A method and system for intelligent weather image acquisition and weather image recognition

Abstract

This invention discloses a method and system for intelligent weather image acquisition and recognition, specifically relating to the field of weather image recognition technology. It involves continuous polarization imaging of the sky region while simultaneously acquiring spatial attitude and temporal information; preprocessing the polarization image data and extracting Stokes parameters to obtain standardized sky polarization light field data; then using a theoretical Rayleigh scattering model to analyze the polarization degree distribution pattern at different wavelengths to obtain theoretical sky polarization pattern characteristics; based on the polarization light field data, identifying the spatial distribution of actual clouds, aerosols, and haze in the sky to form actual atmospheric composition distribution characteristics; extracting anomalous polarization regions through difference analysis between theoretical and actual characteristics, and accurately classifying halo polarization anomalies and scattering polarization anomalies based on geometric and spectral characteristics; generating ultralocal weather change early warning signals based on the classification results; thus improving the accuracy and real-time performance of early warnings for ultralocal anomalous weather phenomena.

Description

Technical Field

[0001] This invention relates to the field of weather image recognition technology, and more specifically, to a method and system for intelligent weather image acquisition and weather image recognition. Background Technology

[0002] Existing sky weather image recognition methods typically use ordinary visible light cameras to capture and analyze sky images. These methods can only acquire basic visual information such as brightness and color of the sky, failing to capture the optical scattering characteristics and spatial distribution of atmospheric components such as clouds, aerosols, and haze. Furthermore, they cannot distinguish the polarization characteristics of different abnormal weather phenomena (such as halos and scattering-type weather phenomena), making it difficult to identify and locate actual areas of weather anomalies and thus hindering accurate early warning of localized weather changes. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for intelligent weather image acquisition and weather image recognition to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for intelligent weather image acquisition and weather image recognition includes the following steps:

[0006] S1: Acquire the original polarization image sequence of the target sky region through a multispectral polarization camera, and simultaneously record the camera's spatial attitude data and timestamp information, and output polarization image data;

[0007] S2: Preprocess the polarization image data and extract the Stokes parameter for each pixel to output standardized sky polarization light field data;

[0008] S3: Based on sky polarization field data, analyze the polarization degree distribution pattern under different wavelengths, and combine the solar azimuth angle to calculate the theoretical Rayleigh scattering model, outputting the theoretical sky polarization pattern characteristics;

[0009] S4: Based on sky polarization light field data, identify the spatial distribution of actual clouds, aerosols and haze, and output the actual atmospheric composition distribution characteristics;

[0010] S5: Compare the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, calculate the difference map in polarization degree and polarization angle dimensions, and extract anomalous polarization regions;

[0011] S6: Based on the geometric and spectral characteristics of the anomalous polarization region, distinguish between halo polarization anomalies and scattering polarization anomalies, and output the atmospheric optical anomaly classification results;

[0012] S7: Generate ultra-local weather change early warning signals based on atmospheric optical anomaly classification results.

[0013] In a preferred embodiment, S1 specifically refers to:

[0014] By continuously capturing images of the target sky region using a multispectral polarization camera, a sequence of original polarization images of the target sky region in different polarization directions under multiple spectral bands is obtained.

[0015] Real-time acquisition of spatial attitude data from a multispectral polarization camera;

[0016] Add unified timestamp information to the original polarization image sequence and the corresponding spatial attitude data respectively;

[0017] The original polarization image sequence data with uniform timestamp information is matched with the corresponding spatial attitude data to output polarization image data.

[0018] In a preferred embodiment, S2 specifically refers to:

[0019] Image denoising is performed on polarization image data to remove random noise and interfering pixels from the polarization image data.

[0020] For each frame of polarization image data after removing random noise and interfering pixels, extract the Stokes parameter data corresponding to each pixel in the polarization image data;

[0021] The Stokes parameter data of all pixels in each frame of polarized image data are normalized and transformed to output normalized sky polarized light field data.

[0022] In a preferred embodiment, S3 specifically refers to:

[0023] Based on sky polarization field data, determine the sky polarization degree distribution pattern for each spectral band;

[0024] For each spectral band, calculate the solar azimuth parameter in the corresponding spectral band based on the sky polarization distribution pattern.

[0025] Based on the solar azimuth angle parameters of each spectral band, and combined with Rayleigh scattering physics theory, the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle are calculated for each spectral band.

[0026] Based on the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle under each spectral band, theoretical sky polarization mode characteristic data are output.

[0027] In a preferred embodiment, S4 specifically refers to:

[0028] Based on sky polarization field data, we can identify the changes in sky polarization characteristics caused by clouds and determine the spatial distribution area of ​​clouds.

[0029] Based on sky polarization light field data, we can identify regions of sky polarization characteristic changes caused by aerosol particle scattering and determine the spatial distribution region of aerosol particles.

[0030] Based on sky polarization light field data, identify the regions of polarization light characteristic changes caused by haze particle scattering and determine the spatial distribution area of ​​haze particles.

[0031] Based on the spatial distribution areas of clouds, aerosol particles, and haze particles, the actual spatial distribution characteristics of atmospheric components are output.

[0032] In a preferred embodiment, S5 specifically refers to:

[0033] Calculate the degree of polarization and the angle of polarization difference between the theoretical sky polarization mode characteristic data and the actual spatial distribution characteristic data of atmospheric composition for each spectral band;

[0034] Based on the difference in polarization degree and the difference in polarization angle, generate a sky polarization degree difference distribution map and a polarization angle difference distribution map for each spectral band;

[0035] For each spectral band, the regions where the difference in polarization degree and the difference in polarization angle exceed a set threshold are extracted to identify abnormal polarization regions.

[0036] In a preferred embodiment, S6 specifically refers to:

[0037] For each abnormal polarization region, extract the boundary contour geometric parameters of the abnormal polarization region;

[0038] For each anomalous polarization region, calculate the average degree of polarization and average polarization angle of the anomalous polarization region in each spectral band;

[0039] Based on the boundary contour geometric parameters of the abnormal polarization region and the average degree of polarization and average polarization angle of the abnormal polarization region in each spectral band, the abnormal polarization type of each abnormal polarization region is determined to be either halo polarization abnormality type or scattering polarization abnormality type.

[0040] Based on the determination of the anomalous polarization type in the anomalous polarization region, the atmospheric optical anomaly classification result is output.

[0041] In a preferred embodiment, S7 specifically refers to:

[0042] Based on the atmospheric optical anomaly classification results, the spatial location parameters of each anomalous polarization region are extracted, including the center coordinate data and coverage data of the anomalous polarization region.

[0043] Based on the spatial position parameters of each anomalous polarization region, calculate the offset of each anomalous polarization region relative to the center position of the target sky region;

[0044] For the offset of each abnormal polarization region, the warning level of each abnormal polarization region is determined according to the preset offset threshold.

[0045] Based on the warning level and spatial location parameters of each anomalous polarization region, a hyperlocal weather change warning signal corresponding to each anomalous polarization region is generated.

