Multi-band combined polar daytime sea fog / low cloud detection method, device and medium

Abstract

The present invention discloses a multi-band combined rapid detection method, device and medium for polar daytime sea fog / low clouds. The method includes: acquiring MODIS data and performing preprocessing; fusing the solar zenith angle data with the preprocessed MODIS data to obtain a MODIS image with the fused solar zenith angle; constructing a sea fog / low cloud detection model SFLDM based on multi-wave combination, wherein the image is divided into daytime and twilight for separate detection using the solar zenith angle; and extracting sea fog / low clouds in the MODIS image using the SFLDM model. The present invention solves the influence of the change of the solar zenith angle on the detection effect of polar sea fog / low clouds, and can accurately, efficiently and rapidly realize the detection of polar daytime sea fog / low clouds.

Description

Technical Field

[0001] The present invention relates to the field of environmental monitoring and meteorological forecasting, and particularly to a method, device, and medium for detecting polar daytime sea fog / low clouds based on multi-band combination. Background Art

[0002] With global warming, the rapid ablation of Arctic sea ice has occurred. Climate models predict that ice-free summers in the Arctic will start from 2034, bringing new opportunities for the opening of the Arctic shipping lane in summer. However, the increase in temperature has led to enhanced sea surface evaporation, and the frequency of sea fog and low clouds on the shipping lane has been increasing continuously, which has a great impact on navigation safety and has become one of the most dangerous weather phenomena in maritime transportation safety. Therefore, carrying out Arctic sea fog / low cloud detection is of great significance for ensuring the safety of Arctic shipping. However, the current lack of monitoring data from Arctic ground observation stations, ships, buoys, etc. makes it difficult to support the monitoring requirements of large-scale sea fog / low clouds on the shipping lane, which has become a bottleneck problem to be solved for safe navigation in the Arctic shipping lane.

[0003] Polar-orbiting remote sensing satellites have a high temporal resolution in this region, and at the same time have characteristics such as a large observation range and low cost, providing a rich potential data source for large-scale detection of Arctic sea fog / low clouds. Traditional remote sensing monitoring algorithms for sea fog / low clouds mostly use visible-light to near-infrared reflection characteristics and infrared radiation characteristics, and are widely used in sea fog detection in mid-latitude and low-latitude regions. However, the above methods still face great challenges in Arctic sea fog / low cloud detection. Compared with mid-latitude and low-latitude regions, the underlying surface conditions of Arctic sea fog / low clouds are more complex, with sea ice and sea water distributed alternately. In addition, due to the high latitude and weak solar radiation in the Arctic region, the reflectivity differences between different land cover types are small and are easily affected by the solar zenith angle, resulting in the reflectivity of sea ice / snow being very close to that of sea fog / low clouds. The infrared band is relatively less affected by sunlight, and the infrared multi-band reflection and radiation differences of each object at different angles are relatively obvious, which is expected to provide another analysis idea for sea fog monitoring. However, relevant technologies are still rarely reported. Summary of the Invention

[0004] The present invention provides a method, device, and medium for detecting polar daytime sea fog / low clouds based on multi-band combination, which can accurately, efficiently, and quickly detect polar daytime sea fog / low clouds.

[0005] To achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A method for quickly detecting polar daytime sea fog / low clouds based on multi-band combination, comprising:

[0007] Step 1, obtaining MODIS data and performing preprocessing;

[0008] Step 2: Fuse the solar zenith angle data with the preprocessed MODIS data to obtain the MODIS image with the fused solar zenith angle.

[0009] Step 3: Construct a sea fog / low cloud detection model SFLDM based on multi-wave combination. Among them, use the solar zenith angle to divide the image into day and twilight for separate detection.

[0010] Step 4: Use the SFLDM model to extract sea fog / low clouds in the MODIS image.

[0011] Furthermore, the preprocessing of the MODIS data includes:

[0012] Step A1.1: Radiometric calibration. Calibrate the 1st - 19th and 26th bands in the acquired original MODIS data to reflectance, and calibrate the 20th - 25th and 27th - 36th bands to brightness temperature.

[0013] Step A1.2: Geometric positioning. Uniformly convert the calibrated MODIS data to the WGS - 84 coordinate system, select the UTM projection for the projection method, and at the same time resample the data spatial resolution to 1 km.

[0014] Step A1.3: Land masking. Use the standard Arctic land boundary SHP file to perform masking processing on the MODIS data, and assign the pixel value of the land part to 0.

