CN117009913B - Sea surface height abnormal data fusion method based on satellite altimeter and tide station - Google Patents

Sea surface height abnormal data fusion method based on satellite altimeter and tide station Download PDF

Info

Publication number
CN117009913B
CN117009913B CN202310499560.5A CN202310499560A CN117009913B CN 117009913 B CN117009913 B CN 117009913B CN 202310499560 A CN202310499560 A CN 202310499560A CN 117009913 B CN117009913 B CN 117009913B
Authority
CN
China
Prior art keywords
data
sea surface
surface height
fusion
day
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
Application number
CN202310499560.5A
Other languages
Chinese (zh)
Other versions
CN117009913A (en
Inventor
周旋
李自强
宋帅
张高英
白志鹏
张芳苒
姚小海
朱科澜
林成鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
61741 Unit Of Pla
Original Assignee
61741 Unit Of Pla
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 61741 Unit Of Pla filed Critical 61741 Unit Of Pla
Priority to CN202310499560.5A priority Critical patent/CN117009913B/en
Publication of CN117009913A publication Critical patent/CN117009913A/en
Application granted granted Critical
Publication of CN117009913B publication Critical patent/CN117009913B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
    • 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
    • G06N3/045—Combinations of networks
    • G06N3/0455—Auto-encoder networks; Encoder-decoder 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/048—Activation 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
    • G06N3/08—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
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The application provides a fusion method of sea surface height abnormal data based on a satellite altimeter and a tide station, which comprises the following steps: acquiring satellite sea surface height abnormal observation data of a target area from a day T-m to a day T, sea surface height abnormal observation data of a tide station on a day T, a sea coverage mask and a data coverage mask; inputting a fusion product generator in the sea surface height abnormal data fusion model; and performing multiple three-dimensional double-mask convolution operations on the satellite sea level height anomaly observation data and the tide station sea level height anomaly observation data by using the ocean coverage mask and the data coverage mask to obtain sea level height anomaly fusion data of the target area T day. In this way, through automatic updating of the data coverage mask and constraint of the ocean coverage mask in the three-dimensional double-mask convolution operation, the space coverage and precision of the sea surface height anomaly fusion data can be improved by combining the observation data of the tide station and carrying out refined reconstruction on the sea surface height anomaly data of an uncovered area, especially the offshore sea area.

