Sea surface wind stripe optical high-resolution remote sensing image detection method

Abstract

The invention relates to a sea surface wind stripe optical high-resolution remote sensing image detection method, which is applied to the field of sea surface wind field remote sensing monitoring, and utilizes wind stripe texture features and imaging geometric parameters of a high-spatial-resolution optical remote sensing image and combines time difference between imaging wavebands of an optical sensor to detect a sea surface wind stripe optical high-resolution remote sensing image. A wind stripe detection method based on a joint phase spectrum is developed, and on this basis, the wind stripe direction is detected according to phase difference asymmetry caused by imaging time difference between different characteristic wavebands. The method is suitable for optical high-spatial-resolution optical remote sensing images, the wind stripe characteristics in the images can be accurately detected, the sea surface wind stripe direction can be effectively estimated based on the wind stripe characteristics, and the problem of directivity blurring in the propagation process is solved. The sea surface wind direction estimation application level can be effectively improved, and a new thought is provided for sea surface wind field monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of sea surface wind field remote sensing, in particular to a sea surface wind streak optical high-resolution remote sensing image detection method. TECHNICAL BACKGROUND The sea surface wind field is the direct driving force of the wave, sea swell and wind-induced sea level change, and is also an important source of regional and global upper ocean circulation. It not only determines the formation and development of wind-generated waves, but also affects the flow structure of the upper ocean and the energy exchange process of the sea-air interface. Changes in sea surface wind speed and direction have a direct impact on regional climate regulation, ocean circulation generation, sea-air coupling processes and the occurrence of extreme weather events. Accurate acquisition of sea surface wind field parameters helps to better understand the evolution mechanism of the sea surface dynamic force and provides key support for marine weather forecasting, marine engineering design and disaster risk assessment.

[0002] Traditional sea surface wind field measurement methods use shore-based observation stations, ships and buoys, etc., which have small measurement range, few measurement points, and are easily limited by weather conditions. With the development of satellite remote sensing technology, especially the widespread use of high spatial resolution optical remote sensing data, it is possible to carry out sea surface wind field inversion based on image texture, wave features and white cap distribution, etc. The Multi Spectral Instrument (MSI) of the European Space Agency's Sentinel-2 satellite has excellent spatial resolution capability and can clearly record fine-scale disturbances on the sea surface, capturing the sea white cap and streak structure features closely related to wind speed. Unlike the point limitation and weather dependence of traditional buoy and ship observation methods, optical satellites provide a high-frequency, large-scale, cost-controlled wind field observation method. Building a direction estimation model based on high-resolution remote sensing images not only expands the spatial scale of wind field monitoring, but also provides a new technical means for the study of wind wave coupling processes and local wind field anomalies. SUMMARY

[0003] The technical problem to be solved by the present application is that the coverage of field wind observation is limited, making it difficult to observe a large area, and based on the time difference between the imaging bands of the optical satellite sensor, the wind direction detection based on Sentinel-2 MSI data is of great significance for sea surface wind field monitoring.

