method: This method corrects atmospheric disturbances by correcting meteorological data. However, since the observation distance of ground-based SAR can reach several thousand meters, if the observation distance is too far, changes in meteorological data will introduce errors and affect the correction effect of atmospheric disturbances. In addition, the atmospheric refraction empirical model usually only considers the impact of atmospheric disturbances on the range direction, while ignoring the errors in the azimuth direction.
[0004] Permanent scatterer technology: This technology uses a set of stable SAR data to find persistent and stable radar target points and analyze their phase changes. By studying the phase changes of stable target points, the impact of atmospheric disturbances on the entire observation area can be estimated, and the delayed phase caused by atmospheric disturbances can be removed from the differential phase. However, this technology has some disadvantages, such as the need for a large amount of SAR image data and complex data processing, requiring a sufficient number of permanent scatterers, and these scatterers have small deformation during the monitoring process.
[0005] Distance function fitting method: This method estimates the correction value of atmospheric changes in the observation area by selecting stable points or deploying corner reflectors, using the distance and phase changes between stable points and ground-based SAR. However, due to the dangerousness of the landslide area, it is not appropriate to deploy corner reflectors manually, and stable points may be difficult to find, which limits the application of this method. Summary of the invention
[0006] In order to solve the problems existing in the above-mentioned prior art, the present invention provides a GB-InSAR atmospheric disturbance compensation method, device, equipment and storage device based on TAU, which solves the problem that the prior art cannot effectively reduce the atmospheric delay error during continuous monitoring during the day when the temperature and water vapor change rapidly.
[0007] GB-I nSAR atmospheric disturbance compensation method based on TAU includes:
[0008] S1, randomly extract multiple pixel sequences from the differentially processed image as TAU sub-model training data;
[0009] S2. construct a TAU sub-model of corresponding structure according to the selected TAU sub-model training data;
[0010] S3, constructing a training framework for the TAU sub-model using k-fold cross validation as the training criterion;
[0011] S4. Use the training framework to train each TAU sub-model, obtain multiple optimal iteration times and perform weighted average as the optimal iteration times;
[0012] S5. Predicting the atmospheric disturbance phase according to the optimal number of iterations combined with the full-time GB-I nSAR differential interferometry data to obtain the atmospheric disturbance phase prediction result;
[0013] S6. Compensation and correction processing is performed on the original differential image according to the atmospheric disturbance phase prediction result to obtain the GB-I nSAR image after atmospheric disturbance compensation and correction.
[0014] Preferably, the S1 comprises:
[0015] S11, subjecting the GB-I nSAR observation results to differential interferometry processing to obtain an original GB-I nSAR image, and then randomly selecting m pixel samples from the original GB-I nSAR image, wherein the time series length of the pixel samples is not less than 100;
[0016] S12. Convert the selected pixel samples into vectors required by the TAU sub-model as TAU sub-model training data.
[0017] Preferably, S2 includes:
[0018] S21. Build the basic TAU sub-model and define the basic information of the model, including: number of neuron layers, number of layers, dropout layer, and activation layer;
[0019] S22, importing the TAU sub-model training data into the constructed TAU sub-model to prepare for training;
[0020] S23. Repeat the method of S21 and S22 to import the TAU sub-model until m TAU sub-models are formed.
[0021] Preferably, S3 includes:
[0022] S31. Before training the TAU sub-model, k-fold cross validation is used to control the training process, specifically: the vector data of the TAU sub-model is evenly divided into k parts, of which k-1 parts are used for model training and the remaining 1 part is used as verification data;
[0023] S32, repeat S31, respectively complete the construction of the training framework for each equal-divided vector of the sub-model, and ensure that each model vector is used as verification data, until the construction of the TAU sub-model training framework is completed k times;
[0024] S33. Repeat S31 and S32 to complete the construction of the training framework for m TAU sub-models.