Description

Sea surface height abnormal data fusion method based on satellite altimeter and tide station
Technical Field
The application relates to the technical field of ocean, in particular to a fusion method of sea surface height abnormal data based on a satellite altimeter and a tide station.
Background
The sea surface height abnormal data refer to deviation of the sea surface relative to the average sea surface, is an important ocean power environment element, can reflect the three-dimensional structure of the ocean subsurface, and plays a vital role in the aspects of ocean temperature salt profile reconstruction, ocean vortex identification, ocean numerical prediction and the like.
In order to ensure the spatial coverage of the sea surface height abnormal data, the prior art generally adopts traditional fusion modes such as space-time objective analysis, optimal interpolation, variation analysis and the like, however, the traditional fusion modes are often based on assumptions such as spatial uniformity, parameter isotropy and the like, so that the generated fusion data has the problems of weak spatial adaptability, incapability of accurately reflecting the spatial characteristics of the sea surface height abnormal data, high calculation cost and the like. Meanwhile, in the prior art, fusion mode is usually only used for fusing the abnormal sea surface height data observed by the multisource satellite altimeter, but the satellite altimeter can only acquire the data of the points below the satellite, so that the space coverage is low, the accuracy and quality of the observed data are obviously reduced due to the influence of land or islands in the offshore area, and further the fusion data are poor in offshore accuracy and even have data loss.
Disclosure of Invention
In view of the above, the present application aims to provide a method for fusing abnormal sea surface height data based on a satellite altimeter and a tide station, which can mine space-time variation characteristics of abnormal sea surface height observation data for a plurality of continuous days, realize fusion of abnormal sea surface height observation data of a multi-source satellite altimeter and a tide station, and generate full-coverage high-precision abnormal sea surface fusion data; the three-dimensional double-mask convolution operation utilizes the constraint of double masks and the automatic updating of the data coverage mask to carry out the fine reconstruction of the uncovered area of the observed data, and the sea coverage mask and the observation data of the tide station are combined to further ensure the fine reconstruction of the sea-surface height abnormal data at the sea-land juncture of the offshore area, so that the space coverage and the accuracy of the fusion data are improved.
The embodiment of the application provides a fusion method of sea surface height abnormal data based on a satellite altimeter and a tide station, which comprises the following steps:
acquiring satellite sea surface height anomaly observation data from T-m days to T days of a target area determined by a multi-source satellite altimeter, tide station sea surface height anomaly observation data from T days of the target area determined by a plurality of tide stations and a sea coverage mask of the target area; wherein the marine coverage mask is used to characterize coverage of the ocean in the target area;
Determining an initial data coverage mask according to the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data and the ocean coverage mask; the data coverage mask is used for representing coverage conditions of effective observation data in the target area from the T-m day to the T day;
inputting the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea surface height abnormal data fusion model; the sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator;
using the ocean coverage mask and the initial data coverage mask, performing multiple three-dimensional double-mask convolution operations on the satellite sea surface height anomaly observation data and the tide station sea surface height anomaly observation data by the fusion product generator to obtain sea surface height anomaly fusion data of the target region T day; the three-dimensional double-mask convolution operation is performed only on effective observed data in the ocean in a three-dimensional space range corresponding to the convolution kernel by superposing the ocean coverage mask and the data coverage mask on the satellite sea surface height abnormal observed data and the tide station sea surface height abnormal observed data, so that the influence of data deletion on the convolution computation is avoided;
After each three-dimensional double-mask convolution operation is executed, the data coverage mask which is used in the next three-dimensional double-mask convolution operation is updated based on the ocean coverage mask and the data coverage mask used in the three-dimensional double-mask convolution operation.
The embodiment of the application also provides a fusion device of sea surface height abnormal data based on a satellite altimeter and a tide station, and the fusion device comprises:
the acquisition module is used for acquiring satellite sea surface height abnormal observation data from the T-m day to the T day of the target area determined by the multi-source satellite altimeter, tide station sea surface height abnormal observation data from the T day of the target area determined by the plurality of tide stations and a sea coverage mask of the target area; wherein the marine coverage mask is used to characterize coverage of the ocean in the target area;
the determining module is used for determining an initial data coverage mask according to the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data and the ocean coverage mask; the data coverage mask is used for representing coverage conditions of effective observation data in the target area from the T-m day to the T day;
The input module is used for inputting the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea surface height abnormal data fusion model; the sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator;
the convolution module is used for performing multiple three-dimensional double-mask convolution operations on the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data by using the ocean coverage mask and the initial data coverage mask by the fusion product generator so as to obtain sea surface height abnormal fusion data of the target area T day; the three-dimensional double-mask convolution operation is performed only on effective observed data in the ocean in a three-dimensional space range corresponding to the convolution kernel by superposing the ocean coverage mask and the data coverage mask on the satellite sea surface height abnormal observed data and the tide station sea surface height abnormal observed data, so that the influence of data deletion on the convolution computation is avoided;
After each three-dimensional double-mask convolution operation is executed, the data coverage mask which is used in the next three-dimensional double-mask convolution operation is updated based on the ocean coverage mask and the data coverage mask used in the three-dimensional double-mask convolution operation.
The embodiment of the application also provides electronic equipment, which comprises: the system comprises a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, the processor and the memory are communicated through the bus when the electronic device is running, and the machine-readable instructions are executed by the processor to perform the steps of the fusion method based on the sea surface height abnormal data of the satellite altimeter and the tide station.
Embodiments of the present application also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of a method of fusion of sea level height anomaly data based on a satellite altimeter and a tide station as described above.
According to the sea surface height anomaly data fusion method based on the satellite altimeter and the tide station, provided by the embodiment of the application, the space-time variation characteristics of continuous multi-day sea surface height anomaly observation data can be mined, fusion of the sea surface height anomaly observation data of the multi-source satellite altimeter and the tide station is realized, and full-coverage high-precision sea surface height anomaly fusion data is generated; the three-dimensional double-mask convolution operation utilizes the constraint of double masks and the automatic updating of the data coverage mask to carry out the fine reconstruction of the uncovered area of the observed data, and the sea coverage mask and the observation data of the tide station are combined to further ensure the fine reconstruction of the sea-surface height abnormal data at the sea-land juncture of the offshore area, so that the space coverage and the accuracy of the fusion data are improved.
In order to make the above objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered limiting the scope, and that other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 shows a flow chart of a method for fusion of sea level anomaly data based on a satellite altimeter and a tide station provided by an embodiment of the present application;
FIG. 2 is a schematic view of a tide station distribution of a target zone according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a training process of a sea level elevation anomaly data fusion model according to an embodiment of the present application;
FIG. 4 illustrates a schematic diagram of a fusion product generator provided in an embodiment of the present application;
FIG. 5 is a schematic diagram of a spatial arbiter according to an embodiment of the present application;
FIG. 6 is a schematic diagram of a time arbiter according to an embodiment of the present disclosure;
FIG. 7 shows a schematic structural diagram of a fusion device based on abnormal sea surface height data of a satellite altimeter and a tide station according to an embodiment of the present application;
fig. 8 shows a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, as provided in the accompanying drawings, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. Based on the embodiments of the present application, every other embodiment that a person skilled in the art would obtain without making any inventive effort is within the scope of protection of the present application.
According to research, the sea surface height abnormal data refer to the deviation of the sea surface relative to the average sea surface, is an important ocean power environment element, can reflect the three-dimensional structure of the ocean subsurface, and plays an important role in the aspects of ocean temperature salt profile reconstruction, ocean vortex identification, ocean numerical prediction and the like.
In order to ensure the spatial coverage of the sea surface height abnormal data, the prior art generally adopts traditional fusion modes such as space-time objective analysis, optimal interpolation, variation analysis and the like, however, the traditional fusion modes are often based on assumptions such as spatial uniformity, parameter isotropy and the like, so that the generated fusion data has the problems of weak spatial adaptability, incapability of accurately reflecting the spatial characteristics of the sea surface height abnormal data, high calculation cost and the like. Meanwhile, in the prior art, fusion mode is usually only used for fusing the abnormal sea surface height data observed by the multisource satellite altimeter, but the satellite altimeter can only acquire the data of the points below the satellite, so that the space coverage is low, the accuracy and quality of the observed data are obviously reduced due to the influence of land or islands in the offshore area, and further the fusion data are poor in offshore accuracy and even have data loss.
Based on the above, the embodiment of the application provides a fusion method of sea surface height abnormal data based on a satellite altimeter and a tide station, which can excavate space-time variation characteristics of continuous multi-day sea surface height abnormal observation data, realize fusion of the sea surface height abnormal observation data of the multi-source satellite altimeter and the tide station, and generate full-coverage high-precision sea surface height abnormal fusion data; the three-dimensional double-mask convolution operation utilizes the constraint of double masks and the automatic updating of the data coverage mask to carry out the fine reconstruction of the uncovered area of the observed data, and the sea coverage mask and the observation data of the tide station are combined to further ensure the fine reconstruction of the sea-surface height abnormal data at the sea-land juncture of the offshore area, so that the space coverage and the accuracy of the fusion data are improved.
Referring to fig. 1, fig. 1 is a flowchart of a method for fusing abnormal sea surface height data based on a satellite altimeter and a tide station according to an embodiment of the present application. As shown in fig. 1, the fusion method provided in the embodiment of the present application includes:
s101, satellite sea surface height anomaly observation data from T-m day to T day of a target area determined by a multi-source satellite altimeter, tide station sea surface height anomaly observation data from T day of the target area determined by a plurality of tide stations and a sea coverage mask of the target area are obtained.