[0046] On the other hand, the present invention provides a system for intelligent weather image acquisition and weather image recognition, comprising:

[0047] Image acquisition module: Acquires raw polarization image sequences of the target sky region through a multispectral polarization camera, and simultaneously records the camera's spatial attitude data and timestamp information, and outputs polarization image data;

[0048] Data processing module: preprocesses polarization image data, extracts Stokes parameters for each pixel, and outputs standardized sky polarization light field data;

[0049] Theoretical Feature Module: Based on sky polarization field data, analyze the polarization degree distribution pattern under different wavelengths, and calculate the theoretical Rayleigh scattering model by combining the solar azimuth angle, and output the theoretical sky polarization pattern characteristics;

[0050] Composition identification module: Based on sky polarization light field data, it identifies the spatial distribution of actual clouds, aerosols and haze, and outputs the actual atmospheric composition distribution characteristics;

[0051] Anomaly Extraction Module: By comparing the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, the module calculates the difference map in the dimensions of polarization degree and polarization angle, and extracts anomalous polarization regions.

[0052] Anomaly Classification Module: Based on the geometric and spectral characteristics of the anomalous polarization region, it distinguishes between halo polarization anomalies and scattering polarization anomalies, and outputs atmospheric optical anomaly classification results;

[0053] Early warning generation module: Based on the atmospheric optical anomaly classification results, it generates ultra-local weather change early warning signals.

[0054] The technical effects and advantages of the intelligent weather image acquisition and weather image recognition method and system of this invention are as follows:

[0055] By simultaneously recording polarization images, spatial attitude, and temporal information using a multispectral polarization camera, high spatiotemporal consistency of sky observation data is achieved. Image denoising and Stokes parameter extraction improve the reliability and accuracy of polarization field data. A Rayleigh scattering model is calculated by combining multi-band polarization degree distribution with solar azimuth angle calculation, providing an accurate theoretical benchmark. Cloud, aerosol, and haze distributions are identified based on polarization field data, enabling rapid and refined classification of atmospheric components. Abnormal polarization regions are accurately extracted through analysis of differences between theoretical and actual polarization characteristics, improving anomaly detection sensitivity. The geometric and spectral characteristics of abnormal regions are used to distinguish between halos and scattering anomalies, enhancing classification accuracy. Hyperlocal early warning signals are generated based on the classification results, enabling real-time and accurate early warning of sudden local weather changes. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of a method for intelligent weather image acquisition and weather image recognition according to the present invention;

[0057] Figure 2 This is a schematic diagram of the structure of a system for intelligent weather image acquisition and weather image recognition according to the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example 1

[0059] Figure 1 This invention provides a method for intelligent weather image acquisition and weather image recognition, which includes the following steps:

[0060] S1: Acquire the original polarization image sequence of the target sky region through a multispectral polarization camera, and simultaneously record the camera's spatial attitude data and timestamp information, and output polarization image data;

[0061] S2: Preprocess the polarization image data and extract the Stokes parameter for each pixel to output standardized sky polarization light field data;

[0062] S3: Based on sky polarization field data, analyze the polarization degree distribution pattern under different wavelengths, and combine the solar azimuth angle to calculate the theoretical Rayleigh scattering model, outputting the theoretical sky polarization pattern characteristics;

[0063] S4: Based on sky polarization light field data, identify the spatial distribution of actual clouds, aerosols and haze, and output the actual atmospheric composition distribution characteristics;

[0064] S5: Compare the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, calculate the difference map in polarization degree and polarization angle dimensions, and extract anomalous polarization regions;

[0065] S6: Based on the geometric and spectral characteristics of the anomalous polarization region, distinguish between halo polarization anomalies and scattering polarization anomalies, and output the atmospheric optical anomaly classification results;

[0066] S7: Generate ultra-local weather change early warning signals based on atmospheric optical anomaly classification results.

[0067] S1: Acquire raw polarization image sequences of the target sky region using a multispectral polarization camera, and simultaneously record the camera's spatial attitude data and timestamp information, outputting polarization image data, including:

[0068] By continuously capturing images of the target sky region using a multispectral polarization camera, a sequence of original polarization images of the target sky region in different polarization directions under multiple spectral bands is obtained.

[0069] A multispectral polarization camera is a high-precision optical image acquisition device that simultaneously possesses multi-band imaging and polarization imaging capabilities. Internally, it is equipped with multiple independent spectral channels, each corresponding to a specific wavelength range, such as the visible light band (e.g., wavelength range of 400 nm to 700 nm) and the near-infrared light band (e.g., wavelength range of 700 nm to 1100 nm). Each spectral channel is equipped with a polarization-sensitive element (e.g., a linear polarizer or a polarizing beam splitter) for simultaneous imaging under different polarization directions (e.g., horizontal polarization, vertical polarization, and ±45 degree polarization directions), thereby obtaining multi-directional polarization images across multiple spectral bands. The multispectral polarization camera is fixedly mounted on a stable support platform or gimbal, and its settings are configured... The lens is pointed towards the sky region to be measured, such as the sky region at an elevation angle of 45 degrees above a specific geographical location. During continuous shooting, the camera shutter speed is set to a fixed parameter (e.g., exposure time of 10 to 100 milliseconds), the shooting frame rate is 20 to 50 frames per second, and shooting is continued for a long period of time (e.g., continuous shooting for 5 to 30 minutes) to form a series of continuous image sequences. Through continuous shooting by a multispectral polarization camera, a raw polarization image sequence with temporal continuity is generated. Each frame of the raw polarization image sequence records the polarization information of the sky region at a specific time, multiple spectral bands, and different polarization directions. For example, between 6:00 AM and 9:00 AM, 25 frames per second are continuously shot to collect a total of 45,000 frames of raw polarization image sequence data of the target sky region.

[0070] Real-time acquisition of spatial attitude data from a multispectral polarization camera;

[0071] To ensure the reliability and accuracy of the original polarization image sequence data, it is necessary to synchronously record the real-time spatial attitude data during the multispectral polarization camera's shooting process. This is achieved by fixing high-precision attitude sensors and positioning devices, such as microelectromechanical systems (MEMS) accelerometers, gyroscopes, and GPS receivers, onto the multispectral polarization camera housing. The attitude sensors are used to acquire the multispectral polarization camera's spatial attitude data in real time during shooting, including the camera's rotation angles (e.g., pitch, roll, and azimuth) around three spatial axes (X, Y, and Z axes) in the spatial coordinate system, with an angle measurement accuracy of 0. The system records real-time three-dimensional spatial coordinate data of the multispectral polarization camera's shooting point (e.g., latitude and longitude coordinates provided by the Global Navigation Satellite System, with latitude and longitude accuracy of ±0.5 meters to ±2 meters and altitude accuracy of ±1 meter to ±3 meters); for example, when the multispectral polarization camera is installed on a ground observation platform, it records spatial attitude data in real time during continuous shooting, including spatial rotation angles (pitch angle 30.05 degrees, roll angle 0.02 degrees, azimuth angle 315.10 degrees) and spatial coordinate data (longitude 120.153576 degrees, latitude 30.287459 degrees, altitude 50 meters).

[0072] Add unified timestamp information to the original polarization image sequence and the corresponding spatial attitude data respectively;

[0073] To ensure a correspondence between polarization image data and spatial attitude data, a unified high-precision synchronous clock source is set up during continuous shooting and spatial attitude data acquisition. For example, using the standard timing signal provided by the Global Navigation Satellite System as a reference, a unified standard time counting method is adopted to add unified timestamp information to each frame of the original polarization image sequence and the simultaneously acquired spatial attitude data. Alternatively, a time synchronization device can be used to uniformly synchronize the time according to the Coordinated Universal Time (UTC) or Beijing Time standard, ensuring that each frame of the original polarization image sequence data obtained through continuous shooting and each attitude data sample acquired in real time can be assigned unified timestamp information. For example, the timestamp corresponding to a certain frame of polarization image sequence data is recorded as 08:35:15:123 milliseconds on March 1, 2025, and at the same time, the timestamp corresponding to the acquired spatial attitude data is also marked as 08:35:15:123 milliseconds on March 1, 2025.

