Method for performing polar region all-day cloud detection by using brightness temperature of FY-3D infrared channel

Abstract

The invention discloses a method for polar region all-day cloud detection by using FY-3D infrared channel brightness temperature, and relates to the technical field of cloud detection, and the method comprises the following steps: obtaining CALIOP data and FY-3D / MERS-II data, specifically CALIOP cloud layer product data and MERS-II observation data and positioning data, carrying out the preprocessing, and carrying out the preprocessing to obtain an FY-3D infrared channel brightness temperature; according to the method, space-time matching is carried out according to transit trajectories of CALIOP and FY-3D satellite MERS-II sensors, matched infrared channel radiance information is extracted, calculation is carried out based on the matched infrared channel radiance information, and the brightness temperature and the brightness temperature difference are calculated. CALIOP data are combined with FY-3D / MERS-II data, the CALIOP is used as a truth value standard of cloud detection, space-time matching is carried out with the FY-3D / MERS-II data, and the brightness temperature difference is calculated. The polar region all-time cloud detection is realized, the accuracy of the cloud detection is improved, and a high-quality data set is provided for the training of a subsequent machine learning model, so that the effects of being not limited by the dependence of visible light and stably working under an all-time condition are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud detection, and particularly relates to a method for all-weather cloud detection in polar regions by using the brightness temperature of the infrared channel of FY-3D. Background Art

[0002] More than 50% of the Earth's surface is covered by clouds. Clouds play a crucial role in the Earth's atmosphere system and are key factors in regulating the Earth's radiation balance and driving the global water vapor cycle. Cloud identification is not only a basic step in the quantitative inversion of atmospheric components such as clouds and aerosols from satellite data but also a prerequisite for further deriving key physical parameters such as cloud characteristics and surface characteristics. The polar region, as an indispensable part of the climate system, is not only a sensitive area of global climate and environmental change but also a key area. The change in cloud cover in this region has a profound impact on regional and even global climate.

[0003] However, due to the perennial heavy snow and ice cover on the polar surface and the low solar altitude angle, cloud detection in these regions faces more severe challenges. Polar cloud systems mainly consist of thin clouds and low-level clouds. Since the temperature difference between clouds and the snow and ice surface is relatively small, the difference in radiation characteristics between clouds and the snow and ice underlying surface is not significant, which increases the difficulty of cloud detection. In addition, during night and polar night periods, due to the lack of observational data in the visible light channel, cloud detection results often have large uncertainties.

[0004] Therefore, under the extreme climatic conditions in the polar region, traditional cloud detection methods often have poor effects. Especially during night, polar night, or low-light conditions, the difficulty of cloud detection increases significantly. To improve the accuracy and stability of cloud detection, a method for all-weather cloud detection in polar regions by using the brightness temperature of the infrared channel of FY-3D is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for all-weather cloud detection in polar regions by using the brightness temperature of the infrared channel of FY-3D to solve the problems raised in the above background art.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for all-weather cloud detection in polar regions by using the brightness temperature of the infrared channel of FY-3D, comprising the following steps:

[0008] Step 1, obtain CALIOP data and FY-3D / MERSI-II data, specifically cloud layer product data, FY-3D satellite MERSI-II observation data, and positioning data, and perform preprocessing;

[0009] Step 2: Perform spatio-temporal matching according to the transit trajectories of the CALIOP and FY-3D satellite MERSI-II sensors, and extract the radiance information of the matched infrared channels;

[0010] Step 3: Calculate based on the radiance information of the matched infrared channels, calculate the brightness temperature and brightness temperature difference, and obtain the surface information, instrument zenith angle, and solar zenith angle of the matched pixels. At the same time, establish a sample data set;

[0011] Step 4: Perform feature encoding on the sample data set, and use the AdaBoost machine learning algorithm to construct a cloud detection model;

[0012] Step 5: Read the test set data in the sample data set and input it into the trained model to evaluate the model performance. At the same time, read the information of the pixels to be detected and combine it with the cloud detection model to obtain the cloud detection result.

