Infrared cloud height inversion method based on multiple optical thickness sets and residual weighting

Abstract

The invention discloses an infrared cloud height inversion method based on multiple optical thickness sets and residual weighting, and the method comprises the steps: generating a simulation curve set changing with zenith angles in advance through a radiation transmission mode under a preset atmosphere profile condition, and carrying out the interpolation of a curve to a zenith angle matrix corresponding to an instrument view field, so as to construct a multi-dimensional radiation lookup table; interpolating the multi-dimensional radiation lookup table according to the total content of the atmospheric water vapor at the current time to generate a cloud height-radiation relation sub-table; for each cloud pixel, a cloud height candidate value of the cloud pixel exists under each optical thickness grade, a residual error between actually measured radiation corresponding to the cloud height candidate value and the cloud height-radiation simulation curve is calculated, an optical thickness residual error of each grade is obtained, an optical thickness weight is constructed, the multiple cloud height candidate values are subjected to weighted fusion, and a cloud height-radiation simulation curve is obtained; and obtaining a cloud height inversion result with cloud pixels. The influence of optical thickness uncertainty on cloud height inversion can be weakened through a multi-light-thickness set and a residual weight mechanism, and the cloud height inversion precision and stability are improved.

Description

Technical Field

[0001] This invention belongs to the field of ground-based infrared remote sensing and cloud parameter inversion technology, specifically involving an infrared cloud high inversion method based on multiple optical thickness sets and residual weighting. Background Technology

[0002] Relying solely on radiative data from the 8-14 μm infrared window band for cloud height retrieval lacks physical constraints. For a given observation direction, the detected infrared radiation value is not determined by cloud height alone, but rather by the combined effects of multiple atmospheric and cloud physical factors, including, but not limited to, cloud optical thickness, cloud base height, atmospheric temperature profile, water vapor content and its vertical distribution, aerosol type and content, cloud facies, and particle size distribution. This makes decoupling cloud height from a single band of radiation inherently uncertain and prone to multiple interpretations. Current techniques typically employ radiative forcing factor sensitivity analysis to fix or parameterize some less influential factors. For example, cloud microphysical parameters, aerosol type, and content are simplified into several typical cases and kept constant in the radiative transfer mode, thereby maximizing the impact of strong factors such as cloud optical thickness, cloud height, and water vapor content. Even so, since cloud optical thickness is usually unknown in actual observations, and the same infrared radiation value can be generated by different combinations of "cloud height + optical thickness", cloud height inversion based on infrared single-channel data still lacks effective constraints, and cloud height solutions are often not unique in physics.

[0003] To simplify the problem, a typical approach is to set the cloud optical thickness to a constant (e.g., cloud optical thickness equals 10), establish a radiation-cloud height lookup table using a radiative transfer model under a preset atmospheric profile, and then use measured radiation values ​​to look up the table to retrieve the cloud height. Since this method essentially assumes that the cloud layer emits near a blackbody, the resulting cloud height is an "optically equivalent cloud height" corresponding to that fixed optical thickness condition. Only when the actual cloud optical thickness is sufficiently large, approaching that of a blackbody, can this equivalent cloud height possibly approximate the actual cloud height. When encountering thin clouds or semi-transparent clouds with small optical thickness, the fixed optical thickness assumption deviates significantly from reality, and the retrieved optically equivalent cloud height often shows a significant deviation from the actual cloud height. Summary of the Invention