[0074] The original polarization image sequence data with uniform timestamp information is matched with the corresponding spatial attitude data to output polarization image data.

[0075] Each frame of the original polarization image sequence data with a unified timestamp is matched with each real-time recorded spatial attitude data with a unified timestamp, based on the same timestamp: the timestamp information of each frame of the original polarization image sequence data is compared with the timestamp information of each set of spatial attitude data acquired in real time. When the timestamp difference is less than 1 millisecond to 5 milliseconds, it is considered that the data was acquired at the same time, and the data matching is completed. Through matching, the data sets corresponding to all timestamps during continuous shooting are summarized and organized into a unified data structure, and the data matching result is output, that is, each frame of image data has an accurately matched spatial attitude data; for example, a total of 45,000 frames of polarization image data are obtained, and each frame of image data corresponds to a set of spatial attitude data, forming a total of 45,000 sets of matched polarization image data.

[0076] S2: Preprocess the polarization image data and extract the Stokes parameter for each pixel, outputting standardized sky polarization light field data, including:

[0077] Image denoising is performed on polarization image data to remove random noise and interfering pixels from the polarization image data.

[0078] Actual shooting processes are often affected by equipment noise and external environmental interference, such as electronic noise generated by the internal circuitry of the multispectral polarization camera, sensor response errors, or random interference from atmospheric turbulence. This results in the acquired polarization image data containing random noise and abnormal interfering pixels. Random noise manifests as random fluctuations in the grayscale or brightness values ​​of image pixels, while interfering pixels appear as abnormally prominent pixels with excessively large grayscale differences from adjacent pixels. To eliminate random noise and interfering pixels, image denoising needs to be performed frame-by-frame on the polarization image data. For example, if there are 45,000 frames of polarization image data, each frame is processed separately. Median filtering is applied to each pixel in each frame to remove random noise. In this process, a 3×3 or 5×5 pixel neighborhood window is established centered on each pixel to be processed. The median of the gray values ​​of all pixels within the neighborhood window is calculated, and the gray value of the pixel to be processed is replaced with the calculated median to effectively remove random noise pixels. For abnormally prominent interfering pixels in the image, a threshold method is used for identification and removal. For example, the pixel gray-level difference threshold is set to 50 gray levels. That is, when the gray value of a pixel differs from the gray value of its neighboring pixels by more than 50 gray levels, it is marked as an interfering pixel and replaced with the average gray value of its neighboring pixels to eliminate abnormal interfering pixels in the image. Through the above methods, random noise and interfering pixels are removed from each frame of polarized image data, generating polarized image data after removing random noise and interfering pixels.

[0079] For each frame of polarization image data after removing random noise and interfering pixels, extract the Stokes parameter data corresponding to each pixel in the polarization image data;

[0080] Stokes parameter data are important parameters describing the state of polarized optical fields, including the first Stokes parameter, the second Stokes parameter, the third Stokes parameter, and the fourth Stokes parameter;

[0081] To extract the above parameters, the polarization image data needs to be processed frame by frame and pixel by pixel. The position of each pixel in each frame of polarization image data needs to be determined. For example, if a frame of polarization image has a resolution of 1024×1024 pixels, all pixels are processed one by one. At each pixel position, the pixel grayscale value for different polarization directions (e.g., 0°, 45°, 90°, and 135° polarization) is read. The first Stokes parameter represents the total intensity of the incident light, which is a linear combination of the grayscale values ​​in the four polarization directions; that is, the first Stokes parameter is equal to the sum of the grayscale values ​​for 0° and 90° polarization. The second Stokes parameter... The Stokes parameter describes the difference in intensity between horizontally and vertically polarized light, which is the difference between the gray values ​​of 0 degrees and 90 degrees of polarization. The third Stokes parameter describes the difference in intensity between 45 degrees and 135 degrees of polarized light, which is the difference between the gray values ​​of 45 degrees and 135 degrees of polarization. The fourth Stokes parameter describes the circular polarization component, which is usually obtained through polarizer rotation and phase modulation measurements, for example, through the circular polarization element built into the camera. The Stokes parameter data for each frame of the image is extracted pixel by pixel. For example, the output of each frame of polarized image data is a data matrix containing all four Stokes parameter values.

[0082] The Stokes parameter data of all pixels in each frame of polarized image data are normalized and transformed to output normalized sky polarization light field data.

[0083] After extracting the Stokes parameters, each frame of polarized image data corresponds to a complete Stokes parameter data matrix. Since the ranges of each parameter may vary significantly, data standardization is required. Standardization is performed on each Stokes parameter matrix in each frame of image data separately. The standardization process uses the max-min standardization method. Through standardization, all Stokes parameter matrix data in each frame of polarized image data are converted into standardized values ​​between 0 and 1, thereby obtaining standardized sky polarized light field data.

[0084] S3: Based on sky polarization field data, analyze the polarization degree distribution patterns at different wavelengths, and combine the solar azimuth angle to calculate the theoretical Rayleigh scattering model, outputting the theoretical sky polarization pattern characteristics, including:

[0085] Based on sky polarization field data, determine the sky polarization degree distribution pattern for each spectral band;

[0086] Sky polarization field data contains multiple spectral bands, each corresponding to a complete normalized Stokes parametric matrix. The data for each band is processed individually. Assuming the obtained sky polarization field data contains three bands, for example, with wavelength centers of 480 nm, 560 nm, and 670 nm, the data for each band is a two-dimensional matrix, where each element represents the normalized polarization characteristic value at the corresponding spatial location in the sky region. When determining the sky polarization degree distribution pattern corresponding to each spectral band, the polarization degree calculation formula is used to process the sky polarization field data for each spectral band. For any band (e.g., wavelength 480 nm), the normalized Stokes parametric matrix corresponding to that band is selected. The Stokes parameter matrix is ​​used to calculate the sky polarization degree pixel by pixel. The formula for calculating the sky polarization degree is: the polarization degree is equal to the ratio of the polarized light intensity to the total light intensity. The polarized light intensity is obtained by taking the square root of the sum of the squares of the second and third Stokes parameters, and the total light intensity is represented by the first Stokes parameter. The polarization degree is calculated for all pixels of the sky polarization field data in the 480 nm band, forming the spatial distribution matrix of the sky polarization degree of the band. Similarly, the same calculation operation is performed for other bands (560 nm, 670 nm), and the corresponding spatial distribution matrix of the sky polarization degree is obtained for each band. The calculated spatial distribution matrix is ​​defined as the sky polarization degree distribution pattern of the band.

[0087] For each spectral band, calculate the solar azimuth parameter in the corresponding spectral band based on the sky polarization distribution pattern.