[0013] In the above Step 1, the acquisition processes of the CALIOP data and the FY-3D / MERSI-II data are as follows:

[0014] Visit the CALIPSO data download page of NASA, download the 1km resolution CLAY cloud layer product data of CALIOP, and extract the time information, longitude and latitude information, and feature classification information of each pixel on the satellite orbit;

[0015] Based on the obtained cloud layer product data, check the feature type and feature type quality assessment flag in the Feature_Classification_Flags variable, and perform cloud detection label marking on all pixels on the CALIOP orbit, which are divided into two categories: cloudy and clear sky. Specifically, when the feature type is marked as cloud and the feature type quality assessment is above medium level (FeatureType == 2 and Feature Type QA >= 2), the scanned pixel is defined as cloudy, otherwise, it is marked as clear sky;

[0016] Obtain from the official website of the FY-3D satellite

[0017] The 1km resolution observation data of MERS-II, FY3D_MERSI_GBAL_L1_YYYYMM

[0018] DD_HHmm_1000M_MS.HDF and the positioning data FY3D_MERSI_GBAL_L1_YYYYMM

[0019] DD_HHmm_GEO1K_MS.HDF, and at the same time perform preprocessing operations on the downloaded transit data.

[0020] A further improvement of the technical solution of the present invention lies in that: in step 2, the process of extracting the infrared channel radiance information is as follows:

[0021] Compare the timestamps in the CALIOP cloud layer product data and the transit trajectory data of the FY-3D satellite MERSI sensor, and find the data points with similar and overlapping times;

[0022] Use the longitude and latitude information to spatially match the pixels in the CALIOP and FY-3D / MERSI-II data to obtain the pixel longitude and latitude;

[0023] According to the matching results of time matching and spatial matching, filter out the matching data pairs that meet both time and space conditions, and obtain the matching pixels;

[0024] In the FY-3D / MERSI-II data, extract the infrared channel radiance information of the matching pixels, and organize the extracted infrared channel radiance information with the CALIOP data.

[0025] A further improvement of the technical solution of the present invention lies in that: the process of obtaining the matching pixels is as follows:

[0026] Based on the CALIOP data, extract the sub-satellite point transit time, and filter the FY-3D satellite MERSI orbits that transit within 5 minutes before and after the CALIOP transit time;

[0027] Centered on the CALIOP sub-satellite point, set a square area of 0.1° each in the up, down, left, and right directions, and intercept all the pixels of the MERSI orbit within this square area;

[0028] Use the Haversine algorithm to calculate the surface distance between all MERSI pixels within the square area and the CALIOP sub-satellite point, and filter the pixels that meet the conditions that the satellite zenith angle of the pixel is less than 40° and the surface distance between the pixel and the CALIOP sub-satellite point does not exceed 1 km as the matching pixels.

[0029] A further improvement of the technical solution of the present invention lies in that: the calculation of the Haversine algorithm is as follows:

[0030]

[0031] where d is the distance between two points, R is the radius of the earth, lat 1 and lat 2 are the latitude coordinates of two points, lon 1 and lon 2 are the longitude coordinates of two points, and both the latitude and longitude coordinates need to be expressed in radians

[0032] A further improvement of the technical solution of the present invention lies in that: in the step 3, the process of establishing the sample data set is as follows:

[0033] Based on the infrared channel radiance information corresponding to the CALIOP-matched data pairs extracted from the FY-3D MERSI data, ensuring that the extracted brightness temperature data has accurate timestamp and geographical location information, calculating the brightness temperature value of each pixel and the brightness temperature difference between different channels, where the brightness temperature difference is the difference between the brightness temperatures of different infrared channels;

[0034] Using the FY-3D / MERSI-II data, extract the surface type of the pixels corresponding to the matched data pairs, and extract the instrument zenith angle and solar zenith angle of each matched pixel from the FY-3D / MERSI-II data, ensuring that the extracted angle information is consistent with the extracted brightness temperature data and surface information in time and space;

[0035] Based on the cloud detection labels, pixel longitude and latitude, infrared channel brightness temperature data and brightness temperature difference, surface type, and instrument zenith angle and solar zenith angle information, integrate them to form a unified sample data set;

[0036] Utilizing the brightness temperature of the infrared channel of the FY-3D satellite, which is not restricted by visible light dependence and can work stably under all-day conditions. Whether it is during the day with sufficient sunlight or during the long polar night, it can continuously obtain cloud detection results.

