A sea fog monitoring method and system based on multi-source satellite remote sensing data

Abstract

The application belongs to the technical field of sea fog monitoring, and particularly relates to a sea fog monitoring method and system based on multi-source satellite remote sensing data, which comprises the following steps: collecting satellite radiation and microwave observation data, combining the response difference of infrared and microwave channels to distinguish the cloud top phase state, dividing clear sky and low cloud blocking scenes, and outputting identification data; combining single-time infrared brightness temperature difference and air pixel parameters to detect sea fog and retrieve optical parameters, forming a scene-specific sea fog physical parameter set; normalizing and coupling the parameter set for reconstruction, constructing a sea fog-specific feature space, separating the fog area and the non-fog area through phase state constraint to obtain an initial judgment feature field, correcting the fog area edge through neighborhood correlation and brightness temperature gradient constraint, strengthening the spatial continuity to generate a spatial distribution field, and locking a reliable area to output the final sea fog detection result. Through the progressive design of scene division, feature optimization and multi-source verification, the accuracy and integrity of sea fog identification are significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of sea fog monitoring technology, specifically a sea fog monitoring method and system based on multi-source satellite remote sensing data. Background Technology

[0002] Existing sea fog monitoring methods mostly rely on single-source infrared data, failing to effectively integrate the complementary characteristics of infrared and microwave data. This makes it difficult to distinguish sea fog from low clouds, resulting in poor scene adaptability. Most methods lack standardized radiometric correction and accurate spatiotemporal registration, leading to large data errors. Furthermore, they only use single-time brightness temperature, making them susceptible to sea surface disturbances and noise. During identification, fixed thresholds and single features are often used, without constructing a multi-dimensional feature space. This results in fragmented fog area edges, high noise levels, and a lack of physical parameter inversion. Additionally, the methods do not incorporate meteorological and visibility data for physical verification, easily leading to false fog areas. Consequently, detection accuracy and reliability are low, making it difficult to meet the high-precision operational monitoring needs of complex nearshore sea areas. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention proposes a sea fog monitoring method based on multi-source satellite remote sensing data. This invention primarily addresses key issues in traditional sea fog monitoring, such as susceptibility to misjudgment and missed detection, distortion in low cloud areas, fragmented fog edges, large data errors, and a lack of physical verification of results.

[0004] The present invention provides a sea fog monitoring method based on multi-source satellite remote sensing data, comprising: S1: collecting satellite observation radiation and microwave observation data, performing radiation correction and spatiotemporal registration, judging cloud top phase by combining infrared and microwave channel response differences, dividing clear sky and low cloud obstruction scenes, and outputting scene identification data.

[0005] S2: Select clear-sky areas based on scene identification data, synthesize multi-time clear-sky sea surface temperatures using satellite observation data, calculate the difference between multi-time clear-sky sea surface temperatures and the current single-time infrared brightness temperature, detect sea fog using spatial pixel analysis parameters, and retrieve optical parameters. Retrieve cloud and water paths in low cloud areas, and output scene-specific sea fog physical parameter sets.

[0006] S3: Perform multi-dimensional feature normalization and coupling reconstruction on the physical parameter set of sea fog in different scenarios to construct a sea fog-specific feature space that includes spectrum, microphysics, and spatial structure. Through phase consistency constraints, accurately separate fog areas from non-fog areas and output the initial judgment feature field of sea fog.

[0007] S4: Based on the initial feature field of sea fog, the spatial correlation of the neighborhood and the extreme value constraint of the brightness temperature gradient are used to adaptively aggregate and correct the broken pixels at the edge of the fog area and the transition region, thereby enhancing the spatial continuity of the fog area and outputting the spatial distribution field of sea fog.

[0008] S5: Physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data, complete the credible area locking through the verification of sea fog generation and dissipation environmental conditions, and output the final sea fog detection results.

[0009] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps for outputting scene identification data in step S1 are as follows:

[0010] S11: Perform radiometric calibration and atmospheric correction on the raw data of satellite and microwave observations to eliminate the effects of sensor errors and atmospheric attenuation, convert the raw observation values ​​into standardized radiometric values, and output a standardized radiometric dataset.

[0011] S12: Based on the standardized radiation dataset, the spatial registration of the infrared and microwave channels is completed using bilinear interpolation, and time synchronization correction is performed to ensure that the spatial resolution and observation time of the two types of data are completely consistent, and the radiation fusion data is output.

[0012] S13: Extract the brightness temperature data of the infrared band and microwave channel from the radiation fusion data, calculate the response difference and characteristic ratio between the two, and construct the cloud top phase state discrimination parameter set.

[0013] S14: Based on the cloud top phase discrimination parameter set and the phase physical characteristics of sea fog and low clouds, a discrimination threshold is set, each observed pixel is classified and judged, and clear sky and low cloud occlusion scenes are divided, and scene identification data is output.

[0014] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps in step S2 of outputting the sea fog physical parameter set for different scenarios are as follows:

[0015] S21: Based on the scene identification data, filter clear sky pixels, extract infrared brightness temperature data from multi-temporal satellite observations, remove outliers, and perform time-series averaging to generate multi-temporal clear sky sea surface temperature spatial distribution data.

[0016] S22: Obtain infrared brightness temperature data for corresponding pixel locations based on multi-time clear sky sea surface temperature spatial distribution data, and calculate the difference between the multi-time clear sky sea surface temperature and the infrared brightness temperature for each pixel to form brightness temperature difference feature field data.

[0017] S23: Based on the brightness temperature difference characteristic field data, use the empty pixel analysis parameters to set the sea fog discrimination threshold, determine the sea fog area range pixel by pixel, derive the optical thickness of the fog area and the effective radius of the particles, and output the sea fog characteristic data of the clear sky area.

[0018] S24: Determine the low cloud cover area based on scene identification data, retrieve the low-level cloud water path parameters based on microwave observation data and radiative transfer relationship, and output low cloud area cloud water characteristic data.

[0019] S25: Merge and match the sea fog feature data in clear sky areas with the cloud and water feature data in low cloud areas according to spatial pixel location to form a set of sea fog physical parameters covering the entire area in different scenarios.

[0020] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps in step S22 for forming brightness temperature difference feature field data are as follows:

[0021] Based on the spatial distribution data of clear sky sea surface temperature over multiple time periods, the corresponding spatial coordinate information is extracted pixel by pixel to construct a coordinate index sequence.