[0088] The solar azimuth parameter refers to the azimuth angle of the sun relative to the location of the observation equipment at the moment of sky observation. The calculation of the solar azimuth parameter is a spatial positioning operation of the sun's position, determined using the sky polarization degree distribution pattern. Based on the calculated sky polarization degree distribution pattern for each spectral band, the center position of the sky region where the polarization degree of the band reaches its minimum is found. For example, for the 480 nm wavelength band, the sky polarization degree distribution pattern is traversed to find the sky region with the polarization degree closest to zero, and the center coordinates of this region are recorded. According to Rayleigh scattering physics and sky polarization theory, the position of minimum polarization degree within the sky region usually corresponds to the actual direction of the sun; therefore, the coordinates of the position of minimum polarization degree are used as the estimated position of the sun. Starting from the spatial coordinates (such as latitude, longitude, and altitude) of the observation equipment's location, and ending at the sky coordinates of the position with the minimum polarization degree, a spatial position vector is established. The relative angle between the vector and the geographic reference coordinate system is calculated in the equipment coordinate system, and the calculated angle is defined as the solar azimuth parameter. Following the above method, the sky polarization degree distribution patterns at wavelengths of 560 nm and 670 nm are calculated respectively to obtain the corresponding solar azimuth parameters. For example, the solar azimuth angle for the 480 nm band is 135.25 degrees, the solar azimuth angle for the 560 nm band is 135.35 degrees, and the solar azimuth angle for the 670 nm band is 135.45 degrees. Finally, the solar azimuth parameter data corresponding to each spectral band is output.

[0089] Based on the solar azimuth angle parameters of each spectral band, and combined with Rayleigh scattering physics theory, the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle are calculated for each spectral band.

[0090] Rayleigh scattering physics describes the physical mechanism by which sunlight is scattered by molecules in a clean atmosphere, resulting in sky polarization. According to Rayleigh scattering physics, once the solar azimuth angle parameter is determined, the ideal sky polarization state can be calculated using theoretical formulas: For each spectral band, the corresponding solar azimuth angle parameter is used as input, and ideal Rayleigh scattering atmospheric conditions are set, assuming that only air molecules exist in the atmosphere and neglecting the scattering effects of clouds, aerosols, etc. The calculation formula is the Rayleigh scattering polarization theory equation, using the solar azimuth angle and the angle between each pixel position in the sky and the sun's position as input, to calculate the theoretical polarization corresponding to each pixel position in the sky. The ideal sky polarization degree and theoretical polarization angle are obtained by traversing the position of each pixel in the sky and substituting the relative spatial angle value with the position of the sun into the Rayleigh scattering polarization theory equation. This yields the ideal sky polarization degree spatial distribution matrix and the ideal polarization angle spatial distribution matrix for all pixel positions in the 480 nm band. Similarly, the above calculations are performed on the 560 nm and 670 nm bands, and the corresponding ideal sky polarization degree spatial distribution matrix and ideal polarization angle spatial distribution matrix are obtained for each band. These constitute the ideal sky polarization characteristics of each spectral band, providing theoretical benchmark data for comparing the actual sky polarization state.

[0091] Based on the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle under each spectral band, theoretical sky polarization mode characteristic data are output.

[0092] The spatial distribution matrix of ideal sky polarization degree and the spatial distribution matrix of ideal polarization angle are combined to output theoretical sky polarization mode feature data. The data structure is as follows: each spectral band corresponds to a combined data unit, which contains two matrices. One matrix records the ideal sky polarization degree at each pixel position in the sky, and the other matrix records the ideal polarization angle at the corresponding position. For example, the data unit of the 480 nm band contains the ideal sky polarization degree matrix and the ideal polarization angle matrix. Each element represents the theoretical value corresponding to a specific position in the sky region. The data units of all spectral bands are combined together to form complete theoretical sky polarization mode feature data.

[0093] S4: Based on sky polarization light field data, identify the spatial distribution of actual clouds, aerosols, and haze, and output the actual atmospheric composition distribution characteristics, including:

[0094] Based on sky polarization field data, we can identify the changes in sky polarization characteristics caused by clouds and determine the spatial distribution area of ​​clouds.

[0095] Sky polarization field data includes normalized Stokes parametric matrices for each spectral band. The presence of clouds in the sky alters the local polarization characteristics because clouds, composed of water droplets or ice crystals, have larger particles and scattering properties different from the Rayleigh scattering properties of air molecules. This causes the polarization characteristics of the sky to deviate from the theoretical sky polarization mode characteristics. The sky polarization field data for each spectral band is analyzed and compared pixel-by-pixel with the corresponding theoretical sky polarization mode characteristics. For example, comparing the 480 nm band data involves calculating the difference between the normalized actual polarization degree and the theoretical polarization degree at each sky pixel location, as well as the difference between the actual polarization angle and the theoretical polarization angle, forming polarization degree difference matrices and polarization angle difference matrices for each band. Based on the typical difference range of known cloud region polarization characteristics, polarization degree difference thresholds and polarization angle difference matrices are set. Cloud identification is performed using polarization angle difference thresholds. For example, the polarization degree difference threshold is set to the range of 0.2 to 0.5, and the polarization angle difference threshold is set to 10 to 20 degrees. The difference matrix of each band is traversed. If a pixel location satisfies both the polarization degree difference and polarization angle difference exceeding the set thresholds, the pixel location is determined to belong to the cloud region. To ensure that the determined cloud regions are continuous and spatially coherent, a spatial clustering algorithm is also required. For example, based on pixel connected component analysis, spatially coherent cloud regions are determined. For example, after calculation by the connected component algorithm, a region with more than 500 consecutive pixels in the 480 nm band is considered a cloud spatial distribution region. The cloud spatial distribution region results of each band are comprehensively superimposed to output a unified cloud spatial distribution region data matrix. Through the above methods, the location and range of cloud spatial distribution regions in the sky are determined.

[0096] Based on sky polarization light field data, we can identify regions of sky polarization characteristic changes caused by aerosol particle scattering and determine the spatial distribution region of aerosol particles.

[0097] Aerosol particles mainly consist of tiny solid particles in the air (such as dust and soot), with particle sizes between those of air molecules and cloud particles. Therefore, the scattering characteristics of aerosol particles differ from Rayleigh scattering and cloud scattering. Aerosol particle regions cause a decrease in the degree of polarization in the sky region, while the polarization angle undergoes a specific, regular change. Analysis is performed based on the difference matrix between sky polarization field data and theoretical sky polarization mode characteristic data. The set polarization degree difference threshold and polarization angle difference threshold differ from those for cloud regions; for example, the polarization degree difference threshold is set to 0.1 to 0.2, and the polarization angle difference threshold is set to 5 to 10 degrees. The difference matrix is ​​analyzed pixel-by-pixel to find values ​​that satisfy these thresholds. The pixel locations within a certain range are identified and marked. Since aerosol particles are distributed in a relatively continuous and detailed manner and typically cover a wide area, a region growing algorithm is used to perform spatial clustering on the pixel locations marked as aerosol particle regions. For example, a continuous region with more than 1000 pixels in the 480 nm band determined by the region growing algorithm is considered a valid aerosol particle region. The spatial distribution region results of aerosols in different bands are then combined and superimposed to form a final unified aerosol particle spatial distribution region matrix. The aerosol particle spatial distribution region data obtained through the above methods fully discloses the spatial distribution information of aerosol particles.

[0098] Based on sky polarization light field data, identify the regions of polarization light characteristic changes caused by haze particle scattering and determine the spatial distribution area of ​​haze particles.