[0037] A further improvement of the technical solution of the present invention lies in that: the calculation formula of the brightness temperature value is:

[0038]

[0039] RAD = RAD0 * Slope + Intercept;

[0040] Where, Tbb is the brightness temperature value, where c 1 = 1.91043 * 10-5 Mw / (m2sr cm-4), c 2 = 1.43878 K, v c is the central wave number of each channel, A and B are channel brightness temperature correction coefficients, RAD0 is the amplified radiance value, RAD is the radiance, Slope is the slope, and Intercept is the intercept.

[0041] A further improvement of the technical solution of the present invention lies in that: in the step 4, the process of constructing the cloud detection model is as follows:

[0042] According to the surface type, divide the sample data set into 6 sub-data sets, including water body, evergreen coniferous forest, open shrubland, grassland, bare land, and ice and snow sub-data sets;

[0043] The integrated sample data set is divided into a training set and a test set. The AdaBoost machine learning model is trained using the training set data to construct a cloud detection model, and the trained cloud detection model is verified using the test set data. The performance of the cloud detection model is evaluated, and model evaluation metrics such as the accuracy and F1 score are calculated.

[0044] A further improvement of the technical solution of the present invention lies in that: the formula for the model evaluation metrics is as follows:

[0045] Accuracy:

[0046]

[0047] F1 score:

[0048]

[0049] Among them, TP is the number of cloudy pixels correctly identified by the model, TN is the number of clear-sky pixels correctly identified by the model, FP is the number of clear-sky pixels misidentified as cloudy by the model, and FN is the number of cloudy pixels misidentified as clear-sky pixels by the model.

[0050] A further improvement of the technical solution of the present invention lies in that: in step 5, the process of outputting the discrimination result is as follows:

[0051] The pixel longitude and latitude, surface information, instrument zenith angle, instrument azimuth angle, and infrared channel brightness temperature are input into the trained model for cloud detection prediction, and the prediction result is output. The prediction result is usually an array containing 0 and 1, where 0 may represent clear sky and 1 represents cloudy.

[0052] The prediction result of the model is compared with the labels in the test set, and the evaluation metrics are calculated and summarized. At the same time, an evaluation report is generated for analyzing and optimizing the model.

[0053] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is as follows:

[0054] 1. The present invention provides a method for polar all-weather cloud detection using the FY-3D infrared channel brightness temperature. By combining CALIOP data with FY-3D / MERSI-II data, using CALIOP as the ground truth standard for cloud detection and performing spatio-temporal matching with FY-3D / MERSI-II data, it not only improves the accuracy of cloud detection but also provides a high-quality data set for the subsequent training of machine learning models. Therefore, by combining the cloud detection ground truth of CALIPSO lidar, the spatio-temporal matching algorithm between the two payloads, and the machine learning algorithm, it is not restricted by the dependence on visible light and can work stably under all-weather conditions, realizing polar all-weather cloud detection and providing strong technical support for fields such as meteorological monitoring and climate change research.

[0055] 2. The present invention provides a method for detecting polar all-day clouds using the brightness temperature of the FY-3D infrared channel. By introducing the AdaBoost machine learning method, a polar cloud detection model is established, and combined with the cloud detection results of the spaceborne lidar CALIOP as the ground truth, cloud detection is realized all day long, providing continuous and reliable data support for polar meteorological observations and climate change research, significantly reducing misjudgment and missed judgment situations, and greatly improving the accuracy of polar cloud detection, providing a more reliable data basis for polar meteorological research.

[0056] 3. The present invention provides a method for detecting polar all-day clouds using the brightness temperature of the FY-3D infrared channel. By combining machine learning algorithms, the cloud detection process of the FY-3D infrared channel brightness temperature data can be highly automated and intelligent. The machine learning model automatically learns the characteristics of clouds and makes accurate classifications based on these characteristics, not only reducing the need for manual intervention but also improving the efficiency of cloud detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0058] Figure 1 is the flowchart of the method of the present invention;

[0059] Figure 2 is the flowchart for extracting the infrared channel radiance data of the present invention;

[0060] Figure 3 is the flowchart of the CALIOP and MERSI pixel spatio-temporal matching algorithm of the present invention;

