Coastal sea ice remote sensing extraction method based on FY-3 / MERSI satellite data

Abstract

The invention discloses a near-shore sea ice remote sensing extraction method based on FY-3 / MERSI satellite data, and the method comprises the following steps: S1, carrying out the preprocessing of an obtained MERSI image, and obtaining an MERSI image only containing sea ice and seawater; s2, establishing a coastal sea ice remote sensing recognition algorithm according to the difference of the spectral characteristics of sea ice and seawater in each channel of the MERSI image, and evaluating the precision and accuracy of the coastal sea ice remote sensing recognition algorithm; s3, according to a coastal sea ice remote sensing recognition algorithm, extracting coastal sea ice information from the MERSI image of each day; and carrying out 10-day averaging and monthly averaging on the daily sea ice data extracted in each month to generate a sea ice ten-day product and a monthly product. According to the invention, automatic identification of high-spatial-resolution sea ice information of a high-latitude offshore area in winter can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite remote sensing applications, and particularly relates to a method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data. Background Art

[0002] High-latitude nearshore waters are high-incidence areas of annual sea ice. For example, sea ice occurs every winter in the Bohai Sea and the northern part of the Yellow Sea in China. Since nearshore waters are important areas for people's production and living, the occurrence of sea ice seriously affects human maritime activities, such as maritime transportation, marine fishery farming, the development of marine resources (oil and gas exploration), port and coastal maritime engineering facilities, etc., and may even cause significant economic losses.

[0003] When sea ice occurs, it is difficult for ships to reach. Traditional sea ice monitoring mainly relies on means such as fixed ocean station observations, coastal ice condition surveys, icebreaker cruising surveys, and ground-based radar observations. These methods are restricted by space and time, and cannot meet the requirements of practical work, especially in terms of obtaining large-scale spatial information, time dynamic monitoring, and data updating. With the development of remote sensing technology, it provides an efficient, fast, and large-scale dynamic observation means for sea ice monitoring. The National Snow and Ice Data Center (NSIDC) in the United States has developed global sea ice remote sensing products for the DMSP satellite passive microwave radiometer and TERRA / AQUA satellite MODIS data. Their spatial resolution and accuracy are relatively low, mainly used for remote sensing monitoring of sea ice in the Arctic and Antarctic regions, and it is difficult to reflect the situation of sea ice occurrence in local sea areas, and the refinement degree is insufficient. The Chinese invention patent "Method for Inverting Arctic Sea Ice Concentration from FY-3 MWRI Data" with the publication number CN 107886473 B proposes a method for inverting Arctic sea ice for the MWRI data of China's Fengyun-3 satellite microwave imager. This invention method is based on the brightness temperature data of horizontal polarization and vertical polarization in the 89 GHz band of MWRI to construct a polarization difference value P for inverting Arctic sea ice concentration. Due to the low sensitivity and spatial resolution of the microwave imager, the error of the inverted sea ice is relatively large. Generally speaking, there is currently a lack of remote sensing products with high spatial resolution that can more precisely reflect the situation of sea ice occurrence in local sea areas, and there is no operational sea ice remote sensing inversion technology for nearshore sea areas. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide a method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data, realizing the automatic recognition of high-spatial-resolution sea ice information in high-latitude nearshore waters in winter, and being able to better meet the requirements of operational sea ice extraction.

[0005] Technical Solution: A method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data includes the following steps: S1. Preprocess the acquired MERSI images to obtain MERSI images containing only sea ice and sea water. S2. Based on the spectral feature differences of sea ice and sea water in each channel of the MERSI images, establish a remote sensing identification algorithm for nearshore sea ice, and evaluate the accuracy and precision of the nearshore sea ice remote sensing identification algorithm. S3. According to the nearshore sea ice remote sensing identification algorithm, extract nearshore sea ice information from the daily MERSI images; then perform 10-day averaging and monthly averaging on the daily sea ice data extracted within each month to generate dekadal sea ice products and monthly sea ice products.