[0025] Preferably, S4 includes:
[0026] S41. Use the training framework to train the TAU sub-model, use the early lystop method to avoid overfitting, and record the number of iterations before overfitting;
[0027] S42, repeat S41 to complete the training of m TAU sub-models respectively, record the m optimal iteration times, and calculate the weighted average of the iteration times, where the weight is the length of the vector time series during the sub-model training; the weighted average iteration times can be expressed as E mean :
[0028]
[0029] Among them, E m represents the optimal number of iterations for the mth TAU, k m Represents the length of the sample data time series when the m-th TAU sub-model is trained.
[0030] Preferably, S5 includes:
[0031] S51, modifying the number of iterations of the TAU sub-model to the optimal number of iterations after integer processing, and then using the TAU sub-model to predict the full time series data of the pixels one by one;
[0032] S52, splicing the full time series data prediction results of all pixels into a complete time series differential image, and obtaining the atmospheric disturbance data of the time period as the atmospheric disturbance phase prediction result.
[0033] Preferably, S6 includes:
[0034] S61, performing geocoding using the atmospheric disturbance phase prediction result obtained in S5 to obtain an accurate atmospheric disturbance value relative to the geographical location;
[0035] S62. Based on the accurate atmospheric disturbance value of the relative geographical location, the original differential image is compensated and corrected to obtain the GB-InSAR image after atmospheric disturbance compensation and correction.
[0036] The GB-InSAR atmospheric disturbance compensation device based on TAU can implement the GB-InSAR atmospheric disturbance compensation method based on TAU, including: a data processing module, a model building module, a model training module, a phase prediction module and an image correction module; the data processing module is used to randomly extract a plurality of pixel sequences with a fixed time series length from the differentially processed image as TAU sub-model training data;
[0037] The model building module is used to build a TAU sub-model of corresponding structure according to the selected TAU sub-model training data;
[0038] The model training module is used to construct a training framework for the TAU sub-model using k-fold cross validation as a training criterion; and train each TAU sub-model to obtain multiple optimal iteration times and perform weighted averaging;
[0039] The phase prediction module is used to predict the atmospheric disturbance phase according to the optimal number of iterations after weighted averaging and in combination with the full-time GB-InSAR differential interferometry data to obtain the atmospheric disturbance phase prediction result;
[0040] The image correction module is used to perform compensation and correction processing on the original differential image according to the atmospheric disturbance phase prediction result to obtain the GB-InSAR image after atmospheric disturbance compensation and correction.
[0041] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements a GB-InSAR atmospheric disturbance compensation method based on TAU when executing the computer program.
[0042] A storage medium stores a computer program, which, when executed by a processor, enables the processor to perform a GB-I nSAR atmospheric disturbance compensation method based on TAU.
[0043] The beneficial effects of the present invention include:
[0044] The present invention adopts the atmospheric disturbance compensation correction method for ground-based radar data based on TAU to perform atmospheric disturbance compensation correction on the original time-series differential interferometry phase image. During the daytime continuous monitoring process with rapid changes in temperature and water vapor, the overall accuracy error is controlled within 0.7mm by cross-validation with the high-precision total station ground observation data of the same period. It can be seen that the atmospheric disturbance compensation correction method for ground-based radar data based on TAU effectively reduces the atmospheric disturbance error and has good adaptability to complex environments and climatic conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 This is the flow chart of atmospheric disturbance compensation correction for ground-based radar data based on TAU.
[0046] Figure 2 This is an image of the monitored object using GB-I nSAR.
[0047] Figure 3 This is the deformation map of the monitoring area along the LOS direction without compensation for atmospheric disturbances.
[0048] Figure 4 The deformation map of the monitoring area along the LOS direction is corrected after atmospheric disturbance compensation.