[0004] In view of the aforementioned shortcomings of existing cloud height inversion methods, the purpose of this invention is to propose an infrared cloud height inversion method based on multiple optical thickness sets and residual weighting.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is: an infrared cloud high inversion method based on multiple optical thickness sets and residual weighting, comprising: Step 1: Acquire images from a ground-based infrared cloud measuring instrument with a working wavelength of 8-14 μm, and convert the images into infrared radiation images with physical quantity attributes based on laboratory radiometric calibration data; obtain the zenith angle corresponding to each pixel based on the zenith angle calibration information; Step 2: Obtain the cloud / clear sky determination result of the infrared radiation image, obtain the cloud pixels, and perform cloud height inversion only on the pixels determined to be clouds; Step 3: Under the preset atmospheric profile conditions, generate a set of simulated curves of radiation variation with zenith angle under multiple optical thickness levels, multiple water vapor contents, and multiple cloud high grid points conditions in advance through the radiative transfer mode. Step 4: Based on the zenith angle of each pixel in Step 1, interpolate and map the zenith angle-radiation simulation curve obtained in Step 3 to construct a multidimensional radiation lookup table. The multidimensional radiation lookup table includes the simulated radiation values ​​corresponding to optical thickness, water vapor content, cloud height, and zenith angle. Step 5: Obtain the total atmospheric water vapor content at the current time. Based on the total atmospheric water vapor content at the current time, interpolate the water vapor dimension of the multidimensional radiation lookup table in Step 4 to generate a cloud height-radiation relationship sub-table that matches the current atmospheric state. Step 6: For each cloud pixel, according to the cloud height-radiation relationship sub-table obtained in Step 5, find the position that is closest to the measured radiation of the cloud pixel at each optical thickness level to obtain multiple cloud height candidate values, and calculate the residual between the measured radiation corresponding to the cloud height candidate value and the cloud height-radiation simulation curve to obtain a multi-level optical thickness residual matrix. Step 7: Construct optical thickness weights based on the multi-level optical thickness residual matrix obtained in Step 6, and perform weighted fusion on the cloud height candidate values ​​corresponding to each optical thickness to obtain the final cloud height inversion result of the clouded pixels.

[0006] Furthermore, in step 3, the optical thickness grades are selected based on radiation sensitivity analysis to form a multi-level optical thickness classification used to characterize the range from thin clouds to near-blackbody thick clouds. The optical thickness is graded in no fewer than five levels. Under different cloud height conditions, the zenith angle-radiation simulation curves corresponding to each optical thickness level show an approximately equidistant separation relationship in terms of radiation as a function of zenith angle, forming a stable and identifiable family of curves.

[0007] Furthermore, step 4 includes: based on the pixel-by-pixel zenith angle matrix obtained in step 1, performing angle-by-angle interpolation on the zenith angle-radiation simulation curves generated in step 3 for each optical thickness, each water vapor content, and each cloud height grid point, so that the multidimensional radiation lookup table is consistent with the imager's field of view in the pixel dimension, thereby constructing a multidimensional radiation lookup table containing four-dimensional parameters: optical thickness, water vapor content, cloud height, and zenith angle.

[0008] Furthermore, step 5 includes: estimating the total atmospheric water vapor content at the current time based on real-time temperature and humidity parameters, and performing linear or piecewise linear interpolation between the radiation simulation results corresponding to adjacent water vapor grid points in the multidimensional radiation lookup table to generate a cloud height-radiation relationship sub-table that matches the current atmospheric temperature and humidity conditions.

[0009] Furthermore, the method for finding candidate cloud height values ​​in step 6 is as follows: under a fixed optical thickness level, find the position closest to the measured radiation along the radiation-cloud height variation curve, and obtain a unique cloud height solution through interpolation; when the measured radiation exceeds the coverage range of the optical thickness curve, output an invalid value as a candidate result for the optical thickness.

[0010] Furthermore, in step 6, the method for finding candidate cloud height values ​​is as follows: try to find the corresponding cloud height solution under all optical thickness levels. When the measured radiation exceeds the coverage range of the cloud height-radiation simulation curve, select the cloud height boundary value corresponding to the endpoint of the cloud height-radiation simulation curve as the candidate result.

[0011] Furthermore, step 7 includes: for each optical thickness level, calculating the residual between the measured radiation corresponding to the candidate cloud height value of the cloud pixel and its corresponding cloud height-radiation simulation curve, and constructing a weight model with monotonically decreasing characteristics based on the residual, so that the optical thickness condition with a smaller residual obtains a higher weight; the weight model can reflect the relative reliability of the matching degree between the optical thickness condition and the measured radiation, and allows the use of exponential decay, piecewise continuous decay or other weight construction methods with monotonically decreasing characteristics to realize the mapping from residual to weight; After obtaining the weights of all optical thickness conditions, the corresponding candidate cloud height values ​​are normalized and weighted to form the final cloud height inversion result of the current pixel. For optical thickness conditions that do not obtain a valid solution, their weights are automatically set to zero and they are removed during the fusion process.