[0022] Based on the coordinate index sequence, the infrared brightness temperature values ​​of the corresponding pixels are extracted point by point in the current single-time satellite infrared observation data, generating single-time brightness temperature data that is strictly aligned with the location of the clear sky sea surface temperature data.

[0023] The difference between the sea surface temperature over multiple clear skies and the brightness temperature data over a single time period is calculated using a pixel-by-pixel point-to-point operation method. The difference results of all pixels are retained and arranged according to the original spatial grid to form brightness temperature difference feature field data.

[0024] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps for outputting the preliminary sea fog feature field in step S3 are as follows:

[0025] S31: Normalize the spectral, microphysical, and spatial structure features of the sea fog physical parameter set for each scene, map each feature to the same numerical range, eliminate dimensional differences, and output a normalized multidimensional feature sequence.

[0026] S32: Based on the normalized multi-dimensional feature sequence, feature-level concatenation and weighted fusion are performed according to the pixel position to complete the multi-dimensional feature coupling reconstruction and generate sea fog-specific feature space data.

[0027] S33: Based on the sea fog-specific feature space data, the feature vector is discriminated and verified pixel by pixel, non-fog feature interference is filtered out, and preliminary separated binary identification data is output.

[0028] S34: Spatial matching is performed between the initially separated binary identifier data and the normalized multidimensional feature sequence. The feature components corresponding to the fog area are retained and arranged according to the original grid. Redundant feature information is removed to form the initial feature field for sea fog judgment.

[0029] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps in step S32 for generating sea fog-specific feature spatial data are as follows:

[0030] Based on the normalized multidimensional feature sequence, the corresponding spectral, microphysical, and spatial structure normalized feature components are extracted according to the position of a single pixel to form a single pixel feature group.

[0031] Single-pixel feature groups are concatenated at the feature level according to their dimensionality to generate a high-dimensional original feature vector with uniform dimension and complete structure.

[0032] Adaptive weights are assigned to different feature components based on their contribution to sea fog identification, and element-wise weighted operations are performed on the high-dimensional original feature vector to generate sea fog-specific feature space data.

[0033] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps for outputting the spatial distribution field of sea fog in step S4 are as follows:

[0034] S41: Based on the initial feature field of sea fog, extract the spatial correlation index in the local neighborhood pixel by pixel, calculate the feature similarity between neighborhood pixels, and output the distribution data of the spatial correlation degree in the neighborhood.

[0035] S42: Construct edge constraints based on the distribution data of correlation degree in the neighborhood space and the extreme values ​​of brightness temperature gradient, identify broken pixels and transition regions at the edge of the fog area, and output the set of pixel locations to be corrected.

[0036] S43: Based on the set of pixel locations to be corrected, an adaptive aggregation algorithm is used to merge and correct fragmented pixels at the edges, optimize the boundary morphology of the fog area, enhance spatial continuity, and generate a spatial distribution field of sea fog.

[0037] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps in step S42 of outputting the set of pixel locations to be corrected are as follows:

[0038] Based on the distribution data of correlation degree in the neighborhood space, the correlation coefficient values ​​of all pixels in the region are extracted. Combined with the brightness temperature gradient amplitude calculated by the initial judgment feature field of sea fog, the gradient extreme value critical value is determined, and a dual edge constraint condition is constructed.

[0039] By comparing the correlation coefficient and gradient magnitude of the neighborhood space on a pixel-by-pixel basis under the dual edge constraint conditions, pixels that do not meet the constraint requirements are screened out, and broken pixels and transition regions at the edge of the fog area are preliminarily identified.

[0040] Neighborhood verification is performed on edge fragmented pixels and transition regions to eliminate misjudged pixels, record the spatial row and column coordinates of valid pixels to be corrected, and output the set of pixel locations to be corrected.

[0041] According to the sea fog monitoring method based on multi-source satellite remote sensing data provided by the present invention, the specific steps for outputting the final sea fog detection result in step S5 are as follows:

[0042] S51: Unify the spatial coordinates of the sea fog spatial distribution field with the meteorological field and visibility observation data to the same coordinate system, perform point-to-point physical matching according to the pixel position, and output the fused dataset.

[0043] S52: Extract key environmental condition parameters for the formation and dissipation of sea fog from the fused dataset, construct environmental verification criteria, verify the rationality of the existence of sea fog pixel by pixel, filter out false fog area pixels that do not meet the formation and dissipation conditions, and obtain reliable spatial distribution data of fog areas.

[0044] S53: Spatial locking is performed on the spatial distribution data of the reliable fog area. Combined with the spatial distribution of sea fog, meteorological environment and visibility correlation information, redundant data is removed and the sea fog detection results are output.

[0045] The present invention also provides a sea fog monitoring system based on multi-source satellite remote sensing data, including: a scene segmentation module, used to collect satellite observation radiation and microwave observation data, perform radiation correction and spatiotemporal registration, determine cloud top phase by combining infrared and microwave channel response differences, segment clear sky and low cloud obstruction scenes, and output scene identification data.

[0046] The parameter inversion module is used to select clear-sky areas based on scene identification data, synthesize multi-time clear-sky sea surface temperatures using satellite observation data, calculate the difference between the multi-time clear-sky sea surface temperatures and the current single-time infrared brightness temperature, detect sea fog using spatial pixel analysis parameters, and invert optical parameters. It also inverts cloud and water paths in low-cloud areas and outputs scene-specific sea fog physical parameter sets.

[0047] The feature construction module is used to normalize and reconstruct the physical parameter set of sea fog in different scenarios in multiple dimensions, and construct a sea fog-specific feature space that includes spectrum, microphysics and spatial structure. Through phase consistency constraints, it achieves accurate separation between fog area and non-fog area and outputs the initial judgment feature field of sea fog.

[0048] The edge correction module is used to adaptively aggregate and correct broken pixels and transition regions at the edge of the fog area based on the initial feature field of the sea fog and by using neighborhood spatial correlation and brightness temperature gradient extreme value constraints. This enhances the spatial continuity of the fog area and outputs the spatial distribution field of the sea fog.

[0049] The result verification module is used to physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data. It completes the credible area locking through the verification of sea fog generation and dissipation environmental conditions and outputs the final sea fog detection results.