[0004] To solve the above technical problems, the technical solution of the present application is: a sea surface wind streak optical high-resolution remote sensing image detection method, comprising the following steps: Step 1, image preprocessing The high-resolution optical remote sensing image is preprocessed, including geographic correction and Rayleigh correction, to obtain the reflectivity image. Step 2, feature band selection For the reflectance images obtained in step 1, two single-band reflectance images with imaging time difference are selected; Step 3, extraction of the region of interest For the two single-band reflectance images extracted in step 2, the pixels containing clouds and land are removed, and the region of interest images suitable for wind streak detection are extracted according to the imaging geometric parameters of the images; Step 4, joint phase spectrum calculation The joint phase spectrum calculation is performed on the two region of interest images described in step 3, and the specific steps are as follows: 4.1, two-dimensional Fourier transform is performed on the two region of interest images respectively to obtain the corresponding Fourier spectra , ; 4.2, cross-spectrum calculation is performed on the Fourier spectra of the two region of interest images to obtain the corresponding inter-spectral correlation and inter-spectral phase difference , the calculation process is as follows: ; ; wherein, is the complex conjugate transform of the spectrum; 4.3, the joint phase difference between the two region of interest images is obtained according to the inter-spectral correlation and the inter-spectral phase difference , the calculation process is as follows: ; Step 4, energy distinction of joint phase difference The preset threshold is used to distinguish the energy of the joint phase difference, the part of the energy value of the joint phase difference is the high-frequency part of the joint phase difference, and the part of the energy value of the joint phase difference is the low-frequency part of the joint phase difference; Step 5, sea surface wind streak direction detection 5.1, energy distribution analysis is performed on the low-frequency part of the joint phase difference to solve the direction of the most concentrated energy of the phase difference, and the direction perpendicular to the direction is the wind streak blur direction; 5.2, phase feature analysis is performed on the high-frequency part of the joint phase difference, and in the wind streak blur direction, the positive phase of the high-frequency part points to the negative phase direction, which is the sea surface wind streak direction.

[0005] ​Further, the high spatial resolution optical remote sensing image in step 1 is a Sentinel-2 MSI multi-spectral image data, and the spatial resolution is 10 m; in step 2, the imaging time difference between the two selected single-band reflectivity images is 0.47 s-1.04 s. As preferred, the two single bands selected in step 2 are: B4 band and B8 band, the wavelength of B4 band is 665 nm, the wavelength of B8 band is 842 nm, and the imaging time difference between the two single-band reflectivity images is 0.74 s.

[0006] Further, in step 3, the condition suitable for wind streak detection is that the included angle between the observation direction of the sensor and the mirror reflection direction of the incident sunlight is satisfies or ; wherein the included angle is obtained according to the imaging geometric parameters of the image: ; wherein is the solar zenith angle, θ is the satellite zenith angle, is the relative azimuth angle of the sun and the satellite.

[0007] The present application has the following beneficial effects: The present application utilizes the structural streak characteristics of the wind field in the high spatial resolution optical remote sensing image, and combines the time difference between the imaging bands of the sensor, to develop a sea surface wind streak optical high-resolution remote sensing image detection method. The actual effect shows that the present application can significantly improve the range of sea surface wind field monitoring, reduce the time, labor and other costs of wind field observation, and improve the efficiency of sea surface wind field monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 is a flowchart of the present application.

[0009] Figure 2 is the Sentinel-2 remote sensing image and Fourier spectrum diagram of Example 1, (a) B8 band remote sensing image; (b) B4 band remote sensing image; (c) Fourier spectrum diagram of B8 band remote sensing image; (d) Fourier spectrum diagram of B4 band remote sensing image.

[0010] Figure 3 is the cross-spectrum calculation result of Example 1, (a) inter-spectrum phase difference image; (b) inter-spectrum correlation image.

[0011] Figure 4 is the joint phase difference energy differentiation and wind direction detection result schematic diagram of Example 1, (a) joint phase difference image; (b) joint phase difference low-frequency part result; (c) joint phase difference high-frequency part result; (d) wind streak direction detection result.

[0012] Figure 5 For example, the regional Sentinel-2 remote sensing images of Example 2 and Distribution map, (a) true-color composite image (B4, B3, B2), (b) corresponding remote sensing image Distribution diagram.

[0013] Figure 6 The results of regional wind stripe direction detection in Example 2 are as follows: (a) wind stripe direction detection results at 5km resolution; (b) wind stripe direction detection results at 0.25° resolution. Detailed Implementation

[0014] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0015] This example applies to two Sentinel-2 MSI multispectral image datasets. The first example image was captured on July 24, 2020, with a spatial resolution of 10 meters, using the B4 (665nm) and B8 (842nm) bands. The second example image was captured on June 28, 2022, with a spatial resolution of 10 meters. This data can be downloaded free of charge from the European Space Agency's Copernicus program website (https: / / browser.dataspace.copernicus.eu / ).