[0012] The aforementioned infrared cloud high inversion method based on multiple optical thickness sets and residual weighting can achieve the following beneficial effects: (1) This invention selects representative and well-discriminative optical thickness levels based on radiation sensitivity experiments, enabling the zenith angle-radiation simulation curves under different optical thickness conditions to maintain a stable and approximately equidistant separation structure in the cloud height segment. This structural "curve family" feature avoids problems such as curve overlap, indistinguishability, or insufficient resolution that may be caused by arbitrarily setting optical thickness levels in the traditional method, and significantly improves the discriminability, interpretability, and physical effectiveness of multiple optical thickness sets in the cloud height inversion process. This classification strategy provides a stable and reliable physical basis for subsequent residual criteria and weighted fusion, which cannot be achieved by the traditional single-optical-thickness lookup table method.

[0013] (2) This invention obtains candidate cloud height values ​​under multiple optical thickness conditions and introduces a residual-based optical thickness credibility modeling and a decreasing weight allocation mechanism to achieve adaptive constraints on the unknown nature of cloud optical thickness. The residual weight model quantifies the consistency between measured radiation and simulated curves with different optical thicknesses into a comparable physical index, and assigns higher contribution to more credible optical thickness conditions through monotonically decreasing weights, thereby improving the cloud height solution from the traditional "single optical thickness mode" to a "multi-optical thickness physical consistency fusion mode". This method effectively avoids the systematic deviation caused by fixed optical thickness assumptions, enabling cloud height inversion to maintain higher physical consistency and stability under thin clouds, high clouds, and complex cloud fields. It is a novel inversion framework that has not been disclosed in the prior art.

[0014] (3) This invention can be achieved solely by radiometrically calibrated infrared radiation data and conventional water vapor observations, without requiring additional multispectral, multi-angle, or active remote sensing methods to provide physical constraints on light thickness uncertainties. The method of this invention is applicable to infrared cloud measurement systems with large, medium, and small fields of view, and can be deployed across sites and climate zones, possessing good engineering feasibility and business scalability.

[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to the present invention. Figure 2 This is an actual cloud height inversion effect diagram of an infrared cloud height inversion method based on multiple optical thickness sets and residual weighting according to the present invention. Detailed Implementation

[0017] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without requiring some of these specific details. The following description of embodiments is merely intended to provide a better understanding of the invention by illustrating examples of the invention.

[0018] Please see Figure 1 This invention provides a high inversion method for infrared clouds based on multiple optical thickness sets and residual weighting, comprising: Step 1: Acquire images from a ground-based infrared cloud measuring instrument with a working wavelength of 8-14 μm, and convert the images into infrared radiation images with physical quantity attributes based on laboratory radiometric calibration data; obtain the zenith angle corresponding to each pixel based on the zenith angle calibration information.

[0019] Specifically, the observation images used in this embodiment are all-sky infrared radiation images acquired by a certain type of ground-based infrared cloud measuring instrument, with a working wavelength of 8-14 μm, covering an effective field of view of approximately 150° × 120°. The observation images have been calibrated in the laboratory to form infrared radiation images in units of W•m⁻²•Sr⁻¹ and contain pixel-by-pixel field-of-view geometric information, used to describe the downward infrared radiation of the atmosphere at different zenith angles across the entire sky. This invention does not involve this calibration process, but only performs subsequent operations based on the calibrated physical quantity images. After acquiring the infrared radiation images, the measured radiation values ​​are modulated to the clear-sky radiation reference of the radiative transfer model using an existing clear-sky consistent modulation method (this method can be implemented using any existing radiation modulation technique based on a model clear-sky reference) to ensure the comparability of radiation values ​​during the inversion process.

[0020] Furthermore, the imaging system of this embodiment, such as an infrared cloud measuring instrument, needs to have pixel-by-pixel zenith angle calibration information and pixel-by-pixel radiation intensity physical quantity to ensure accurate mapping of the multidimensional radiation lookup table in pixel space and consistency of radiation value; the method is applicable to large field of view or medium and small field of view infrared cloud measuring systems, and the infrared cloud measuring instrument can provide pixel-by-pixel field of view geometric parameters and radiation intensity information across the entire field of view.

[0021] Step 2: Obtain the cloud / clear sky determination result of the infrared radiation image, obtain the cloud pixels, and perform cloud height inversion only on the pixels determined to be clouds.