[0015] Furthermore, the specific process of fusing the solar zenith angle data with the preprocessed MODIS data is as follows:

[0016] Step B1: Resampling of the solar zenith angle data. Resample the solar zenith angle data in the originally acquired MODIS data to make the size of the solar zenith angle data the same as that of the preprocessed MODIS data.

[0017] Step B2: Fuse the solar zenith angle data after Step B1 with the preprocessed MODIS data. The specific steps are as follows:

[0018] Step B2.1: Geometric registration. Select the GCP ground control point method for the registration method, and perform pixel - by - pixel matching and correspondence between the pixel points in the solar zenith angle data and the pixel points in the MODIS image.

[0019] Step B2.2: Land masking of the solar zenith angle data. Use the standard Arctic land boundary SHP file to perform masking processing on the solar zenith angle data, and assign the pixel value of the land part to 0.

[0020] Step B2.3, data fusion: fuse the solar zenith angle data after step B2.2 with the pre-processed MODIS data to obtain a MODIS image with fused solar zenith angle.

[0021] Furthermore, the sea fog / low cloud detection model SFLDM adopts different detection methods according to daytime and dawn and dusk;

[0022] For daytime sea fog / low cloud detection, the specific steps are as follows:

[0023] Step C2.1, seawater, sea ice / snow removal: use R B7(2.130μm) Fixed threshold and BTD B20-B31 Fixed thresholds for removing seawater, sea ice / snow;

[0024] Step C2.2, medium and high cloud removal: SFLCRI dynamic threshold and BT are used in turn. B31(11.030μm) Fixed threshold to remove medium and high clouds;

[0025] For sea fog / low cloud detection at dawn and dusk, the specific steps are as follows:

[0026] Step C3.1, removal of seawater, sea ice / snow, and medium-high clouds: Use the SFLCRI fixed threshold to preliminarily remove seawater, sea ice / snow, and medium-high clouds;

[0027] Step C3.2, remove the medium and high clouds: use BT B31(11.030μm) The brightness temperature threshold is used to perform secondary removal of medium and high clouds.

[0028] Furthermore, it is characterized in that the method of dividing dawn and dusk and daytime is: using the solar zenith angle value to divide the daytime MODIS images into two categories: daytime and dawn and dusk, among which the solar zenith angle of [0°, 70°] is divided into daytime images, and the solar zenith angle of [70°, 90°] is divided into dawn and dusk images.

[0029] Furthermore, SFLCRI is the sea fog / low cloud detection index, and the calculation formula is:

[0030]

[0031] Where R B5(1.240μm) and R B18(0.936μm) They are the band reflectances of MODIS images MODIS B5 and B18 respectively.

[0032] Furthermore, the specific process of removing medium and high clouds in step C2.2 includes:

[0033] Step C2.2.1, preliminary removal of medium and high clouds:

[0034] The OTSU method is used to determine the SFLCRI extraction threshold T for sea fog / low clouds. When the determined SFLCRI extraction threshold T is outside the range of (1.7, 2.3), the fixed threshold for distinguishing middle and high clouds from sea fog / low clouds is set to 1.8, and the pixels with SFLCRI ≤ 1.8 are classified as middle and high clouds for removal. Otherwise, the pixels with SFLCRI ≤ T are classified as middle and high clouds for removal;

[0035] Step C2.2.2, further removal of middle and high clouds:

[0036] Use the brightness temperature threshold BT B31(11.030μm) = 267K to perform secondary optimization on the detection results of sea fog / low clouds after Step C2.2.1. The pixels with BT B31(11.030μm) ≤ 267K in the sea fog / low cloud results are classified as middle and high clouds for further removal;

[0037] After Step C2.2.1, Step C2.2.2 and Step C2.2.3 are in a parallel relationship;

[0038] Step C2.2.3, extraction of thin sea fog / low clouds: Use the brightness temperature threshold BT B31(11.030μm) = 269K to perform secondary optimization on the detection results of middle and high clouds after Step C2.2.1. The pixels with BT B31(11.030μm) > 269K in the middle and high cloud detection results are classified as thin sea fog / low clouds;

[0039] Step C2.2.4: Take the union of the sea fog / low cloud detection results after Step C2.2.2 and the sea fog / low cloud detection results after Step C2.2.3 to merge and obtain the final sea fog / low cloud detection results.