Wherein T represents a certain date, and m is a positive integer; in the specific implementation, satellite sea surface height abnormal observation data of a certain target date (T day) and the previous day to the previous m days of the target date and sea surface height abnormal observation data of a tide stand of the target date can be selected for subsequent generation of sea surface height abnormal fusion data of the target date; the ocean coverage mask is used to characterize the coverage of the ocean in the target area. The ocean coverage mask may be expressed in a matrix form, and when a certain matrix element is 1, the ocean coverage mask indicates that the position corresponding to the matrix element in the target area is ocean; when a certain matrix element is 0, it indicates that the position corresponding to the matrix element in the target area is land.
In specific implementation, the embodiment of the application can select a range of 90-160 DEG E-5-55 DEG N as a target area, and can select SARAL, cryosat-2, HY-2A, HY-2B, jason-2, jason-3 and Sentinel-3A, sentinel-3B as a multi-source satellite altimeter. The data is derived from CMEMS (https:// marine.copernicus. Eu /), the space resolution along the orbit is 7km, and satellite sea surface height anomaly observation data can be directly obtained from the data.
However, the satellite altimeter can only observe the point under the satellite, the space coverage of the obtained satellite sea surface height anomaly observation data is low, for example, the day of 20 days of 9 months in 2022, the day space coverage of the satellite altimeter along the orbit sea surface height anomaly observation data of SARAL, cryosat-2, HY-2B, jason-3 and Sentinel-3A, sentinel-3B is 2.1%, 1.9%, 2.0%, 1.8% and 2.6%, and the day space coverage of the satellite altimeter combination is only 11.7%.
Considering that sea surface height observation data provided by the tide station has the advantages of high accuracy, strong continuity and the like, the embodiment of the application introduces the observation data of the tide station; meanwhile, the tide station is mainly distributed on the near shore or islands, so that the sea surface height abnormal observation data determined by the multisource satellite altimeter and the tide station are fused through a fusion processing technology, the advantages of the multisource satellite altimeter and the tide station are brought into play, and the sea surface height abnormal fusion product with full coverage and high space-time resolution is generated, so that the requirements of sea temperature salt profile reconstruction, sea vortex identification and sea numerical forecasting are met.
Referring to fig. 2, fig. 2 is a schematic view of a tide station distribution diagram of a target area according to an embodiment of the present application. As shown in fig. 2, 105 tide stations can be selected in the target area to obtain daily sea level altitude observation data, and the dots represent tide station positions.
In a possible implementation manner, acquiring the abnormal observation data of the sea level height of the tide station on the day T of the target area determined by the plurality of tide stations in step S101 may include: s1011, acquiring daily sea surface tide level observation data of the target area observed by each tide station, and extracting daily sea surface height observation data from the daily sea surface tide level observation data. S1012, determining a sea surface height average value based on the daily sea surface height observation data, and determining daily sea surface height abnormal observation data by subtracting the sea surface height average value from the daily sea surface height observation data. S1013, sea level air pressure correction and average sliding filtering processing are carried out on the daily sea level height abnormal observation data, and the sea level height abnormal observation data of the tide station on the T day is extracted from the processing result.
For the above steps S1011 to S1013, the daily sea surface level observation data of the target area observed by each tide station includes tidal signals and the like in addition to the daily sea surface height observation data. Therefore, after removing the tidal signal from the daily sea surface tide level observation data and extracting the daily sea surface height observation data, subtracting the average value of the sea surface heights from the daily sea surface height observation data to determine the daily sea surface height anomaly observation data, and performing sea level air pressure correction processing and average sliding filtering processing for a certain time length (for example, 19 days) on the daily sea surface height anomaly observation data to further reduce the influence of tides on the sea surface height anomaly observation data determined by the tide station.
S102, determining an initial data coverage mask according to the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data and the ocean coverage mask.
The data coverage mask is used for representing coverage conditions of effective observation data in the target area from T-m days to T days, wherein the effective observation data comprise observation data of a multi-source satellite altimeter and a tide station. The data coverage mask can be expressed in a matrix form, and when a certain matrix element is 1, the position corresponding to the matrix element in the data matrix is expressed as effective data; when a certain matrix element is 0, it indicates that the position corresponding to the matrix element in the data matrix is invalid data.
In the implementation, the satellite sea surface height abnormal observation data from the T-m day to the T day and the tide station sea surface height abnormal observation data from the T day can be simply overlapped and combined, for example, the tide station sea surface height abnormal observation data from the T day and the satellite sea surface height abnormal observation data from the T-m day to the T day are combined into the same three-dimensional matrix, so that the three-dimensional sea surface height abnormal observation data is obtained; the three-dimensional dimension of the three-dimensional sea surface height anomaly observation data is a two-dimensional space dimension corresponding to the target area and a time dimension corresponding to the time from T-m days to T days.
Then, determining an initial data coverage mask according to the three-dimensional sea surface height anomaly observation data and the ocean coverage mask; specifically, for a certain matrix element in the matrix of the initial data coverage mask, when the corresponding position of the matrix element in the target area is ocean (i.e. the matrix element in the corresponding position in the matrix of the ocean coverage mask is 1), and the observed data exists in the corresponding position of the three-dimensional sea surface height abnormal observed data, the value of the matrix element is 1; the value of the matrix element is 0 in the rest of the cases.
S103, inputting the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea surface height abnormal data fusion model.
The sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator.
As mentioned above, satellite altimeters can only observe the points under the satellite, and the space coverage is low; the tide station observation data has high accuracy and strong continuity, but is mainly distributed at the offshore-land junction. Therefore, according to the space-time distribution characteristics of satellite altimeter and tide station observation data, the embodiment of the application provides a sea surface height abnormal data fusion model based on three-dimensional double-mask convolution, wherein the fusion model is constructed based on a generation countermeasure network and comprises a fusion product generator based on three-dimensional double-mask convolution, a space discriminator and a time discriminator.
In a model training stage, the sea surface height abnormal data fusion model provided by the embodiment of the application is used for realizing fusion of satellite sea surface height abnormal observation data of a T-m-T day multisource satellite altimeter and sea surface height abnormal observation data of a T day tide station, and generating sea surface height abnormal fusion data of a T day. During training, satellite sea surface height abnormal observation data of a certain historical date (T day) and the previous day to the previous m days of the historical date and sea surface height abnormal observation data of a tide station on the historical date can be selected to be used as a training data set for model training; it should be noted that, the selected historical date during training and the target date during fusion may be different, but satellite sea surface height anomaly observation data m days before the corresponding date are required to be acquired for training or fusion. Wherein, the generator of the sea surface height anomaly data fusion model can be expressed as:
Wherein X is T Represents the abnormal fusion data of the sea surface height on the T day,satellite sea surface altitude anomaly observation data representing T, T-1, …, T-m days, < >>The sea surface height abnormality observation data of the tide station on the T day is represented by A T ,A T-1 ,…,A T-m The data coverage masks of the three-dimensional sea level height anomaly data obtained by combining T, T-1, … and T-m days are respectively represented, and S represents the sea coverage mask.
The discriminator is used for carrying out true and false discrimination on the generated T-day sea surface height abnormal fusion data, and the embodiment of the application adopts a double-discriminator structure, and comprises a space discriminator and a time discriminator, so that the rationality and the authenticity of the sea surface height abnormal fusion data are respectively ensured in the time dimension and the space dimension.
Referring to fig. 3, fig. 3 is a schematic diagram illustrating a training process of a sea level height anomaly data fusion model according to an embodiment of the present application. As shown in fig. 3, in one possible embodiment, the sea level height anomaly data fusion model is trained by:
step 1, acquiring sample satellite sea surface height anomaly observation data from T-m days to T days of the target area determined by a multi-source satellite altimeter, sample tide station sea surface height anomaly observation data from T days of the target area determined by a plurality of tide stations, a sea coverage mask of the target area, an initial sample data coverage mask from T-m days to T days and multi-source fusion sea surface height anomaly data of the target area.
It should be noted that, because the space coverage rate of the sea surface height anomaly observation data acquired by the single-day multisource satellite altimeter and the tide station is low, in order to realize the fusion of the multisource sea surface height anomaly data and generate a fully covered sea surface height anomaly fusion product, the time-space variation characteristics of the sea surface height anomaly observation data are mined by utilizing the sea surface height anomaly observation data observed by continuous multiple days, and the problems of data reconstruction of a large-scale uncovered area and data refinement reconstruction of a near-shore area are solved on the basis.
Here, the manner of acquiring the abnormal observation data of the sea level height of the sample satellite from T-m day to T day, the abnormal observation data of the sea level height of the sample tide station from T day, the ocean coverage mask and the sample data coverage mask from T-m day to T day may refer to the descriptions in S101 and S102, and the same technical effects may be achieved, and will not be described here again.
And for the multisource fusion sea surface height abnormal data of the target area, the multisource fusion sea surface height abnormal data refers to the existing sea surface height abnormal data product which can be directly obtained and fused in other modes during specific implementation. The multisource fusion sea surface height anomaly data obtained by the embodiment of the application are derived from AVISO (https:// www.aviso.altimetry.fr), the spatial resolution is 0.25 degrees, the time resolution is 24 hours, and the multisource fusion sea surface height anomaly data of SARAL, cryosat-2, HY-2A, HY-2B, jason-2, jason-3, sentinel-3A, sentinel-3B and other satellite altimeters are fused by adopting an optimal interpolation algorithm.
And 2, fusing the multisource fusion sea surface height abnormal data with the sample tide station sea surface height abnormal observation data by using a variation analysis method to obtain near shore correction sea surface height abnormal fusion data from the T-m day to the T day.
In the step, in order to improve the accuracy of multisource fusion sea surface height anomaly data on the near shore, in the embodiment of the application, an AVISO sea surface height anomaly fusion product and sample tide station sea surface height anomaly observation data are fused again through a variation analysis method, so that near shore correction sea surface height anomaly fusion data are generated and used as true values of a deep learning training set in a subsequent model training process.
And step 3, inputting the abnormal sea surface height observation data of the sample satellite, the abnormal sea surface height observation data of the sample tide station, the ocean coverage mask and the data coverage mask of the initial sample from the T-m day to the T day into a fusion product generator in an initial abnormal sea surface height data fusion model to obtain initial abnormal sea surface height fusion data of the target area on the T day.