[0050] This invention provides a sea fog monitoring method based on multi-source satellite remote sensing data. The beneficial effects of this invention are as follows:

[0051] 1. This invention effectively addresses the shortcomings of traditional methods that suffer from the aliasing of clear sky, low cloud, and sea fog signals by dividing the scene into clear sky and low cloud-covered scenarios and employing differentiated inversion strategies for each scenario. In the clear sky region, multi-timescale clear sky sea surface temperatures are synthesized, and sea fog is detected and its optical thickness and effective particle radius are retrieved by combining brightness temperature difference values ​​and spatial pixel analysis. The stability of Fengyun-4 satellite data is utilized to improve the accuracy of parameter inversion. In the low cloud region, cloud water paths are retrieved through microwave observation and the RTTOV model, avoiding interference from low cloud cover on infrared signals. This scenario-based processing mode ensures the accurate extraction of sea fog-related parameters in different regions, avoids the limitations of a single inversion method, and makes the output scenario-based sea fog physical parameter set more reliable, providing high-quality data support for subsequent sea fog identification.

[0052] 2. This invention constructs a sea fog-specific feature space. By normalizing and coupling the spectral, microphysical, and spatial structural features, it overcomes the limitations of traditional single-feature recognition and strengthens the feature differences between sea fog and non-fog regions. Through phase consistency constraints and neighborhood verification, it initially achieves accurate separation between fog and non-fog areas. Based on this, it utilizes neighborhood spatial correlation and brightness temperature gradient extreme value constraints to adaptively aggregate and correct fragmented pixels at the fog region edges, effectively solving the problems of blurred fog region edges and numerous isolated noise points, strengthening the spatial continuity of the fog region, and making the output sea fog spatial distribution field boundary clear and structurally complete, significantly improving the accuracy and completeness of sea fog recognition.

[0053] 3. This invention integrates multi-source data from satellite remote sensing, meteorological fields, and visibility observations. By unifying spatial coordinates and performing point-to-point matching, it eliminates spatiotemporal biases in the data. Based on the environmental conditions of sea fog formation and dissipation, a verification criterion is constructed, filtering out false fog pixels pixel by pixel to ensure that the detection results conform to physical laws. Confidential fog areas are identified and multi-source information is integrated to output a final result with complete parameters. The fusion and verification of multi-source data not only compensates for the shortcomings of a single data source but also further enhances the reliability of the results through environmental condition verification. This ensures that the final sea fog detection results accurately reflect the actual distribution and core characteristics of sea fog, and can be directly applied to operational scenarios such as sea fog monitoring and early warning, possessing strong practical value and promotional significance. Attached Figure Description

[0054] The invention will now be further described with reference to the accompanying drawings.

[0055] Figure 1 This is a flowchart illustrating the steps of a sea fog monitoring method based on multi-source satellite remote sensing data provided in an embodiment of the present invention.

[0056] Figure 2 This is a flowchart of a sea fog monitoring method based on multi-source satellite remote sensing data provided in an embodiment of the present invention;

[0057] Figure 3This is a block diagram of a sea fog monitoring system based on multi-source satellite remote sensing data provided in an embodiment of the present invention. Detailed Implementation

[0058] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below according to specific embodiments.

[0059] like Figures 1 to 3 As shown in the figure, an embodiment of the present invention provides a sea fog monitoring method based on multi-source satellite remote sensing data, the method comprising:

[0060] S1: Collect satellite observation radiation and microwave observation data, perform radiation correction and spatiotemporal registration, determine cloud top phase by combining infrared and microwave channel response differences, classify clear sky and low cloud obscured scenes, and output a preprocessed dataset with scene labels.

[0061] S11: Perform radiometric calibration and atmospheric correction on the raw data of satellite and microwave observations to eliminate the effects of sensor errors and atmospheric attenuation, convert the raw observation values ​​into standardized radiometric values, and output a standardized radiometric dataset.

[0062] Radiometric calibration is performed on the raw radiation data observed by satellite. The observed DN values ​​are converted into radiance information based on the sensor calibration coefficients, and sensor gain, bias and nonlinearity errors are removed.

[0063] The formula for converting observed DN values ​​into radiance information based on sensor calibration coefficients is expressed as follows:

[0064]

[0065] In the formula, DN is the entrance pupil radiance, which is dimensionless. Gain and Offset are scaling factors.

[0066] The formula for brightness temperature calibration of microwave observation data is expressed as follows:

[0067]

[0068] In the formula, For microwave brightness temperature, The voltage is the voltage being observed, and C and D are the instrument coefficients.

[0069] The formula for correction using the atmospheric radiative transfer model is expressed as follows:

[0070]

[0071] In the formula, To standardize radiance, Atmospheric transmittance, , The upward and downward radiance of the atmosphere must satisfy the conservation of radiative energy.

[0072] Brightness temperature calibration and instrument bias correction were performed on the raw microwave observation data, and the observation signals were uniformly converted into standard brightness temperature form. Then, atmospheric transmittance and upward radiation components were calculated using the atmospheric radiative transfer model, and atmospheric attenuation correction was applied to the radiance and brightness temperature data. All raw observations were converted into equivalent standardized radiation data of the Earth's surface, forming a standardized radiation dataset.

[0073] S12: Based on the standardized radiation dataset, the spatial registration of the infrared and microwave channels is completed using bilinear interpolation, and time synchronization correction is performed to ensure that the spatial resolution and observation time of the two types of data are completely consistent, and the radiation fusion data is output.

[0074] Based on the standardized radiation dataset, the pixel row and column coordinates and geographic latitude and longitude information of the infrared and microwave channels are extracted to establish the spatial coordinate mapping relationship between the multi-source data. The bilinear interpolation algorithm is used to spatially resample the data at different resolutions to achieve pixel-level spatial matching of all channels. At the same time, linear time interpolation is performed based on the observation timestamp to eliminate the observation time difference between the multi-source data, so that all data are aligned at the same time and under the same spatial grid, generating radiation fusion data that is completely consistent in time and space.

[0075] S13: Extract the brightness temperature data of the infrared band and microwave channel from the radiation fusion data, calculate the response difference and characteristic ratio between the two, and construct the cloud top phase state discrimination parameter set.