[0022] Specifically, the cloud / clear sky determination result for the current time is obtained from the modulated infrared radiation image, so that each pixel simultaneously possesses the modulated radiation value and the corresponding cloud / clear sky identifier. This invention does not involve the specific implementation process of the cloud detection method; any feasible infrared radiation-based cloud detection technology in this field can be used as preprocessing.

[0023] Step 3: Under the preset atmospheric profile conditions, generate a set of simulated curves of radiation variation with zenith angle under multiple optical thickness levels, multiple water vapor contents, and multiple cloud high grid points conditions in advance through the radiative transfer mode.

[0024] Specifically, under preset atmospheric profile conditions, this embodiment utilizes radiative transfer modes to construct a set of simulated curves showing the variation of radiation with zenith angle under multiple optical thicknesses, multiple water vapor contents, and multiple cloud height grid points. The set of simulated curves includes zenith angle-radiation simulation curves and cloud height-radiation simulation curves. The grading method for each parameter is determined based on the results of previous sensitivity analysis of the main influencing factors to ensure that the generated curves have good discriminative power and engineering applicability in the radiation space.

[0025] Specifically, keeping other atmospheric conditions constant, the radiation response caused by changes in total water vapor content mainly exhibits an overall amplitude of approximately linear change, and can therefore be categorized using several representative water vapor content levels. Cloud height parameters also show a similar variation pattern, making them suitable for modeling using equally spaced height grids. In contrast, the influence of cloud optical thickness on radiation is nonlinear, and optical thickness levels are selected based on radiation sensitivity analysis, forming multiple optical thickness gradations to characterize the range from thin clouds to near-blackbody thick clouds. The number of optical thickness gradations is no less than five, and under different cloud height conditions, the zenith angle-radiation simulation curves corresponding to each optical thickness level show an approximately equally spaced separation relationship in numerical terms, forming a stable and identifiable family of curves, thereby ensuring continuous distinguishability between optical thickness dimensions. In this embodiment, optical thickness is preferably divided into typical levels with clear distinction, such as 0.5, 1, 2, 3, 5, and 10, so that the radiation variation curves with zenith angle under different optical thickness conditions maintain good separation within the range of thin to thick clouds. Simultaneously, considering that when the optical thickness is greater than approximately 10, the cloud layer is approximately a blackbody, and its radiation variation tends to saturate, setting the optical thickness to 10 is sufficient to meet the modeling requirements. Through the above parameter division, the set of simulated curves with multiple optical thicknesses, multiple water vapor levels, and multiple cloud heights constructed in this embodiment can cover the main influence range of radiation variation under typical atmospheric and cloud conditions, and has the advantages of high distinction, stable structure, and moderate computational load.

[0026] Step 4: Based on the zenith angle of each pixel in Step 1, interpolate and map the zenith angle-radiation simulation curve obtained in Step 3 to construct a multidimensional radiation lookup table. The multidimensional radiation lookup table includes the simulated radiation values ​​corresponding to optical thickness, water vapor content, cloud height, and zenith angle.

[0027] Specifically, after obtaining a set of zenith angle-radiation simulation curves showing the variation of radiation with zenith angle under various optical thicknesses, water vapor contents, and cloud heights, the simulation results need to be mapped to the actual field of view of a ground-based infrared cloud meter to construct a multidimensional radiation lookup table consistent with the instrument's pixel-by-pixel imaging geometry. Specifically, the infrared cloud meter has already obtained the zenith angle information corresponding to each pixel through field-of-view geometry calibration. In this embodiment, the radiation-zenith angle simulation curves generated by the radiative transfer mode are interpolated according to this zenith angle matrix, so that each pixel can obtain the corresponding simulated radiation value at each optical thickness, water vapor content, and cloud height grid point. The resulting lookup table is consistent with the imager output in the pixel dimension and includes information such as optical thickness, water vapor content, cloud height, and radiation in the physical dimension, forming a complete multidimensional radiation lookup table that corresponds one-to-one with the imaging data.

[0028] Step 5: Obtain the total atmospheric water vapor content at the current time. Based on the total atmospheric water vapor content at the current time, interpolate the water vapor dimension of the multidimensional radiation lookup table in Step 4 to generate a cloud height-radiation relationship sub-table that matches the current atmospheric state.