In the step, three-dimensional double-mask convolution is established in the fusion product generator, and when the space-time change characteristics of sea surface height abnormal observation data of continuous multiple days are mined, the sea surface coverage mask and the data coverage mask are overlapped on satellite sea surface height abnormal observation data and tide station sea surface height abnormal observation data, so that each convolution calculation of a convolution kernel is only executed for effective observation data in the sea in a three-dimensional space range corresponding to the convolution kernel, the influence of data deletion on the convolution calculation is avoided, and fine reconstruction is realized. The data coverage mask is used for helping to realize data reconstruction of a large-range uncovered area, and the ocean coverage mask is used for helping to realize fine reconstruction of sea-surface height abnormal observation data at the offshore-land junction of an offshore area by combining with observation data of a tide station. After each three-dimensional double-mask convolution operation, the precise data reconstruction of the uncovered area is realized by fixing the ocean coverage mask and updating the data coverage mask, and meanwhile, the data reconstruction accuracy of the near-shore area is ensured.
Thus, the fusion product generator outputs initial sea surface height abnormal fusion data of T days after a series of mining analysis of space-time characteristics and data reconstruction of uncovered areas.
And 4, inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data of the target area T day into a spatial discriminator in the initial sea surface height abnormal data fusion model to obtain a spatial discrimination result of the spatial discriminator on the spatial characteristics of the initial sea surface height abnormal fusion data of the target area T day.
And 5, inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data from the T-m day to the T day into a time discriminator in the initial sea surface height abnormal data fusion model to obtain a time discrimination result of the time discriminator aiming at the time characteristic of the initial sea surface height abnormal fusion data of the target area T day.
Aiming at the step 4 and the step 5, the space discriminator discriminates true and false space discrimination results through the space feature difference between the initial sea surface height abnormal fusion data of the T day and the near shore correction sea surface height abnormal fusion data of the T day; the time discriminator composes the first time series data from the initial sea surface height abnormal fusion data of the T day and the near shore correction sea surface height abnormal fusion data from the T-m day to the T-1 day, takes the near shore correction sea surface height abnormal fusion data from the T-m day to the T day as the second time series data, and discriminates the true and false time discrimination result of the T day sea surface height abnormal fusion data generated by the fusion product generator through the time characteristic difference between the first time series data and the second time series data. And then, the feedback after the connection of the two is fed to a fusion product generator, so that the generator is further improved. The improved T-day sea surface height abnormal fusion data regenerated by the generator is more in line with the actual situation compared with the previous sea surface height abnormal fusion data, so that the double discriminators can further adjust discrimination capability. By repeating the process, the generator and the double discriminators are mutually opposed and finally reach balance, so that the generated T-day sea surface height abnormal fusion data can have better global consistency and time-space continuity.
And step 6, performing iterative training on the initial sea surface height abnormal data fusion model based on the initial sea surface height abnormal fusion data of the target area on the T day, correcting the sea surface height abnormal fusion data from the T-m day to the near shore on the T day, and performing iterative training on the initial sea surface height abnormal data fusion model according to the space discrimination result and the time discrimination result to obtain the trained sea surface height abnormal data fusion model.
In the step, the difference between the T-day sea surface height abnormal fusion product and the near shore correction sea surface height abnormal fusion data can be calculated by using a loss function based on the T-day initial sea surface height abnormal fusion data, the near shore correction sea surface height abnormal fusion data from the T-m day to the T day, the space discrimination result and the time discrimination result, and parameters of a fusion product generator can be adjusted according to the difference.
When the initial sea surface height abnormal data fusion model is subjected to iterative training, the loss function of the fusion product generator consists of a countermeasure loss and a content loss. Countering losses includes two parts: the countermeasures of the generator and the spatial discriminator, and the countermeasures of the generator and the temporal discriminator. Thus, the loss function of the fusion product generator is expressed as:
L G ＝L adv (G)+δL content
Wherein L is G Representing a loss function of the fusion product generator; l (L) adv (G) Representing a challenge loss function of the fusion product generator; l (L) content Representing a content loss function of the fusion product generator.
The fight loss function can be expressed as:
content loss is used to measure the difference between the initial sea level height anomaly fusion data on day T and the corrected sea level height anomaly fusion data on day T, where l is used 1 Distance, correcting the sea surface height abnormal fusion data by using the initial sea surface height abnormal fusion data of the T day and the near shore of the T dayIs minimized. Since the investigation region covers only the ocean, the ocean mask s is increased when the content loss is calculated. Notably, ground current causes sea level anomalies due to interactions between ground current and sea level, and the highly anomalous sea surface reacts to the flow field. Therefore, in order to ensure the accuracy of the flow field characteristics reflected by the abnormal data of the sea surface height and simultaneously ensure the accuracy of the abnormal data of the sea surface height under the action of the flow field characteristics, the embodiment of the application designs a flow field loss item L current To constrain the flow field characteristics. Content loss L content Can be expressed as:
L content ＝E[||L⊙(x f -x a )|| 1 ]+αL current
wherein L is content Representing a content loss function of the fusion product generator; l (L) current Representing a flow field loss term; delta is a constant, preferably 80 in the examples herein; g represents a fusion product generator; d (D) S Representing a spatial arbiter; d (D) T Representing a time discriminator; p (P) G The distribution of the sea surface height abnormal fusion data of the target area T day is represented, and X represents the sea surface height abnormal fusion data of the target area T day; x is X a Correcting the sea surface height abnormal fusion data on the near shore of the T day; α is a constant, preferably 4 in the examples herein; II … II 1 Representation l 1 A distance;a flow field characteristic representative of the target region, associated with ground diversion; u, v represent the east and north components, respectively.
The loss function of the spatial arbiter is expressed as:
wherein P is data1 Representing the distribution of the offshore correction sea surface height abnormal fusion data on the T day;
the loss function of the time arbiter is expressed as:
wherein P is data2 The distribution of the time series of the offshore correction sea surface height anomaly fusion data from day T-m to day T is shown.
Referring now to fig. 4, fig. 4 is a schematic structural diagram of a fusion product generator according to an embodiment of the present application. As shown in fig. 4, the fusion product generator includes an encoder and a decoder; the encoder comprises an input unit and a p-layer encoding layer, and the decoder comprises a p-layer decoding layer and an output unit; the coding layers and decoding layers with the same layer number are connected in a jumping way and form a U-shaped network structure so as to fuse the characteristic data output by the coding layers with the same layer number with the characteristic data output by the decoding layers; wherein p is a positive integer; illustratively, in FIG. 4 p is 5 and m is 4.
Each coding layer in the coder comprises a three-dimensional double-mask convolution module and a three-dimensional maximum pooling module; each decoding layer in the decoder comprises a three-dimensional double-mask convolution module and a three-dimensional up-sampling unit; the input unit in the encoder is connected with a first layer coding layer; the bottom coding layer is connected with a bottom decoding layer in the decoder through a three-dimensional double-mask convolution module; each three-dimensional double-mask convolution module includes a plurality of three-dimensional double-mask convolution units, a group normalization unit, and a ReLU activation function unit.
The encoder can decompose the three-dimensional sea surface height abnormal data, the data coverage mask and the sea coverage mask into data of different layers, extract the characteristics of the three-dimensional sea surface height abnormal data on different time-space scales and reconstruct the data of an unobserved area. Meanwhile, the data size processed by each layer is reduced, so that the training speed of the network and the processing speed of the model are increased.
The decoder is used for restoring the characteristics of each layer and outputting a sea surface height abnormal fusion product. The jump connection connects the encoder features and the decoder features of the same level, so that the shallow features on the encoder and the deep features on the decoder can be effectively fused, the fusion product generator can obtain better feature extraction capability and information fusion capability, and the loss generated during decoding is reduced.
Referring to fig. 5, fig. 5 is a schematic structural diagram of a spatial discriminator according to the embodiment of the application. The spatial discriminator provided by the embodiment of the application can discriminate the true or false of the generated T day sea surface height abnormal fusion data from the spatial characteristics. The sea surface height abnormal fusion data only covers the sea area, and in order to reduce the influence of land and islands and better distinguish the T day near shore corrected sea surface height abnormal fusion data from the T day sea surface height abnormal fusion data generated by the generator, a space discriminant based on two-dimensional mask convolution is built. By taking reference to the idea of three-dimensional double-mask convolution, a fixed ocean coverage mask is introduced into two-dimensional convolution, so that each convolution operation extracts the spatial distribution characteristics of the sea surface height abnormal fusion product, and the interference of land and islands is avoided. As shown in fig. 5, the spatial arbiter has a depth of 5 layers and includes a two-dimensional mask convolution unit and an output unit. The two-dimensional mask convolution unit consists of two-dimensional mask convolution, batch normalization and LeakyRelu activation functions, the convolution kernel size is 4 multiplied by 4, the step size is 2, the data dimension is reduced, the receptive field is increased, and the network captures a larger range of spatial features. The output unit is slightly different from the two-dimensional mask convolution unit, and the adopted Sigmoid activates the function. Unlike the original arbiter for generating the countermeasure network, the arbiter divides the input data into 14×16 data blocks, discriminates all the data blocks except the land (including the data blocks at the sea-land junction and the data blocks at the sea position), and finally outputs true or false according to the discrimination result average value of the blocks.
Referring to fig. 6, fig. 6 is a schematic structural diagram of a time discriminator according to the embodiment of the application. The time discriminator provided by the embodiment of the application can discriminate the true or false of a time sequence formed by the T-day sea surface height abnormal fusion data generated by the generator and the T-m-T-1 day near shore correction sea surface height abnormal fusion data from the time characteristic. As shown in fig. 7, the temporal arbiter includes a spatial dimension reduction unit, a two-dimensional/three-dimensional convolution residual block, an averaging pooling unit, and an output unit. The space dimension reduction unit rearranges the space data dimension to the depth dimension, so as to realize downsampling and improve the operation efficiency. The two-dimensional/three-dimensional convolution residual block is used for realizing the time feature extraction of the T-m-T daily sea surface height abnormal fusion data, and the gradient vanishing problem caused by increasing the network depth is relieved through the residual block. The output unit consists of a full connection layer and a ReLU activation function, and outputs a continuous value representing the probability of approaching a real time sequence of fused data.
