Sea surface temperature space-time sequence acquisition method and system based on sea surface temperature inversion

Abstract

The invention discloses a sea surface temperature space-time sequence acquisition method and system based on sea surface temperature inversion, and the method comprises the steps: obtaining a GF-5A satellite image and an MODIS remote sensing image, and obtaining a sea surface temperature (SST) inversion image of the corresponding MODIS remote sensing image and the brightness temperature of a thermal infrared band 3 and a thermal infrared band 4 of the GF-5A satellite image; obtaining a sea surface temperature inversion image of the GF-5A satellite image; establishing an SRCNN super-resolution reconstruction model, taking the sea surface temperature inversion image of the MODIS remote sensing image as input, training to obtain a sea surface temperature super-resolution reconstruction image of the MODIS remote sensing image, and performing verification by using the sea surface temperature inversion image of the GF-5A satellite image; and obtaining a sea surface temperature space-time sequence based on the super-resolution reconstruction result of the sea surface temperature inversion image of the MODIS remote sensing image. According to the invention, high-resolution sea surface temperature inversion and space-time sequence acquisition are realized.

Description

Technical Field

[0001] This invention relates to the field of marine environmental element inversion, and in particular to a method for high-resolution sea surface temperature inversion and spatiotemporal sequence acquisition based on GF-5A satellite imagery and SRCNN super-resolution reconstruction model. Background Technology

[0002] Sea surface temperature (SST) is a fundamental physical quantity that influences various dynamic and thermodynamic processes in the coastal zone. Obtaining high-precision SST data is an important data foundation for further developing and utilizing the natural resources of the coastal zone and achieving sustainable development of the coastal zone.

[0003] Traditional SST inversion is mostly performed using thermal infrared imagery from foreign satellites such as the Landsat series and the Earth Observation System series. Thermal infrared imagery from domestic satellites such as the Environmental Satellite Series and the Fengyun series has begun to be applied to SST inversion, and researchers have proposed corresponding SST inversion algorithms applicable to various domestic satellite thermal infrared imagery. However, there are currently few studies using high-resolution thermal infrared imagery for SST inversion. Furthermore, the GF-5A satellite has only been in service for a short time and has a long observation period, making it difficult to analyze the spatiotemporal variations of SST in the study area. This is a technical problem that this invention urgently needs to solve. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for obtaining sea surface temperature spatiotemporal sequence based on sea surface temperature inversion. This method combines the sea surface temperature inversion method based on the split-window algorithm of high-resolution satellite imagery with the high-resolution imagery acquisition based on the SRCNN super-resolution reconstruction model of low-resolution satellite imagery, thereby achieving high-resolution sea surface temperature inversion and spatiotemporal sequence acquisition.

[0005] To achieve the above-mentioned objectives, the present invention proposes the following technical solution:

[0006] In a first aspect, the present invention proposes a method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion, comprising:

[0007] Acquire GF-5A satellite imagery and MODIS remote sensing imagery, and obtain the corresponding sea surface temperature inversion imagery from MODIS remote sensing imagery and the brightness temperature of visible and mid-infrared light from GF-5A satellite imagery.

[0008] Based on the brightness temperature of visible and mid-infrared light from the GF-5A satellite, sea surface temperature inversion images from GF-5A satellite images are obtained.

[0009] An SRCNN super-resolution reconstruction model was established, using the sea surface temperature inversion image of the MODIS remote sensing image as input, and trained to obtain the sea surface temperature super-resolution reconstruction image of the MODIS remote sensing image. The sea surface temperature inversion image of the GF-5A satellite image was then used for verification.

[0010] Based on the super-resolution reconstruction results of the sea surface temperature inversion image from the MODIS remote sensing image, the spatiotemporal sequence of sea surface temperature is obtained.

[0011] In some implementations, the sea surface temperature inversion imagery from the GF-5A satellite is expressed as follows:

[0012] ;

[0013] in, For sea surface temperature inversion images, and These are the brightness temperatures for visible light and mid-infrared light, respectively. , and All of these are parameters of the split-window algorithm, and they are all obtained through regression coefficients. and and two related values and That is, the correlation value and include , , , ;

[0014] Among them, two correlation values The calculation formula is as follows:

[0015] ;

[0016] in, To improve atmospheric transmittance, Surface emissivity, This is an intermediate parameter, used to select between visible light and mid-infrared light.