[0029] Specifically, since the infrared radiation transmission process is significantly affected by atmospheric water vapor content, this embodiment requires adaptive adjustment of the water vapor dimension of the multidimensional radiation lookup table based on the current water vapor observation value during actual cloud height inversion, in order to obtain a subset of the lookup table consistent with the current atmospheric conditions. Specifically, the total atmospheric water vapor content at the current time is estimated based on real-time temperature and humidity parameters, and linear or piecewise linear interpolation is performed between the radiation simulation results corresponding to adjacent water vapor grid points in the multidimensional radiation lookup table to generate a cloud height-radiation relationship sub-table that matches the current atmospheric temperature and humidity conditions. That is, the total atmospheric water vapor content at the current time is obtained using conventional meteorological observations or reanalysis data, and this is used as the input parameter for the water vapor content dimension of the lookup table. Based on this water vapor content value, interpolation calculations are performed on the radiation simulation results corresponding to adjacent water vapor grid points in the lookup table to generate a cloud height-radiation relationship sub-table that matches the current atmospheric water vapor conditions. This cloud height-radiation relationship sub-table retains the optical thickness and cloud height dimensions, only replacing the radiation values ​​related to water vapor, making it closer to the actual atmospheric background. Through the above processing, the lookup table in this embodiment can be dynamically updated with changes in atmospheric humidity, thereby improving the matching degree between various physical parameters in the cloud height inversion process and enhancing the physical consistency and stability of the inversion results.

[0030] Step 6: For each cloud pixel, according to the cloud height-radiation relationship sub-table obtained in Step 5, find the position that is closest to the measured radiation of the cloud pixel at each optical thickness level to obtain multiple cloud height candidate values, and calculate the residual between the measured radiation corresponding to the cloud height candidate value and the cloud height-radiation simulation curve to obtain a multi-level optical thickness residual matrix.

[0031] Specifically, after obtaining a cloud height-radiance relationship sub-table that matches the current atmospheric water vapor conditions, this embodiment searches for corresponding cloud height candidate values ​​for each cloud pixel under various optical thickness conditions. Specifically, for a given pixel, its radiation variation curve with cloud height exhibits monotonic variation at a given optical thickness level. Therefore, the corresponding cloud height solution can be obtained by finding the position on the cloud height-radiance simulation curve that is closest to the measured radiation value. In one implementation, when the measured radiation value is within the effective range of the optical thickness curve, a unique cloud height candidate value can be obtained through interpolation. When the measured radiation exceeds the curve range, resulting in no effective solution, an invalid value can be output to indicate that no cloud height candidate value was obtained under that optical thickness condition. In another implementation, the corresponding cloud height solution is searched at all optical thickness levels. When the measured radiation exceeds the coverage of the cloud height-radiance simulation curve, the cloud height boundary value corresponding to the endpoint of the cloud height-radiance simulation curve can be selected as a candidate result to improve the continuity of cloud height results in engineering applications. This invention covers both of the above methods, and the specific implementation can be selected according to business needs. Through the above processing, a set of cloud height candidate values ​​can be obtained for each cloud pixel at multiple optical thickness levels, providing input for subsequent residual calculation and weighted fusion.

[0032] Step 7: Construct optical thickness weights based on the multi-level optical thickness residual matrix obtained in Step 6, and perform weighted fusion on the cloud height candidate values ​​corresponding to each optical thickness to obtain the final cloud height inversion result of the clouded pixels.

[0033] Specifically, after obtaining candidate cloud height values ​​and corresponding residuals for a pixel under multiple optical thickness conditions, this embodiment constructs optical thickness weights based on the residual magnitudes and performs weighted fusion on multiple candidate cloud heights to obtain the final cloud height inversion result. For each optical thickness level, the residual between the measured radiation corresponding to the candidate cloud height value of the clouded pixel and its corresponding cloud height-radiation simulation curve is calculated, and a weight model with monotonically decreasing characteristics is constructed based on this residual, so that optical thickness conditions with smaller residuals receive higher weights; the weight model can reflect the relative reliability of the matching degree between optical thickness conditions and measured radiation, and allows the use of exponential decay, piecewise continuous decay, or other weight construction methods with monotonically decreasing characteristics to achieve the mapping from residuals to weights; After obtaining the weights of all optical thickness conditions, the corresponding candidate cloud height values ​​are normalized and weighted to form the final cloud height inversion result of the current pixel. For optical thickness conditions that do not obtain a valid solution, their weights are automatically set to zero and they are removed during the fusion process.