[0076] Based on radiative fusion data, the observed brightness temperature value corresponding to the infrared band is extracted pixel by pixel. A two-dimensional matrix of infrared brightness temperature spatial distribution field is constructed using pixel row and column indices to ensure a one-to-one correspondence between each pixel location and its geographic coordinates, outputting the infrared brightness temperature spatial distribution field data. Based on the same radiative fusion data, the brightness temperature values ​​corresponding to the microwave channel are extracted point-by-point at identical pixel locations. A microwave brightness temperature spatial distribution field is constructed using a grid structure consistent with the infrared brightness temperature field to ensure strict matching of pixel positions, and the microwave brightness temperature spatial distribution field data is output. Using the infrared and microwave brightness temperature spatial distribution fields, the brightness temperature difference characteristics are calculated pixel-by-pixel. Following the conservation of radiant energy differences, the brightness-temperature difference characteristic field data is output. Within the same pixel grid, the brightness-temperature ratio characteristic is calculated point-by-point. The ratio is a dimensionless quantity that reflects the difference in response to cloud top phase state across different spectral bands, outputting brightness-temperature ratio feature field data. Based on the brightness-temperature difference feature field, the spatial gradient components in the row and column directions are calculated pixel by pixel, and the gradient magnitude is synthesized. This is used to characterize the intensity of boundary changes between foggy and non-foggy areas, and outputs brightness temperature gradient feature field data. The brightness temperature difference, brightness temperature ratio, and brightness temperature gradient corresponding to each pixel are superimposed according to spatial location to form a multi-dimensional feature vector combination. This combination has significant distinguishing ability for water clouds, ice clouds, clear-sky sea surfaces, and sea fog. It is integrated in pixel order into a cloud top phase discrimination parameter set that can be directly used for phase identification.

[0077] S14: Based on the cloud top phase discrimination parameter set and the phase physical characteristics of sea fog and low clouds, a discrimination threshold is set, each observed pixel is classified and judged, and clear sky and low cloud occlusion scenes are divided, and scene identification data is output.

[0078] Based on the cloud top phase discrimination parameter set, and combined with the physical differences between sea fog and low clouds in terms of phase, altitude, and brightness temperature characteristics, multi-dimensional dynamic discrimination conditions are set to perform point-by-point discrimination for each pixel. Pixels that meet the characteristics of clear sky and ocean are marked as clear sky scenes, and pixels that meet the characteristics of low-level cloud occlusion are marked as low-level cloud occlusion scenes. Isolated noise points and abnormal pixels are removed through spatial connectivity analysis, and scene labeling data with clear classification results are output.

[0079] S2: Based on the preprocessed dataset with scene labels, multi-time clear-sky sea surface temperature (SST) and single-time infrared brightness temperature (IRB) differences are synthesized using Fengyun-4 data in clear-sky areas. This is supplemented by spatial pixel analysis parameters to detect sea fog in the Yellow Sea and retrieve optical parameters. Cloud and water paths are retrieved in low-cloud areas, outputting scene-specific sea fog physical parameter sets. Satellite observation data was obtained under clear-sky conditions using my country's second-generation geostationary meteorological satellite, Fengyun-4, taking advantage of the stable and slow-changing characteristics of sea surface temperature.

[0080] S21: Based on the pixel identifiers marked as clear sky in the scene identifier data, filter out all clear sky pixels in the entire monitoring area row by row and column by column, record the spatial row and column coordinates and geographical latitude and longitude information of each clear sky pixel, ensure that no clear sky pixel is missed and no low cloud-obscured pixels are mixed in, and output a list of clear sky pixel coordinates and corresponding identifier data.

[0081] Based on this clear-sky pixel coordinate list, the original satellite infrared brightness temperature data for multiple consecutive observation times corresponding to each clear-sky pixel are accurately extracted, ensuring that each clear-sky pixel has a corresponding brightness temperature observation value at each consecutive time, forming a three-dimensional data matrix of pixel coordinates-observation time-infrared brightness temperature.

[0082] Outlier detection is performed on the temporal brightness temperature data of each clear-sky pixel in the three-dimensional data matrix. The 3σ criterion is used to calculate the mean μ and standard deviation σ of the temporal brightness temperature data of each pixel. An outlier judgment threshold is set, and brightness temperature values ​​exceeding the threshold range are judged as outliers and removed. The effective temporal brightness temperature data sequence of each clear-sky pixel is output.

[0083] For each clear-sky pixel, after removing outliers, the effective temporal brightness temperature data is averaged separately according to the pixel position. That is, the arithmetic mean of the effective brightness temperature data of a single pixel is calculated to eliminate random errors and small fluctuations in single-time observations, so that the calculation results are closer to the real clear-sky sea surface temperature.

[0084] The temporal average brightness temperature values ​​of all clear-sky pixels are filled into the spatial grid of the monitoring area according to their original spatial coordinates. The low-cloud-obscured pixel areas that were not selected are marked blank. Finally, multi-time clear-sky sea surface temperature spatial distribution data with uniform spatial grid, stable values ​​and strict correspondence with the geographical coordinates of the observation area are generated. This data can be directly used for subsequent brightness temperature difference calculation.

[0085] S22: Obtain infrared brightness temperature data for corresponding pixel locations based on multi-time clear sky sea surface temperature spatial distribution data, and calculate the difference between the multi-time clear sky sea surface temperature and the infrared brightness temperature for each pixel to form brightness temperature difference feature field data.

[0086] Based on multi-time clear sky sea surface temperature spatial distribution data, all valid clear sky pixels contained therein are traversed, and the corresponding spatial row and column coordinates (x, y, z) are extracted pixel by pixel. i y i (i=1,2,...,N, where N is the total number of pixels in clear sky) and geographic latitude and longitude coordinates (l oni , l ati The coordinate information is arranged according to the cell number to construct the coordinate index sequence Index={(x i y i , l oni , l ati The coordinate information of each clear sky pixel is unique and verifiable, providing a precise positioning benchmark for subsequent brightness temperature matching.

[0087] Based on the coordinate index sequence Index, extract the spatial row and column coordinates (x, y, y). i y i Using this as a matching benchmark, the corresponding (x) is searched point by point in the current single-time satellite infrared observation data. i y i Infrared brightness temperature value T at position ) b,IR (x i y i ), remove invalid pixels with mismatched coordinates, and generate a single-time brightness temperature data sequence TIR={T b,IR (x i y i The sequence is defined as |i=1,2,...,N}, ensuring that the pixel positions of the sequence strictly correspond one-to-one with the pixel positions of the multi-time clear sky sea surface temperature data, without offset or misalignment.