[0006] Furthermore, the preprocessing of the MERSI image data includes the following steps: S11. Image radiometric calibration; Convert the data of the visible-near infrared-shortwave infrared bands and the thermal infrared band in the MERSI images into radiance. The radiometric calibration algorithms for the two bands are as follows: Visible-near infrared-shortwave infrared band: L1 = a + b·DN + c·DN 2 Thermal infrared band: L2 = a + b·DN + c·DN 2 + d·DN 3 In the formula, a, b, c, and d are calibration coefficients; L1 is the radiance of the visible-near infrared-shortwave infrared band, L2 is the radiance of the thermal infrared band; DN is the pixel value of the original image. For the MERSI visible-near infrared-shortwave infrared band, calculate the apparent reflectance R at the top of the atmosphere according to the radiance L1: , In the formula, E0 is the solar irradiance at the top of the atmosphere in the visible-near infrared-shortwave infrared band, is the solar zenith angle; For the MERSI thermal infrared band, calculate the brightness temperature T from the radiance L2 according to Planck's law rad : , In the formula, c1 and c2 are two coefficients in Planck's law; is the wavelength of the thermal infrared band, with the unit of ; S12. Image geometric calibration; Correspond the longitude and latitude information of the MERSI image at certain intervals with the image row and column numbers to construct a geometric control point matrix. With the help of the remote sensing image processing software ENVI / IDL, call the geometric correction function envi_register_doit through IDL language programming to achieve the geometric correction of the image, so that each pixel of the image has geographical positioning information; S13, Image mosaicking; After geometric correction of multiple scenes of images, according to the geographical positions of the multiple scenes of images, with the help of the remote sensing image processing software ENVI / IDL, call the image mosaicking function mosaic_doit through IDL language programming to achieve the splicing of multiple images; S14, Image cropping; Crop out the study sea area from the mosaicked MERSI image according to the longitude and latitude range of the study sea area; S15, Image resampling; Perform pixel resampling on the image with a spatial resolution of 250 meters so that the image spatial resolution is 1000 meters to ensure that the spatial resolutions of all channels are consistent; S16, Image land-sea-cloud separation; According to the spectral curve graphs of different ground objects in the MERSI image, select the reflectances of the green light and near-infrared bands to construct a normalized difference index: (R2 - R4) / (R2 + R4), and set a threshold to remove land and clouds from the MERSI image, where R2 represents the reflectance of the green light band and R4 represents the reflectance of the near-infrared band.

[0007] Furthermore, the detailed steps for establishing a remote sensing identification algorithm for nearshore sea ice include: S21, According to the characteristic that the reflectance of sea ice in the red light band is higher than that of sea water, adopt the adaptive threshold method for the reflectance image of the 3rd band of MERSI, and perform histogram statistics on the entire sea area image; at the right inflection point of the peak, where the slope of the reflectance change is the largest, determine the threshold between ice and water, and identify the pixels with reflectance greater than the threshold as sea ice; S22, Perform principal component analysis on the reflectance data of 19 bands of visible-near-infrared-shortwave infrared of MERSI, and identify sea ice for the first principal component component according to the adaptive threshold method. At the left inflection point of the peak of the first principal component image and where the numerical change slope is the largest, determine the threshold between ice and water, and identify the pixels with values less than the threshold in the first principal component image as sea ice; S23, According to the characteristic that the brightness temperature of sea ice is lower than that of sea water in the thermal infrared band, adopt the adaptive threshold method for the brightness temperature of the thermal infrared band of MERSI, and perform histogram statistics on the thermal infrared band image of the entire sea area. At the left inflection point of the peak, determine the threshold between ice and water, and identify the pixels with brightness temperature less than the threshold as sea ice; S24, based on the MERSI red light band, the first principal component and the thermal infrared band, three kinds of sea ice image information are obtained, and the pixel value of sea ice is marked as 1, and the pixel value of non-sea ice is marked as 0; when fusing the multi-band sea ice information, the sea ice identified by the red light M3 band and the first principal component is taken into an "and" relationship: if all the pixels in the sea ice image extracted by the two bands are 1, it is judged as sea ice, and if all the pixels are 0, it is judged as non-sea ice; then the sea ice image identified by the thermal infrared band is taken into an "or" relationship: if one of the two sea ice images has a sea ice image pixel identified as sea ice, the pixel is finally sea ice; By counting the number of pixels in the sea ice image and then calculating the area of ​​sea ice based on the area corresponding to each pixel.

