Method for constructing rapid initial correction of ultraviolet limb cut height based on machine learning
Through machine learning, the ultraviolet edge cutting correction model is constructed, which solves the problem of insufficient ultraviolet edge detection accuracy in the existing technology, and achieves a fast and universal correction effect, improves the accuracy of atmospheric component inversion, and is suitable for business-oriented applications.
Patent Information
- Application Number
- CN202510380535.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-28
- Publication Date
- 2025-08-01
AI Technical Summary
The existing ultraviolet edge detection technology has a high deviation in atmospheric component inversion, resulting in insufficient accuracy, and the existing correction methods are complex in calculations and poor universality, which is not suitable for business applications.
A machine learning method is adopted to build a nonlinear cutting correction model based on inputs such as date, latitude and longitude, geometric parameters, etc., and quickly correct the MLS environment parameters, and combine it with the machine learning model XGBoost with extreme gradient enhancement for training and evaluation.
It realizes rapid and universal ultraviolet edge cutting correction, improves the accuracy of atmospheric component profile inversion, and meets business needs.
Smart Images
Figure CN120408067A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical fields of machine learning and satellite remote sensing technology, and in particular to a method for quickly initially correcting the ultraviolet limb tangent height based on machine learning. Background Art
[0002] In the fields of atmospheric composition detection using ultraviolet spectra, etc., when the limb sounding technology obtains the atmospheric profile, it has the advantages of high vertical resolution, high sensitivity, strong coverage ability, and being independent of ground control. During the detection process, due to the influence of atmospheric refraction, the limb detection spectrum has a height deviation, which in turn affects the accuracy of atmospheric composition inversion. Common methods for correcting the tangent height of the ultraviolet limb payload include the correction method based on the atmospheric refractive index model, the reference standard atmospheric model correction method, and the correction method using multi-angle observation data. These three methods use the vertical distribution data of atmospheric temperature, pressure, humidity, etc., calculate the atmospheric refractive index of each height layer according to the corresponding formula, and calculate the height correction amount caused by the deflection of light rays based on the geometric relationship of light propagation and Snell's law to correct the original observed height, without considering the requirements for speed in operational operation. The machine learning algorithm learns the internal laws and representation levels of various variables related to the measured spectrum and the tangent height, obtains the non-linear relationship of tangent height correction, thereby completing the precise correction of the ultraviolet limb tangent height and improving the inversion accuracy of the atmospheric composition profile. Summary of the Invention
[0003] The present invention provides a method for quickly initially correcting the ultraviolet limb tangent height based on machine learning. Aiming at the disadvantages of complex calculation, poor universality, and inapplicability to operations in the correction of the ultraviolet limb tangent height, a more simple and fast calculation method based on a machine learning compensation model is provided.
[0004] To achieve the above object, the present invention provides a method for quickly initially correcting the ultraviolet limb tangent height based on machine learning, including the following steps:
[0005] Obtain the ultraviolet spectrum detection data file, and extract the date, longitude and latitude tangent height sequence, and geometric parameters;
[0006] According to the extracted date, longitude and latitude, and geometric parameters, extract the environmental parameters of the MLS (Microwave Limb Sounder) in real time, and calculate the deflection angle of the atmospheric refraction light ray;
[0007] Perform preliminary correction on the tangent height in the ultraviolet spectrum detection data according to the deflection angle of the atmospheric refraction light ray to obtain the preliminarily corrected height value;
[0008] Use the date, longitude and latitude, and geometric parameters as the input of machine learning, and the preliminarily corrected tangent height value as the output label to perform model training, verification, and testing to obtain the non-linear tangent height correction model.
[0009] It also includes: evaluating the non-linear tangent height correction model, using the Shapley evaluation model to make an unbiased estimate of the contribution of each feature.
[0010] Furthermore, the geometric parameters include: solar zenith angle, observation angle, azimuth angle, field of view angle, and tangent height sequence.
[0011] Furthermore, the steps for calculating the atmospheric density are as follows:
[0012] Obtain data, and collect atmospheric temperature, pressure, and humidity data at different altitude levels from the environmental parameters of the MLS extracted in real time;
[0013] Unit conversion, ensuring that the temperature unit is K and the pressure unit is P a ; if the humidity data is not in the specific humidity form, convert it to specific humidity;
[0014] Calculate the density, substitute p, T, and q of each altitude level into the wet air density calculation formula to calculate the atmospheric density of each altitude level.