[0088] Retrieves sea surface temperature values ​​for corresponding pixels from multi-time clear-sky sea surface temperature spatial distribution data. The corresponding values ​​in the single-time brightness temperature data sequence TIR generated above. The difference is calculated using a pixel-by-pixel point-to-point operation method, expressed by the formula as follows: The difference results retain all pixels with clear skies. According to its original spatial row and column coordinates The corresponding data is filled into the spatial grid of the monitoring area, and non-clear sky pixel areas are marked as blank, ultimately forming brightness temperature difference feature field data. .

[0089] S23: Based on brightness temperature difference characteristic field data First, statistical analysis was performed on the empty pixels in the entire monitoring area to extract the empty pixel analysis parameters, including the mean gray value of the empty pixels. Spatial variance and neighborhood gradient mean By combining the typical brightness temperature response characteristics of Yellow Sea fog in the infrared channel, a dynamic fog discrimination threshold is constructed.

[0090] The formula is expressed as:

[0091]

[0092] Wherein, k is an adaptive adjustment coefficient, ranging from 1.2 to 1.5, calibrated according to actual observations in the sea area. Subsequently, brightness temperature difference feature field data are compared pixel by pixel. With discrimination threshold ,when ≥ When this occurs, the pixel is determined to be a foggy pixel. < At that time, pixels were identified as non-fog area pixels, and spatial mask data of fog areas in the Yellow Sea was generated. Based on this fog area mask, the brightness temperature difference values ​​of all fog area pixels were selected. By combining the atmospheric radiative transfer forward model, the relationship between the brightness temperature difference and the optical thickness τ of sea fog and the effective particle radius is established. The quantitative correlation is obtained by iteratively calculating and deriving the optical thickness and effective particle radius of each fog pixel. The parameters such as fog pixel coordinates, fog range, optical thickness, and effective particle radius are integrated according to spatial location, and finally the sea fog feature data in clear sky area is output. This data only contains the physical parameters related to sea fog in clear sky scene and strictly corresponds to the clear sky area in scene identification data.

[0093] S24: Based on the scene identification data, traverse all pixels in the entire monitoring area, filter out the set of pixels marked as low-cloud obstruction, record the spatial row and column coordinates of each low-cloud pixel, determine the spatial range, boundary outline, and geographic latitude and longitude range of the low-cloud obstruction area, and output a list of low-cloud area coordinates. Based on this coordinate list, extract multi-channel microwave brightness temperature data corresponding to the low-cloud pixel locations from the microwave observation data that has completed radiometric correction and spatiotemporal registration. The physical correlation between microwave brightness temperature and low-level cloud water path (LWP) is constructed by combining the RTTOV model, and an objective functional is established. in, To observe microwave brightness temperature, To simulate microwave brightness temperature in the model, an iterative optimization algorithm was used for inversion calculation to minimize the residual between the simulated and observed brightness temperatures until the results stabilized and converged. Low-level cloud water path parameters were then obtained pixel by pixel. The low-cloud pixel coordinates, low-cloud area range, and corresponding cloud-water path parameters are organized and archived to output low-cloud area cloud-water feature data. This data is accurately matched with the low-cloud occlusion area in the scene identification data, providing core parameters for subsequent scene-by-scene fusion.

[0094] S25: Project the clear-sky sea fog feature data and the low-cloud cloud water feature data to the same map projection system and spatial grid resolution to ensure spatial consistency between the two types of data. Using the pixel row and column number as the sole matching criterion, precisely match the fog area parameters in the clear-sky sea fog feature data with the cloud water path parameters in the low-cloud cloud water feature data point by point according to the pixel spatial coordinates. This ensures seamless connection of the sea fog parameters in the clear-sky area and the cloud water parameters in the low-cloud area on the same spatial grid, without overlap, misalignment, or missing parameters. Subsequently, integrate the parameters of the two types of data, fusing all sea fog-related physical parameters such as fog area identification, optical thickness, effective particle radius, and low-level cloud water path according to the pixel dimension to form a three-dimensional data matrix of pixel coordinates-scene type-physical parameters. For blank pixels outside the coverage area of ​​the two types of data, use spatial constraint interpolation to reasonably fill them, ensuring parameter integrity. Finally, a set of physical parameters for sea fog in different scenarios is generated, which covers the entire monitoring area, is strictly distinguished by clear sky and low cloud scenarios, and has complete and spatially continuous parameters. This dataset can be directly used for the optimization and output of subsequent sea fog detection results.

[0095] S3: Perform multi-dimensional feature normalization and coupling reconstruction on the physical parameter set of sea fog in different scenarios to construct a sea fog-specific feature space that includes spectrum, microphysics, and spatial structure. Through phase consistency constraints, accurately separate fog areas from non-fog areas and output the initial judgment feature field of sea fog.

[0096] S31: Based on the physical parameter set of sea fog in different scenarios, three core feature parameters are first separated: spectral features, microphysical features, and spatial structure features. Min-max normalization is then performed on each type of feature parameter. This operation maps features with different dimensions and numerical ranges to the [0,1] interval, completely eliminating interference caused by differences in dimensions and numerical levels, ensuring balanced weights for all types of features in subsequent fusion. Finally, all normalized features are organized according to pixel position, outputting an ordered normalized multi-dimensional feature sequence, with each pixel corresponding to a normalized feature vector containing the three types of features.

[0097] S32: Based on the normalized multi-dimensional feature sequence, extract the three types of normalized feature vectors corresponding to each pixel: spectral, microphysical, and spatial structure. Concatenate these feature vectors at the pixel level according to their position, and then combine them into a single high-dimensional feature vector. Based on the priority requirements for sea fog recognition, assign adaptive weights to the three types of features, and perform multi-dimensional feature coupling and reconstruction using a weighted fusion formula. Arrange the fused high-dimensional feature vectors of all pixels according to a spatial grid to construct a dedicated sea fog feature space data, containing three-dimensional features of spectral, microphysical, and spatial structure, specifically suitable for sea fog recognition. This feature space effectively highlights the feature differences between sea fog and non-fog areas, providing core support for subsequent accurate separation.