[0008] Furthermore, the results of sea ice remote sensing extraction from MERSI images were compared with the results of manual visual interpretation of MERSI images. The nearshore sea ice remote sensing recognition algorithm was evaluated by constructing a confusion matrix and based on the overall classification accuracy, Kappa coefficient, mapping accuracy of various types of features, and user accuracy; The results of visual interpretation of sea ice in the Bohai and Yellow Seas from FY-3 satellite visible light infrared scanning radiometer images in winter for a period of time were selected as the true value, and the accuracy of sea ice identified by the nearshore sea ice remote sensing identification algorithm from MERSI images of the same period was tested. The accuracy of the nearshore sea ice remote sensing identification algorithm was evaluated by calculating the correlation coefficient, root mean square error and mean absolute percentage error as evaluation indicators.

[0009] Compared with the prior art, the present invention has the following significant effects: The present invention aims at the images of the Medium Resolution Spectral Imager (MERSI) of my country's second-generation Fengyun polar-orbiting satellite series (FY-3), and utilizes its visible light, near-infrared, short-wave infrared and thermal infrared data to detect sea ice through an adaptive threshold method of multi-band images, and then fuses the multi-band sea ice information to realize automatic recognition of high spatial resolution sea ice information in high-latitude nearshore waters in winter. The present invention is easy to implement on various computer platforms and can better meet the needs of automatic extraction of commercial sea ice remote sensing. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is a flow chart of the present invention; Figure 2 Spectral curves of different objects in the MERSI image of the present invention; Figure 3 MERSI red light band reflectivity histogram in an embodiment of the present invention; Figure 4 It is a brightness temperature histogram of the MERSI thermal infrared band in an embodiment of the present invention; Figure 5Histogram of the first principal component in the visible-shortwave infrared band of MERSI in the embodiments of the present invention; Figure 6 Scatter plot comparing the sea ice area extracted by the inshore sea ice algorithm with the visually interpreted sea ice area in the embodiments of the present invention; Figure 7 Schematic diagram of the inshore sea ice remote sensing dekadal product in the Bohai Sea and the Yellow Sea in late January 2022 in the embodiments of the present invention; Figure 8 Schematic diagram of the inshore sea ice remote sensing monthly product in the Bohai Sea and the Yellow Sea in January 2022 in the embodiments of the present invention. Detailed implementation manners

[0011] The present invention will be further described in detail below in conjunction with the accompanying drawings of the specification and the specific implementation manners.

[0012] The present invention provides a method for remotely sensing and extracting inshore sea ice based on FY-3 / MERSI satellite data. In order to verify the method proposed by the present invention, taking the Bohai Sea and the northern Yellow Sea of China as the study area and the data in January 2022 as an example, dekadal and monthly inshore sea ice remote sensing products are generated. The implementation process is as follows Figure 1 shown, including the following steps: Step 1, preprocessing of FY-3 / MERSI images; The images of the Medium Resolution Spectral Imager (MERSI) on the second-generation Fengyun polar-orbiting series satellites (FY-3) are usually L1B-level raw images, and preprocessing is required for the images when remotely sensing and extracting sea ice. The preprocessing of MERSI image data mainly includes operations such as image radiometric correction, geometric correction, image mosaicking, image cropping, image resampling, land-sea separation, and cloud removal.

[0013] (11) Image radiometric correction; The MERSI images contain data in the visible-near infrared-shortwave infrared band and the thermal infrared band. The central wavelengths of each band are shown in Table 1. The pixel values (DN) of the raw images are digital signals and need to be converted into radiance with physical meaning according to the radiometric calibration coefficients. The radiometric correction algorithms for the two bands are as follows: Visible-near infrared-shortwave infrared band: L1 = a + b·DN + c·DN 2 Thermal infrared band: L2 = a + b·DN + c·DN 2 + d·DN 3 In the formula: a, b, c, and d are calibration coefficients; L1 is the radiance in the visible-near infrared-shortwave infrared band, and L2 is the radiance in the thermal infrared band.

[0014] The calibration coefficients in the visible-near infrared-shortwave infrared bands are from the attribute data "VIR_Cal_Coeff" in the image file, and the calibration coefficients in the thermal infrared band are from the attribute data "IR_Cal_Coeff" in the image file.