[0016] like Figure 1 The diagram shown is a flowchart of the high-resolution optical remote sensing image detection method for sea surface wind stripes according to the present invention. The method of the present invention includes the following steps: Step 1: Image Preprocessing High-resolution optical remote sensing images undergo preprocessing, including georeferencing and Rayleigh correction, to obtain reflectance images. The Rayleigh correction formula is as follows: ; in, This represents the Rayleigh-corrected reflectance; Indicates pixel radiance; Indicates Rayleigh scattering radiance; Indicates solar incident irradiance; It represents the solar zenith angle.

[0017] Step 2: Selecting characteristic bands For the reflectance image obtained in step 1, select two single-band reflectance images with an imaging time difference. (As in Example 1) Figure 2 (a) B8 band image and Figure 2(b) B4 band image, the imaging interval between the bands is 0.74 s. In addition to this, other two bands with imaging time difference can be selected, and in the embodiment, only B4 band and B8 band are taken as examples for illustration. If other bands are selected, the imaging time difference between the two selected single-band reflectivity images should be within the range of 0.47 s-1.04 s.

[0018] Step 3, extraction of the region of interest For the two single-band reflectivity images extracted in step 2, the pixels containing clouds and land are removed, and the region of interest images suitable for wind streak detection are extracted according to the imaging geometry parameters of the images. In this step, the condition suitable for wind streak detection is that the angle between the sensor observation direction and the mirror reflection direction of the incident sunlight is satisfied or .

[0019] When this step is specifically implemented, first, the angle between the sensor observation direction and the mirror reflection direction of the incident sunlight is calculated according to the following formula : ; wherein, is the solar zenith angle, θ is the satellite zenith angle, is the relative azimuth angle of the sun and the satellite.

[0020] Then, the extraction of the region of interest is performed according to the condition or , so as to obtain two region of interest images. In the two region of interest images, the pixel value of each pixel is the reflectivity value after Rayleigh correction, such as the region of interest images in Figure 2 .

[0021] Step 4, joint phase spectrum calculation The joint phase spectrum calculation is performed on the two region of interest images described in step 3, and the specific steps are as follows: 4.1, two-dimensional Fourier transform is respectively performed on the two region of interest images to obtain Fourier spectrum , and the Fourier spectrum is converted into wave number coordinate representation, and the specific calculation process is as follows: ; ; wherein, is the image coordinate of the region of interest image, is the reflectivity of the th region of interest image, is the Fourier spectrum coordinate,​ the Fourier spectrum energy of the Fourier spectrum corresponding to the image of the first region of interest in the wave number coordinate, is a two-dimensional Fourier transform, is a converted wave number coordinate, is the Fourier spectrum energy of the Fourier spectrum corresponding to the image of the first region of interest in the wave number coordinate, the Fourier spectrum energy of the Fourier spectrum corresponding to the image of the first region of interest in the wave number coordinate, is the pixel number of the transverse coordinate of the image of the region of interest, is the pixel number of the longitudinal coordinate of the image of the region of interest, is the sensor resolution, is the serial number of the image of the region of interest. The Fourier spectra of the B8 band and the B4 band are shown in (c) and Figure 2 (d). Figure 2

[0022] 4.2, cross spectrum calculation is performed on the Fourier spectra of the two images of the region of interest to obtain the corresponding inter-spectral correlation and the inter-spectral phase difference , the calculation process is as follows: ; ; Specifically, in this embodiment, the Fourier spectrum of the image of the region of interest in the wave number coordinate form is subjected to cross spectrum calculation to obtain the corresponding inter-spectral correlation and the inter-spectral phase difference , the calculation process is as follows: ; .