[0015] Furthermore, the steps for calculating the deflection angle of the atmospheric refracted light are as follows:
[0016] Assume that the atmosphere is divided into N layers, which are sequentially labeled as the 1st layer, the 2nd layer,..., the Nth layer from the outer layer to the inner layer; each layer of the atmosphere has a uniform refractive index n i (i = 1, 2,... n), and the interfaces of each layer are parallel planes;
[0017] Interface between the 1st layer and the 2nd layer:
[0018] Assume that the light enters the 1st layer of the atmosphere from outer space with a refractive index approximately n0 = 1 at an incident angle θ 01 , and according to Snell's law, the refraction angle θ 11 in the 1st layer of the atmosphere satisfies n0sinθ 01 = n1sinθ 11 , then
[0019] When the light enters the 2nd layer from the 1st layer, the incident angle θ 12 is equal to the refraction angle θ 11 in the 1st layer. Let the refraction angle in the 2nd layer be θ 22 , and apply Snell's law again n1sinθ 12 = n2sinθ 22 ,
[0020] The deflection angle Δθ1 of the light at the interface between the 1st layer and the 2nd layer is θ 11 - θ22 ;
[0021] Interface from layer i to layer i+1: i=1, 2, 3...N-1;
[0022] The light propagates in the i-th layer with an incident angle θ i,i+1 Arriving at the interface between the i-th layer and the i+1-th layer, let the refraction angle in the i-th layer be θ i,i , the refraction angle in the i+1th layer is θ i+1,i+1 ;
[0023] According to Snell's law i sinθ i,i =n i+1 sinθ i+1,i+1 ,have to
[0024] Deflection angle Δθ i =θ i,i -θ i+1,i+1 ;
[0025] Layer N-1 to Layer N interface:
[0026] The light enters the N-1th layer at an incident angle θ N,N-1 Entering the Nth layer, let the refraction angle in the N-1th layer be θ N-1,N-1 , the refraction angle in the Nth layer is θ N,N ;
[0027] By Snell's law N-1 sinθ N-1,N-1 =n N sinθ N,N ,
[0028] Deflection angle Δθ N =θ N-1,N-1 -θ N,N .
[0029] Furthermore, the steps for preliminary height correction are as follows:
[0030] Assume that the satellite is located at point S, the actual position of the target object is A, and due to the deflection of light, the target position detected by the satellite is B; let the light deflection angle be θ, the vertical height from the satellite to the ground be H, the uncorrected height measurement value is h0, and the corrected height value is h; if it is approximately assumed that the light is linearly deflected in the atmosphere, in the right triangle, Where α is the viewing angle of the satellite observing the target, d is the projection of the satellite to the target in the horizontal direction through the trigonometric function relationship Will Substitute the expressions of and tanα and simplify to obtain the relationship between h and h0. When the deflection angle θ is small, tanθ≈θ, in radians, and simplify the calculation formula:
[0031] Furthermore, the model selection area for model training is based on the extreme gradient boosting machine learning model XGBoost.
[0032] According to one aspect of the present invention, there is provided a storage medium in which instructions are stored. When a computer reads the instructions, the computer is caused to execute the method for constructing a fast initial correction of ultraviolet limb tangent height based on machine learning described in any one of the above.
[0033] According to another aspect of the present invention, there is provided an electronic device including a processor and the above storage medium, and the processor executes the instructions in the storage medium.
[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0035] The present invention adopts a machine learning method, which can quickly correct the tangent height of the ultraviolet limb spectrum at different times and spaces, has a fast calculation speed, strong universality, is not affected by the ultraviolet limb detection spectrum, only depends on the atmospheric temperature, humidity and pressure profiles, and has the characteristics of fast correction. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.
[0037] Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings as understood by those of ordinary skill in the art to which the present invention belongs.
[0038] Figure 1 It is a schematic diagram of the method flow of the present invention;
[0039] Figure 2 It is a schematic diagram of the generation of the preliminary tangent height correction training samples provided by a preferred embodiment of the present invention;
[0040] Figure 3 It is a schematic diagram of using machine learning for model training in a preferred embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0041] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
[0042] Example 1: As Figures 1-3 shown, an embodiment of the present invention discloses a method for constructing a rapid initial correction of ultraviolet limb tangent height based on machine learning, including the following steps:
[0043] Step 1: Obtain an ultraviolet spectral detection data file, and extract the date, longitude and latitude tangent height sequence, and geometric parameters. The geometric parameters include the following data: solar zenith angle, observation angle, azimuth angle, field of view angle, tangent height sequence h0.