[0098] S33: Based on the sea fog-specific feature space data, extract the high-dimensional fused feature vector for each pixel. Perform pixel-by-pixel discrimination and verification of the feature vector according to the phase consistency constraint. Calculate the similarity between the feature vector and the standard sea fog feature template. When the similarity is higher than a set threshold and the phase consistency constraint is met, it is initially determined to be a fog zone pixel. Otherwise, it is determined to be a non-fog zone pixel. Filter out non-fog feature interference pixels with abnormal feature vectors or spatially isolated features. Use 3×3 neighborhood verification to further eliminate misjudged pixels, ensuring the reliability of the discrimination results. Finally, output the binary identifier data that initially separates the fog zone from the non-fog zone, achieving the initial separation of fog and non-fog areas.

[0099] S34: The initially separated binarized identifier data and normalized multidimensional feature sequences are imported into the same spatial coordinate system. Using the pixel row and column number as the unique matching benchmark, precise spatial matching is performed pixel by pixel to ensure that the binarized identifier and normalized features are completely aligned and correspond one-to-one on the spatial grid. Based on the binarized identifier data, all fog area pixels marked as 1 are selected, and the corresponding normalized multidimensional feature components are retained. The feature components corresponding to non-fog area pixels marked as 0 are removed, eliminating redundant non-fog feature information. The retained fog area feature components are filled into the spatial grid of the monitoring area according to their original spatial row and column coordinates. Blank pixels at the edge of the fog area are supplemented using neighborhood feature interpolation to ensure the spatial continuity of the feature field. Finally, a spatially continuous, feature-stable preliminary sea fog judgment feature field containing only the core features of the fog area is formed. This feature field can be directly used for subsequent fog area edge correction and precision optimization.

[0100] S4: Based on the initial feature field of sea fog, the spatial correlation of the neighborhood and the extreme value constraint of the brightness temperature gradient are used to adaptively aggregate and correct the broken pixels at the edge of the fog area and the transition region, thereby enhancing the spatial continuity of the fog area and outputting the spatial distribution field of sea fog.

[0101] S41: Based on the initial feature field of sea fog, a 3×3 or 5×5 local sliding neighborhood is constructed with each pixel as the center. Multi-dimensional information such as fog area identification, brightness temperature difference, and microphysical features within the neighborhood is extracted pixel by pixel. The feature similarity and spatial correlation coefficient between neighboring pixels are calculated, and a spatial correlation measurement matrix is ​​constructed. The degree of neighborhood spatial correlation of each pixel is obtained through global traversal operation. Finally, the spatially continuous, pixel-by-pixel labeled neighborhood spatial correlation distribution data is output.

[0102] S42: Based on the distribution data of spatial correlation in the neighborhood, and combined with the brightness temperature gradient amplitude and gradient direction information simultaneously calculated in the initial sea fog judgment feature field, a lower threshold for the correlation coefficient and an upper threshold for the gradient extremum are set to construct a dual constraint condition for the fog area edge. The entire region's pixels are traversed, and pixels that do not meet the spatial correlation constraint and are located at gradient abrupt change positions are identified as edge fragmented pixels or transitional pixels. Their row and column coordinates and geographic coordinates are recorded, and a set of pixel locations to be corrected is output.

[0103] S43: Based on the set of pixel locations to be corrected, an adaptive aggregation algorithm is used to merge and relabel edge fragmented pixels, isolated noise pixels, and blurred transition pixels, with the correlation degree of the neighborhood space as the weight. The target pixel category is adaptively corrected according to the proportion of effective fog area pixels in the neighborhood, smoothing the jagged and fragmented structure of the fog area boundary, enhancing the spatial connectivity and geometric regularity of the fog area, eliminating the interference of false edges and isolated points, and generating a sea fog spatial distribution field with complete spatial structure, clear boundaries, and strong continuity.

[0104] S5: Physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data, complete the credible area locking through the verification of sea fog generation and dissipation environmental conditions, and output the final sea fog detection results.

[0105] S51: Based on the spatial distribution field of sea fog, extract its spatial row and column coordinates and geographic latitude and longitude information to determine its original map projection system. Use this coordinate system as a unified benchmark to calibrate the coordinates of meteorological field data and visibility observation data. Through coordinate transformation algorithms, the grid coordinates of the meteorological field and the latitude and longitude coordinates of the visibility observation stations are uniformly transformed to the WGS84 coordinate system, ensuring that the spatial benchmarks of the three types of data are completely consistent and eliminating coordinate offsets and projection differences. Subsequently, using the pixel row and column numbers of the sea fog spatial distribution field as the matching benchmark, point-to-point physical matching is performed according to the pixel position to match the sea fog identifiers, meteorological element values, and visibility observation values ​​under the same spatial coordinates. Invalid data with mismatched coordinates are eliminated, and a small amount of missing observation data is supplemented. Finally, a three-dimensional fusion dataset containing sea fog spatial information, meteorological environment information, and visibility information is formed, providing complete and aligned data support for subsequent environmental verification.

[0106] S52: Based on the fused dataset, key environmental parameters for sea fog formation and dissipation are extracted pixel-by-pixel. Core parameters include: relative humidity in the lower atmosphere, near-sea wind speed, sea-air temperature difference, and visibility. These parameters are essential for sea fog formation and maintenance. Based on these key parameters, environmental verification criteria for the reasonableness of sea fog existence are constructed, setting effective threshold ranges for each parameter and employing multi-condition joint verification logic. The fog area determination results in the fused dataset are verified pixel-by-pixel, comparing the corresponding environmental parameters with the threshold ranges. If any parameter exceeds the threshold range, the fog area pixel is determined to be a false fog area pixel and filtered out. If all parameters meet the threshold requirements, the fog area pixel is retained, ultimately obtaining spatially continuous and reliable spatial distribution data of fog areas that conforms to the physical laws of sea fog formation and dissipation.

[0107] S53: Based on the spatial distribution data of reliable fog areas, a spatial contour extraction algorithm is used to spatially lock all reliable fog areas that meet the verification conditions. The boundary contour, spatial range, and core area of ​​each reliable fog area are clearly defined. The coordinate information of the fog area edges and internal pixels is marked, and isolated noise points and residual pixels from pseudo-fog areas around the fog areas are removed. Subsequently, combining the details of sea fog spatial distribution, meteorological environmental parameters, and visibility correlation information from the fused dataset, the data is integrated by pixel dimension. Duplicate parameter information and redundant non-fog area data are removed, the data format is standardized, and annotation information of the core fog area parameters is supplemented. Accurate, reliable, and parameter-complete sea fog detection results are output, which can be directly used for sea fog monitoring, early warning, and subsequent operational applications.