[0040] Furthermore, in Step C2.1, the fixed threshold of R B7(2.130μm) is set to 0.02, and the fixed threshold of BTD B20-B31 is set to 4K. The pixels with R B7(2.130μm) ≤ 0.02 are initially removed as sea water, sea ice / snow, and the pixels with BTD B20-B31 ≤ 4K are further removed as sea water, sea ice / snow;

[0041] In Step C2.2, the fixed threshold of BT B31(11.030μm) is set to 267K. The pixels with BT B31(11.030μm) ≤ 267K are removed as middle and high clouds;

[0042] In Step C3.1, the fixed threshold of SFLCRI is set to 2.3. The pixels with SFLCRI ≤ 2.3 are removed as sea water, sea ice / snow and middle and high clouds;

[0043] In Step C3.2, the brightness temperature threshold of BT B31(11.030μm) is set to 267K. The pixels with BTB31(11.030μm) Pixels ≤ 267K are removed twice as mid - high clouds.

[0044] An electronic device includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor implements the multi - band joint polar daytime sea fog / low cloud rapid detection method described in any one of the above technical solutions.

[0045] A computer - readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the multi - band joint polar daytime sea fog / low cloud rapid detection method described in any one of the above technical solutions.

[0046] Beneficial effects

[0047] The technical solution of the present invention provides a sea fog / low cloud detection model for polar daytime using comprehensive dynamic and static thresholds - SFLDM, and constructs a multi - band joint polar daytime sea fog / low cloud detection method.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: Aiming at the blank in the research on Arctic sea fog / low cloud detection, a sea fog / low cloud detection algorithm applicable to Arctic daytime is proposed, which solves the influence of the change of solar zenith angle on the sea fog / low cloud detection effect. A sea fog / low cloud detection index SFLCRI is constructed based on the B18 (0.936μm) water vapor absorption band and the B5 (1.240μm) atmospheric window, which solves the problem that it is difficult to separate the sea surface from sea fog / low clouds at dawn and dusk in previous algorithms. At the same time, the coupling of SFLCRI and the BT B31(11.030μm) threshold effectively improves the detection accuracy of sea fog / low clouds, and solves the problem of missed detection of mid - high clouds by the BT B31(11.030μm) threshold when there is an inversion layer in the past. In terms of sea surface elimination, the present invention proposes to use the radiance difference BTD B20-B31 for sea surface elimination. By coupling R B7(2.130μm) , the elimination accuracy of the sea surface is improved. At the same time, a solar zenith angle constraint condition is introduced to avoid the problem of false detection of the sea surface caused by the change of the solar zenith angle. The verification results of CALIOP sounding radar data show that the overall accuracy of the present invention is relatively high, and the detection accuracy, false alarm rate and reliability of sea fog / low clouds are 91.02%, 14.22% and 78.52% respectively. Description of the drawings

[0049] Figure 1 It is a flowchart of a rapid detection algorithm for Arctic summer daytime sea fog / low clouds based on multi - band joint.

[0050] Figure 2The research area concerned by the present invention includes the sea-land distribution, remote sensing images and CALIOP transit maps, where NEP is the Northeast Passage, TSR is the Central Passage, and NWP is the Northwest Passage;

[0051] Figure 3 It is a flow chart of the fusion of solar zenith angle data and MODIS images;

[0052] Figure 4 It is a flow chart of the SFLDM model;

[0053] Figure 5 It is a curve of the reflectance of the observed object in the study area changing with time;

[0054] Figure 6 It is for the BT B20(3.750μm) 、BT B31(11.030μm) and BTD B20-B31 value curves changing with time;

[0055] Figure 7 It is for R B7(2.130μm) and BTD B20-B31 scatter plot and fixed threshold detection results

[0056] Figure 8 It is a curve of the SFLCRI value of the observed object changing with time;

[0057] Figure 9 It is a MODIS true color image and sea fog / low cloud detection result map at 04:00, 07:15, 11:45, 15:00 and 21:30 on July 15, 2016;

[0058] Figure 10 It is the sea fog / low cloud detection area at 13 moments from 02:20 to 21:30. Specific implementation manner

[0059] The following is a detailed description of the embodiments of the present invention. Based on the technical solution of the present invention, detailed implementation manners and specific operation processes are given, and the technical solution of the present invention is further explained.

[0060] This embodiment provides a multi-band joint rapid detection method for Arctic summer daytime sea fog / low clouds. Referring to Figure 1 shown below, it includes the following steps:

[0061] Step 1: Preprocessing of raw data, perform preprocessing operations on the used MODIS data and CALIOP data.