[0044] Step 2: According to the date, longitude and latitude, and geometric parameters in Step 1, extract real-time MLS temperature, humidity, air pressure and other parameters for calculating the deflection angle of atmospheric refracted light.
[0045] Step 3: Based on the atmospheric refraction model, simulate and calculate the bending of atmospheric refracted light to obtain the actual propagation path of light in the atmosphere. According to the atmospheric refraction theoretical model, combined with the atmospheric parameters during satellite detection (such as vertical distribution data of atmospheric density, temperature, humidity, etc.) and the satellite's observation geometric parameters (such as observation angle, satellite altitude, etc.), calculate the light deflection angle through corresponding mathematical formulas; perform preliminary tangent height correction on the height information in the original detection data according to the deflection angle to obtain the preliminarily corrected height value h1.
[0046] The steps for calculating atmospheric density are as follows:
[0047] Obtain data: Collect atmospheric temperature T, pressure P, and humidity data humidity at different altitude levels. These data can be obtained from the MLS level 2 product set.
[0048] Unit conversion: Ensure that the temperature unit is K and the pressure unit is P a . If the humidity data is not in the form of specific humidity, it needs to be converted to specific humidity according to the corresponding relationship.
[0049] Calculate density: Substitute p, T, and q at each altitude level into the wet air density calculation formula to calculate the atmospheric density at each altitude level.
[0050] Preferably, the calculation formula for the light deflection angle is as follows:
[0051] Assume that the atmosphere is divided into N layers, which are sequentially labeled as the 1st layer, the 2nd layer,, the Nth layer from the outer layer to the inner layer. Each layer of the atmosphere has a uniform refractive index n i (i = 1, 2,...n), and the interfaces of each layer are parallel planes.
[0052] (1) Interface between the 1st layer and the 2nd layer:
[0053] Assume that light enters from outer space (refractive index approximately n0 = 1) at an incident angle θ 01Entering the first layer of the atmosphere, according to Snell's law, the refraction angle θ within the first layer of the atmosphere 11 satisfies n0sinθ 01 = n1sinθ 11 , then
[0054] When the light ray enters the second layer from the first layer, the incident angle θ 12 is equal to the refraction angle θ within the first layer 11 . Let the refraction angle within the second layer be θ 22 . Applying Snell's law again, n1sinθ 12 = n2sinθ 22 .
[0055] The deflection angle Δθ1 of the light ray at the interface between the first layer and the second layer is Δθ1 = θ 11 - θ 22 .
[0056] (2) Interface between the i-th layer and the (i + 1)-th layer (i = 1, 2, 3... N - 1):
[0057] The light ray propagates within the i-th layer and reaches the interface between the i-th layer and the (i + 1)-th layer at the incident angle θ i,i+1 . Let the refraction angle within the i-th layer be θ i,i , and the refraction angle within the (i + 1)-th layer be θ i+1,i+1 .
[0058] According to Snell's law, n i sinθ i,i = n i+1 sinθ i+1,i+1 , we can obtain
[0059] The deflection angle Δθ i = θ i,i - θ i+1,i+1 .
[0060] (3) Interface between the (N - 1)-th layer and the N-th layer:
[0061] The light ray enters the N-th layer from the (N - 1)-th layer at the incident angle θ N,N-1 . Let the refraction angle within the (N - 1)-th layer be θ N-1,N-1 , and the refraction angle within the N-th layer be θ N,N .
[0062] From Snell's law, n N-1 sinθ N-1,N-1 = n N sinθ N,N ,
[0063] The deflection angle ΔθN = θ N-1,N-1 -θ N,N 。
[0064] Preferably, the formula for tangent height correction calculation is as follows:
[0065] Assume that the satellite is located at point S, the actual position of the target object is A, and due to the deflection of light, the detected position of the target by the satellite is B. Let the light deflection angle be θ, the vertical height from the satellite to the ground be H, the uncorrected height measurement value be h0, and the corrected height value be h. If it is approximately considered that the light is linearly deflected in the atmosphere, in a right triangle, where α is the viewing angle of the satellite observing the target, and d is the horizontal projection from the satellite to the target. Through trigonometric relations Substitute and the expression of tanα into it and simplify, and the relationship between h and h0 can be obtained. In some simplified cases, when the deflection angle θ is small, tanθ≈θ (in radians), the calculation formula can be further simplified
[0066] Step 4: Use the date, longitude and latitude, solar zenith angle, azimuth angle, observation angle, field of view angle, temperature, humidity, air pressure, and original tangent height as the input of machine learning, and the preliminarily corrected tangent height as the output label to perform model training, verification, and testing to obtain a non-linear tangent height correction model.