[0034] Specifically, let Z be the candidate value of the cloud height for a certain cloud pixel under the j-th optical thickness condition. j The corresponding radiation residual is ε jThe residual is defined as the minimum distance between the measured radiance value corresponding to the candidate cloud height point and the simulated curve, reflecting the degree of matching between the optical thickness condition and the measured radiance. To give higher weight to optical thickness conditions with smaller residuals, this embodiment constructs the weighting coefficients in the following form: Where k is the weight decay parameter, whose value can be configured according to business needs or actual data characteristics to adjust the sensitivity of the residual to the weight distribution. As the residual increases, the corresponding weight decays exponentially, thus highlighting the optical thickness conditions that are more consistent with the measured radiation. After obtaining the weights, the candidate cloud height values ​​under each optical thickness condition are normalized and weighted and fused to obtain the final cloud height inversion result: Where N represents the number of optical thickness levels. In this embodiment, the weight of optical thickness conditions without a valid solution can be set to zero, thereby automatically eliminating such optical thickness conditions during the fusion process. By constructing weights based on residuals and performing weighted fusion, this embodiment can adaptively select optical thickness conditions that are most consistent with the measured radiation, thereby effectively reducing the impact of optical thickness uncertainty on cloud height inversion and making the inversion results more stable, consistent, and physically uniform.

[0035] This embodiment has the following beneficial effects: (1) This invention selects representative and well-discriminative optical thickness levels based on radiation sensitivity experiments, enabling the zenith angle-radiation simulation curves under different optical thickness conditions to maintain a stable and approximately equidistant separation structure in the cloud height segment. This structural "curve family" feature avoids problems such as curve overlap, indistinguishability, or insufficient resolution that may be caused by arbitrarily setting optical thickness levels in the traditional method, and significantly improves the discriminability, interpretability, and physical effectiveness of multiple optical thickness sets in the cloud height inversion process. This classification strategy provides a stable and reliable physical basis for subsequent residual criteria and weighted fusion, which cannot be achieved by the traditional single-optical-thickness lookup table method.

[0036] (2) This invention obtains candidate cloud height values ​​under multiple optical thickness conditions and introduces a residual-based optical thickness credibility modeling and a decreasing weight allocation mechanism to achieve adaptive constraints on the unknown nature of cloud optical thickness. The residual weight model quantifies the "consistency between measured radiation and simulated curves with different optical thicknesses" into a comparable physical index, and assigns higher contribution to more credible optical thickness conditions through monotonically decreasing weights, thereby improving the cloud height solution from the traditional "single optical thickness mode" to a "multi-optical thickness physical consistency fusion mode". This method effectively avoids the systematic bias caused by the fixed optical thickness assumption, enabling cloud height inversion to maintain higher physical consistency and stability under thin clouds, high clouds, and complex cloud fields. It is a novel inversion framework that has not been disclosed in the prior art.

[0037] (3) This invention can be achieved solely by radiometrically calibrated infrared radiation images and conventional water vapor observations, without requiring additional multispectral, multi-angle, or active remote sensing methods to provide physical constraints on light thickness uncertainties. The method of this invention is applicable to infrared cloud measurement systems with large, medium, and small fields of view, and can be deployed across sites and climate zones, possessing good engineering feasibility and business expansion capabilities.

[0038] The above description is merely a preferred embodiment of the present invention. Any simple modifications, equivalent changes, and alterations made by those skilled in the art to the above embodiments without departing from the scope of the present invention and based on the technical essence of the present invention shall still fall within the scope of the present invention.