[0108] like Figure 3As shown, the present invention also provides a sea fog monitoring system based on multi-source satellite remote sensing data, comprising:

[0109] The scene segmentation module is used to collect satellite observation radiation and microwave observation data, perform radiation correction and spatiotemporal registration, determine cloud top phase by combining the response differences of infrared and microwave channels, segment clear sky and low cloud obscured scenes, and output scene identification data.

[0110] The parameter inversion module is used to select clear-sky areas based on scene identification data, synthesize multi-time clear-sky sea surface temperatures using satellite observation data, calculate the difference between the multi-time clear-sky sea surface temperatures and the current single-time infrared brightness temperature, detect sea fog in the Yellow Sea using spatial pixel analysis parameters, and invert optical parameters. It also inverts cloud and water paths in low-cloud areas and outputs scene-specific sea fog physical parameter sets.

[0111] The feature construction module is used to normalize and reconstruct the physical parameter set of sea fog in different scenarios in multiple dimensions, and construct a sea fog-specific feature space that includes spectrum, microphysics and spatial structure. Through phase consistency constraints, it achieves accurate separation between fog area and non-fog area and outputs the initial judgment feature field of sea fog.

[0112] The edge correction module is used to adaptively aggregate and correct broken pixels and transition regions at the edge of the fog area based on the initial feature field of the sea fog and by using neighborhood spatial correlation and brightness temperature gradient extreme value constraints. This enhances the spatial continuity of the fog area and outputs the spatial distribution field of the sea fog.

[0113] The result verification module is used to physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data. It completes the credible area locking through the verification of sea fog generation and dissipation environmental conditions and outputs the final sea fog detection results.

[0114] In summary, this embodiment provides a sea fog monitoring method and system based on multi-source satellite remote sensing data. By constructing a sea fog-specific feature space, it normalizes and couples the spectral, microphysical, and spatial structure features for reconstruction, overcoming the limitations of traditional single-feature recognition and enhancing the feature differences between sea fog and non-fog areas. Through phase consistency constraints and neighborhood verification, it initially achieves accurate separation between fog and non-fog areas. Furthermore, by utilizing neighborhood spatial correlation and brightness temperature gradient extreme value constraints, it adaptively aggregates and corrects fragmented pixels at the fog area edges, effectively solving the problems of blurred fog area edges and numerous isolated noise points, enhancing the spatial continuity of the fog area, and making the output sea fog spatial distribution field clear and structurally complete, significantly improving the accuracy and completeness of sea fog identification.

[0115] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring sea fog based on multi-source satellite remote sensing data, characterized in that, include: S1: Collect satellite observation radiation and microwave observation data, perform radiation correction and spatiotemporal registration, combine the differences in infrared and microwave channel responses to determine cloud top phase, classify clear sky and low cloud obstruction scenes, and output scene identification data; S2: Select clear sky areas based on the scene identification data, synthesize multi-time clear sky sea surface temperature using satellite observation data, calculate the difference between the multi-time clear sky sea surface temperature and the current single-time infrared brightness temperature, use spatial pixel analysis parameters to detect sea fog and invert optical parameters; invert cloud water paths in low cloud areas and output scene-specific sea fog physical parameter sets; S21: Based on the scene identification data, filter clear sky pixels, extract infrared brightness temperature data from multi-temporal satellite observations, remove outliers, perform time-series averaging, and generate multi-temporal clear sky sea surface temperature spatial distribution data. S22: Obtain infrared brightness temperature data for the corresponding pixel position based on the spatial distribution data of clear sky sea temperature over multiple time periods, and calculate the difference between the clear sky sea temperature over multiple time periods and the infrared brightness temperature pixel by pixel to form brightness temperature difference feature field data. S23: Based on the brightness temperature difference feature field data, use the empty pixel analysis parameters to set the sea fog discrimination threshold, determine the sea fog area range pixel by pixel, deduce the optical thickness of the fog area and the effective radius of the particles, and output the sea fog feature data of the clear sky area. S24: Determine the low cloud cover area based on the scene identification data, retrieve the low cloud water path parameters based on microwave observation data and radiative transfer relationship, and output the cloud water characteristic data of the low cloud area. S25: Merge and match the sea fog feature data in clear sky areas with cloud and water feature data in low cloud areas according to spatial pixel location to form a set of sea fog physical parameters covering the entire area in different scenarios; S3: Perform multi-dimensional feature normalization and coupling reconstruction on the physical parameter set of sea fog in the sub-scene to construct a sea fog-specific feature space containing spectrum, microphysics, and spatial structure. Complete the accurate separation of fog area and non-fog area through phase consistency constraint and output the initial judgment feature field of sea fog. S4: Based on the initial feature field of sea fog, the spatial correlation of the neighborhood and the extreme value constraint of the brightness temperature gradient are used to adaptively aggregate and correct the broken pixels and transition regions at the edge of the fog area, enhance the spatial continuity of the fog area, and output the spatial distribution field of sea fog. S41: Based on the initial feature field of sea fog, extract the spatial correlation index in the local neighborhood pixel by pixel, calculate the feature similarity between neighborhood pixels, and output the distribution data of the spatial correlation degree in the neighborhood. S42: Construct edge constraints based on the neighborhood spatial correlation distribution data and brightness temperature gradient extreme values, identify broken pixels and transition regions at the edge of the fog area, and output the set of pixel locations to be corrected. S43: Based on the set of pixel locations to be corrected, an adaptive aggregation algorithm is used to merge and correct the fragmented pixels at the edges, optimize the boundary morphology of the fog area, enhance spatial continuity, and generate a sea fog spatial distribution field. S5: Physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data, complete the credible area locking through the verification of sea fog generation and dissipation environmental conditions, and output the final sea fog detection result.

2. The sea fog monitoring method based on multi-source satellite remote sensing data according to claim 1, characterized in that: In step S1, the specific steps for outputting scene identification data are as follows: S11: Perform radiometric calibration and atmospheric correction on the raw data of satellite and microwave observations to eliminate sensor errors and atmospheric attenuation, convert the raw observation values ​​into standardized radiometric values, and output a standardized radiometric dataset. S12: Based on the standardized radiation dataset, use bilinear interpolation to complete the spatial registration and time synchronization correction of the infrared and microwave channels, so that the spatial resolution and observation time of the two types of data are completely consistent, and output radiation fusion data. S13: Extract the brightness temperature data of the infrared band and microwave channel from the radiation fusion data, calculate the response difference and feature ratio between the two, and construct the cloud top phase state discrimination parameter set; S14: Based on the cloud top phase discrimination parameter set and the phase physical characteristics of sea fog and low clouds, a discrimination threshold is set, each observed pixel is classified and judged, clear sky and low cloud occlusion scenes are divided, and scene identification data is output.