[0049] Figure 5 This is the time diagram of the deformation error of the characteristic point along the LOS after correction based on TAU atmospheric disturbance compensation. DETAILED DESCRIPTION
[0050] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0051] The following is combined with Figure 1-5 The specific embodiments of the present invention are described in detail;
[0052] The GB-I nSAR atmospheric disturbance compensation method based on TAU includes:
[0053] S1, randomly extract multiple pixel sequences of fixed time series length from the differenced image as TAU sub-model training data;
[0054] S11, firstly, the electromagnetic wave signal returned by GB-I nSAR observation is subjected to differential interference processing to obtain the original GB-I nSAR image, and then m pixel samples are randomly selected from the original GB-I nSAR image, and the time series length of the pixel samples is not less than 100;
[0055] S12. Convert the selected pixel samples into vectors required by the TAU sub-model as TAU sub-model training data.
[0056] S2. Construct a TAU sub-model with the same structure for the selected TAU sub-model training data; wherein the TAU sub-model structure and hyperparameter information are set based on experience.
[0057] S21. Build the basic TAU sub-model and define the basic information of the model, such as the number of neuron layers, dropout layers, activation layers, etc.
[0058] S22, importing the TAU sub-model vector obtained in S12 into the constructed TAU sub-model for training;
[0059] S23. Repeat the method of S21 and S22 to import the TAU sub-model until m TAU sub-models are formed.
[0060] S3, constructing a training framework for the TAU sub-model using k-fold cross validation as the training criterion;
[0061] S31. Before training the TAU sub-model, k-fold cross validation is used to control the training process, specifically: the vector data of the TAU sub-model is evenly divided into k parts, of which k-1 parts are used for model training and the remaining 1 part is used as verification data;
[0062] S32. Repeat S31 to construct a training framework for each equally divided vector of the sub-model, ensuring that each model vector is used as verification data, until the construction of the TAU sub-model training framework is completed k times.
[0063] S33. Repeat S31 and S32 to complete the construction of the training framework for m TAU sub-models.
[0064] S4, train each TAU sub-model to obtain the optimal number of iterations;
[0065] S41. Run the constructed training framework to train the TAU sub-model, use the earlystop method to avoid overfitting, and record the number of iterations before overfitting;
[0066] S42, repeat S41 to complete the training of m TAU sub-models respectively, record the m optimal iteration times, and calculate the weighted average of the iteration times as the optimal iteration times, where the weight is the length of the vector time series during the sub-model training. The weighted average iteration times can be expressed as E mean :
[0067]
[0068] Among them, E m represents the optimal number of iterations for the mth TAU, k m Represents the length of the sample data time series when the m-th TAU sub-model is trained.
[0069] S5. According to the optimal number of iterations obtained through training, the atmospheric disturbance phase is predicted by combining the full-time GB-InSAR differential interferometry data.
[0070] S51, modifying the number of iterations of the TAU sub-model to the optimal number of iterations after integer processing, and then using the TAU sub-model to predict the full time series data of the pixels one by one;
[0071] S52, splicing the full time series data prediction results of all pixels into a complete time series differential image, and obtaining the atmospheric disturbance data of the time period as the atmospheric disturbance phase prediction result.
[0072] S6. According to the result of atmospheric disturbance phase prediction, the original differential image is compensated and corrected to obtain the GB-I nSAR image after atmospheric disturbance compensation and correction.
[0073] S61, using the atmospheric disturbance phase prediction result predicted in S5 to perform geocoding to obtain an accurate atmospheric disturbance value relative to the geographical location;
[0074] S62, using the accurate atmospheric disturbance value relative to the geographical location obtained in S61, the original differential image is compensated and corrected to obtain a GB-I nSAR image after atmospheric disturbance compensation and correction.
[0075] In order to verify the effect of the atmospheric disturbance compensation correction method based on TAU ground-based radar data, a landslide area was selected as the research object for experiment. The experimental results are shown in the attached figure. Figures 2 to 5 shown.