[0061] Figure 4 is the flowchart for establishing the sample data set of the present invention;

[0062] Figure 5 is the flowchart for constructing the model of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0064] Example 1, as Figures 1 to 5 shown, the present invention provides a method for detecting polar all-weather clouds using the brightness temperature of the FY-3D infrared channel. Step 1: Obtain CALIOP data and FY-3D / MERSI-II data, specifically CALIOP cloud layer product data, MERSI-II observation data and positioning data, and perform preprocessing; the process of obtaining CALIOP data and FY-3D / MERSI-II data is as follows: Access the CALIPSO data download page of NASA, download the 1km resolution CLAY cloud layer product data of CALIOP, extract the time information, longitude and latitude information, and feature classification information of each pixel on the satellite orbit. Based on the obtained cloud layer product data, check the feature type and feature type quality assessment flag in the Feature_Classification_Flags variable, and perform cloud detection label marking on all pixels on the CALIOP orbit, which are divided into two categories: cloudy and clear sky. Specifically, when the feature type is marked as cloud and the feature type quality assessment is above medium level (Feature Type == 2 and Feature Type QA >= 2), the scanned pixel is defined as cloudy, otherwise, it is marked as clear sky; Obtain the 1km resolution observation data FY3D_MERSI_GBAL_L1_YYYYMM, DD_HHmm_1000M_MS.HDF and positioning data FY3D_MERSI_GBAL_L1_YYYYMM, DD_HHmm_GEO1K_MS.HDF of MERS-II from the official website of the FY-3D satellite, and at the same time perform preprocessing operations on the downloaded transit data;

[0065] Step 2: Perform spatio-temporal matching according to the transit trajectories of the CALIOP and FY-3D satellite MERSI-II sensors, and extract the radiance information of the infrared channels after matching. The process of extracting the radiance information of the infrared channels is as follows: Compare the timestamps in the CALIOP cloud product data and the transit trajectory data of the FY-3D satellite MERSI sensor, find the data points with similar and overlapping times, use the longitude and latitude information to spatially match the pixels in the CALIOP and FY-3D / MERSI-II data to obtain the pixel longitude and latitude. According to the matching results of time matching and spatial matching, filter out the matching data pairs that meet both time and space conditions, and obtain the matching pixels. The process of obtaining the matching pixels is as follows: Based on the CALIOP data, extract the sub-satellite point transit time, filter the FY-3D satellite MERSI orbits that transit within 5 minutes before and after the CALIOP transit time. Taking the CALIOP sub-satellite point as the center, set a square area of 0.1° in each of the up, down, left, and right directions, and intercept all the pixels of the MERSI orbit within this square area. Use the Haversine algorithm to calculate the surface distance between all the MERSI pixels in the square area and the CALIOP sub-satellite point, and filter out the pixels that meet the conditions that the satellite zenith angle of the pixel is less than 40° and the surface distance between the pixel and the CALIOP sub-satellite point does not exceed 1 km as the matching pixels. In the FY-3D / MERSI-II data, extract the radiance information of the infrared channels of the matching pixels, and organize the extracted radiance information of the infrared channels with the CALIOP data.

[0066] Step 3: Based on the radiance information of the infrared channels after matching, calculate the brightness temperature and the brightness temperature difference, and obtain the surface information, instrument zenith angle, and solar zenith angle of the matching pixels, and establish a sample data set at the same time. The process of establishing the sample data set is as follows: Based on the radiance information of the infrared channels corresponding to the CALIOP matching data pairs extracted from the FY-3D / MERSI-II data, calculate the brightness temperature value of each pixel and the brightness temperature difference between different channels, ensuring that the extracted brightness temperature data has accurate timestamp and geographical location information. Calculate the brightness temperature difference between the infrared channels and the brightness temperature value of each pixel. Among them, the brightness temperature difference is the difference between the brightness temperatures of two infrared channels. Use the FY-3D / MERSI-II data to extract the surface type of the matching pixels, and extract the instrument zenith angle and solar zenith angle of each matching pixel from the FY-3D MERSI data. Integrate based on the cloud detection label, pixel longitude and latitude, infrared channel brightness temperature data and brightness temperature difference, surface type, and instrument zenith angle and solar zenith angle information to form a unified sample data set.