[0062] Step A1: MODIS data is remote sensing image data obtained by the Moderate-resolution Imaging Spectroradiometer carried by the polar-orbiting meteorological satellites Terra and Aqua, abbreviated as MODIS data. Its preprocessing steps include:

[0063] Step A1.1: Radiometric calibration. There are 36 bands in the original MODIS data. Among them, bands 1-19 and 26 are calibrated to reflectance, and bands 20-25 and 27-36 are calibrated to brightness temperature;

[0064] Step A1.2: Geometric positioning. The calibrated MODIS data is uniformly converted to the WGS-84 coordinate system. The projection method is selected as UTM projection, and at the same time, the data spatial resolution is resampled to 1 km;

[0065] Step A1.3: Land masking. The MODIS data is masked by means of the standard Arctic land boundary SHP file, and the pixel values of the land part are assigned 0.

[0066] Step A2: CALIOP data is lidar data jointly developed by CNES and NASA. The vertical feature layer distribution data (VFM) in the CALIOP data level 2 product is selected to verify the sea fog / low cloud detection results. In the VFM data, the targets with the detection target being Cloud, the cloud top height less than 2000 m, and clear air above the cloud top are defined as sea fog / low clouds.

[0067] Figure 2 Shows the land-sea distribution, remote sensing image, and CALIOP radar transit map of the study area concerned in this embodiment.

[0068] Step 2: Obtain the solar zenith angle value of each pixel of the satellite image, and fuse the solar zenith angle data with the preprocessed MODIS data. The specific process is as follows:

[0069] Step B1: Resampling of solar zenith angle data. The solar zenith angle data is provided in the MODIS data, but the size of this data set (271*406) is different from the size of the preprocessed MODIS data (data size 1354*2030). Therefore, it is necessary to resample the solar zenith angle data. The sampling method is selected as cubic convolution resampling, and the size of the solar zenith angle data is uniformly sampled to 1354*2030.

[0070] Step B2: Fuse the solar zenith angle data after step B1 with the MODIS data after step A1. The specific steps are as follows:

[0071] Step B2.1: Geometric registration. The GCP ground control point method is selected for registration, and pixel-by-pixel matching and correspondence are carried out between the pixel points in the solar zenith angle data and the pixel points in the MODIS image.

[0072] Step B2.2: Land masking of solar zenith angle data. The solar zenith angle data is masked with the help of the standard Arctic land boundary SHP file, and the pixel values of the land part are set to 0.

[0073] Step B2.3: Data fusion. The solar zenith angle data after Step B2.2 is fused with the MODIS image after Step A1 to obtain a MODIS image with fused solar zenith angle.

[0074] Figure 3 The specific process of fusing the solar zenith angle and the MODIS image is shown.

[0075] Step 3: Construct a sea fog / low cloud detection model - SFLDM. The Arctic sea area consists of four types of objects: sea water, sea ice / snow, sea fog / low cloud, and mid-high cloud. To conduct sea fog / low cloud detection, appropriate spectral indices and models need to be constructed to sequentially separate and distinguish each object. According to the different solar zenith angles, the model is divided into two parts: daytime and twilight. The specific process is as follows:

[0076] Step C1: Division of twilight and daytime images. The daytime images are divided into daytime and twilight categories using the solar zenith angle value. Among them, the solar zenith angle of the daytime image is defined as [0°, 90°], and the images with a solar zenith angle of [0°, 70°] are divided into daytime images, while the images with a solar zenith angle of [70°, 90°] are divided into twilight images.

[0077] Step C2: Detection of sea fog / low cloud during daytime (solar zenith angle [0°, 70°]). The specific steps are as follows:

[0078] Step C2.1: Removal of sea water, sea ice / snow, using the R B7(2.130μm) fixed threshold and BTD B20-B31 fixed threshold to remove sea water, sea ice / snow. The specific process is as follows:

[0079] Step C2.1.1: Preliminary removal of sea water, sea ice / snow. As Figure 5 (b) shows, the reflectance of sea water, sea ice / snow in R B7(2.130μm) is relatively low. Therefore, R B7(2.130μm) = 0.02 is selected to remove sea water and sea ice / snow. The pixels with R B7(2.130μm) ≤ 0.02 are excluded as sea water, sea ice / snow, that is, the pixel values are set to 0.

[0080] Step C2.1.2: Further removal of sea water, sea ice / snow. As Figure 6As shown in (c), the bright temperature difference value on seawater, sea ice / snow BTD B20-B31 is relatively low, so BTD B20-B31 = 4K is selected for seawater and sea ice / snow removal, and the pixels with BTD B20-B31 ≤ 4K are excluded as seawater, sea ice / snow.