Referring back to fig. 1, S104, using the ocean coverage mask and the initial data coverage mask, the fusion product generator performs a plurality of three-dimensional double-mask convolution operations on the satellite sea surface height anomaly observation data and the tide station sea surface height anomaly observation data to obtain sea surface height anomaly fusion data of the target area T day.
It is worth noting that the traditional three-dimensional convolution can carry out indiscriminate convolution on observed data and missing data of an uncovered area, so that the mined space-time variation characteristics of the three-dimensional convolution are severely interfered by the missing data, serious distortion exists when the missing data of the uncovered area is reconstructed, and the problems of artifacts and the like exist in a near-shore area. Therefore, the embodiment of the application provides three-dimensional double-mask convolution operation, by superposing the sea coverage mask and the data coverage mask on the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data, the shape of the convolution kernel is changed according to the coverage condition of the effective observation data in the sea, so that each convolution calculation of the convolution kernel is only executed for the effective observation data in the sea in the three-dimensional space range corresponding to the convolution kernel, the influence of data deletion on the convolution calculation and the data reconstruction is avoided, and the data reconstruction is further ensured, in particular, the accuracy of the data reconstruction at the sea-land junction of the offshore area is ensured.
In one possible implementation, S104 may include: s1041, merging the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data by the fusion product generator to obtain three-dimensional sea surface height abnormal observation data; the three-dimensional dimension of the three-dimensional sea surface height anomaly observation data is a two-dimensional space dimension and a time dimension; s1042, using the ocean coverage mask and the initial data coverage mask, performing a plurality of three-dimensional double-mask convolution operations on the three-dimensional sea surface height anomaly data by the fusion product generator; wherein, for the data of each position point (i, j, k) in the three-dimensional sea surface height abnormal observation data, the formula of convolution calculation of the position point by any one three-dimensional double-mask convolution operation is expressed as follows:
Where, represents multiplication; the "" indicates a matrix dot product; u represents the convolution kernel width of the three-dimensional double-mask convolution operation, and V represents the convolution kernel height of the three-dimensional double-mask convolution operation; t represents the convolution kernel depth of the three-dimensional double-mask convolution operation; w represents a convolution kernel weight; w (u, v, t) represents the weight value of any point in the convolution kernel; x represents the three-dimensional sea surface height anomaly observation data; a represents the data coverage mask; s represents the marine coverage mask; b represents bias;is a scale factor used for representing the proportion of the convolution kernel in the three-dimensional sea surface height anomaly observation data, which corresponds to the effective observation data in the sea in the three-dimensional space range, when the convolution calculation is carried out on the position point by the three-dimensional double-mask convolution operation.
After each double-mask three-dimensional convolution operation is executed, based on the ocean coverage mask and the data coverage mask used by the three-dimensional double-mask convolution operation, updating to obtain a corresponding mask element value of each position point (i, j, k) (corresponding to the center point of the current convolution kernel) in the data coverage mask to be used by the next three-dimensional double-mask convolution operation through the following formula:
the above shows that when effective data exists in the ocean within the three-dimensional space range and the current position point (i, j, k) is in the ocean area in each convolution operation, the mask element value in the data coverage mask is updated to be 1, and then the convolution result of the current position point (i, j, k) can be calculated according to the formula in the next three-dimensional double-mask convolution operation; when the three-dimensional double-mask convolution check is supposed to be that effective data does not exist in the ocean in the three-dimensional space range or the current position point (i, j, k) is in the land area, the mask element value of the current position point (i, j, k) is 0, and the convolution result in the next three-dimensional double-mask convolution operation is 0. Thus, through multiple three-dimensional double-mask convolution operations, the number of 0 s in the data coverage mask of the ocean region gradually decreases, i.e., the data missing portion gradually becomes smaller, and finally, the data coverage mask and the ocean coverage mask tend to be consistent.
In an experiment, an experiment hardware platform CPU is Intel Xeon Silver 4212R, GPU NVIDIA Tesla V100s, and a CUDA parallel framework and a cuDNN acceleration library are configured. The deep learning framework used for training was PyTorch, batch size 16, epoch 2000 times. ADAM optimizers are used for generating the countermeasure network by the fusion product, the learning rate of the generated model is 0.0002, the learning rate of the discrimination model is 0.0002, and the two models are optimized alternately.
The time of training data set is from 2016 1 month 1 day to 2021 12 month 31 day, the time of verifying data set is from 2022 1 month 1 day to 2022 6 month 30 day, and the time of testing data set is from 2022 7 month 1 day to 2022 12 month 31 day. The experiments used Root Mean Square Error (RMSE) and Bias (Bias) to evaluate the effect of sea level height anomaly fusion products. The fusion data sources are SARAL, cryosat-2, HY-2B, jason-3 and Sentinel-3A, sentinel-3B satellite altimeter sea surface height abnormal data and tide station data, and the test data sources are sea surface height abnormal data which do not participate in fusion of the HY-2C satellite altimeter.
The experimental results show that: the spatial coverage of the fusion data produced by the fusion method provided by the embodiment of the application can reach 100%, which shows that the sea surface height abnormal data fusion model based on three-dimensional double-mask convolution can realize the full coverage of the multi-source sea surface height abnormal fusion product, solves the reconstruction problem of uncovered area data, fully considers the contribution of tide station data especially in the coastal area, and achieves better effect.
Furthermore, in order to verify the advancement of the fusion method and the precision of the generated fusion data, the fusion data generated in the embodiment of the application are checked and analyzed by utilizing the abnormal sea surface height data of the HY-2C satellite altimeter which does not participate in fusion in the period from 7 months 1 to 12 months 31 years 2022, and a comparison experiment is carried out on the fusion data and the precision and the running time of the abnormal fusion data of the optimal interpolation sea surface height and the three-dimensional convolution sea surface height.
Experimental results show that firstly, from the aspect of program execution efficiency, the generation time of a three-dimensional convolution and three-dimensional double-mask convolution sea surface height anomaly fusion product on a single day is 0.03 seconds, the generation time of an optimal interpolation fusion product is 427 seconds, and the sea surface height anomaly fusion model adopting deep learning has obvious advantages compared with the traditional optimal interpolation.
Secondly, from the aspect of the spatial coverage of the fusion data, the optimal interpolation and the fusion data generated by the embodiment of the application realize the full coverage of a research area, the number of matched data points is 114241, the three-dimensional convolution sea surface height abnormal fusion data cannot realize the full coverage, and particularly, the coverage of a coastal area is lower, and the number of the matched data points is 91020.
From the aspect of deviation, the deviation of the optimal interpolation, the three-dimensional convolution and the fusion data generated by the embodiment of the application is slightly smaller than the sea surface height abnormal data of the HY-2C satellite altimeter, wherein the absolute deviation of the fusion data generated by the embodiment of the application is minimum and is 0.027 meter.
From the aspect of matching data distribution, the data of the three-dimensional convolution sea surface height abnormal fusion product are most dispersed, and the root mean square error reaches 0.191 m, which indicates that the space-time characteristics of the sparse sea surface height abnormal data are difficult to be extracted well by adopting a convolution network, and the reconstruction of uncovered areas is realized; the fusion data generated by the embodiment of the application have better consistency with the sea surface height abnormal data of the HY-2C satellite altimeter, wherein the root mean square error of the fusion data generated by the embodiment of the application is slightly smaller and is 0.158 meter.
In sum, the root mean square error, the absolute deviation and the daily running time of the sea surface height abnormal fusion data generated based on the three-dimensional double-mask convolution are respectively 0.158 meter, 0.027 meter and 0.03 second, and the root mean square error and the deviation are smaller than the optimal interpolation and the three-dimensional convolution sea surface height abnormal fusion data, so that the sea surface height abnormal data fusion model based on the three-dimensional double-mask convolution is superior to the reference model based on the three-dimensional convolution, and the fusion data has higher precision; and meanwhile, the daily running time is obviously superior to that of the optimal interpolation sea surface height anomaly fusion data, which shows that the calculation cost is obviously superior to that of the traditional algorithm.
According to the sea surface height abnormal data fusion method based on the satellite altimeter and the tide station, through automatic updating of the data coverage mask and constraint of the ocean coverage mask in three-dimensional double-mask convolution operation, the sea surface height abnormal data of an uncovered observation area, especially a coastal sea area, are subjected to fine reconstruction by combining the tide station observation data, so that the sea surface height abnormal data fusion of the multi-source satellite altimeter and the tide station is realized, and the space coverage and the precision of finally generated sea surface height abnormal fusion data can be improved.
Furthermore, the full-coverage and high-precision sea surface height abnormal fusion data has important significance for sea temperature salt profile reconstruction, sea vortex identification and sea numerical prediction. By utilizing fully-covered and high-precision sea surface height abnormal fusion data, the three-dimensional structure field of the ocean subsurface layer can be mastered, the space-time distribution characteristics of ocean jump layers, the three-dimensional structure and the motion characteristics of ocean vortex are analyzed, and important data support is provided for underwater weapon equipment application.
Referring to fig. 7, fig. 7 is a schematic structural diagram of a fusion device for sea level height anomaly data based on a satellite altimeter and a tide station according to an embodiment of the present application. As shown in fig. 7, the fusion device 700 includes:
An acquisition module 710, configured to acquire satellite sea surface height anomaly observation data from T-m days to T days of a target area determined by a multi-source satellite altimeter, tide station sea surface height anomaly observation data from T days of the target area determined by a plurality of tide stations, and a sea coverage mask of the target area; wherein the marine coverage mask is used to characterize coverage of the ocean in the target area;
a determining module 720, configured to determine an initial data coverage mask according to the satellite sea level height anomaly observation data, the tide station sea level height anomaly observation data, and the sea coverage mask; the data coverage mask is used for representing coverage conditions of effective observation data in the target area from the T-m day to the T day;
the input module 730 is configured to input the satellite sea level height anomaly observation data, the tide station sea level height anomaly observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea level height anomaly data fusion model; the sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator;