[0017] In some implementations, the step of establishing an SRCNN super-resolution reconstruction model, using the sea surface temperature inversion image of the MODIS remote sensing image as input, to train and obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image further includes: the SRCNN super-resolution reconstruction model includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in series; the input sea surface temperature inversion image of the MODIS remote sensing image is passed through the first convolutional layer to obtain a low-resolution rock image feature vector, then through the second convolutional layer to obtain a high-resolution remote sensing image feature vector, and finally through the third convolutional layer to obtain a high-resolution remote sensing image.

[0018] Secondly, the sea surface temperature spatiotemporal sequence acquisition system based on sea surface temperature inversion proposed in this invention includes:

[0019] The acquisition module is used to acquire GF-5A satellite imagery and MODIS remote sensing imagery, and obtain the corresponding sea surface temperature inversion imagery from MODIS remote sensing imagery and the brightness temperature of visible and mid-infrared light from GF-5A satellite imagery.

[0020] The inversion module is used to obtain sea surface temperature inversion images from GF-5A satellite images based on the brightness temperature of visible and mid-infrared light from the GF-5A satellite.

[0021] The reconstruction module is used to establish an SRCNN super-resolution reconstruction model. It uses the sea surface temperature inversion image of the MODIS remote sensing image as input data to train and obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image. The model is then verified using the sea surface temperature inversion image of the GF-5A satellite image.

[0022] The spatiotemporal sequence output module 400 is used to obtain the spatiotemporal sequence of sea surface temperature based on the super-resolution reconstructed sea surface temperature image of the MODIS remote sensing image.

[0023] In some implementations, the sea surface temperature inversion imagery from the GF-5A satellite is expressed as follows:

[0024] ;

[0025] in, For sea surface temperature inversion images, and These are the brightness temperatures for visible light and mid-infrared light, respectively. , and All of these are parameters of the split-window algorithm, and they are all obtained through regression coefficients. and and two related values and Calculated, i.e., the relevant value and include , , , ;

[0026] Among them, two correlation values The calculation formula is as follows:

[0027] ;

[0028] in, To improve atmospheric transmittance, Surface emissivity, This is an intermediate parameter, used to select between visible light and mid-infrared light.

[0029] In some implementations, the step of establishing an SRCNN super-resolution reconstruction model, using the sea surface temperature inversion image of the MODIS remote sensing image as input, to train and obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image further includes: the SRCNN super-resolution reconstruction model includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in series; the input sea surface temperature inversion image of the MODIS remote sensing image is passed through the first convolutional layer to obtain a low-resolution rock image feature vector, then through the second convolutional layer to obtain a high-resolution remote sensing image feature vector, and finally through the third convolutional layer to obtain a high-resolution remote sensing image.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] 1) The sea surface temperature inversion method based on the split-window algorithm acquires high-resolution satellite images from MODIS remote sensing images and high-resolution images from low-resolution satellite images (GF-5A satellite) based on the SRCNN super-resolution reconstruction model. The combination of the two achieves high-resolution sea surface temperature inversion and spatiotemporal sequence acquisition. It has significant advantages in terms of SST resolution, the number of high-resolution satellite images required, and SST inversion accuracy.

[0032] 2) The process of training a super-resolution reconstruction model using SRCNN to obtain super-resolution sea surface temperature (SST) images from MODIS remote sensing imagery includes the following steps: First, feature extraction maps low-resolution SST image patches to feature vectors; second, nonlinear mapping nonlinearly maps low-resolution features to high-resolution features in the feature space. This process learns the complex, underlying correspondence between "low-resolution SST mode" and "high-resolution SST mode"; third, reconstruction reconstructs the final high-resolution SST image from the mapped high-resolution features, restoring more realistic physical details and enhancing monitoring accuracy. Attached Figure Description

[0033] Figure 1 This is a flowchart of a method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion according to Embodiment 1 of the present invention.

[0034] Figure 2 Reconstructing the SRCNN super-resolution model structure diagram;

[0035] Figure 3 This is a schematic diagram of a system for obtaining spatiotemporal sequences of sea surface temperature based on sea surface temperature inversion, according to Embodiment 2 of the present invention. Detailed Implementation

[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0037] Example 1: As Figure 1 As shown, the present invention provides a method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion, comprising the following steps:

[0038] Step 1: Acquire GF-5A satellite imagery and MODIS remote sensing imagery within a certain spatiotemporal range, and obtain the corresponding sea surface temperature (SST) inversion imagery from the MODIS remote sensing imagery and the brightness temperature from the GF-5A satellite thermal infrared bands 3 and 4; where:

[0039] Further steps include preprocessing the acquired GF-5A satellite imagery and MODIS remote sensing imagery:

[0040] The preprocessing of MODIS remote sensing images includes: first, geometric correction of MODIS remote sensing images; then, reprojection and resampling to make the coordinate system and resolution of MODIS remote sensing images consistent with GF-5A satellite data; then, cloud removal and sea-land separation; and finally, sea surface temperature inversion using a mature split-window algorithm to obtain sea surface temperature inversion images from MODIS remote sensing images.