[0067] Select the machine learning model XGBoost based on extreme gradient boosting.
[0068] Select the learner type as gbtree, set max_depth to [3, 5, 7], set gamma to [0.1, "0.2", "0.3"], set subsample to 0.8, and set colsamplebytree to [0.8, 0.9, 1]. Set the learning task Objectives to reg:squarederror, and the evaluation metric eval_metric loss function to rmse.
[0069] Perform feature selection and feature transformation. By setting the threshold to 0.05, select the features that have a significant impact on the target variable, remove irrelevant or redundant features to reduce the amount of calculation and avoid overfitting; feature transformation uses data normalization. Scale the numerical features to a specified range, usually the interval [0, 1]. Use the formula for transformation, where x max and x min are the maximum and minimum values of each feature respectively.
[0070] The GridSearch method is used for model training and evaluation, and finally the parameter combination with the best performance on the validation set is selected. The parameters are set as'max_depth': [3, 5, 7], 'learning_rate': [0.1, 0.01, 0.005], 'n_estimators': [50, 100, 150, 200, 250, 300], and a total of 54 parameter combinations are searched.
[0071] Step Five: Evaluate the model.
[0072] For the evaluation of the model, the Shapley evaluation model is used to make an unbiased estimate of the contribution of each feature.
[0073] Preferably, the Shapley evaluation model includes the following steps:
[0074] (1) Determine the dataset of the feature matrix X (with multiple features x1, x2, x3,...., x n ) and the target variable y. The feature matrix X of this model includes (date, longitude, latitude, solar zenith angle, azimuth angle, observation angle, field of view angle, original tangent height, atmospheric profile), and the target variable y is the corrected tangent height h1.
[0075] (2) Calculate the marginal contribution of the feature combination: Consider all possible combinations of feature subsets. For the matrix X with 22 feature vectors, there are a total of 2 22 feature subset combinations, and calculate the marginal contribution of each feature subset combination.
[0076] (3) Calculate the Shapley value: The Shapley value is obtained by taking the weighted average of the marginal contributions of each feature under all possible orders of feature subsets.
[0077] Example 2:
[0078] The computer-readable storage medium of this embodiment stores a computer program, and when the program is executed by a processor, it implements the steps in the method for quickly and initially correcting the ultraviolet limb tangent height based on machine learning in Example 1.
[0079] The computer-readable storage medium of this embodiment can be the internal storage unit of the terminal, such as the hard disk or memory of the terminal; the computer-readable storage medium of this embodiment can also be the external storage device of the terminal, such as the plug-in hard disk, smart memory card, secure digital card, flash card, etc. equipped on the terminal; further, the computer-readable storage medium can also include both the internal storage unit and the external storage device of the terminal.
[0080] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store the data that has been output or will be output.
[0081] Embodiment 3:
[0082] The computer device of this embodiment includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the method for constructing a fast initial correction of ultraviolet limb tangent height based on machine learning in Embodiment 1.
[0083] In this embodiment, the processor can be a central processing unit, or can also be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.; the memory can include a read-only memory and a random access memory, and provides instructions and data to the processor. A part of the memory can also include a non-volatile random access memory. For example, the memory can also store information about the device type.
[0084] Those skilled in the art should understand that the content disclosed in the embodiments can be provided as a method, a system, or a computer program product. Therefore, this solution can be implemented in the form of a hardware embodiment, a software embodiment, or a form combining software and hardware embodiments. Moreover, this solution can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories and optical memories, etc.) containing computer-usable program codes.
[0085] This solution is described with reference to the flowcharts and / or block diagrams of the method and computer program products according to the embodiments of this solution. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions; these computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0086] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the function specified in one process Figure 1 or more processes and / or blocks Figure 1 or more blocks specified in the block.
[0087] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the function specified in one process Figure 1 or more processes and / or blocks Figure 1 or more blocks specified in the block.
[0088] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.
[0089] The examples described in the present invention are only descriptions of the preferred embodiments of the present invention, and do not limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various deformations and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope of the present invention.