[0081] Among them, as Figure 7 shown, the removal accuracy of seawater and sea ice / snow with R B7(2.130μm) ≤ 0.02 and BTD B20-B31 ≤ 4K is 98.83%.

[0082] Step C2.2: Mid-high cloud removal, using SFLCRI dynamic threshold and BT B31(11.030μm) fixed threshold to remove mid-high clouds.

[0083] Step C2.2.1: Preliminary mid-high cloud removal. Based on the reflectance of two water vapor absorption near-infrared bands of MODIS B5 (1.240μm) and B18 (0.936μm), a sea-fog and low-cloud ratio index (SFLCRI) is constructed to remove mid-high clouds. The calculation formula is:

[0084]

[0085] In the formula, R B5(1.240μm) and R B18(0.936μm) are the reflectances of MODIS B5 (1.240μm) and B18 (0.936μm) bands respectively. The SFLCRI time series statistical results of different observation objects are as Figure 8 shown. Among them, the SFLCRI values of seawater, sea ice / snow, and mid-high clouds are relatively low and stable, while the SFLCRI value of sea-fog and low-clouds is relatively high. The sample statistical results show that the SFLCRI extraction threshold T interval of sea-fog and low-clouds is generally (1.7, 2.3). The threshold dynamically obtained according to the actual situation can better perform separation detection. The OTSU method is used to determine the optimal segmentation threshold T. OTSU mainly uses the minimum square distance from the pixel to various representative elements (centers) as the value condition, and divides an image into background and foreground parts through the threshold T. The foreground image is sea-fog and low-clouds. Let the proportion of the foreground image in the whole image be P1(t), the gray mean be μ1(t), the proportion of the background image be P0(t), the gray mean be μ0(t), and the between-class variance σ 2 can be expressed as:

[0086] σ 2 = P0(t)P1(t)(μ0(t) - μ1(t)) 2

[0087] Traverse the threshold T from the minimum gray value to the maximum gray value. When T makes the between-class variance σ 2 reach the maximum, select T as the optimal segmentation threshold, and divide SFLCRI ≤ T into medium and high clouds. However, when the content of sea fog / low clouds in the image of the study area is relatively low or high, the segmentation threshold T obtained by OTSU is likely to be outside the interval (1.7, 2.3). At this time, SFLCRI = 1.8 is used as the fixed threshold to distinguish medium and high clouds from sea fog / low clouds.

[0088] Step C2.2.2: Further removal of medium and high clouds. The temperature of medium and high clouds is relatively low, quite different from other observed objects, and not affected by the solar zenith angle. In this algorithm, the brightness temperature threshold BT B31(11.030μm) = 267K is used to perform secondary optimization on the detection results of sea fog / low clouds after step C2.2.1, and the pixels with BT B31(11.030μm) ≤ 267K in the sea fog / low cloud results are divided into medium and high clouds.

[0089] Among them, after step C2.2.1, steps C2.2.2 and C2.2.3 are in a parallel relationship.

[0090] Step C2.2.3: Extraction of thin sea fog / low clouds. When the thickness of sea fog / low clouds is relatively thin, due to its limited radiation shielding of the sea surface, its R B5(1.240μm) and R B18(0.936μm) are close to the sea surface. Therefore, its SFLCRI value is lower than the threshold T, and using the SFLCRI threshold is likely to misjudge it as medium and high clouds. However, its brightness temperature is significantly different from that of medium and high clouds. During the day, its BT B31(11.030μm) is greater than 269K. In this algorithm, the brightness temperature threshold BT B31(11.030μm) = 269K is used to perform secondary optimization on the detection results of medium and high clouds after step C2.2.1, and the pixels with BT B31(11.030μm) > 269K in the medium and high cloud detection results are divided into thin sea fog / low clouds.

[0091] Step C2.2.4: Take the union of the sea fog / low cloud detection results after step C2.2.2 and the sea fog / low cloud detection results after step C2.2.3, and merge to obtain the final sea fog / low cloud detection results.

[0092] Step C3: Detection of sea fog / low clouds at dawn and dusk (solar zenith angle [70°, 90°]), and the specific steps are as follows:

[0093] Step C3.1: Removal of sea water, sea ice / snow and medium and high clouds. At dawn and dusk, the SFLCRI value of sea fog / low clouds is relatively high, quite different from the SFLCRI values of sea water, sea ice / snow and medium and high clouds. Select the fixed threshold SFLCRI ≤ 2.3 of SFLCRI to remove sea water, sea ice / snow and medium and high clouds.