[0041] The preprocessing of GF-5A satellite imagery includes: firstly, separating land and sea images from the GF-5A satellite imagery; then, calculating the radiance of GF-5A remote sensing data bands 3 (visible light) and 4 (mid-infrared light) through radiometric calibration; next, calculating the effective wavelengths of GF-5A remote sensing data bands 3 and 4 using the spectral response function of the GF-5A satellite; finally, calculating the brightness temperature of bands 3 and 4 using the effective wavelengths and radiance of the two bands to obtain the brightness temperature information of GF-5A satellite bands 3 and 4; and finally, obtaining the zenith angle of the satellite sensor from the metadata file of the GF-5A satellite imagery. .

[0042] The radiometric calibration formula for the GF-5A satellite is as follows:

[0043] (1);

[0044] in, For emissivity, A0, A1, and A2 are the grayscale values ​​of the pixels, and they are radiometric calibration coefficients, which can all be obtained from the metadata files of the GF-5A satellite.

[0045] The formulas for calculating the effective wavelengths of each thermal infrared band of the GF-5A satellite are as follows:

[0046] (2);

[0047] in, For effective wavelength, For wavelength, This represents the spectral response function value at the corresponding wavelength.

[0048] The formulas for calculating the brightness temperature of GF-5A satellite bands 3 and 4 are as follows:

[0049] (3);

[0050] in, For brightness temperature, At the speed of light, is Planck's constant. Boltzmann's constant, For emissivity, The effective wavelength.

[0051] Step 2: Based on the brightness temperature information of GF-5A satellite bands 3 and 4 obtained in Step 1, obtain the sea surface temperature inversion image from the GF-5A satellite imagery. Specifically, sea surface temperature inversion is achieved through a split-window algorithm. The sea surface temperature inversion process for GF-5A satellite imagery is as follows:

[0052] Step 2.1: Calculate the atmospheric transmittance based on the effective wavelength to obtain the improved atmospheric transmittance;

[0053] The improved formula for calculating atmospheric transmittance is as follows:

[0054] (4);

[0055] ;

[0056] ;

[0057] ;

[0058] ;

[0059] in, Atmospheric transmittance, This refers to the atmospheric water vapor content. , , and To pass through the effective wavelength The calculated parameters, Effective wavelength;

[0060] Brightness temperature based on GF-5A satellite imagery in bands 3 and 4 and zenith angle from satellite sensors The improved zenith angle value and temperature correction value of the atmospheric transmittance of the GF-5A satellite sensor were obtained. The brightness temperature value of the thermal infrared band of the satellite sensor and the zenith angle value of the satellite sensor (i.e., the angle between the line connecting the satellite sensor and the pixel and the normal to the ground plane at the pixel) both have a certain impact on the calculation of atmospheric transmittance. The specific formula for the zenith angle value of the satellite sensor is as follows:

[0061] (5);

[0062] in, and These are the brightness temperatures of GF-5A satellite imagery in bands 3 and 4, respectively. The zenith angle of the satellite sensor. and These are the angular improvements in atmospheric transmittance for bands 3 and 4, respectively. The zenith angle of the satellite sensor. and These are the temperature correction values ​​for atmospheric transmittance in bands 3 and 4, respectively.

[0063] Step 2.2: Set the improved regression coefficients. and Temperature parameter With brightness and temperature The linear relationship coefficient between them, temperature parameters It is the ratio of radiance to its rate of change with temperature. .

[0064] Step 2.3: Use the split-window algorithm to retrieve sea surface temperature and obtain the sea surface temperature retrieval image from the GF-5A satellite;

[0065] The sea surface temperature inversion imagery from the GF-5A satellite is expressed as follows:

[0066] (6);

[0067] in, For sea surface temperature inversion images, and These are the brightness temperatures for bands 3 and 4, respectively. , and All of these are parameters of the split-window algorithm, and they are all obtained through regression coefficients. and and related values and Calculated, i.e., the relevant value and include , , , .

