A method for correcting line-of-sight errors caused by thermal deformation of optical loads in medium and low orbits

Abstract

This invention relates to the fields of optical sensor calibration and deep learning technology, and discloses a method for correcting thermal deformation line-of-sight errors of medium- and low-orbit optical payloads, comprising the following steps: S1, constructing an on-orbit camera imaging model based on stars, and obtaining the camera line-of-sight vector according to the imaging model; S2, analyzing the causes of errors in the camera line-of-sight vector caused by sunlight and constructing characteristic parameters; S3, constructing a camera thermal deformation error model, and calculating the thermal deformation error matrix R under different characteristic parameters. STD S4: Improve the BP neural network by adjusting its network structure using a Newton-Raphson optimizer and initializing its network parameters using a rime optimization algorithm; S5: Set the feature parameters and thermal deformation error matrix R... STD As input data, an improved BP neural network is used for training to predict thermal deformation errors. This invention can reduce camera line-of-sight offset errors and solve the problem of frequent on-orbit camera line-of-sight correction.

Description

Technical Field

[0001] This invention relates to the fields of optical sensor calibration and deep learning technology, specifically to a method for correcting line-of-sight errors caused by thermal deformation of optical loads in medium and low orbits. Background Technology

[0002] In recent years, low- and medium-Earth orbit (LEO) cameras have been widely used in remote sensing, Earth observation, astronomical observation, and space surveillance. LEO optical payloads offer new capabilities not covered by geostationary orbit observation missions, such as the ability to continuously monitor and detect moving targets over a wide area, making them an important means of early warning and monitoring of space targets. The target monitoring and positioning accuracy of space optical payloads depends on the pointing accuracy of the camera's line of sight. Due to the short orbital period of LEO, the incident light and the space thermal environment of the optical payload change drastically and irregularly within a short period of time due to the influence of sunlight. This causes optical distortion within the camera and changes in the satellite-based mounting matrix. These changes ultimately affect the camera's imaging process and line-of-sight accuracy, significantly hindering high-precision target indication and positioning. Correcting the line-of-sight accuracy problems of optical payloads caused by space thermal deformation is of great significance for the detection, tracking, positioning, and indication of space targets.

[0003] Low- and medium-Earth orbit (LEO) satellites have short orbital periods, and their operating temperature variations are relatively stable compared to high-Earth orbit (HEO) satellites, but their patterns are less consistent. Therefore, correcting the line-of-sight (LOS) error caused by thermal deformation in space on the optical payloads of LEO satellites is more difficult. Currently, there is little research on correcting LEO camera LOS pointing errors. In 1996, a six-state extended Kalman filter algorithm based on stellar observation data was proposed to correct spacecraft attitude errors. In 2008, Leprince et al. proposed a LOS correction method for pushbroom cameras in LEO. This method is based on modeling the Satellite Pour l'Observation de la Terre (SPOT) 4-HRV1 sensor, but mainly addresses the mechanical strain caused by satellite launch, rather than the effects of thermal deformation in space. For narrow-field-of-sight tracking sensors in space-based optoelectronic systems, Clemons et al. proposed an integrated target tracking and positioning algorithm. While tracking the target, background stellar observations are used to correct the LOS pointing in real time. This method can eliminate the influence of dynamic changes in LOS pointing errors in real time, effectively improving the timeliness of the entire system and the accuracy of target tracking and positioning. Furthermore, regarding camera distortion caused by jitter and attitude fluctuations in low- and medium-Earth orbit (LEO) satellite platforms, the geometric pointing accuracy of the ZY-3 satellite using bundle adjustment for regional network adjustment reaches 3.5 pixels for planar measurement and 1.8 pixels for elevation without ground control points (GCPs); with high-precision GCPs, the accuracy reaches 1.5 pixels for planar measurement and 0.8 pixels for elevation. However, these methods for correcting camera geometric positioning errors rely on the premise that error variations follow certain patterns and are not entirely suitable for addressing camera line-of-sight (LOS) pointing offset issues caused by complex spatial thermal deformation. A significant gap exists in the field of LEO optical payload thermal deformation LOS pointing error correction. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a method for correcting thermal deformation line-of-sight errors of medium- and low-orbit optical payloads. This method utilizes an improved BP neural network to map the relationship between the space thermal environment and the thermal deformation error matrix, thereby reducing camera line-of-sight offset and significantly improving the accuracy of camera line-of-sight correction and resolving the issue of frequent on-orbit corrections.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: a method for correcting the line-of-sight error caused by thermal deformation of optical loads in medium and low orbits, comprising the following steps:

[0006] S1, Construct an on-orbit camera imaging model based on stars, and obtain the camera line-of-sight vector based on the imaging model;

[0007] S2, Analyze the reasons for the error caused by sunlight on the camera's line-of-sight vector and construct feature parameters;

[0008] S3. Construct a camera thermal deformation error model and calculate the thermal deformation error matrix R under different characteristic parameters. STD ;

[0009] S4. Improve the BP neural network by using a Newton-Raphson optimizer to adjust the network structure of the BP neural network and using the rime optimization algorithm to initialize the network parameters of the BP neural network.