3. The sea fog monitoring method based on multi-source satellite remote sensing data according to claim 1, characterized in that: In step S22, the specific steps for forming the brightness temperature difference feature field data are as follows: Based on the spatial distribution data of clear sky sea surface temperature over multiple time periods, the corresponding spatial coordinate information is extracted pixel by pixel to construct a coordinate index sequence. Based on the coordinate index sequence, the infrared brightness temperature values ​​of the corresponding pixels are extracted point by point in the current single-time satellite infrared observation data to generate single-time brightness temperature data that is strictly aligned with the location of the clear sky sea surface temperature data. The difference between the sea surface temperature over multiple clear skies and the brightness temperature data over a single time period is calculated using a pixel-by-pixel point-to-point operation method. The difference results of all pixels are retained and arranged according to the original spatial grid to form brightness temperature difference feature field data.

4. The sea fog monitoring method based on multi-source satellite remote sensing data according to claim 1, characterized in that: In step S3, the specific steps for outputting the initial sea fog feature field are as follows: S31: Normalize the spectral, microphysical and spatial structure features of the physical parameter set of sea fog in the sub-scene, map each feature to the same numerical range, eliminate the difference in dimensions, and output a normalized multi-dimensional feature sequence. S32: Based on the normalized multi-dimensional feature sequence, feature-level concatenation and weighted fusion are performed according to pixel position to complete the multi-dimensional feature coupling reconstruction and generate sea fog-specific feature space data. S33: Based on the sea fog-specific feature space data, the feature vector is discriminated and verified pixel by pixel, non-fog feature interference is filtered out, and preliminary separated binary identification data is output. S34: Spatial matching is performed between the initially separated binary identifier data and the normalized multi-dimensional feature sequence, retaining the feature components corresponding to the fog area and arranging them according to the original grid, removing redundant feature information, and forming a preliminary sea fog judgment feature field.

5. A sea fog monitoring method based on multi-source satellite remote sensing data according to claim 4, characterized in that: In step S32, the specific steps for generating sea fog-specific feature space data are as follows: Based on the normalized multidimensional feature sequence, the corresponding spectral, microphysical, and spatial structure normalized feature components are extracted according to the position of a single pixel to form a single pixel feature group; The single-pixel feature groups are concatenated at the feature level according to the dimensional order to generate a high-dimensional original feature vector with uniform dimension and complete structure. Adaptive weights are assigned to different feature components based on their contribution to sea fog identification, and element-wise weighted operations are performed on the high-dimensional original feature vector to generate sea fog-specific feature space data.

6. The sea fog monitoring method based on multi-source satellite remote sensing data according to claim 1, characterized in that: In step S42, the specific steps for outputting the set of pixel locations to be corrected are as follows: Based on the correlation distribution data of the neighborhood space, the correlation coefficient values ​​of all pixels in the region are extracted. Combined with the brightness temperature gradient amplitude calculated by the initial judgment feature field of sea fog, the gradient extreme value critical value is determined, and a dual edge constraint condition is constructed. Based on the dual edge constraint conditions, the correlation coefficient and gradient magnitude of the neighborhood space are compared pixel by pixel to filter out pixels that do not meet the constraint requirements, and the broken pixels and transition regions at the edge of the fog area are preliminarily identified. Neighborhood verification is performed on the edge-broken pixels and transition regions to eliminate misjudged pixels, the spatial row and column coordinates of the valid pixels to be corrected are recorded, and the set of pixel locations to be corrected is output.

7. The sea fog monitoring method based on multi-source satellite remote sensing data according to claim 1, characterized in that: In step S5, the specific steps for outputting the final sea fog detection result are as follows: S51: The spatial coordinates of the sea fog spatial distribution field are uniformly calibrated to the same coordinate system with the meteorological field and visibility observation data, and point-to-point physical matching is performed according to the pixel position to output the fused dataset. S52: Extract key environmental condition parameters for sea fog formation and dissipation from the fused dataset, construct environmental verification criteria, verify the rationality of sea fog existence pixel by pixel, filter out false fog area pixels that do not meet the formation and dissipation conditions, and obtain reliable fog area spatial distribution data. S53: Spatial locking is performed on the spatial distribution data of the reliable fog area. Combined with the spatial distribution of sea fog, meteorological environment and visibility correlation information, redundant data is removed and the sea fog detection results are output.

8. A sea fog monitoring system based on multi-source satellite remote sensing data, comprising a sea fog monitoring method based on multi-source satellite remote sensing data as described in any one of claims 1 to 7, characterized in that, The monitoring system includes: The scene segmentation module is used to collect satellite observation radiation and microwave observation data, perform radiation correction and spatiotemporal registration, determine cloud top phase by combining the response differences of infrared and microwave channels, segment clear sky and low cloud obstruction scenes, and output scene identification data. The parameter inversion module is used to select clear sky areas based on the scene identification data, synthesize multi-time clear sky sea surface temperature using satellite observation data, calculate the difference between the multi-time clear sky sea surface temperature and the current single-time infrared brightness temperature, detect sea fog using spatial pixel analysis parameters and invert optical parameters; invert cloud water paths in low cloud areas and output scene-specific sea fog physical parameter sets. The feature construction module is used to perform multi-dimensional feature normalization and coupling reconstruction on the physical parameter set of sea fog in different scenarios, construct a sea fog-specific feature space including spectrum, microphysics, and spatial structure, and achieve accurate separation of fog area and non-fog area through phase consistency constraint, and output the initial judgment feature field of sea fog. The edge correction module is used to adaptively aggregate and correct the broken pixels and transition regions at the edge of the fog area based on the initial feature field of the sea fog and by using neighborhood spatial correlation and brightness temperature gradient extreme value constraints. This enhances the spatial continuity of the fog area and outputs the spatial distribution field of the sea fog. The result verification module is used to physically match and fuse the spatial distribution field of sea fog with meteorological field and visibility observation data, complete the credible area locking through the verification of sea fog generation and dissipation environmental conditions, and output the final sea fog detection result.

Images