[0068] Two correlation values and The calculation formula is as follows:

[0069] (7);

[0070] in, The atmospheric transmittance after the improvement of formula (4) , Surface emissivity, These are intermediate parameters; when calculating the intermediate parameters for band 3, The value is 1, meaning all correlation values ​​are the same as those for band 3; when calculating the intermediate parameters for band 4, The value is 2, meaning that all relevant values ​​are values ​​for band 4;

[0071] Brightness temperature, regression coefficients, and atmospheric transmittance have been obtained, while the surface emissivity of water is typically taken as 0.995. Substituting these values ​​into the formula yields the sea surface temperature inversion image.

[0072] Step 3: Establish a super-resolution reconstruction model based on Super-Resolution Convolutional Neural Network (SRCNN). Using the sea surface temperature inversion image from the MODIS remote sensing imagery obtained in Step 1 (i.e., the SST inversion imagery of the low-resolution MODIS remote sensing imagery) as input data, train the model to obtain a super-resolution reconstructed image of the MODIS remote sensing imagery. Validate this model using the sea surface temperature inversion imagery from the GF-5A satellite imagery obtained in Step 2 (i.e., the SST inversion imagery of the high-resolution GF-5A satellite imagery). Train the SRCNN super-resolution reconstruction model using the training set, and perform super-resolution reconstruction using the trained SRCNN super-resolution reconstruction model, outputting the super-resolution reconstruction result of the SST inversion imagery of the MODIS remote sensing imagery. Use the sea surface temperature inversion imagery from the GF-5A satellite imagery obtained in Step 2 (i.e., the SST inversion imagery of the high-resolution GF-5A satellite imagery) as validation information for the super-resolution reconstruction result of the SRCNN super-resolution reconstruction model. Then, use a test set to verify the accuracy of the SRCNN model's super-resolution reconstruction.

[0073] Specifically, such as Figure 2 As shown, the SRCNN super-resolution reconstruction model includes three concatenated convolutional layers: the first, second, and third convolutional layers. The input MODIS remote sensing image, representing sea surface temperature inversion, is processed through the first convolutional layer to obtain a low-resolution rock image feature vector, then through the second convolutional layer to obtain a high-resolution remote sensing image feature vector, and finally through the third convolutional layer to obtain the high-resolution remote sensing image. The sizes of the three convolutional kernels are set accordingly. =9、 =1、 =5; Number of the first convolutional kernels =64; Number of the second convolutional kernels =32; Since the input images are all SST inversion results, therefore = =1, meaning the number of input and output remote sensing image channels is the same, where, This represents the number of low-resolution remote sensing image channels input. The number of channels for the output high-resolution remote sensing image is set to 50; the batch size is set to 8; the initial learning rate is set to 0.001, which is optimized using the Adam optimizer; the loss function is the mean squared error (MSE), which is also optimized using the Adam optimizer.

[0074] Step 4: Based on the super-resolution reconstruction results of the sea surface temperature inversion image of the MODIS remote sensing image obtained in Step 3, obtain the sea surface temperature, arrange it into a spatiotemporal sequence, and construct the sea surface temperature spatiotemporal sequence.

[0075] Example 2, as follows Figure 3 As shown, the present invention provides a system for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion, comprising:

[0076] The acquisition module 100 is used to acquire GF-5A satellite images and MODIS remote sensing images within a certain spatiotemporal range, and obtain the corresponding sea surface temperature (SST) inversion images of the MODIS remote sensing images and the brightness temperature of visible light (band 3) and mid-infrared light (band 4) of the GF-5A satellite.

[0077] The inversion module 200 is used to obtain sea surface temperature inversion images of GF-5A satellite images based on the brightness and temperature information of visible light (band 3) and mid-infrared light (band 4) of the GF-5A satellite obtained by the acquisition module 100.

[0078] The reconstruction module 300 is used to establish an SRCNN super-resolution reconstruction model. It uses the sea surface temperature inversion image of the MODIS remote sensing image in step 1 as input data to train and obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image. The model is then verified using the sea surface temperature inversion image of the GF-5A satellite image in step 2.

[0079] The spatiotemporal sequence output module 400 is used to obtain the spatiotemporal sequence of sea surface temperature based on the super-resolution reconstructed sea surface temperature image of the MODIS remote sensing image from the super-resolution reconstruction module 300.

[0080] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0081] The above specific embodiments illustrate the principles and implementation methods of the present invention using specific examples. The descriptions of the embodiments are merely for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention. Any modifications, equivalent substitutions, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention. Additionally, any improvements and modifications made without departing from the principles of the present invention should also be considered to fall within the protection scope of the present invention.