[0067] Utilizing the brightness temperature of the infrared channel of the FY-3D satellite, which is not restricted by the dependence on visible light, it can operate stably under all-day conditions. Whether it is during the day with sufficient sunlight or during the long polar night, it can continuously obtain cloud detection results. This characteristic brings continuity and integrity to polar meteorological observations. Scientists can thereby track the dynamic changes of polar clouds at different times and study the mechanism of the role of clouds in the polar day-night cycle and seasonal alternation;

[0068] Step 4: Perform feature encoding on the sample data set and use the AdaBoost machine learning algorithm to construct a cloud detection model. The process of constructing the cloud detection model is as follows: According to the surface type, the sample data set is divided into 6 sub-data sets, including water body, evergreen coniferous forest, open shrubland, grassland, bare land, and ice-snow sub-data sets. The integrated sample data set is divided into a training set and a test set. Use the training set data to train the AdaBoost machine learning model to construct a cloud detection model, and use the data of the test set to evaluate the trained cloud detection model, evaluate the performance of the cloud detection model, and calculate the model evaluation indicators such as accuracy and F1 score. By introducing the AdaBoost machine learning method, a polar cloud detection model is established, and combined with the cloud detection results of the spaceborne lidar CALIOP as the ground truth, cloud detection can be carried out all day long, providing continuous and reliable data support for polar meteorological observations and climate change research, significantly reducing the situations of misjudgment and missed judgment, and greatly improving the accuracy of polar cloud detection, providing a more reliable data basis for polar meteorological research;

[0069] Step 5: Read the test set data in the sample data set and input it into the trained model to evaluate the model performance. At the same time, read the information of the pixel to be detected and combine it with the cloud detection model to obtain the cloud detection result. The process of outputting the discrimination result is as follows: Input the pixel longitude and latitude, surface information, instrument zenith angle, instrument azimuth angle, and infrared channel brightness temperature into the trained model for cloud detection prediction and output the prediction result. The prediction result is usually an array containing 0 and 1, where 0 may represent clear sky and 1 represents cloudy. Compare the prediction result of the model with the label in the test set, calculate and summarize the evaluation indicators, and generate an evaluation report for analyzing and optimizing the model;

[0070] Combining the accurate cloud detection results of CALIOP as the ground truth, and comprehensively considering various auxiliary data, such as surface information, longitude and latitude information, etc., the AdaBoost machine learning model is trained through a large number of sample data. The model can learn the complex feature patterns of clouds in different polar environments. For example, in ice and snow covered areas, the model can accurately distinguish clouds from ice and snow based on the subtle differences in the brightness temperature of the infrared channels and specific surface type features; under different solar altitude angles and seasonal variations, precise judgments are made using longitude and latitude information and the data laws accumulated over a long time. Through actual tests, the misjudgment and missed judgment situations are significantly reduced, and there is a substantial improvement in the polar cloud detection accuracy, providing a more reliable data basis for polar meteorological research and helping to deeply understand the distribution, evolution of polar clouds and their impact on the climate system.

[0071] Example 2, as Figures 1 to 5 shown, based on Example 1, the present invention provides a technical solution: Preferably, the Haversine algorithm is calculated as:

[0072]

[0073] where d is the distance between two points, R is the radius of the earth, lat 1 and lat 2 are the latitude coordinates of the two points, lon 1 and lon 2 are the longitude coordinates of the two points, and both the latitude and longitude coordinates need to be expressed in radians;

[0074] The calculation process of the brightness temperature difference is as follows: Based on the MERSI II 1KM resolution observation data file FY3D_MERSI_GBAL_L1_YYYYMM, DD_HHmm_1000M_MS.HDF, calculate the brightness temperature of MERSI channels 20, 21, 22, 23, 24, 25 at the matching pixels, and further calculate the brightness temperature and brightness temperature difference using the formula, including the brightness temperature difference between channel 20 and 25 (B20_25), the brightness temperature difference between channel 24 and 20 (B24_20), the brightness temperature difference between channel 23 and 24 (B23_24), and the brightness temperature difference between channel 20 and 21. Among them, the variables EV_1KM_Emissive and EV_250_Aggr.1KM_Emissive in the MERSI II 1KM resolution observation data file are the amplified radiance values RAD0 of the 1km emission channels (CH20~23) and 250m emission channels (CH24~25), respectively;