[0099] Haze particles are fine and highly concentrated suspended particulate matter, typically smaller in size and higher in concentration than general aerosol particles. Their scattering characteristics show a significant reduction in sky polarization, even approaching zero, while the polarization angle exhibits highly consistent changes. A new threshold was set for the difference matrix between sky polarization field data and theoretical sky polarization mode characteristic data, specifically for haze particles. For example, the polarization degree difference threshold was set between 0.5 and 0.8 degrees, and the polarization angle difference threshold was set within a narrow range of 0 to 5 degrees. Then, the difference matrix was traversed pixel by pixel to mark the pixel positions that met the above threshold conditions. Due to the high spatial continuity of haze particle regions, a more stringent spatial clustering method was employed, such as morphological image analysis algorithms, to determine continuous spatial distribution regions. For example, a continuous region of haze particles in the 480 nm band, determined by morphological algorithms, must exceed 1500 pixels to be considered a valid haze particle region. The spatial distribution regions of haze particles in different bands were comprehensively superimposed to output a unified haze particle spatial distribution region data matrix. Through these methods, the spatial distribution regions of haze particles were determined.

[0100] Based on the spatial distribution areas of clouds, aerosol particles, and haze particles, output the spatial distribution characteristics data of actual atmospheric components.

[0101] The obtained spatial distribution regions of clouds, aerosol particles, and haze particles are stored in matrix form. Each matrix data element marks whether the sky pixel location belongs to the corresponding atmospheric component region. The actual atmospheric component spatial distribution feature data is the final output result. The three matrices are merged pixel by pixel to form a comprehensive matrix. Each matrix element records the atmospheric component type of the corresponding pixel location using an encoding method. For example, a pixel location belonging to the cloud region is encoded as a value 1, belonging to the aerosol particle region is encoded as a value 2, belonging to the haze particle region is encoded as a value 3, and belonging to none of the above regions is encoded as a value 0. The merged matrix determines the actual atmospheric component classification of each pixel location in the sky region, which is the actual atmospheric component spatial distribution feature data, providing a complete data output result of the spatial distribution of atmospheric components in the sky region.

[0102] S5: Compare the theoretical sky polarization pattern characteristics with the actual atmospheric composition distribution characteristics, calculate the difference map in the dimensions of polarization degree and polarization angle, and extract anomalous polarization regions, including:

[0103] Calculate the degree of polarization and the angle of polarization difference between the theoretical sky polarization mode characteristic data and the actual spatial distribution characteristic data of atmospheric composition for each spectral band;

[0104] The theoretical sky polarization mode feature data includes spatial distribution matrices of ideal sky polarization degree and ideal polarization angle corresponding to multiple spectral bands. These matrices record the theoretical polarization degree and theoretical polarization angle of each pixel location in the sky under ideal cloudless, aerosol-free, and haze-free environments in different spectral bands. For example, the theoretical sky polarization mode feature data for the 480 nm band contains 1024×1024 elements. Each element records the theoretical polarization degree value between 0 and 1 and the theoretical polarization angle value between 0 and 180 degrees.

[0105] The actual spatial distribution characteristics of atmospheric components include a cloud spatial distribution region matrix, an aerosol particle spatial distribution region matrix, and a haze particle spatial distribution region matrix. Each element of each matrix marks whether each pixel position in the sky belongs to the corresponding atmospheric component region. For example, a pixel position belonging to the cloud region is marked as 1, belonging to the aerosol particle region is marked as 2, belonging to the haze particle region is marked as 3, and not belonging to the above regions is marked as 0.

[0106] When calculating the difference in polarization degree and polarization angle between the theoretical sky polarization mode feature data and the actual atmospheric composition spatial distribution feature data for each spectral band, the actual atmospheric composition spatial distribution feature data is first converted to correspond to the true polarization data of the actual sky region. For example, for pixel positions marked as cloud, aerosol, and haze regions, the actual polarization degree and actual polarization angle data matrix of the corresponding positions are extracted based on the actual sky polarization light field data.

[0107] For each spectral band, the difference between the theoretical polarization degree spatial distribution matrix and the actual polarization degree matrix is ​​calculated pixel by pixel. For example, in the 480 nm band, the theoretical polarization degree of each pixel position is subtracted from the actual polarization degree of the corresponding pixel position to obtain the polarization degree matrix of the 480 nm band. Each matrix element represents the difference between the theoretical polarization degree and the actual polarization degree of the corresponding sky position. Similarly, the same polarization degree difference calculation is performed on the 560 nm band and the 670 nm band to obtain the polarization degree difference matrix of each band.

[0108] The difference between the theoretical polarization angle spatial distribution matrix and the actual polarization angle matrix for each band is calculated pixel by pixel. For example, in the 480 nm band, the difference between the theoretical polarization angle and the actual polarization angle at each sky position is calculated to obtain the polarization angle difference matrix for the 480 nm band. The matrix elements represent the angle difference between the theoretical and actual polarization angles at the corresponding positions. The same polarization angle difference calculation is performed for the 560 nm band and the 670 nm band to obtain their respective polarization angle difference matrices.

[0109] Based on the difference in polarization degree and the difference in polarization angle, generate a sky polarization degree difference distribution map and a polarization angle difference distribution map for each spectral band;

[0110] Based on the calculated polarization degree difference matrix and polarization angle difference matrix, a difference distribution map corresponding to each spectral band is generated to present the spatial distribution characteristics of the difference data in a visual form.

[0111] The method for generating the polarization degree difference distribution map is as follows: Select the polarization degree difference matrix of the 480 nm band and map the difference values ​​of the matrix elements to a visual color table; for example, using a heatmap mode, areas with large differences are mapped to red, areas with medium differences are mapped to yellow or green, and areas with small differences or close to zero are mapped to blue; after the above mapping process, the 480 nm band polarization degree difference matrix is ​​transformed into a visual sky polarization degree difference distribution map to intuitively display the location, shape, and area of ​​polarization degree anomalies in the sky; the same method is used to generate sky polarization degree difference distribution maps for the 560 nm and 670 nm bands respectively.

[0112] The polarization angle difference distribution map uses the same generation method. Taking the polarization angle difference matrix of the 480 nm band as an example, the difference range (e.g., 0 degrees to 45 degrees) is color-mapped. Large difference areas are represented by red, medium difference areas by yellow or green, and small difference areas close to zero by blue. Similarly, polarization angle difference distribution maps are generated for the 560 nm and 670 nm bands respectively.

[0113] The above methods transform complex difference data into visual images, showing the spatial distribution of the differences between theoretical and actual polarization characteristics within the sky region.

[0114] For each spectral band, the regions where the difference in polarization degree and the difference in polarization angle exceed a set threshold are extracted to identify abnormal polarization regions.

[0115] If the difference in the polarization degree and polarization angle distribution map shows a region where the difference exceeds the range of the general natural background, it can be regarded as an abnormal polarization region. In order to identify the abnormal region, a threshold is set for region extraction.

[0116] For example, taking the polarization degree difference distribution map of the 480 nm band as an example, if the polarization degree difference threshold is set to 0.3, the area where the polarization degree difference of all pixel positions in the map exceeds 0.3 is defined as the polarization degree abnormal area; similarly, if the polarization angle difference distribution map of the 480 nm band is set to 15 degrees, the area where the polarization angle difference of all pixel positions in the map exceeds 15 degrees is defined as the polarization angle abnormal area.

[0117] Morphological image analysis methods are used to extract spatial regions of polarization degree anomalies and polarization angle anomalies, respectively. For example, a binarized threshold segmentation algorithm is used to mark the anomalies, and then a connected component detection algorithm is used to determine the region boundaries, extracting the spatial location and shape of the anomalies. For example, if more than 500 pixels are detected in a continuous region of polarization degree anomalies in the 480 nm band, it is defined as a valid anomalous polarization region. The same method is applied to the 560 nm and 670 nm bands to extract the valid anomalous polarization regions within each band.