Claims

1. A method for obtaining spatiotemporal series of sea surface temperature based on sea surface temperature inversion, characterized in that, include: Acquire GF-5A satellite imagery and MODIS remote sensing imagery, and obtain the corresponding sea surface temperature inversion imagery from MODIS remote sensing imagery and the brightness temperature of visible and mid-infrared light from GF-5A satellite imagery. Based on the brightness temperature of visible and mid-infrared light from the GF-5A satellite, sea surface temperature inversion images from GF-5A satellite images are obtained. An SRCNN super-resolution reconstruction model was established, using the sea surface temperature inversion image of the MODIS remote sensing image as input, and trained to obtain the sea surface temperature super-resolution reconstruction image of the MODIS remote sensing image. The model was then validated using the sea surface temperature inversion image of the GF-5A satellite image. Based on the super-resolution reconstruction results of sea surface temperature from the MODIS remote sensing image, a spatiotemporal sequence of sea surface temperature is obtained.

2. The method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion according to claim 1, characterized in that, The sea surface temperature inversion imagery from the GF-5A satellite is expressed as follows: ； in, For sea surface temperature inversion images, and These are the brightness temperatures for visible light and mid-infrared light, respectively. , and All of these are parameters of the split-window algorithm, and they are all obtained through regression coefficients. and and two related values and Calculated, i.e., the relevant value and include , , , ; Among them, two correlation values and The calculation formula is as follows: ； in, To improve atmospheric transmittance, Surface emissivity, This is an intermediate parameter, used to select between visible light and mid-infrared light.

3. The method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion according to claim 1, characterized in that, The establishment of the SRCNN super-resolution reconstruction model, using the sea surface temperature inversion image of the MODIS remote sensing image as input, and training to obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image, further includes: the SRCNN super-resolution reconstruction model includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in series. The input sea surface temperature inversion image of the MODIS remote sensing image is passed through the first convolutional layer to obtain a low-resolution rock image feature vector, then through the second convolutional layer to obtain a high-resolution remote sensing image feature vector, and finally through the third convolutional layer to obtain a high-resolution remote sensing image.

4. A system for obtaining spatiotemporal series of sea surface temperature based on sea surface temperature inversion, comprising: The acquisition module is used to acquire GF-5A satellite imagery and MODIS remote sensing imagery, and obtain the corresponding sea surface temperature inversion imagery of MODIS remote sensing imagery and the brightness temperature of visible and mid-infrared light of GF-5A satellite imagery. The inversion module is used to obtain the sea surface temperature inversion image of the GF-5A satellite image based on the brightness temperature of the visible and mid-infrared light of the GF-5A satellite image. The reconstruction module is used to establish an SRCNN super-resolution reconstruction model. It uses the sea surface temperature inversion image of the MODIS remote sensing image as input data to train and obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image. The model is then verified using the sea surface temperature inversion image of the GF-5A satellite image. The spatiotemporal sequence output module 400 is used to obtain the spatiotemporal sequence of sea surface temperature based on the super-resolution reconstructed sea surface temperature image of the MODIS remote sensing image.

5. The method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion according to claim 1, characterized in that, The sea surface temperature inversion imagery from the GF-5A satellite is expressed as follows: ； in, For sea surface temperature inversion images, and These are the brightness temperatures for visible light and mid-infrared light, respectively. , and All of these are parameters of the split-window algorithm, and they are all obtained through regression coefficients. and and two related values and Calculated, i.e., the relevant value and include , , , ; Among them, two correlation values and The calculation formula is as follows: ； in, Atmospheric transmittance, Surface emissivity, This is an intermediate parameter, used to select between thermal visible light and mid-infrared light.

6. The method for obtaining the spatiotemporal sequence of sea surface temperature based on sea surface temperature inversion according to claim 1, characterized in that, The establishment of the SRCNN super-resolution reconstruction model, using the sea surface temperature inversion image of the MODIS remote sensing image as input, and training to obtain the super-resolution reconstruction image of the sea surface temperature of the MODIS remote sensing image, further includes: the SRCNN super-resolution reconstruction model includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in series. The input sea surface temperature inversion image of the MODIS remote sensing image is passed through the first convolutional layer to obtain a low-resolution rock image feature vector, then through the second convolutional layer to obtain a high-resolution remote sensing image feature vector, and finally through the third convolutional layer to obtain a high-resolution remote sensing image.

Images