A convolution module 740, configured to perform multiple three-dimensional double-mask convolution operations on the satellite sea level height anomaly observation data and the tide station sea level height anomaly observation data by using the ocean coverage mask and the initial data coverage mask by using the fusion product generator, so as to obtain sea level height anomaly fusion data of the target area T day; the three-dimensional double-mask convolution operation is performed only on effective observed data in the ocean in a three-dimensional space range corresponding to the convolution kernel by superposing the ocean coverage mask and the data coverage mask on the satellite sea surface height abnormal observed data and the tide station sea surface height abnormal observed data, so that the influence of data deletion on the convolution computation is avoided;
after each three-dimensional double-mask convolution operation is executed, the data coverage mask which is used in the next three-dimensional double-mask convolution operation is updated based on the ocean coverage mask and the data coverage mask used in the three-dimensional double-mask convolution operation.
Further, when the acquiring module 710 is configured to acquire abnormal observation data of sea level height of the tide station on the day T of the target area determined by a plurality of tide stations, the acquiring module 710 is configured to:
Acquiring daily sea surface tide level observation data of the target area observed by each tide station, and extracting daily sea surface height observation data from the daily sea surface tide level observation data;
determining a sea surface height average value based on the daily sea surface height observation data, and determining daily sea surface height abnormal observation data by subtracting the sea surface height average value from the daily sea surface height observation data;
and carrying out sea level air pressure correction and average sliding filtering treatment on the daily sea level height abnormal observation data, and extracting the sea level height abnormal observation data of the tide station on the T day from the treatment result.
Further, when the convolution module 740 is configured to perform, by the fusion product generator, a plurality of three-dimensional double-mask convolution operations on the satellite sea level height anomaly observation data and the tide station sea level height anomaly observation data using the ocean coverage mask and the initial data coverage mask, to obtain sea level height anomaly fusion data for the target region T day, the convolution module 740 is configured to:
combining the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data by the fusion product generator to obtain three-dimensional sea surface height abnormal observation data; the three-dimensional dimension of the three-dimensional sea surface height anomaly observation data is a two-dimensional space dimension and a time dimension;
Performing, by the fusion product generator, a plurality of three-dimensional double-mask convolution operations on the three-dimensional sea level height anomaly data using the sea coverage mask and the initial data coverage mask; wherein, for the data of each position point (i, j, k) in the three-dimensional sea surface height abnormal observation data, the formula of convolution calculation of the position point by any one three-dimensional double-mask convolution operation is expressed as follows:
where, represents multiplication; the "" indicates a matrix dot product; u represents the convolution kernel width of the three-dimensional double-mask convolution operation, and V represents the convolution kernel height of the three-dimensional double-mask convolution operation; t represents the convolution kernel depth of the three-dimensional double-mask convolution operation; w represents a convolution kernel weight; w (u, v, t) represents the weight value of any point in the convolution kernel; x represents the three-dimensional sea surface height anomaly observation data; a represents the data coverage mask; s represents the marine coverage mask; b represents bias;is a scale factor used for representing the proportion of the convolution kernel in the three-dimensional sea surface height anomaly observation data, which corresponds to the effective observation data in the sea in the three-dimensional space range, when the convolution calculation is carried out on the position point by the three-dimensional double-mask convolution operation.
Further, after each three-dimensional double-mask convolution operation is performed, the convolution module 740 updates, based on the ocean coverage mask and the data coverage mask used by the three-dimensional double-mask convolution operation, the corresponding mask element value of each position point (i, j, k) in the data coverage mask to be used by the next three-dimensional double-mask convolution operation by using the following formula:
further, the fusion device further comprises a training module; the training module is used for training to obtain the sea surface height abnormal data fusion model by the following modes:
acquiring sample satellite sea surface height anomaly observation data from T-m day to T day of the target area determined by a multi-source satellite altimeter, sample tide station sea surface height anomaly observation data from T day of the target area determined by a plurality of tide stations, a sea coverage mask of the target area, an initial sample data coverage mask from T-m day to T day and multi-source fusion sea surface height anomaly data of the target area;
fusion is carried out on the multisource fusion sea surface height abnormal data and the sample tide station sea surface height abnormal observation data by using a variation analysis method, so that near shore correction sea surface height abnormal fusion data from T-m day to T day are obtained; wherein the offshore correction sea surface height anomaly fusion data is used as a true value in a model training process;
Inputting the abnormal sea surface height observation data of the sample satellite, the abnormal sea surface height observation data of the sample tide station, the ocean coverage mask and the data coverage mask of the initial sample from the T-m day to the T day into a fusion product generator in an initial abnormal sea surface height data fusion model to obtain abnormal sea surface height fusion data of the target area from the T day;
inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data of the target area T day into a spatial discriminator in the initial sea surface height abnormal data fusion model to obtain a spatial discrimination result of the spatial discriminator on the spatial characteristics of the initial sea surface height abnormal fusion data of the target area T day;
inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data from the T-m day to the T day into a time discriminator in the initial sea surface height abnormal data fusion model to obtain a time discrimination result of the time discriminator aiming at the time characteristic of the initial sea surface height abnormal fusion data of the target area T day;
and carrying out iterative training on the initial sea surface height abnormal data fusion model based on the initial sea surface height abnormal fusion data of the target area on the T day, the near shore correction sea surface height abnormal fusion data from the T-m day to the T day, the space discrimination result and the time discrimination result to obtain the trained sea surface height abnormal data fusion model.
When the initial sea surface height abnormal data fusion model is iteratively trained based on the initial sea surface height abnormal fusion data of the target area on the T day, the near shore correction sea surface height abnormal fusion data from the T-m day to the T day, the spatial discrimination result and the time discrimination result, the loss function of the fusion product generator is expressed as:
L G ＝L adv (G)+δL content
L content ＝E[‖S⊙(X-X a )‖ 1 ]+αL current
wherein L is G Representing a loss function of the fusion product generator; l (L) adv (G) Representing a fight loss function of the fusion product generator; l (L) content Representing a content loss function of the fusion product generator; l (L) current Representing a flow field loss term; delta is a constant; g represents a fusion product generator; d (D) S Representing a spatial arbiter; d (D) T Representing a time discriminator; p (P) G A distribution of sea surface height anomaly fusion data representing the target region T day, X representing the target region T daySea surface height anomaly fusion data; x is X a Correcting the sea surface height abnormal fusion data on the near shore of the T day; alpha is a constant; II … II 1 Representation l 1 A distance;a flow field characteristic representative of the target region, associated with ground diversion; u and v represent the east and north components, respectively;
the loss function of the spatial arbiter is expressed as:
wherein P is data1 Representing the distribution of the offshore correction sea surface height abnormal fusion data on the T day;
The loss function of the time arbiter is expressed as:
wherein P is data2 The distribution of the time series of the offshore correction sea surface height anomaly fusion data from day T-m to day T is shown.
Further, the fusion product generator includes an encoder and a decoder; the encoder comprises an input unit and a p-layer encoding layer, and the decoder comprises a p-layer decoding layer and an output unit; the coding layers and decoding layers with the same layer number are connected in a jumping way and form a U-shaped network structure so as to fuse the characteristic data output by the coding layers with the same layer number with the characteristic data output by the decoding layers; wherein p is a positive integer;
each coding layer in the coder comprises a three-dimensional double-mask convolution module and a three-dimensional maximum pooling module; each decoding layer in the decoder comprises a three-dimensional double-mask convolution module and a three-dimensional up-sampling unit; the input unit in the encoder is connected with a first layer coding layer; the bottom coding layer is connected with a bottom decoding layer in the decoder through a three-dimensional double-mask convolution module; each three-dimensional double-mask convolution module includes a plurality of three-dimensional double-mask convolution units, a group normalization unit, and a ReLU activation function unit.
Referring to fig. 8, fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As shown in fig. 8, the electronic device 800 includes a processor 810, a memory 820, and a bus 830.
The memory 820 stores machine-readable instructions executable by the processor 810, when the electronic device 800 is running, the processor 810 communicates with the memory 820 through the bus 830, and when the machine-readable instructions are executed by the processor 810, the steps of the fusion method in the method embodiment shown in fig. 1 can be executed, and the specific implementation can be referred to the method embodiment and will not be described herein.
The embodiment of the present application further provides a computer readable storage medium, where a computer program is stored on the computer readable storage medium, and when the computer program is executed by a processor, the steps of the fusion method in the embodiment of the method shown in fig. 1 may be executed, and a specific implementation manner may refer to the embodiment of the method and will not be repeated herein.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Finally, it should be noted that: the foregoing examples are merely specific embodiments of the present application, and are not intended to limit the scope of the present application, but the present application is not limited thereto, and those skilled in the art will appreciate that while the foregoing examples are described in detail, the present application is not limited thereto. Any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or make equivalent substitutions for some of the technical features within the technical scope of the disclosure of the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application, and are intended to be included in the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A fusion method of sea surface height anomaly data based on a satellite altimeter and a tide station, the fusion method comprising:
acquiring satellite sea surface height anomaly observation data from T-m days to T days of a target area determined by a multi-source satellite altimeter, tide station sea surface height anomaly observation data from T days of the target area determined by a plurality of tide stations and a sea coverage mask of the target area; wherein the marine coverage mask is used to characterize coverage of the ocean in the target area;
Determining an initial data coverage mask according to the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data and the ocean coverage mask; the data coverage mask is used for representing coverage conditions of effective observation data in the target area from the T-m day to the T day;
inputting the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea surface height abnormal data fusion model; the sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator;
using the ocean coverage mask and the initial data coverage mask, performing multiple three-dimensional double-mask convolution operations on the satellite sea surface height anomaly observation data and the tide station sea surface height anomaly observation data by the fusion product generator to obtain sea surface height anomaly fusion data of the target region T day; the three-dimensional double-mask convolution operation is performed only on effective observed data in the ocean in a three-dimensional space range corresponding to the convolution kernel by superposing the ocean coverage mask and the data coverage mask on the satellite sea surface height abnormal observed data and the tide station sea surface height abnormal observed data, so that the influence of data deletion on the convolution computation is avoided;