[0118] By combining the effective abnormal polarization region extraction results of each band, the results of each band are spatially superimposed to determine the complete spatial location and range of the abnormal polarization region.

[0119] S6: Based on the geometric and spectral characteristics of the anomalous polarization region, distinguish between halo polarization anomalies and scattering polarization anomalies, and output the atmospheric optical anomaly classification results, including:

[0120] For each abnormal polarization region, extract the boundary contour geometric parameters of the abnormal polarization region;

[0121] The anomalous polarization region is located within the sky region, and each region records pixel position information in the form of a two-dimensional matrix. For example, in a 480 nm band image, an anomalous polarization region obtained through threshold segmentation and connected component detection is recorded as a two-dimensional binary matrix, where a pixel value of 1 represents a pixel inside the anomalous region, and a pixel value of 0 represents a pixel outside the anomalous region. In order to accurately describe the shape of the anomalous polarization region, it is necessary to extract geometric contour parameters from the anomalous polarization region, including area, perimeter, major axis, and minor axis length.

[0122] The area of ​​the anomalous polarization region is calculated by statistically counting the number of pixels marked as 1 in the binary matrix. For example, if the matrix data of the anomalous polarization region is 1024×1024 pixels, the total number of anomalous pixels marked within the region is 1500. The actual sky area corresponding to each pixel is calculated based on the field of view and pixel resolution of the camera. For example, assuming that the actual sky area corresponding to each pixel is 1 square meter, the area of ​​the region is accurately calculated to be 1500 square meters. If the sky areas corresponding to different pixels are different, the areas of each pixel can be summed to ensure the accuracy of the area calculation.

[0123] The perimeter of the anomalous polarization region is calculated using a boundary contour extraction algorithm. This includes performing an edge detection algorithm on the binary matrix, such as using the Sobel or Canny edge detection operator to determine the region boundary, resulting in a closed contour line. All boundary pixels on the closed contour line are then connected point-by-point for calculation, and the perimeter is obtained by summing the actual distances between every two consecutive adjacent pixels. For example, after boundary extraction, the anomalous polarization region has a total of 500 boundary pixels. Each pixel's side length corresponds to an actual sky field of view length of 1 meter. After pixel-by-pixel distance calculation, the perimeter of the region is found to be 500 meters.

[0124] The major and minor axis lengths of the anomalous polarization region are calculated using an ellipse fitting algorithm. The least squares ellipse fitting method is applied to the obtained closed contour line to fit an ellipse contour that is closest to the shape of the region boundary. The major axis of the fitted ellipse is defined as the major axis length of the anomalous region, and the minor axis is defined as the minor axis length of the anomalous region. For example, the major axis length of the fitted ellipse is calculated to be 70 meters and the minor axis length is 30 meters using the least squares fitting algorithm.

[0125] For each anomalous polarization region, calculate the average degree of polarization and average polarization angle of the anomalous polarization region in each spectral band;

[0126] Average degree of polarization and average angle of polarization are used to describe the overall polarization characteristics of an anomalous polarization region across different spectral bands. To calculate the average degree of polarization and average angle of polarization of the region in each spectral band, calculations based on sky polarization field data are required.

[0127] First, based on the binary matrix corresponding to the anomalous polarization region in each spectral band, the positions of all pixels belonging to the anomalous region are determined; for example, in a 480 nm band image, the anomalous polarization region contains 1500 pixel positions, and each of the 1500 pixels corresponds to a specific degree of polarization and polarization angle in the standardized sky polarization field data.

[0128] For each determined pixel location, the corresponding polarization degree is extracted. The sum of all extracted polarization degree values ​​is then divided by the total number of pixels to obtain the average polarization degree of the abnormal region in the 480 nm band. For example, if the sum of the polarization degree values ​​of 1500 pixels is 900, then the average polarization degree of the region in the 480 nm band is calculated as 900 divided by 1500, resulting in 0.6. The average polarization degree in the 560 nm and 670 nm bands is calculated using the same method.

[0129] A similar method is used to extract the polarization angle corresponding to each pixel position in the abnormal region and perform statistical averaging. For example, the sum of the polarization angle values ​​of all pixels in the 480 nm band is 9000 degrees, so the average polarization angle is 9000 degrees divided by 1500 pixels, resulting in an average polarization angle of 6 degrees. The same calculation method is used for the 560 nm and 670 nm bands to determine the average polarization angle of each band.

[0130] Based on the boundary contour geometric parameters of the abnormal polarization region and the average degree of polarization and average polarization angle of the abnormal polarization region in each spectral band, the abnormal polarization type of each abnormal polarization region is determined to be either halo polarization abnormality type or scattering polarization abnormality type.

[0131] Halo polarization anomalies and scattering polarization anomalies have different geometric contours and polarization characteristics. Halo polarization anomalies are usually elliptical or nearly circular, with the ratio of the major axis to the minor axis being close to 1, and the average degree of polarization is low, with the average polarization angle varying little within the region. Scattering polarization anomalies generally have more irregular shapes, with the ratio of the major axis to the minor axis being significantly greater than 1, the average degree of polarization being high, and the polarization angle varying significantly.

[0132] The geometric parameters of the boundary contour of the anomalous region and the average degree of polarization and average polarization angle values ​​of each band are calculated comprehensively. For example, the threshold for the ratio of the major and minor axes of halo polarization anomalies is set to 1 to 1.2, the threshold for the average degree of polarization is less than 0.4, and the change in polarization angle is less than 5 degrees. If a certain anomalous polarization region has a major axis of 70 meters and a minor axis of 68 meters, the ratio is 1.03, the average degree of polarization is 0.35, and the change in average polarization angle is 2 degrees, which meets the definition of halo polarization anomaly type.

[0133] Another anomalous region has a major axis of 70 meters and a minor axis of 30 meters, with a ratio of 2.3, an average degree of polarization of 0.65, and an average polarization angle variation of 20 degrees, thus satisfying the definition of a scattering polarization anomaly type. Using the above determination method, each anomalous region can be classified as either a halo polarization anomaly type or a scattering polarization anomaly type.

[0134] Based on the determination of the abnormal polarization type in the abnormal polarization region, the atmospheric optical anomaly classification result is output.

[0135] Output the results of the abnormal polarization region type determination in the form of a data matrix or list; for example, create an abnormal region classification result matrix, record the serial number of each abnormal region, spatial location parameters, boundary contour geometric parameters, average degree of polarization and average polarization angle of each spectral band, and the corresponding abnormal polarization type.

[0136] For example, the first anomalous area is recorded as: "Area 1, center coordinates (100, 200), area 1500 square meters, major axis 70 meters, minor axis 68 meters, average polarization degree of 480 nm 0.35, average polarization degree of 560 nm 0.33, average polarization degree of 670 nm 0.34, polarization type: halo polarization anomaly type"; the second anomalous area is recorded as: "Area 2, center coordinates (300, 400), area 2000 square meters, major axis 70 meters, minor axis 30 meters, average polarization degree of 480 nm 0.65, average polarization degree of 560 nm 0.62, average polarization degree of 670 nm 0.60, polarization type: scattering polarization anomaly type".