[0015] For the MERSI visible-near infrared-shortwave infrared bands, in order to reflect the reflection characteristics of sea ice and sea water, it is necessary to calculate the top-of-atmosphere apparent reflectance R from the radiance L1: , where: E0 is the solar irradiance at the top of the atmosphere in the visible-near infrared-shortwave infrared bands, is the solar zenith angle of incidence.

[0016] For the MERSI thermal infrared band, in order to better characterize the thermal radiation characteristics of sea ice and sea water, it is necessary to calculate the radiance L2 into the brightness temperature T according to Planck's law rad : , where: c1 and c2 are two coefficients in Planck's law, c1 takes the value of 1.191×10 8 , c2 takes the value of 1.439×10 4 , is the wavelength in the thermal infrared band, with the unit of micrometer ( ).

[0017] (12) Image geometric correction; In order to make each pixel of the MERSI image correspond to the ground position one by one and eliminate the geometric distortion of the image during satellite imaging, it is necessary to perform geometric correction on the MERSI image. Since the MERSI image contains longitude and latitude information, the longitude and latitude information at a certain interval of the image is corresponding to the image row and column number information to construct a geometric control point matrix. With the help of the remote sensing image processing software ENVI / IDL, the geometric correction function envi_register_doit is called through IDL language programming to achieve the geometric correction of the image, so that each pixel of the image has geographic positioning information.

[0018] (13) Image mosaicking; Due to the deviation of the daily orbital operation of the FY-3 satellite, the image scanning is incomplete, and the study area will appear in multi-track images. It is necessary to splice multiple images. After performing geometric correction on multiple images, according to the geographical positions of multiple images, with the help of the remote sensing image processing software ENVI / IDL, the image mosaicking function mosaic_doit is called through IDL language programming to achieve the splicing of multiple images.

[0019] (14)Image cropping; After mosaicking multiple MERSI images, it usually includes a large spatial range around the study sea area. To better highlight the study sea area, it is necessary to crop the study sea area from the mosaicked MERSI image according to the longitude and latitude range of the study sea area. This can be accomplished by using the remote sensing image processing software ENVI / IDL and programming in IDL language to call the image subset function ENVI_SUBSET_VIA_ROI_DOIT.

[0020] (15)Image resampling; Since the MERSI multispectral image has two spatial resolutions of 250 meters and 1000 meters, in order to facilitate the extraction of sea ice information, it is necessary to make the spatial resolutions of all channels consistent. Therefore, it is necessary to resample the pixels of the image with a spatial resolution of 250 meters so that the spatial resolution of the image is 1000 meters. This can be achieved by using the remote sensing image processing software ENVI / IDL and programming in IDL language to call the image resampling function resize_doit.

[0021] (16)Image separation of sea, land and cloud; From Figure 2 the reflectance spectrograms of sea water, sea ice, land and cloud in the visible-near infrared-shortwave infrared bands of the MERSI image, it can be seen that the reflectance of cloud in the visible-near infrared bands of MERSI is very high. Land and cloud have similar variation characteristics in the green light (M2) and near infrared (M4) bands, that is, the reflectance in the M2 band is lower than that in the M4 band; while the reflectance of sea ice and sea water in these two bands is R2>R4. Therefore, the normalized difference index: (R2-R4) / (R2+R4) is constructed using the reflectance of these two bands, and a threshold is set to remove land and cloud from the MERSI image, where R2 represents the reflectance in the green light (M2) band and R4 represents the reflectance in the near infrared (M4) band.

[0022] Step 2, construct a remote sensing identification algorithm for nearshore sea ice; For the MERSI image of the nearshore sea area after preprocessing such as sea-land separation and cloud removal, there are mainly two substances, sea ice and sea water, in the image. A remote sensing identification algorithm for nearshore sea ice is established based on the spectral characteristics differences of the two in each channel of MERSI.

[0023] (21)Sea ice identification based on the red light band of MERSI; According to the characteristic that the reflectance of sea ice in the red light band is higher than that of sea water, the adaptive threshold method is adopted for the reflectance image of the 3rd band (M3) of MERSI, that is, the histogram of the entire sea area image is statistically analyzed, and at the right inflection point of the peak where the slope of the reflectance change is the largest, it is determined as the threshold between ice and water (as Figure 3 shown), and the pixels with reflectance greater than the threshold are identified as sea ice.