After each three-dimensional double-mask convolution operation is executed, the data coverage mask which is used in the next three-dimensional double-mask convolution operation is updated based on the ocean coverage mask and the data coverage mask used in the three-dimensional double-mask convolution operation.
2. The fusion method of claim 1, wherein obtaining tide station sea level anomaly observations for the target zone T day determined by a plurality of tide stations comprises:
acquiring daily sea surface tide level observation data of the target area observed by each tide station, and extracting daily sea surface height observation data from the daily sea surface tide level observation data;
determining a sea surface height average value based on the daily sea surface height observation data, and determining daily sea surface height abnormal observation data by subtracting the sea surface height average value from the daily sea surface height observation data;
and carrying out sea level air pressure correction and average sliding filtering treatment on the daily sea level height abnormal observation data, and extracting the sea level height abnormal observation data of the tide station on the T day from the treatment result.
3. The fusion method of claim 1, wherein performing, by the fusion product generator, a plurality of three-dimensional double-mask convolution operations on the satellite sea level anomaly observation data and the tide station sea level anomaly observation data using the sea coverage mask and the initial data coverage mask to obtain sea level anomaly fusion data for the target region T day comprises:
Combining the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data by the fusion product generator to obtain three-dimensional sea surface height abnormal observation data; the three-dimensional dimension of the three-dimensional sea surface height anomaly observation data is a two-dimensional space dimension and a time dimension;
performing, by the fusion product generator, a plurality of three-dimensional double-mask convolution operations on the three-dimensional sea level height anomaly data using the sea coverage mask and the initial data coverage mask; wherein, for the data of each position point (i, j, k) in the three-dimensional sea surface height abnormal observation data, the formula of convolution calculation of the position point by any one three-dimensional double-mask convolution operation is expressed as follows:
where, represents multiplication; the "" indicates a matrix dot product; u represents the convolution kernel width of the three-dimensional double-mask convolution operation, and V represents the convolution kernel height of the three-dimensional double-mask convolution operation; t represents the convolution kernel depth of the three-dimensional double-mask convolution operation; w represents a convolution kernel weight; w (u, v, t) represents the weight value of any point in the convolution kernel; x represents the three-dimensional sea surface height anomaly observation data; a represents the data coverage mask; s represents the marine coverage mask; representing the bias; Is a scale factor used for representing the proportion of the convolution kernel in the three-dimensional sea surface height anomaly observation data, which corresponds to the effective observation data in the sea in the three-dimensional space range, when the convolution calculation is carried out on the position point by the three-dimensional double-mask convolution operation.
4. A fusion method according to claim 3, characterized in that after each three-dimensional double-mask convolution operation is performed, based on the marine coverage mask and the data coverage mask used by the three-dimensional double-mask convolution operation, the corresponding mask element values in the data coverage mask to be used by each position point (i, j, k) in the next three-dimensional double-mask convolution operation are updated by the following formula:
5. the fusion method of claim 1, wherein the sea level height anomaly data fusion model is trained by:
acquiring sample satellite sea surface height anomaly observation data from T-m day to T day of the target area determined by a multi-source satellite altimeter, sample tide station sea surface height anomaly observation data from T day of the target area determined by a plurality of tide stations, a sea coverage mask of the target area, an initial sample data coverage mask from T-m day to T day and multi-source fusion sea surface height anomaly data of the target area;
Fusion is carried out on the multisource fusion sea surface height abnormal data and the sample tide station sea surface height abnormal observation data by using a variation analysis method, so that near shore correction sea surface height abnormal fusion data from T-m day to T day are obtained; wherein the offshore correction sea surface height anomaly fusion data is used as a true value in a model training process;
inputting the abnormal sea surface height observation data of the sample satellite, the abnormal sea surface height observation data of the sample tide station, the ocean coverage mask and the data coverage mask of the initial sample from the T-m day to the T day into a fusion product generator in an initial abnormal sea surface height data fusion model to obtain abnormal sea surface height fusion data of the target area from the T day;
inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data of the target area T day into a spatial discriminator in the initial sea surface height abnormal data fusion model to obtain a spatial discrimination result of the spatial discriminator on the spatial characteristics of the initial sea surface height abnormal fusion data of the target area T day;
inputting the initial sea surface height abnormal fusion data of the target area T day and the near shore correction sea surface height abnormal fusion data from the T-m day to the T day into a time discriminator in the initial sea surface height abnormal data fusion model to obtain a time discrimination result of the time discriminator aiming at the time characteristic of the initial sea surface height abnormal fusion data of the target area T day;
And carrying out iterative training on the initial sea surface height abnormal data fusion model based on the initial sea surface height abnormal fusion data of the target area on the T day, the near shore correction sea surface height abnormal fusion data from the T-m day to the T day, the space discrimination result and the time discrimination result to obtain the trained sea surface height abnormal data fusion model.
6. The fusion method according to claim 5, wherein when the initial sea level height anomaly data fusion model is iteratively trained based on the target region T-day initial sea level height anomaly fusion data, the near shore correction sea level height anomaly fusion data from T-m day to T-day, the spatial discrimination result, and the temporal discrimination result, a loss function of the fusion product generator is expressed as:
L e ＝L adv (G)+δL content
L content ＝E[||s⊙(X-X a )|| 1 ]+αL current
wherein L is G Representing a loss function of the fusion product generator; l (L) adv (G) Representing a fight loss function of the fusion product generator; l (L) content Representing a content loss function of the fusion product generator; l (L) current Representing a flow field loss term; delta is a constant; g represents a fusion product generator; d (D) S Representing a spatial arbiter; d (D) T Representing a time discriminator; G the distribution of the sea surface height abnormal fusion data of the target area T day is represented, and the sea surface height abnormal fusion data of the target area T day is represented; x is X a Correcting the sea surface height abnormal fusion data on the near shore of the T day; alpha is a constant; II … II 1 Representation l 1 A distance;a flow field characteristic representative of the target region, associated with ground diversion; v represents the east and north components, respectively;
the loss function of the spatial arbiter is expressed as:
wherein P is data1 Representing the distribution of the offshore correction sea surface height abnormal fusion data on the T day;
the loss function of the time arbiter is expressed as:
in the method, in the process of the invention, data2 the distribution of the time series of the offshore correction sea surface height anomaly fusion data from day T-m to day T is shown.
7. The method of claim 1, wherein the fusion product generator comprises an encoder and a decoder; the encoder comprises an input unit and a p-layer encoding layer, and the decoder comprises a p-layer decoding layer and an output unit; the coding layers and decoding layers with the same layer number are connected in a jumping way and form a U-shaped network structure so as to fuse the characteristic data output by the coding layers with the same layer number with the characteristic data output by the decoding layers; wherein p is a positive integer;
each coding layer in the coder comprises a three-dimensional double-mask convolution module and a three-dimensional maximum pooling module; each decoding layer in the decoder comprises a three-dimensional double-mask convolution module and a three-dimensional up-sampling unit; the input unit in the encoder is connected with a first layer coding layer; the bottom coding layer is connected with a bottom decoding layer in the decoder through a three-dimensional double-mask convolution module; each three-dimensional double-mask convolution module includes a plurality of three-dimensional double-mask convolution units, a group normalization unit, and a ReLU activation function unit.
8. A fusion device for sea surface height anomaly data based on a satellite altimeter and a tide station, the fusion device comprising:
the acquisition module is used for acquiring satellite sea surface height abnormal observation data from the T-m day to the T day of the target area determined by the multi-source satellite altimeter, tide station sea surface height abnormal observation data from the T day of the target area determined by the plurality of tide stations and a sea coverage mask of the target area; wherein the marine coverage mask is used to characterize coverage of the ocean in the target area;
the determining module is used for determining an initial data coverage mask according to the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data and the ocean coverage mask; the data coverage mask is used for representing coverage conditions of effective observation data in the target area from the T-m day to the T day;
the input module is used for inputting the satellite sea surface height abnormal observation data, the tide station sea surface height abnormal observation data, the ocean coverage mask and the initial data coverage mask into a fusion product generator in a pre-trained sea surface height abnormal data fusion model; the sea surface height abnormal data fusion model comprises a fusion product generator, a space discriminator and a time discriminator; the sea surface height abnormal data fusion model is obtained by mutual countermeasure training of the fusion product generator and the space discriminator and mutual countermeasure training of the fusion product generator and the time discriminator;
The convolution module is used for performing multiple three-dimensional double-mask convolution operations on the satellite sea surface height abnormal observation data and the tide station sea surface height abnormal observation data by using the ocean coverage mask and the initial data coverage mask by the fusion product generator so as to obtain sea surface height abnormal fusion data of the target area T day; the three-dimensional double-mask convolution operation is performed only on effective observed data in the ocean in a three-dimensional space range corresponding to the convolution kernel by superposing the ocean coverage mask and the data coverage mask on the satellite sea surface height abnormal observed data and the tide station sea surface height abnormal observed data, so that the influence of data deletion on the convolution computation is avoided;
after each three-dimensional double-mask convolution operation is executed, the data coverage mask which is used in the next three-dimensional double-mask convolution operation is updated based on the ocean coverage mask and the data coverage mask used in the three-dimensional double-mask convolution operation.
9. An electronic device, comprising: a processor, a memory and a bus, said memory storing machine readable instructions executable by said processor, said processor and said memory communicating via said bus when the electronic device is operating, said machine readable instructions when executed by said processor performing the steps of the satellite altimeter and tide station based sea surface altitude anomaly data fusion method according to any one of claims 1 to 7.
10. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, which when executed by a processor performs the steps of the fusion method of satellite altimeter and tide station based sea surface altitude anomaly data according to any one of claims 1 to 7.
CN202310499560.5A 2023-05-05 2023-05-05 Sea surface height abnormal data fusion method based on satellite altimeter and tide station Active CN117009913B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310499560.5A CN117009913B (en) 2023-05-05 2023-05-05 Sea surface height abnormal data fusion method based on satellite altimeter and tide station