[0075] The calculation formula for the brightness temperature value is:

[0076]

[0077] where c 1= 1.91043 * 10-5 Mw / (m2srcm-4), c 2 = 1.43878 K, v c Represents the central wavenumber of each channel. Then, using the channel brightness temperature correction coefficients A and B, calculate the blackbody brightness temperature Tbb of each channel, and the brightness temperature difference is the difference between the brightness temperatures of two channels;

[0078] RAD = RAD0 * Slope + Intercept;

[0079] Among them, RAD0 is the amplified radiance value, RAD is the radiance, Slope represents the slope in the calibration formula, which is a coefficient used to convert the original radiance value into the calibrated radiance value, and Intercept represents the intercept in the calibration formula, which is also a coefficient used together with the slope for calibration calculation;

[0080] Among them, Table 1 shows the equivalent central wavenumbers and channel brightness temperature correction coefficients of the FY-3D MERSI thermal emission channels (CH20 - CH25);

[0081]

[0082] Table 1

[0083] The formula for the model evaluation index is:

[0084] Accuracy: The ratio of the number of correctly classified samples to the total number of samples;

[0085]

[0086] F1_score: The higher the F1 score, the better the comprehensive evaluation performance;

[0087]

[0088] Among them, Accuracy is the accuracy rate, Recall is the recall rate, TP is the number of cloudy pixels correctly identified by the model, TN is the number of clear sky pixels correctly identified by the model, FP is the number of clear sky pixels misidentified as cloudy by the model, FN is the number of cloudy pixels misidentified as clear sky by the model. Among them, Table 2 shows the evaluation results of the cloud detection AdaBoost machine learning algorithm in the Arctic region;

[0089] Surface type Accuracy F1 score Water body 0.904 0.881 Bush 0.910 0.899 Ice and snow 0.925 0.925 Bare land 0.896 0.889 Grassland 0.881 0.867 Evergreen coniferous forest 0.932 0.922

[0090] Table 2

[0091] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims described above.

Claims

1. A method for polar all-day cloud detection using FY-3D infrared channel brightness temperature, characterized in that: The following steps are involved: Step 1: Acquire CALIOP data and FY-3D / MERSI-II data, specifically CALIOP cloud product data and MERSI-II observation data and positioning data, and perform preprocessing; Step 2: Perform spatiotemporal matching based on the transit trajectories of CALIOP and the MERSI-II sensor of the FY-3D satellite, and extract the radiance information of the matched infrared channel; Step 3: Calculate the brightness temperature and brightness temperature difference based on the matched infrared channel radiance information, obtain the surface information of the matched pixels and the instrument zenith angle and solar zenith angle, and establish a sample data set; Step 4: Encode the features of the sample data set and use the AdaBoost machine learning algorithm to build a cloud detection model; Step 5: Read the test set data in the sample data set and input it into the trained model to evaluate the model performance. At the same time, read the pixel information to be detected and combine it with the cloud detection model to obtain the cloud detection result.

2. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 1, characterized in that: In step 1, the acquisition process of CALIOP data and FY-3D / MERSI-II data is as follows: Visit NASA's CALIPSO data download page, download CALIOP's 1km resolution CLAY cloud product data, and extract the time information, longitude and latitude information, and feature classification information of each pixel on the satellite orbit; Based on the acquired cloud product data, the feature type and feature type quality assessment flag in the Feature_Classification_Flags variable are checked, and all pixels on the CALIOP track are marked with cloud detection labels and classified into two categories: cloudy and clear sky. Obtained from the official website of FY-3D satellite MERS-II 1km resolution observation data FY3D_MERSI_GBAL_L1_YYYYMM DD_HHmm_1000M_MS.HDF and positioning data FY3D_MERSI_GBAL_L1_YYYYMM DD_HHmm_GEO1K_MS.HDF, and preprocess the downloaded transit data.

3. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 2, characterized in that: In step 2, the process of extracting the infrared channel radiance information is as follows: Compare the timestamps in the CALIOP cloud product data and the FY-3D satellite MERSI sensor transit track data to find data points with similar and overlapping times; Use the longitude and latitude information to spatially match the pixels in the CALIOP and FY-3D / MERSI-II data to obtain the longitude and latitude of the pixels; According to the matching results of time matching and space matching, matching data pairs that meet both time and space conditions are screened out, and matching pixels are obtained; In the FY-3D / MERSI-II data, the infrared channel radiance information of the matching pixels is extracted, and the extracted infrared channel radiance information is collated with the CALIOP data.

4. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 3 is characterized in that: The process of obtaining the matching pixels is as follows: Based on CALIOP data, the subsatellite point transit time was extracted, and the FY-3D satellite MERSI orbits that transited within 5 minutes before or after the CALIOP transit time were selected; With the CALIOP subsatellite point as the center, a square area with an area of ​​0.1° in the upper, lower, left and right directions is set, and all pixels in the MERSI orbit within this square area are intercepted; The Haversine algorithm is used to calculate the surface distance between all MERSI pixels and the CALIOP subsatellite point in the square area, and the pixels whose satellite zenith angle is less than 40° and whose surface distance between the pixel and the CALIOP subsatellite point is no more than 1 km are selected as matching pixels.

5. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 4, characterized in that: The Haversine algorithm is calculated as: Where d is the distance between two points, R is the radius of the Earth, and are the latitude coordinates of the two points, and These are the longitude coordinates of the two points. Both latitude and longitude coordinates must be expressed in radians.

6. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 5, characterized in that: In step 3, the process of establishing the sample data set is as follows: Based on the infrared channel radiance information corresponding to the CALIOP matching data pair extracted from the FY-3D / MERSI-II data, the brightness temperature value of each pixel and the brightness temperature difference of different channels are calculated; Using FY-3D / MERSI-II data, extract the surface type of the corresponding pixels of the matching data; Extract the instrument zenith angle and solar zenith angle of each matching pixel from FY-3D / MERSI-II data; A unified sample data set is formed by integrating cloud detection labels, pixel longitude and latitude, infrared channel brightness temperature data and brightness temperature difference, surface type, instrument zenith angle, and solar zenith angle information.

7. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 6, characterized in that: The calculation formula of the brightness temperature value is: RAD=RAD0*Slope+Intercept; Among them, Tbb is the brightness temperature value, c_1=1.91043*10-5Mw / (m2 sr cm-4), c_2=1.43878K, v_c is the central wavenumber of each channel, A and B are the channel brightness temperature correction coefficients, RAD0 is the amplified radiance value, RAD is the radiance, Slope is the slope, and Intercept is the intercept.

8. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 7, characterized in that: In step 4, the process of building a cloud detection model is as follows: According to the land surface type, the sample dataset is divided into six sub-datasets, including water bodies, evergreen coniferous forests, open shrubs, grasslands, bare land, and ice and snow sub-datasets; The integrated sample data set is divided into a training set and a test set. The AdaBoost machine learning model is trained using the training set data to build a cloud detection model. The trained cloud detection model is verified using the test set data to evaluate the performance of the cloud detection model, and the model evaluation indicators of accuracy, recall, and F1 score are calculated.

9. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 8, characterized in that: The model evaluation index formula is: Accuracy calculation formula: Accuracy=(TP+TN) / (TP+FN+FP+TN); F1 score calculation formula: F1 score=2(TP / (TP+FP)×TP / (TP+FN)) / (TP / (TP+FP)+TP / (TP+FN)); Among them, TP is the number of cloud pixels correctly identified by the model, TN is the number of clear sky pixels correctly identified by the model, FP is the number of clear sky pixels incorrectly identified by the model as having clouds, and FN is the number of clear sky pixels incorrectly identified by the model as having clouds.

10. The method for polar all-day cloud detection using FY-3D infrared channel brightness temperature according to claim 9, characterized in that: In step 5, the process of obtaining the cloud detection result is: Input the pixel longitude and latitude, surface information, instrument zenith angle, solar zenith angle, and infrared channel brightness temperature in the test data set into the trained model to perform cloud detection prediction and output the prediction results; Compare the model's predictions with the labels in the test set, calculate and summarize evaluation metrics, and generate an evaluation report to analyze and optimize the model.

Images