[0024] (22) Sea ice identification based on the principal components of the visible-near infrared-shortwave infrared bands of MERSI; According to the differences in reflectance between thin sea ice and turbid water bodies in the near infrared, blue, and red bands, as well as the differences in the absorption and reflection characteristics of sea ice and sea water in the shortwave infrared band, in order to make full use of the multi-band information of MERSI, principal component analysis is performed on the reflectance data of 19 bands of MERSI's visible-near infrared-shortwave infrared. Since the first principal component (PC1) contains more than 80% of the variance information of all the original bands, the first principal component is used for sea ice identification according to the adaptive threshold method (as Figure 4 shown), that is, the threshold between ice and water is determined at the left inflection point of the peak of the PC1 image histogram and the point where the numerical change slope is the largest, and the pixels with values less than the threshold in the PC1 image are identified as sea ice.

[0025] (23) Sea ice identification based on the thermal infrared band of MERSI; According to the characteristic that the brightness temperature of sea ice is lower than that of sea water in the thermal infrared band, the adaptive threshold method is used for the brightness temperature of the thermal infrared band of MERSI (11.25 or 12.0 ), that is, the histogram of the thermal infrared band image of the entire sea area is statistically analyzed, and the threshold between ice and water is determined at the left inflection point of the peak (the point where the brightness temperature change slope is the largest) (as Figure 5 shown), and the pixels with brightness temperature less than the threshold are identified as sea ice.

[0026] Table 1 Performance indicators of the Medium Resolution Spectral Imager MERSI on FY-3 series satellites

[0027] (24)Fusion of multi-band sea ice information; In this embodiment, due to the daily changes in the atmospheric conditions, the atmospheric transmittance of the daily MERSI images in the visible, near-infrared, shortwave infrared, and thermal infrared bands is different, and the thresholds for extracting sea ice from the corresponding band images each day are also different. Sea ice is extracted by statistically analyzing the reflectance or brightness temperature values of the images and automatically determining the thresholds. Based on the red band, the first principal component component, and the thermal infrared band of MERSI, three types of sea ice image information can be identified. The pixel values of sea ice are marked as 1, and the pixel values of non-sea ice are marked as 0. When fusing multi-band sea ice information, the sea ice identified by the red M3 band and the first principal component component is taken in an "and" relationship, that is, if all the pixels in the sea ice images extracted by the two bands are 1, it is judged as sea ice; otherwise, it is 0, non-sea ice. Then, an "or" relationship is performed with the sea ice image identified by the thermal infrared band, that is, as long as the pixels in one of the two sea ice images are identified as sea ice, the final pixel is sea ice, ensuring that the sea ice information identified by the visible-near-infrared-shortwave infrared bands and the thermal infrared band complements each other.

[0028] By counting the number of pixels in the sea ice image and then according to the area size corresponding to each pixel, the area where sea ice occurs can be calculated.

[0029] Step 3, accuracy evaluation of the nearshore sea ice remote sensing identification algorithm; (31) Algorithm evaluation based on the confusion matrix; For the results of remote sensing image classification, it is usually necessary to compare the field investigation or visual interpretation results with the remote sensing classification images, construct a confusion matrix of the classified ground objects, calculate the overall classification accuracy, Kappa coefficient, mapping accuracy, and user accuracy of various ground objects to evaluate the accuracy of the nearshore sea ice remote sensing identification algorithm. Since the images of the entire sea area mainly have two types of ground objects, sea ice and sea water, and the sea ice in the Bohai Sea and the Yellow Sea is in a melting state in March every year, in this embodiment, December 31, 2020, and December 27, 2021, January 9, 2021, and January 27, 2022, February 16, 2022, and February 24, 2022 are selected to represent the results of sea ice remote sensing extraction from MERSI images in December, January, and February of each year respectively, and compared with the results of manual visual interpretation of MERSI images. By constructing a confusion matrix and according to the calculation methods of evaluation indicators such as the overall classification accuracy, the accuracy results of classifying and extracting sea ice and sea water from 6 MERSI images are comprehensively calculated (Table 2). As can be seen from Table 2, the overall accuracy of remote sensing extraction of sea ice and sea water from multiple MERSI images is 90.73%, the Kappa coefficient is 0.732, the mapping accuracy and user accuracy of sea ice are 80.42% and 78.78% respectively, indicating that the sea ice remote sensing algorithm of the present invention has good accuracy.