Claims

1. A high-inversion method for infrared clouds based on multiple optical thickness sets and residual weighting, characterized in that, include: Step 1: Acquire images from a ground-based infrared cloud measuring instrument with a working wavelength of 8-14 μm, and convert the images into infrared radiation images with physical quantity attributes based on laboratory radiometric calibration data; obtain the zenith angle corresponding to each pixel based on the zenith angle calibration information; Step 2: Obtain the cloud / clear sky determination result of the infrared radiation image, obtain the cloud pixels, and perform cloud height inversion only on the pixels determined to be clouds; Step 3: Under the preset atmospheric profile conditions, generate a set of simulated curves of radiation variation with zenith angle under multiple optical thickness levels, multiple water vapor contents, and multiple cloud high grid points conditions in advance through the radiative transfer mode. Step 4: Based on the zenith angle of each pixel in Step 1, interpolate and map all the zenith angle-radiation simulation curves obtained in Step 3 to construct a multidimensional radiation lookup table. The multidimensional radiation lookup table includes the simulated radiation values ​​corresponding to optical thickness, water vapor content, cloud height, and zenith angle. Step 5: Obtain the total atmospheric water vapor content at the current time. Based on the total atmospheric water vapor content at the current time, interpolate the water vapor dimension of the multidimensional radiation lookup table in Step 4 to generate a cloud height-radiation relationship sub-table that matches the current atmospheric state. Step 6: For each cloud pixel, according to the cloud height-radiation relationship sub-table obtained in Step 5, find the position that is closest to the measured radiation of the cloud pixel at each optical thickness level to obtain multiple cloud height candidate values, and calculate the residual between the measured radiation corresponding to the cloud height candidate value and the cloud height-radiation simulation curve to obtain a multi-level optical thickness residual matrix. Step 7: Construct optical thickness weights based on the multi-level optical thickness residual matrix obtained in Step 6, and perform weighted fusion on the cloud height candidate values ​​corresponding to each optical thickness to obtain the final cloud height inversion result of the clouded pixels.

2. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, In step 3, the optical thickness grades are selected based on radiation sensitivity analysis to form a multi-level optical thickness classification used to characterize the range from thin clouds to near-blackbody thick clouds. The optical thickness is graded in no fewer than five levels. Under different cloud height conditions, the zenith angle-radiation simulation curves corresponding to each optical thickness level show an approximately equidistant separation relationship in terms of radiation as a function of zenith angle, forming a stable and identifiable family of curves.

3. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, Step 4 includes: based on the pixel-by-pixel zenith angle matrix obtained in Step 1, performing angle-by-angle interpolation on the zenith angle-radiation simulation curves generated in Step 3 for each optical thickness, each water vapor content, and each cloud height grid point, so that the multidimensional radiation lookup table is consistent with the imager's field of view in the pixel dimension, thereby constructing a multidimensional radiation lookup table containing four-dimensional parameters of optical thickness, water vapor content, cloud height, and zenith angle.

4. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, Step 5 includes: estimating the total atmospheric water vapor content at the current time based on real-time temperature and humidity parameters, and performing linear or piecewise linear interpolation between the radiation simulation results corresponding to adjacent water vapor grid points in the multidimensional radiation lookup table to generate a cloud height-radiation relationship sub-table that matches the current atmospheric temperature and humidity conditions.

5. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, The method for finding candidate cloud height values ​​in step 6 is as follows: under a fixed optical thickness level, find the position closest to the measured radiation along the cloud height-radiation simulation curve, and obtain a unique cloud height solution through interpolation; when the measured radiation exceeds the coverage range of the optical thickness curve, output an invalid value as a candidate result for the optical thickness.

6. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, The method for finding candidate cloud height values ​​in step 6 is as follows: try to find the corresponding cloud height solution under all optical thickness levels. When the measured radiation exceeds the coverage range of the cloud height-radiation simulation curve, select the cloud height boundary value corresponding to the endpoint of the cloud height-radiation simulation curve as the candidate result.

7. The infrared cloud high inversion method based on multiple optical thickness sets and residual weighting according to claim 1, characterized in that, Step 7 includes: For each optical thickness level, the residual between the measured radiation corresponding to the candidate cloud height value of the cloud pixel and its corresponding cloud height-radiation simulation curve is calculated, and a weight model with monotonically decreasing characteristics is constructed based on the residual, so that the optical thickness condition with the smaller residual gets a higher weight. After obtaining the weights of all optical thickness conditions, the corresponding cloud height candidate values ​​are normalized and weighted to form the final cloud height inversion result of the current pixel. For optical thickness conditions that do not obtain a valid solution, their weights are automatically set to zero and removed during the weighted fusion process.

Images