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310499560.5A CN117009913B (en) 2023-05-05 2023-05-05 Sea surface height abnormal data fusion method based on satellite altimeter and tide station

Publications (2)

Publication Number Publication Date
CN117009913A CN117009913A (en) 2023-11-07
CN117009913B true CN117009913B (en) 2024-01-30

Family

ID=88560781

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310499560.5A Active CN117009913B (en) 2023-05-05 2023-05-05 Sea surface height abnormal data fusion method based on satellite altimeter and tide station

Country Status (1)

Country Link
CN (1) CN117009913B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114494811A (en) * 2022-02-07 2022-05-13 国家海洋环境预报中心 Method and system for fusing abnormal height data of satellite along sea level
WO2022126480A1 (en) * 2020-12-17 2022-06-23 深圳先进技术研究院 High-energy image synthesis method and device based on wasserstein generative adversarial network model
CN115345238A (en) * 2022-08-17 2022-11-15 中国人民解放军61741部队 Method and device for generating seawater transparency fusion data
CN115393540A (en) * 2022-08-30 2022-11-25 无锡九方科技有限公司 Intelligent fusion method and system of three-dimensional marine environment field based on deep learning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022126480A1 (en) * 2020-12-17 2022-06-23 深圳先进技术研究院 High-energy image synthesis method and device based on wasserstein generative adversarial network model
CN114494811A (en) * 2022-02-07 2022-05-13 国家海洋环境预报中心 Method and system for fusing abnormal height data of satellite along sea level
CN115345238A (en) * 2022-08-17 2022-11-15 中国人民解放军61741部队 Method and device for generating seawater transparency fusion data
CN115393540A (en) * 2022-08-30 2022-11-25 无锡九方科技有限公司 Intelligent fusion method and system of three-dimensional marine environment field based on deep learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于GNSS-R的海面风速探测技术研究;周旋 等;电子与信息学报(第07期);全文 *
联合卫星重力、卫星测高和海洋资料研究中国南海海平面变化;冯伟 等;中国科学:地球科学;全文 *

Also Published As

Publication number Publication date
CN117009913A (en) 2023-11-07

Similar Documents

Publication Publication Date Title
Sandwell et al. Global marine gravity from retracked Geosat and ERS‐1 altimetry: Ridge segmentation versus spreading rate
US10436924B2 (en) Denoising seismic data
CN106814378A (en) A kind of GNSS location time series cyclophysis method for digging
Chen et al. ARU-net: Reduction of atmospheric phase screen in SAR interferometry using attention-based deep residual U-net
CN115345238B (en) Method and device for generating seawater transparency fusion data
Zhang et al. Satellite derived bathymetry based on ICESat-2 diffuse attenuation signal without prior information
CN115980745A (en) Underground mining subsidence area intelligent identification method and device, electronic equipment and storage medium
Ferrando et al. Coastal modification in relation to sea storm effects: Application of 3D remote sensing survey in Sanremo Marina (Liguria, NW Italy)
Karamvasis et al. Performance analysis of open source time series InSAR methods for deformation monitoring over a broader mining region
Zeiger et al. SNR-based water height retrieval in rivers: Application to high amplitude asymmetric tides in the Garonne river
CN117009913B (en) Sea surface height abnormal data fusion method based on satellite altimeter and tide station
Gruszczynski et al. Orthogonal transformation in extracting of common mode errors from continuous GPS networks
Tang et al. Changes of Chinese coastal regions induced by land reclamation as revealed through TanDEM-X DEM and InSAR analyses
Khalili et al. Mapping of mean deformation rates based on APS-corrected InSAR data using unsupervised clustering algorithms
Qiao et al. Vertical land motion monitored with satellite radar altimetry and tide gauge along the Texas coastline, USA, between 1993 and 2020
CN116068511B (en) Deep learning-based InSAR large-scale system error correction method
Jiang et al. Bayesian inversion of wrapped satellite interferometric phase to estimate fault and volcano surface ground deformation models
Juda et al. Comparison of three recent discrete stochastic inversion methods and influence of the prior choice
CN103389073A (en) Method for selecting satellite-to-ground matched data through water-color remote sensing
Hao et al. GRDL: A New Global Reservoir Area‐Storage‐Depth Data Set Derived Through Deep Learning‐Based Bathymetry Reconstruction
CN115905877A (en) Depth sounding model inversion method and system based on VGGNet
CN116187168A (en) Method for improving water depth inversion accuracy based on neural network-gravity information wavelet decomposition
Peters et al. Geological seabed stability model for informing Irish offshore renewable energy opportunities
CN113516764B (en) Lake and reservoir underwater three-dimensional terrain simulation method and device based on digital elevation model
Peschoud et al. Fusion of sun-synchronous and geostationary images for coastal and ocean color survey application to OLCI (Sentinel-3) and FCI (MTG)

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