[0094] Step C3.2: Further removal of mid- and high-level clouds, using the brightness temperature threshold BT B31(11.030μm) ≤267K to perform a secondary removal of mid- and high-level clouds.

[0095] Figure 4 The SFLDM model process is shown.

[0096] Figure 5 The curve of the reflectance of the observed object in the study area changing with time is shown.

[0097] Figure 6 The BT of the observed object is shown B20(3.750μm) , BT B31(11.030μm) and BTD B20-B31 value curves changing with time.

[0098] Figure 7 The R of 1204 sea surface sample points is shown B7(2.130μm) and BTD B20-B31 scatter plot and the fixed threshold detection results.

[0099] Figure 8 The curve of the SFLCRI value of the observed object changing with time is shown.

[0100] Step 4: Based on the new algorithm, input the MODIS detection dataset after fusing the solar zenith angle data to obtain the sea fog / low cloud detection results.

[0101] Step D1: Before the sea fog / low cloud detection, the data preprocessing step A1 and the solar zenith angle fusion step 2 need to be carried out. After completion in sequence, the image is input into the SFLDM model to obtain the sea fog / low cloud detection results.

[0102] Next, the accuracy verification of the detection results of the SFLDM model of the present invention is carried out:

[0103] Select the CALIOP active sounding radar data to qualitatively and quantitatively evaluate the reliability of the algorithm for the satellite sea fog / low cloud detection results at similar times in the summer of 2016 (June - August).

[0104] Figure 9 The MODIS true color images and the sea fog / low cloud detection results at 04:00, 07:15, 11:45, 15:00 and 21:30 on July 15, 2016 are shown. The remote sensing detection results of the sea fog / low cloud show that this sea fog belongs to typical advection fog at sea. The warm and moist air mass is at the front of the cyclone. Affected by the cyclone's movement, the warm and moist air mass gradually moves southeastward ( Figure 9(a), (c), (e) rectangular and circular areas), during the movement of the air mass, when it encounters the cold sea surface, it gradually transforms into sea fog. Sea fog appears in the Chukchi Sea area at 4:00 (UTC-11). Subsequently, the warm and humid air mass continues to transform into sea fog, and the fog area gradually expands towards the Beaufort Sea area. At 7:15 (UTC-11), the covered area of the sea fog is about 115,045 km 2 , at 11:45 (UTC-11), the warm and moist air mass continues to transform into sea fog, and the area of the sea fog expands to 248,021 km 2 , at 15:00 (UTC-11), the warm and humid air mass has completely transformed into sea fog, with a clear boundary and a stable shape. At this time, the area of the sea fog is as high as 301,416 km 2 . After that, affected by the wind, the sea fog enters the dissipation period and gradually drifts towards the southeast. By 21:30 (UTC-11), the fog area has drifted to the coastal area of Alaska. There is still some sea fog in the Beaufort Sea area, but the sea fog in the Chukchi Sea area has gradually dissipated. SFLDM can better reflect the generation, development and dissipation process of the sea fog in the study area. Among them, in the initial stage of the formation of the sea fog, due to the existence of the warm and moist air mass, Figure 9 (b), (d) the detection results of the fog area are relatively fragmented, but with the continuous movement and transformation of the air mass, Figure 9 (f) the detection results of the fog area have tended to be continuous. The detection results of the sea fog are highly consistent with the spatial changes of the warm and humid air mass and highly coincide with the detection results of the active radar CALIOP.

[0105] Figure 10 Shows the detected area of the heavy fog at 13 moments from 02:20 to 21:30. From the perspective of time and space, sea fog has formed at 2:20 in the morning. At this time, the area of the fog area is about 75,237 km 2 , after the sea fog has developed for 9 hours, the area has expanded 2.5 times and gradually reached the peak value. At 15:00 in the afternoon, the area reaches the peak value of 301,416 km 2 , then the fog gradually dissipates, but the dissipation speed is slow. By 21:30 at dusk, the area of the sea fog is still 195,615 km 2 , compared with the peak moment, it only decreases by 1 / 3. This advection fog has the characteristics of long duration, high intensity, rapid spatial change of the fog area and large coverage. In terms of spatial resolution, spectral resolution and time resolution, MODIS data can better meet the sea fog monitoring needs of the study area.