[0023] wherein, is the correlation spectrum between the characteristic bands, is the phase difference spectrum, is the complex conjugate transform of the spectrum. The inter-spectral correlation is symmetric about the midpoint, and the inter-spectral phase difference is not symmetric about the midpoint. The value range of the inter-spectral correlation is 0 to 1, and the value range of the inter-spectral phase difference is -180° to 180°. The inter-spectral phase difference between the images of the region of interest corresponding to the characteristic bands B4 and B8 is shown in Figure 3 (a), which is not symmetric about the midpoint, and the inter-spectral correlation is shown in Figure 3 (b), which is symmetric about the midpoint.

[0024] 4.3, the joint phase difference between the two images of the region of interest is obtained according to the inter-spectral correlation and the inter-spectral phase difference , the calculation process is as follows:​ ; The joint phase difference between the feature bands B4 and B8 and the image of the region of interest is shown in Figure 4 (a).

[0025] Step 4, energy distribution of joint phase difference A preset threshold is used The energy value of the joint phase difference is energy-distributed, and the part of the energy value of the joint phase difference is the high-frequency part of the joint phase difference, and the part of the energy value of the joint phase difference is the low-frequency part of the joint phase difference. In this embodiment, the preset threshold .

[0026] Step 5, sea surface wind streak detection 5.1, energy distribution analysis is performed on the low-frequency part of the joint phase difference, and the low-frequency part of the joint phase difference represents the wind streak energy in the remote sensing image, as shown by the red dashed line part in Figure 4 (b), the direction with the most concentrated energy of the phase difference is solved, and the direction perpendicular to the direction is the wind streak blur direction.

[0027] 5.2, phase feature analysis is performed on the high-frequency part of the joint phase difference, and the high-frequency part of the joint phase difference represents the wave energy in the remote sensing image, as shown by the white dashed line part in Figure 4 (c), in the wind streak blur direction, the positive phase of the high-frequency part points to the negative phase direction, which is the sea surface wind streak direction. The wind streak direction detection result of the first embodiment is shown in Figure 4 (d), the gray dashed line is the most concentrated direction, and the white direction in the figure is the wind streak direction.

[0028] Figure 5 (a) is the regional Sentinel-2 MSI sea surface true color synthesis image (B4 (665 nm), B3 (560 nm), B2 (490 nm)) of the second embodiment, Figure 5 (b) is the distribution diagram calculated according to the imaging geometric parameters, and it can be seen that the region satisfies , which meets the sea surface wind streak detection range. Figure 6 (a) is the regional wind streak direction detection result using the present application, and the black arrow direction in the figure is the wind streak direction. The spatial resolution of the black arrow is 5 km, Figure 6 (b) is the regional wind streak direction detection result using the present application, and the black arrow in Figure 6 (a) is down-sampled to a spatial resolution of 0.25°.

[0029] In combination with the above two embodiments, it can be seen that the present application has good effects in wide-area wind streak remote sensing detection and direction inversion.

[0030] The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, which is made by using the content of the present application specification and drawings, is also included in the patent protection scope of the present application.

Claims

1. A method for detecting wind stripes in high-resolution optical remote sensing images of the sea surface, comprising the following steps: Step 1: Image Preprocessing Preprocessing of high-resolution optical remote sensing images, including geo-correction and Rayleigh correction, is performed to obtain reflectance images; Step 2: Selecting characteristic bands For the reflectance image obtained in step 1, select two single-band reflectance images with an imaging time difference; Step 3: Region of Interest Extraction For the two single-band reflectance images extracted in step 2, pixels containing clouds and land are removed, and regions of interest suitable for wind stripe detection are extracted based on the imaging geometry parameters of the images. Step 4: Joint Phase Spectrum Calculation The joint phase spectrum calculation is performed on the two region-of-interest images described in step 3. The specific steps are as follows: 4.1 Perform two-dimensional Fourier transforms on the images of the two regions of interest respectively to obtain the corresponding Fourier spectra. , ; 4.