[0076] After GB-SAR image data preprocessing, Figure 2 The radar intensity, phase coherence, time coherence, and spatial coherence of the landslide monitoring area are shown in the figure. The red area in the figure is the landslide monitoring area. It can be seen that most radars in the monitoring area have high intensity, good phase stability, high temporal and spatial coherence, and clear images of the study area.
[0077] Since continuous observations exclude non-zero baselines, the deformation results of landslide points are affected by atmospheric disturbances. First, the landslide deformation map of the study area is obtained by conventional time-series differential interferometry processing ( Figure 3 ).
[0078] The TAU method was used to predict the original differential interferometry image, and finally the atmospheric disturbance information of the area was obtained. The original differential interferometry image was compensated and corrected, and the landslide deformation map along the LOS direction after atmospheric compensation and correction by the TAU method was obtained ( Figure 4 ).
[0079] Four monitoring points (P1, P2, P3, P4) were randomly selected to explore the landslide error information. The results are as follows: Figure 5 As shown, analysis Figure 5 It can be obtained that the atmospheric compensation correction method of TAU can make the GB-I nSAR monitoring accuracy error less than 0.7mm, thus verifying the reliability and feasibility of TAU's ground-based radar data atmospheric disturbance compensation correction method in GB-I nSAR field monitoring.
[0080] In summary, the method proposed in this scheme performs atmospheric disturbance compensation correction on the original time-series differential interferometry phase image. During the daytime continuous monitoring process with rapid changes in temperature and air humidity, the overall accuracy error is controlled within 0.8 mm, indicating that the atmospheric disturbance compensation correction method based on TAU ground-based radar data effectively reduces the measurement error caused by atmospheric disturbances and has good adaptability to complex environments and climatic conditions.
[0081] In another embodiment, it relates to a GB-I nSAR atmospheric disturbance compensation device based on TAU, which can implement a GB-I nSAR atmospheric disturbance compensation method based on TAU, including: a data processing module, a model building module, a model training module, a phase prediction module and an image correction module;
[0082] The data processing module is used to randomly extract a plurality of pixel sequences of fixed time series length from the differentially processed image as TAU sub-model training data;
[0083] The model building module is used to build a TAU sub-model of corresponding structure according to the selected TAU sub-model training data;
[0084] The model training module is used to construct a training framework for the TAU sub-model using k-fold cross validation as a training criterion; and train each TAU sub-model to obtain multiple optimal iteration times and perform weighted averaging;
[0085] The phase prediction module is used to predict the atmospheric disturbance phase according to the optimal number of iterations after weighted averaging and combine the full-time GB-I nSAR differential interferometry data to obtain the atmospheric disturbance phase prediction result;
[0086] The image correction module is used to perform compensation and correction processing on the original differential image according to the atmospheric disturbance phase prediction result to obtain the GB-I nSAR image after atmospheric disturbance compensation and correction.
[0087] In another embodiment, it relates to an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a GB-I nSAR atmospheric disturbance compensation method based on TAU when executing the computer program.
[0088] In another embodiment, a storage medium is provided, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs a GB-I nSAR atmospheric disturbance compensation method based on TAU.
[0089] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.
Claims
1. GB-InSAR atmospheric disturbance compensation method based on TAU, characterized in that: include: S1, randomly extract multiple pixel sequences from the differentially processed image as TAU sub-model training data; S2. construct a TAU sub-model of corresponding structure according to the selected TAU sub-model training data; S3, constructing a training framework for the TAU sub-model using k-fold cross validation as the training criterion; S4. Use the training framework to train each TAU sub-model, obtain multiple optimal iteration times and perform weighted average as the optimal iteration times; S5. Predict the atmospheric disturbance phase according to the optimal number of iterations combined with the full-time GB-InSAR differential interferometry data to obtain the atmospheric disturbance phase prediction result; S6. Compensation and correction are performed on the original differential image according to the atmospheric disturbance phase prediction result to obtain the GB-InSAR image after atmospheric disturbance compensation and correction.
2. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S1 includes: S11, subjecting the GB-InSAR observation results to differential interferometry processing to obtain an original GB-InSAR image, and then randomly selecting m pixel samples from the original GB-InSAR image, wherein the time series length of the pixel samples is not less than 100; S12. Convert the selected pixel samples into vectors required by the TAU sub-model as TAU sub-model training data.
3. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S2 includes: S21. Build the basic TAU sub-model and define the basic information of the model, including: number of neuron layers, number of layers, dropout layer, and activation layer; S22, importing the TAU sub-model training data into the constructed TAU sub-model to prepare for training; S23. Repeat the method of S21 and S22 to import the TAU sub-model until m TAU sub-models are formed.
4. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S3 includes: S31. Before training the TAU sub-model, k-fold cross validation is used to control the training process, specifically: the vector data of the TAU sub-model is evenly divided into k parts, of which k-1 parts are used for model training and the remaining 1 part is used as verification data; S32, repeat S31, respectively complete the construction of the training framework for each equal-divided vector of the sub-model, and ensure that each model vector is used as verification data, until the construction of the TAU sub-model training framework is completed k times; S33. Repeat S31 and S32 to complete the construction of the training framework for m TAU sub-models.
5. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S4 includes: S41. Use the training framework to train the TAU sub-model, use the earlystop method to avoid overfitting, and record the number of iterations before overfitting; S42, repeat S41 to complete the training of m TAU sub-models respectively, record the m optimal iteration times, and calculate the weighted average of the iteration times, where the weight is the length of the vector time series during the sub-model training; the weighted average iteration times can be expressed as E mean : Among them, E m represents the optimal number of iterations for the mth TAU, k m Represents the length of the sample data time series when the m-th TAU sub-model is trained.
6. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S5 includes: S51, modifying the number of iterations of the TAU sub-model to the optimal number of iterations after integer processing, and then using the TAU sub-model to predict the full time series data of the pixels one by one; S52, splicing the full time series data prediction results of all pixels into a complete time series differential image, and obtaining the atmospheric disturbance data of the time period as the atmospheric disturbance phase prediction result.
7. The GB-InSAR atmospheric disturbance compensation method based on TAU according to claim 1, characterized in that: The S6 includes: S61, performing geocoding using the atmospheric disturbance phase prediction result obtained in S5 to obtain an accurate atmospheric disturbance value relative to the geographical location; S62. Based on the accurate atmospheric disturbance value of the relative geographical location, the original differential image is compensated and corrected to obtain the GB-InSAR image after atmospheric disturbance compensation and correction.
8. GB-InSAR atmospheric disturbance compensation device based on TAU, characterized in that: A GB-InSAR atmospheric disturbance compensation method based on TAU capable of implementing any one of claims 1 to 7, comprising: a data processing module, a model building module, a model training module, a phase prediction module and an image correction module; The data processing module is used to randomly extract a plurality of pixel sequences of fixed time series length from the differentially processed image as TAU sub-model training data; The model building module is used to build a TAU sub-model of corresponding structure according to the selected TAU sub-model training data; The model training module is used to construct a training framework for the TAU sub-model using k-fold cross validation as a training criterion; and train each TAU sub-model to obtain multiple optimal iteration times and perform weighted averaging; The phase prediction module is used to predict the atmospheric disturbance phase according to the optimal number of iterations after weighted averaging and in combination with the full-time GB-InSAR differential interferometry data to obtain the atmospheric disturbance phase prediction result; The image correction module is used to perform compensation and correction processing on the original differential image according to the atmospheric disturbance phase prediction result to obtain the GB-InSAR image after atmospheric disturbance compensation and correction.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the GB-InSAR atmospheric disturbance compensation method based on TAU according to any one of claims 1 to 7 is implemented.
10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the GB-InSAR atmospheric disturbance compensation method based on TAU according to any one of claims 1 to 7.