[0137] S7: Based on atmospheric optical anomaly classification results, generate ultra-local weather change early warning signals, including:

[0138] Based on the atmospheric optical anomaly classification results, the spatial location parameters of each anomalous polarization region are extracted, including the center coordinate data and coverage data of the anomalous polarization region.

[0139] The atmospheric optical anomaly classification results are anomaly region classification datasets composed of halo polarization anomaly types or scattering polarization anomaly types. These datasets record the sequence number, boundary geometry parameters, and polarization characteristics of each anomalous polarization region across different wavelengths. For example, in the 480 nm band, the area of ​​the anomalous polarization region is determined to be 1500 square meters, with a major axis of 70 meters, a minor axis of 68 meters, an average degree of polarization of 0.35, and an average polarization angle of 6 degrees. To provide weather change warnings for anomalous polarization regions, it is necessary to extract the spatial location parameters of each region, including center coordinates and coverage area data.

[0140] The center position coordinate data of an anomalous polarization region refers to the centroid of the spatial region surrounded by the outline boundary of each anomalous region in a two-dimensional coordinate system of the sky region. The calculation process of the center position coordinate data is as follows: A weighted average is calculated using the horizontal and vertical coordinates of all pixels marked as anomalous within the sky region; the horizontal coordinates of each pixel within the anomalous region in the two-dimensional binary matrix are summed, and the sum is divided by the total number of pixels within the anomalous region; for example, an anomalous polarization region contains 1500 pixels in the 480 nm band. For each pixel location marked as abnormal, if the sum of the horizontal coordinates is 150,000, then the horizontal center coordinate is calculated as 150,000 divided by 1500, which equals a horizontal coordinate value of 100. The vertical coordinates are calculated similarly by dividing the sum of the vertical coordinates by the total number of pixels. For example, if the sum of the vertical coordinates is 300,000, then the vertical center coordinate is calculated as 300,000 divided by 1500, which equals a vertical coordinate value of 200. Using this method, the center coordinates of the abnormal polarization region can be extracted, for example, (100, 200), in pixel coordinates.

[0141] Coverage data refers to the spatial area and range of an anomalous polarization region in the sky. Coverage data is based on the boundary contour coordinates of the anomalous polarization region. The coverage range of the anomalous polarization region is determined using the boundary pixel coordinates of the anomalous region. For example, if an anomalous polarization region in the 480 nm band has a boundary composed of 500 boundary pixels, the coordinates of each boundary pixel are recorded to form a closed boundary contour coordinate set. This closed boundary contour coordinate set defines the spatial coverage range of the anomalous polarization region. Coverage data can also be represented using a bounding rectangle, i.e., finding the smallest bounding rectangle that completely covers the anomalous region and recording the coordinates of its upper left and lower right corners. For example, the upper left corner coordinates are (70, 170), and the lower right corner coordinates are (130, 230). All of these methods can be used to obtain the coverage data of the anomalous polarization region.

[0142] Based on the spatial position parameters of each anomalous polarization region, calculate the offset of each anomalous polarization region relative to the center position of the target sky region;

[0143] The center position of the target sky region is the reference point set by the observation equipment when photographing the target sky region. It is usually the coordinates of the center of the sky region corresponding to the center point of the camera's field of view. For example, if the camera image resolution is 1024×1024 pixels, the center position of the target sky region is the coordinates of the image center point (512,512).

[0144] The offset of the anomalous polarization region relative to the center of the target sky region represents the spatial distance between the coordinates of the center of the anomalous polarization region and the center of the target sky region. A two-dimensional coordinate calculation formula is used: subtract the center coordinates of the target sky region from the center coordinates of the anomalous region, and then calculate the absolute value or take the square root of the result to obtain the offset. For example, the horizontal offset of the anomalous region's center coordinates (100, 200) from the target sky region's center coordinates (512, 512) is 512 minus 100, resulting in 412; the vertical offset is 512 minus 200, resulting in 312. The Euclidean distance between the horizontal and vertical offsets is then calculated as the square root of (412² + 312²). Through calculation, the offset of the anomalous polarization region relative to the center of the target sky region is approximately 516.3 pixels.

[0145] For the offset of each abnormal polarization region, the warning level of each abnormal polarization region is determined according to the preset offset threshold.

[0146] The warning level is a classification standard indicating the severity of weather changes caused by abnormal polarization regions. Pre-set offset thresholds are the basis for dividing the offset values ​​of abnormal regions. For example, the offset threshold can be set into multiple threshold ranges. The setting method is as follows: an offset of less than 200 pixels is defined as a Level 1 warning, indicating that the abnormal region is close to the center of the sky and may cause direct impact; an offset between 200 and 500 pixels is defined as a Level 2 warning, indicating that the abnormal region is far from the center of the sky and may cause indirect or secondary impact; an offset exceeding 500 pixels is defined as a Level 3 warning, indicating that the abnormal region is far from the center and the resulting weather changes are relatively weak.

[0147] For example, if the offset of the abnormal area is 516.3 pixels, then the abnormal area belongs to the level three warning level according to the offset threshold.

[0148] Based on the warning level and spatial location parameters of each anomalous polarization region, a hyperlocal weather change warning signal corresponding to each anomalous polarization region is generated;

[0149] Ultra-local weather change warning signals provide local weather anomaly alerts based on the spatial location parameters of abnormal polarization regions and warning levels.

[0150] The generation method involves encoding the center coordinates, coverage area, and warning level information of each abnormal area to form a digital and visualized weather change warning signal. For example, a level 1 warning area is marked with a red area, level 2 with a yellow area, and level 3 with a blue area. The center coordinates are displayed on the weather monitoring screen. For example, if the center coordinates of an abnormal area are (100, 200), the offset is 516.3 pixels, and the warning level is 3, then the warning signal will be marked in blue and the relevant location and warning information will be clearly displayed.

[0151] Ultra-local weather change warning signals can be sent to the meteorological service system via electronic warning messages, displaying the location of the abnormal area and warning information, so as to realize the functions of real-time weather monitoring and local weather change warning. Example 2

[0152] The difference between Embodiment 2 and Embodiment 1 is that this embodiment introduces a system for intelligent weather image acquisition and weather image recognition.

[0153] Figure 2 A schematic diagram of the structure of an intelligent weather image acquisition and weather image recognition system according to the present invention is provided. The intelligent weather image acquisition and weather image recognition system includes:

[0154] Image acquisition module: Acquires raw polarization image sequences of the target sky region through a multispectral polarization camera, and simultaneously records the camera's spatial attitude data and timestamp information, and outputs polarization image data;

[0155] Data processing module: preprocesses polarization image data, extracts Stokes parameters for each pixel, and outputs standardized sky polarization light field data;

[0156] Theoretical Feature Module: Based on sky polarization field data, analyze the polarization degree distribution pattern under different wavelengths, and calculate the theoretical Rayleigh scattering model by combining the solar azimuth angle, and output the theoretical sky polarization pattern characteristics;

[0157] Composition identification module: Based on sky polarization light field data, it identifies the spatial distribution of actual clouds, aerosols and haze, and outputs the actual atmospheric composition distribution characteristics;

[0158] Anomaly Extraction Module: By comparing the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, the module calculates the difference map in the dimensions of polarization degree and polarization angle, and extracts anomalous polarization regions.