[0030] Table 2 Accuracy of remote sensing extraction of sea ice and sea water from MERSI images of FY-3 satellite of the present invention

[0031] (32) Accuracy evaluation of sea ice area based on long time series; The results of visual interpretation of sea ice in the Bohai and Yellow Seas from the FY-3 satellite Visible Infrared Scanning Radiometer (VIRR) images in the winter from December 2008 to March 2015 were taken as the true values. The accuracy of the sea ice identified by the nearshore sea ice remote sensing identification algorithm from the MERSI images of the same period was tested. The accuracy of the nearshore sea ice remote sensing identification algorithm was evaluated by calculating the correlation coefficient, root mean square error and mean absolute percentage error as evaluation indicators. Figure 6 The scatter plot is a comparison of the sea ice area extracted by the nearshore sea ice algorithm and the sea ice area interpreted visually. The correlation coefficient between the two is as high as 0.982, and the average absolute percentage error is 11.89%, indicating that the nearshore sea ice algorithm of the present invention has good accuracy.

[0032] Step 4: Generation of sea ice remote sensing products; According to the nearshore sea ice remote sensing identification algorithm of the present invention, nearshore sea ice information is extracted from the daily MERSI images, and the pixel values ​​of the sea ice images are 1 and 0, respectively representing sea ice and non-sea ice. The daily sea ice data extracted in each month are averaged for 10 days and monthly. Due to the influence of cloud cover over the ocean, effective sea ice information cannot be extracted from the daily MERSI images. Through the analysis of sea ice information obtained by remote sensing extraction for many years, the threshold of the mid-term average in this embodiment is 0.35, and the threshold of the monthly average is 0.20, and the sea ice ten-day product and monthly product are further generated. Figure 7 and Figure 8 They are the sea ice products in late January 2022 in the Bohai Sea and the Yellow Sea and the monthly sea ice products in January 2022 respectively.

Claims

1. A method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data, characterized in that, The steps are as follows: S1. Preprocess the acquired MERSI images to obtain MERSI images containing only sea ice and sea water; S2. Based on the spectral feature differences of sea ice and sea water in each channel of the MERSI images, establish a remote sensing identification algorithm for nearshore sea ice, and evaluate the accuracy and precision of the nearshore sea ice remote sensing identification algorithm; S3. According to the nearshore sea ice remote sensing identification algorithm, extract nearshore sea ice information from the daily MERSI images; then perform 10-day averaging and monthly averaging on the daily sea ice data extracted within each month to generate dekadal sea ice products and monthly sea ice products.

2. The method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data according to claim 1, wherein The preprocessing of MERSI image data includes the following steps: S11. Image radiometric calibration; Convert the data of the visible-near infrared-shortwave infrared bands and the thermal infrared band in the MERSI images into radiance. The radiometric calibration algorithms for the two bands are as follows: Visible - Near Infrared - Shortwave Infrared Band: L1 = a + b·DN + c·DN 2 Thermal infrared band: L2 = a + b·DN + c·DN 2 + d·DN 3 where a, b, c, and d are calibration coefficients; L1 is the radiance of the visible-near infrared-shortwave infrared band, L2 is the radiance of the thermal infrared band; DN is the pixel value of the original image; For the MERSI visible-near infrared-shortwave infrared band, calculate the apparent reflectance R at the top of the atmosphere according to the radiance L1: ， where E0 is the solar irradiance at the top of the atmosphere in the visible-near infrared-shortwave infrared band, is the solar zenith angle of incidence; For the MERSI thermal infrared band, the radiance L2 is calculated into brightness temperature T according to Planck's law rad : ， where c1 and c2 are two coefficients in Planck's law; is the wavelength in the thermal infrared band, with the unit of ; S12. Image geometric correction; Correspond the longitude and latitude information at certain intervals in the MERSI images with the image row and column number information to construct a geometric control point matrix. With the help of the remote sensing image processing software ENVI / IDL, call the geometric correction function envi_register_doit through IDL language programming to realize the geometric correction of the images, so that each pixel of the images has geographic positioning information; S13. Image mosaicking; After geometric correction of multiple scenes of images, according to the geographical locations of the multiple scenes of images, with the help of the remote sensing image processing software ENVI / IDL, call the image mosaicking function mosaic_doit through IDL language programming to realize the splicing of multiple images; S14. Image cropping; Crop the study area from the mosaicked MERSI images according to the longitude and latitude range of the study area; S15. Image resampling; Perform pixel resampling on the images with a spatial resolution of 250 meters to make the spatial resolution of the images 1000 meters, ensuring that the spatial resolutions of all channels are consistent; S16. Image land-sea-cloud separation; According to the spectral curve graphs of different ground objects in the MERSI images, select the reflectances of the green light and near infrared bands to construct a normalized difference index: (R2 - R4) / (R2 + R4), and set a threshold to remove land and clouds from the MERSI images, where R2 represents the reflectance of the green light band and R4 represents the reflectance of the near infrared band.