[0106] Select CALIOP data to quantitatively verify the detection results of the algorithm. The detection evaluation indicators include the probability of detection POD, the false alarm rate FAR and the critical success index CSI. Among them, POD represents the detection accuracy. The higher the POD value, the higher the accuracy of the algorithm. CSI represents the reliability degree. The higher the CSI value, the more reliable the algorithm. The calculation methods are as follows:

[0107] Table 1 Detection Results and CALIOP Data Matrix

[0108]

[0109]

[0110] Where: N H represents the number of pixels where both the CALIOP detection result and the MODIS detection result are sea fog / low cloud; N F represents the misdetected pixels (i.e., the number of pixels where CALIOP is non-sea fog / low cloud while MODIS is sea fog / low cloud), and conversely, N M represents the undetected pixels, that is, the number of pixels where CALIOP is sea fog / low cloud while MODIS detects non-sea fog / low cloud.

[0111] Table 2 shows the detection accuracy of sea fog / low cloud in 9 time periods during the summer (June - August) of 2016. It can be seen from Table 2 that the detection method of sea fog / low cloud in the summer of 2016 has a relatively high accuracy. The average values of POD, FAR, and CSI are 91.02%, 14.22%, and 78.52% respectively, indicating that the method of the present invention has relatively high stability and reliability for detecting summer fog in the Arctic shipping lane.

[0112] Table 2 Accuracy of Sea Fog / Low Cloud Detection Results (UTC-11)

[0113]

[0114] Further analysis shows that except on July 4th, POD is higher than 90% (average 94.46%) at other times. The reason for the low POD value on July 4th is the appearance of very thin sea fog / low cloud, making its BT MIR2.130μm and BTD MIR-TIR very close to the sea surface. FAR is generally lower than 20% (average value is 14.22%). The main reason is that there is a slight difference in the transit times of CALIOP and MODIS, up to 39 minutes. And sea fog / low cloud changes with time to a certain extent, especially when weather conditions are unstable or a cyclone exists, the change of sea fog / low cloud is larger. The asynchronous time of the verification data and the detection data leads to a slightly higher FAR value. For example, the misdetection rate on July 24th is as high as 80.95%. The main reason is the movement of clouds by the cyclone. Some mid-high clouds at the CALIOP detection time (11:22) are converted into sea fog / low cloud at the MODIS image imaging time (11:40), resulting in a large area of sea fog / low cloud with a true value of mid-high cloud in the CALIOP detection data. In addition, except on July 4th and July 24th, the CSI value remains stable at other times (average is 84.73%), further confirming the reliability of the method of the present invention.

[0115] The above embodiments are the preferred embodiments of the present application. Those of ordinary skill in the art can also make various transformations or improvements on this basis. Without departing from the general concept of the present application, these transformations or improvements should fall within the scope of protection required by the present application.

Claims

1. A rapid detection method for multi - band combined polar daytime sea fog / low clouds, characterized in that, Including: Step 1: Obtain MODIS data and perform preprocessing. Step 2: Fuse the solar zenith angle data with the preprocessed MODIS data to obtain a MODIS image with fused solar zenith angle. Step 3: Construct a sea fog / low cloud detection model SFLDM based on multi-wave combination. Among them, use the solar zenith angle to divide the image into day and twilight for separate detection. For the sea fog / low cloud detection model SFLDM, different detection methods are adopted according to day and twilight. For the detection of sea fog / low cloud during the day, the specific steps are as follows: Step C2.1, seawater, sea ice / snow removal: successively adopt the R B7(2.130μm) fixed threshold and BTD B20-B31 fixed threshold to remove seawater, sea ice / snow; Step C2.2, mid- and high-level cloud removal: successively use the SFLCRI dynamic threshold and the BT B31(11.030μm) fixed threshold to remove mid- and high-level clouds; SFLCRI is the sea fog / low cloud detection index; For the detection of sea fog / low cloud at twilight, the specific steps are as follows: Step C3.1: Removal of sea water, sea ice / snow and mid-high clouds: Adopt the SFLCRI fixed threshold to preliminarily remove sea water, sea ice / snow and mid-high clouds. Step C3.2, further removal of mid- and high-level clouds: Use the BT B31(11.030μm) brightness temperature threshold to perform secondary removal of mid- and high-level clouds; Step 4: Use the SFLDM model to extract sea fog / low cloud in the MODIS image.

2. The rapid detection method of sea fog / low cloud according to claim 1, characterized in that, The preprocessing of MODIS data includes: Step A1.1: Radiometric calibration: Calibrate the 1st - 19th and 26th bands in the acquired original MODIS data to reflectance, and calibrate the 20th - 25th and 27th - 36th bands to brightness temperature. Step A1.2: Geometric positioning: Uniformly convert the calibrated MODIS data to the WGS-84 coordinate system, select the UTM projection for the projection method, and resample the data spatial resolution to 1 km at the same time. Step A1.3: Land masking: Use the standard Arctic land boundary SHP file to perform masking processing on the MODIS data, and assign the pixel value of the land part to 0.