[0159] Anomaly Classification Module: Based on the geometric and spectral characteristics of the anomalous polarization region, it distinguishes between halo polarization anomalies and scattering polarization anomalies, and outputs atmospheric optical anomaly classification results;

[0160] Early warning generation module: Based on the atmospheric optical anomaly classification results, it generates ultra-local weather change early warning signals.

[0161] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0162] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0163] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0164] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0165] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0166] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0167] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0168] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0169] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0170] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent weather image acquisition and weather image recognition, characterized in that, Includes the following steps: S1: Acquire raw polarization image sequences of the target sky region using a multispectral polarization camera, and simultaneously record the camera's spatial attitude data and timestamp information, outputting polarization image data, specifically: By continuously capturing images of the target sky region using a multispectral polarization camera, a sequence of original polarization images of the target sky region in different polarization directions under multiple spectral bands is obtained. Real-time acquisition of spatial attitude data from a multispectral polarization camera; Add unified timestamp information to the original polarization image sequence and the corresponding spatial attitude data respectively; The original polarization image sequence data with unified timestamp information is matched with the corresponding spatial attitude data to output polarization image data; the polarization image data is formed by matching the original polarization image sequence with unified timestamp information with the real-time acquired camera spatial attitude data. S2: Preprocess the polarization image data and extract the Stokes parameter for each pixel, outputting standardized sky polarization light field data, specifically: Image denoising is performed on polarization image data to remove random noise and interfering pixels from the polarization image data. For each frame of polarization image data after removing random noise and interfering pixels, extract the Stokes parameter data corresponding to each pixel in the polarization image data; The Stokes parameter data of all pixels in each frame of polarized image data are normalized to output normalized sky polarization light field data; the sky polarization light field data is a set of normalized Stokes parameter matrices. S3: Based on sky polarization field data, analyze the polarization degree distribution pattern at different wavelengths, and combine the solar azimuth angle and the theoretical Rayleigh scattering model to calculate and output the theoretical sky polarization pattern characteristics, specifically: Based on sky polarization field data, determine the sky polarization degree distribution pattern for each spectral band; For each spectral band, calculate the solar azimuth parameter in the corresponding spectral band based on the sky polarization distribution pattern. Based on the solar azimuth angle parameters of each spectral band, and combined with Rayleigh scattering physics theory, the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle are calculated for each spectral band. Based on the spatial distribution characteristics of the ideal sky polarization degree and the spatial distribution characteristics of the ideal polarization angle under each spectral band, theoretical sky polarization mode characteristic data are output. The theoretical Rayleigh scattering model is a computational model based on Rayleigh scattering physics theory and solar azimuth angle parameters. It is used to output the spatial distribution matrix of ideal polarization degree and the spatial distribution matrix of ideal polarization angle for each spectral band. The theoretical sky polarization mode characteristics are theoretical feature data composed of the combination of the spatial distribution matrix of ideal polarization degree and the spatial distribution matrix of ideal polarization angle calculated by the theoretical Rayleigh scattering model. S4: Based on sky polarization light field data, identify the spatial distribution of actual clouds, aerosols and haze, and output the actual atmospheric composition distribution characteristics; S5: Compare the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, calculate the difference map in the dimensions of polarization degree and polarization angle, and extract anomalous polarization regions; S6: Based on the geometric and spectral characteristics of the anomalous polarization region, distinguish between halo polarization anomalies and scattering polarization anomalies, and output the atmospheric optical anomaly classification results; S7: Generate ultra-local weather change early warning signals based on atmospheric optical anomaly classification results.

2. The method for intelligent weather image acquisition and weather image recognition according to claim 1, characterized in that, S4, specifically: Based on sky polarization field data, we can identify the changes in sky polarization characteristics caused by clouds and determine the spatial distribution area of ​​clouds. Based on sky polarization light field data, we can identify regions of sky polarization characteristic changes caused by aerosol particle scattering and determine the spatial distribution region of aerosol particles. Based on sky polarization light field data, identify the regions of polarization light characteristic changes caused by haze particle scattering and determine the spatial distribution area of ​​haze particles. Based on the spatial distribution areas of clouds, aerosol particles, and haze particles, the actual spatial distribution characteristics of atmospheric components are output.

3. The method for intelligent weather image acquisition and weather image recognition according to claim 2, characterized in that, S5, specifically: Calculate the degree of polarization and the angle of polarization difference between the theoretical sky polarization mode characteristic data and the actual spatial distribution characteristic data of atmospheric composition for each spectral band; Based on the difference in polarization degree and the difference in polarization angle, generate a sky polarization degree difference distribution map and a polarization angle difference distribution map for each spectral band; For each spectral band, the regions where the difference in polarization degree and the difference in polarization angle exceed a set threshold are extracted to identify abnormal polarization regions.

4. The method for intelligent weather image acquisition and weather image recognition according to claim 3, characterized in that, S6, specifically: For each abnormal polarization region, extract the boundary contour geometric parameters of the abnormal polarization region; For each anomalous polarization region, calculate the average degree of polarization and average polarization angle of the anomalous polarization region in each spectral band; Based on the boundary contour geometric parameters of the abnormal polarization region and the average degree of polarization and average polarization angle of the abnormal polarization region in each spectral band, the abnormal polarization type of each abnormal polarization region is determined to be either halo polarization abnormality type or scattering polarization abnormality type. Based on the determination of the abnormal polarization type in the abnormal polarization region, the atmospheric optical anomaly classification result is output.

5. The method for intelligent weather image acquisition and weather image recognition according to claim 4, characterized in that, S7, specifically: Based on the atmospheric optical anomaly classification results, the spatial location parameters of each anomalous polarization region are extracted, including the center coordinate data and coverage data of the anomalous polarization region. Based on the spatial position parameters of each anomalous polarization region, calculate the offset of each anomalous polarization region relative to the center position of the target sky region; For the offset of each abnormal polarization region, the warning level of each abnormal polarization region is determined according to the preset offset threshold. Based on the warning level and spatial location parameters of each anomalous polarization region, a hyperlocal weather change warning signal corresponding to each anomalous polarization region is generated.

6. A system for intelligent weather image acquisition and weather image recognition, used to implement the method for intelligent weather image acquisition and weather image recognition as described in any one of claims 1-5, characterized in that, include: Image acquisition module: Acquires raw polarization image sequences of the target sky region through a multispectral polarization camera, and simultaneously records the camera's spatial attitude data and timestamp information, and outputs polarization image data; Data processing module: preprocesses polarization image data, extracts Stokes parameters for each pixel, and outputs standardized sky polarization light field data; Theoretical Feature Module: Based on sky polarization field data, analyze the polarization degree distribution pattern under different wavelengths, and combine the solar azimuth angle to calculate the theoretical Rayleigh scattering model, outputting the theoretical sky polarization pattern characteristics; Composition identification module: Based on sky polarization light field data, it identifies the spatial distribution of actual clouds, aerosols and haze, and outputs the actual atmospheric composition distribution characteristics; Anomaly Extraction Module: By comparing the characteristics of theoretical sky polarization patterns with the actual atmospheric composition distribution characteristics, the module calculates the difference map in the dimensions of polarization degree and polarization angle, and extracts anomalous polarization regions. Anomaly Classification Module: Based on the geometric and spectral characteristics of the anomalous polarization region, it distinguishes between halo polarization anomalies and scattering polarization anomalies, and outputs atmospheric optical anomaly classification results; Early warning generation module: Based on the atmospheric optical anomaly classification results, it generates ultra-local weather change early warning signals.

Images