3. The nearshore sea ice remote sensing extraction method based on FY-3 / MERSI satellite data according to claim 1, wherein The detailed steps for establishing the remote sensing identification algorithm for nearshore sea ice include: S21. According to the characteristic that the reflectance of sea ice in the red light band is higher than that of sea water, adopt the adaptive threshold method for the reflectance image of the MERSI Band 3, and perform histogram statistics on the entire sea area image; at the right inflection point of the peak where the slope of the reflectance change is the largest, determine the threshold between ice and water, and identify the pixels with reflectance greater than the threshold as sea ice; S22, perform principal component analysis on the reflectance data of MERSI's 19 bands of visible, near infrared and short-wave infrared, identify sea ice on the first principal component according to the adaptive threshold method, determine the threshold between ice and water at the left turning point of the first principal component image histogram peak and the point where the slope of the value change is the largest, and identify the pixels in the first principal component image with values ​​less than the threshold as sea ice; S23, based on the characteristic that the brightness temperature of sea ice in the thermal infrared band is lower than that of sea water, an adaptive threshold method is used for the brightness temperature of the MERSI thermal infrared band. Histogram statistics are performed on the thermal infrared band images of the entire sea area. At the left inflection point of the peak, the threshold between ice and water is determined, and pixels with brightness temperatures less than the threshold are identified as sea ice. S24, based on the MERSI red light band, the first principal component and the thermal infrared band, three kinds of sea ice image information are obtained, and the pixel value of sea ice is marked as 1, and the pixel value of non-sea ice is marked as 0; when fusing the multi-band sea ice information, the sea ice identified by the red light M3 band and the first principal component is taken into an "and" relationship: if all the pixels in the sea ice image extracted by the two bands are 1, it is judged as sea ice, and if all the pixels are 0, it is judged as non-sea ice; then the "or" relationship is taken with the sea ice image identified by the thermal infrared band: if one of the two sea ice images has a sea ice image pixel identified as sea ice, the pixel is finally sea ice; By counting the number of pixels in the sea ice image and then calculating the area of ​​sea ice based on the area corresponding to each pixel.

4. The method for remotely sensing and extracting nearshore sea ice based on FY-3 / MERSI satellite data according to claim 1, wherein The results of sea ice remote sensing extraction from MERSI images were compared with those of manual visual interpretation of MERSI images. The nearshore sea ice remote sensing recognition algorithm was evaluated by constructing a confusion matrix and based on the overall classification accuracy, Kappa coefficient, mapping accuracy of various types of features, and user accuracy. The results of visual interpretation of sea ice in the Bohai and Yellow Seas from FY-3 satellite visible light infrared scanning radiometer images in winter for a period of time were selected as the true value, and the accuracy of sea ice identified by the nearshore sea ice remote sensing identification algorithm from MERSI images of the same period was tested. The accuracy of the nearshore sea ice remote sensing identification algorithm was evaluated by calculating the correlation coefficient, root mean square error and mean absolute percentage error as evaluation indicators.

Images