3. The rapid detection method of sea fog / low cloud according to claim 1, wherein The specific process of fusing the solar zenith angle data with the preprocessed MODIS data is as follows: Step B1: Resampling of solar zenith angle data: Resample the solar zenith angle data in the originally acquired MODIS data to make the size of the solar zenith angle data the same as that of the preprocessed MODIS data. Step B2: Fuse the solar zenith angle data after Step B1 with the preprocessed MODIS data. The specific steps are as follows: Step B2.1: Geometric registration: Select the GCP ground control point method for the registration method, and match and correspond the pixel points in the solar zenith angle data with the pixel points in the MODIS image pixel by pixel. Step B2.2: Land masking of solar zenith angle data: Use the standard Arctic land boundary SHP file to perform masking processing on the solar zenith angle data, and assign the pixel value of the land part to 0. Step B2.3: Data fusion: Fuse the solar zenith angle data after Step B2.2 with the preprocessed MODIS data to obtain a MODIS image with fused solar zenith angle.

4. The sea fog / low cloud rapid detection method according to claim 1, wherein The division method of twilight and day is: Use the solar zenith angle value to divide the daytime MODIS image into two categories: day and twilight. Among them, the image with a solar zenith angle of [0°, 70°] is divided into day images, and the image with a solar zenith angle of [70°, 90°] is divided into twilight images.

5. The rapid detection method of sea fog / low cloud according to claim 1, characterized in that The calculation formula of SFLCRI is: where R B5(1.240μm) and R B18(0.936μm) are the band reflectances of MODIS B5 and B18, respectively.

6. The rapid detection method of sea fog / low cloud according to claim 1, characterized in that The specific process of removing mid-high clouds in Step C2.2 includes: Step C2.2.1: Preliminary removal of mid-high clouds: The OTSU method is used to determine the SFLCRI extraction threshold T for sea fog / low clouds. When the determined SFLCRI extraction threshold T is outside the interval (1.7, 2.3), the fixed threshold for distinguishing middle and high clouds from sea fog / low clouds is set to 1.8, and the pixels with SFLCRI ≤ 1.8 are classified as middle and high clouds for removal. Otherwise, the pixels with SFLCRI ≤ T are classified as middle and high clouds for removal; Step C2.2.2, further removal of middle and high clouds: Using the brightness temperature threshold BT B31(11.030μm) = 267K to perform secondary optimization on the detection results of sea fog / low clouds after step C2.2.1, and classify the pixels with BT B31(11.030μm) ≤ 267K in the sea fog / low cloud results as mid-high clouds for further removal; After step C2.2.1, step C2.2.2 and step C2.2.3 are in a parallel relationship; Step C2.2.3, Thin sea fog / low cloud extraction: Use the brightness temperature threshold BT B31(11.030μm) = 269K to perform secondary optimization on the detection results of middle and high clouds after Step C2.2.1, and divide the pixels with BT B31(11.030μm) > 269K in the middle and high cloud detection results into thin sea fog / low clouds; Step C2.2.4: Take the union of the sea fog / low cloud detection results after step C2.2.2 and the sea fog / low cloud detection results after step C2.2.3 to obtain the final sea fog / low cloud detection result.

7. The rapid sea fog / low cloud detection method according to claim 1, wherein In step C2.1, R B7(2.130μm) The fixed threshold is set to 0.02, and BTD B20-B31 The fixed threshold is set to 4K. For R B7(2.130μm) Those with R ≤ 0.02 are preliminarily removed as seawater, sea ice / snow, and for BTD B20-B31 Pixels with BTD ≤ 4K are further removed as seawater, sea ice / snow; In step C2.2, BT B31(11.030μm) The fixed threshold is set to 267K, and the pixels of BT B31(11.030μm) ≤ 267K are removed as mid-high clouds; In step C3.1, the SFLCRI fixed threshold is set to 2.3, and the pixels with SFLCRI ≤ 2.3 are removed as sea water, sea ice / snow, and middle and high clouds; In step C3.2, BT B31(11.030μm) The brightness temperature threshold is set to 267K, and pixels where BT B31(11.030μm) ≤ 267K are used as mid- and high-level clouds for secondary removal.

8. An electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that, When the computer program is executed by the processor, the processor implements the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1 to 7.

Images