Claims
1. A method for quickly and initially correcting the ultraviolet limb tangent height based on machine learning, characterized in that It includes the following steps: Obtain the ultraviolet spectrum detection data file, and extract the date, longitude and latitude altitude sequence, and geometric parameters; According to the extracted date, longitude and latitude, and geometric parameters, extract the environmental parameters of MLS in real time, and calculate the deflection angle of the atmospheric refracted light; Preliminarily correct the altitude in the ultraviolet spectrum detection data according to the deflection angle of the atmospheric refracted light to obtain the preliminarily corrected altitude value; Use the date, longitude and latitude, and geometric parameters as the input of machine learning, and the preliminarily corrected altitude value as the output label to perform model training, verification, and testing to obtain a non-linear altitude correction model.
2. The method according to claim 1, characterized in that, It also includes: Evaluate the non-linear altitude correction model, use the Shapley evaluation model to perform an unbiased estimate of the contribution of each feature.
3. The method according to claim 1, wherein The geometric parameters include: solar zenith angle, observation angle, azimuth angle, field of view angle, and altitude sequence.
4. The method according to claim 1, wherein The steps for calculating the atmospheric density are as follows: Obtain data, and collect the atmospheric temperature, pressure, and humidity data of different altitude layers from the environmental parameters of MLS extracted in real time; Unit conversion to ensure the temperature unit is K and the pressure unit is P a ; If the humidity data is not in specific humidity form, convert it to specific humidity; Calculate the density by substituting p, T, and q at each altitude level into the wet air density calculation formula to calculate the atmospheric density at each altitude level.
5. The method according to claim 1, characterized in that, The steps for calculating the deflection angle of the atmospheric refracted light are as follows: Suppose the atmosphere is divided into N layers, which are sequentially labeled as the 1st layer, the 2nd layer, ……, the Nth layer from the outer layer to the inner layer; each layer of the atmosphere has a uniform refractive index n i (i = 1, 2,... n), and the interfaces of each layer are parallel planes; Interface from the first layer to the second layer: Assume that light comes from outer space with a refractive index approximately \(n_0 = 1\) at an incident angle \(\theta\). 01 It enters the first layer of the atmosphere. According to Snell's law, the refraction angle \(\theta\) in the first layer of the atmosphere 11 satisfies \(n_0\sin\theta\) 01 \(= n_1\sin\theta\) 11 , then When light enters from the first layer into the second layer, the incident angle is θ 12 which is equal to the refraction angle θ in the first layer 11 . Let the refraction angle in the second layer be θ 22 . Applying Snell's law again, n1sinθ 12 = n2sinθ 22 . The deflection angle Δθ1 of the light at the interface between the first layer and the second layer is θ 11 -θ 22 ; Interface from the i-th layer to the (i + 1)-th layer: i = 1, 2, 3... N - 1; Light travels within the i-th layer at an incident angle θ i,i+1 and reaches the interface between the i-th layer and the (i + 1)-th layer. Let the refraction angle within the i-th layer be θ i,i , and the refraction angle within the (i + 1)-th layer be θ i+1,i+1 ; According to Snell's law n i sinθ i,i =n i+1 sinθ i+1,i+1 , we get Deflection angle Δθ i = θ i,i - θ i+1,i+1 ; Interface from the (N - 1)-th layer to the N-th layer: Light enters the Nth layer at an incident angle θ from the (N - 1)th layer N,N-1 Let the refraction angle in the (N - 1)th layer be θ N-1,N-1 , and the refraction angle in the Nth layer be θ N,N ; From Snell's law n N-1 sinθ N-1,N-1 = n N sinθ N,N , Deflection angle Δθ N = θ N-1,N-1 - θ N,N .
6. The method according to claim 1, wherein The steps for preliminary altitude correction are as follows: Suppose the satellite is located at point S, the actual position of the target object is A, and due to light deflection, the target position detected by the satellite is B; let the light deflection angle be θ, the vertical height of the satellite from the ground be H, the uncorrected height measurement be h0, and the corrected height value be h; if it is approximately considered that the light is linearly deflected in the atmosphere, in a right triangle, where α is the viewing angle of the satellite observing the target, and d is the horizontal projection of the satellite to the target. Through trigonometric relations Substitute and the expressions of tanα into it and simplify to obtain the relationship between h and h0; when the deflection angle θ is small, tanθ≈θ, in radians, simplify the calculation formula:
7. The method according to claim 1, characterized in that The model selection area for model training is based on the machine learning model XGBoost of extreme gradient boosting.
8. A storage medium, characterized in that, The storage medium stores instructions, and when the computer reads the instructions, the computer executes the method for quickly and preliminarily correcting the ultraviolet limb altitude based on machine learning as described in any one of claims 1-7.
9. An electronic device, characterized in that, It includes a processor and the storage medium described in claim 9, and the processor executes the instructions in the storage medium.