2. Perform cross-spectral calculation on the Fourier spectra of the two regions of interest images to obtain the corresponding inter-spectral correlation. Phase difference between spectra The calculation process is as follows: ; ; in, This is the complex conjugate transform of the spectrum; 4.

3. Based on the spectral correlation between two region-of-interest images Phase spectral difference Find the joint phase difference between two region of interest images. The calculation process is as follows: ; Step 4: Combined phase difference energy differentiation Use preset threshold Energy differentiation is performed on the joint phase difference; the energy value of the joint phase difference. The part that represents the high-frequency portion of the joint phase difference is the energy value of the joint phase difference. The portion that represents the low-frequency part of the joint phase difference; Step 5: Detection of wind stripe direction on the sea surface 5.1 Perform energy distribution analysis on the low-frequency part of the joint phase difference, and solve for the direction in which the energy of the phase difference is most concentrated. The direction perpendicular to this direction is the fuzzy direction of the wind stripes. 5.

2. Perform phase characteristic analysis on the high-frequency part of the joint phase difference. In the direction of wind stripe ambiguity, the positive phase of the high-frequency part pointing to the negative phase direction is the direction of the sea surface wind stripe.

2. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: In step 1, Rayleigh correction is calculated using the following formula: ; in, Indicates Rayleigh-corrected reflectance; Indicates pixel radiance; Indicates Rayleigh scattering radiance; Indicates solar incident irradiance; It represents the solar zenith angle.

3. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: The high spatial resolution optical remote sensing image in step 1 is Sentinel-2 MSI multispectral image data with a spatial resolution of 10m. In step 2, the imaging time difference between the two selected single-band reflectance images is 0.47s-1.04s.

4. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 3, characterized in that: In step 2, the two single bands selected are B4 band and B8 band. The wavelength of B4 band is 665nm and the wavelength of B8 band is 842nm. The imaging time difference between the two single band reflectance images is 0.74s.

5. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: In step 3, the condition suitable for wind stripe detection is: the angle between the sensor's observation direction and the direction of the solar incident light reflected by the specular surface. satisfy or Among them, the included angle Determined based on the imaging geometry parameters of the image: ; in, The zenith angle of the sun. θ For the satellite zenith angle, This represents the relative azimuth angle between the sun and the satellite.

6. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: In step 4.1, a two-dimensional Fourier transform is performed on the two region-of-interest images to convert the Fourier spectrum into wavenumber coordinates. The calculation process is as follows: ; ; in, The image coordinates of the region of interest. For the first The reflectance of an image of a region of interest. For Fourier spectrum coordinates, For the first Fourier spectral energy in Fourier spectral coordinates corresponding to the region of interest image. For two-dimensional Fourier transform, The transformed wavenumber coordinates, For the first Fourier spectral energy in wavenumber coordinates corresponding to the image of a region of interest. The x-coordinate of the region of interest image is the number of pixels. The vertical coordinate of the region of interest image is the number of pixels. For sensor resolution, The image number represents the region of interest.

7. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 6, characterized in that: In step 4.2, cross-spectral calculation is performed on the Fourier spectrum of the region of interest image in wavenumber coordinate form to obtain the corresponding inter-spectral correlation. Phase difference between spectra The calculation process is as follows: ; 。 8. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: In step 4.2, the inter-spectral correlation is centrally symmetric about the midpoint, while the inter-spectral phase difference is not centrally symmetric; the inter-spectral correlation ranges from 0 to 1, and the inter-spectral phase difference ranges from -180° to 180°.

9. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: The preset threshold in step 4 .

10. The method for detecting high-resolution optical remote sensing images of sea surface wind stripes according to claim 1, characterized in that: In step 5, the low-frequency portion of the combined phase difference represents the wind stripe energy in the remote sensing image, and the high-frequency portion represents the wave energy in the remote sensing image.

Images