[0010] S5: Combine the characteristic parameters in S2 and the thermal deformation error matrix R in S3 STD As input data, the input data is trained using the improved BP neural network in S4 to predict thermal deformation error and complete the correction of camera line-of-sight error.

[0011] Furthermore, the above-mentioned method for correcting the line-of-sight error due to thermal deformation of optical loads in low-to-medium orbit orbits, specifically S1, is as follows:

[0012] S11: Construct the internal positioning model of the low-to-medium orbit camera imaging model, and use the internal positioning model to transform the pixel coordinate system into the camera coordinate system to obtain the camera outgoing vector in the camera coordinate system.

[0013] S12: Construct an external positioning model for the imaging model of the low-to-medium orbit camera. Use the external positioning model to transform the camera coordinate system into the celestial coordinate system to obtain the camera exit vector in the celestial coordinate system. The camera exit vector in the celestial coordinate system is the camera line-of-sight vector.

[0014] Furthermore, the above-mentioned method for correcting the line-of-sight error due to thermal deformation of optical loads in medium and low orbits, specifically S2, is as follows:

[0015] S21: Calculate the angle θ between the solar vector and the satellite vector in the celestial coordinate system. Sun-Sat included angle θ Sun-Sat These are characteristic parameters representing the influence of space thermal environment temperature changes caused by sunlight on load thermal deformation.

[0016] S22: Calculate the angle θ between the solar vector and the camera line-of-sight vector in the celestial coordinate system. Sun-LOS included angle θ Sun-LOS These are the characteristic parameters that affect the camera's optical distortion model due to sunlight.

[0017] Furthermore, in the above-mentioned method for correcting the thermal deformation line-of-sight error of medium- and low-orbit optical loads, the specific method for constructing the camera thermal deformation error model in S3 is as follows:

[0018] S31, Assume the thermal deformation error matrix R under the space thermal environment. STD Only with two feature parameters θ Sun-Sat and θ Sun-LOS Related to, and under a fixed space thermal environment, the thermal deformation error matrix R STD same;

[0019] S32, Construct the characteristic parameters and thermal deformation error matrix R STD The mapping relationship between them.

[0020] Furthermore, the improvement method in S4 of the above-mentioned method for correcting the line-of-sight error of optical loads under thermal deformation is as follows:

[0021] S41: Use the Newton-Raphson optimizer to adjust the learning rate, convergence error, and number of hidden layer nodes in the BP neural network;

[0022] S42: Use the rime optimization algorithm to initialize the weights and biases of the BP neural network. The weight values ​​include the weight values ​​from the input layer to the hidden layer and between each hidden layer. The bias values ​​include the bias values ​​of each hidden layer.

[0023] Furthermore, the above-mentioned method for correcting the line-of-sight error due to thermal deformation of optical loads in low-to-medium orbit, specifically S5, is as follows:

[0024] S51: Combine the two characteristic parameters θ Sun-Sat and θ Sun-LOS The thermal deformation error matrix R is used as the independent variable. STD Create a dataset as the dependent variable;

[0025] S52: Divide the dataset into a training set and a test set in a 7:3 ratio. Input the training set into the improved BP neural network for training to obtain a network that represents the mapping relationship between feature parameters and thermal deformation error matrix.

[0026] S53: Input the independent variable portion of the test set into the trained network to predict the dependent variable thermal deformation error matrix R. STD ;

[0027] S54: Using the thermal deformation error matrix R STD The camera line-of-sight vector is corrected using the formula (1):

[0028] LOS J-correct =R STD ·LOS J-cal (1);

[0029] Among them, LOS J-cal LOS is the calculated value of the camera's line-of-sight vector. J-correct This is the correction value for the camera's line-of-sight vector.

[0030] The beneficial effects of this invention are as follows: This invention proposes to use the angular relationship between the solar vector, satellite vector, and camera line-of-sight vector to characterize the thermal environment of the payload, which provides the possibility for quantitative analysis of the complex factor of the space thermal environment and overcomes the problem of irregular line-of-sight errors of medium and low orbit payloads that require frequent correction.

[0031] This invention uses deep learning networks to address the camera's line-of-sight correction problem. By leveraging the complexity of neural networks, it solves the mapping relationship between the spatial thermal environment and the camera's line-of-sight offset, thereby significantly improving the accuracy of the camera's line-of-sight correction.

[0032] This invention improves the BP neural network by using a Newton-Raphson optimizer network structure for optimization, and uses a rime optimization algorithm to initialize the network parameters, thereby improving network performance and reducing the possibility of overfitting. Attached Figure Description

[0033] Figure 1 This is a flowchart of the method;

[0034] Figure 2 This is a schematic diagram of the camera's interior orientation model;

[0035] Figure 3 Schematic diagrams of the focal plane coordinate system and the ideal focal plane coordinate system;

[0036] Figure 4 This is a schematic diagram of the camera's external orientation model;

[0037] Figure 5 This is a schematic diagram showing the angle between the solar vector and the satellite vector;

[0038] Figure 6 This is a schematic diagram showing the angle between the solar vector and the camera's line-of-sight vector. Detailed Implementation

[0039] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0040] like Figure 1-6 As shown, this embodiment provides a method for correcting the line-of-sight error caused by thermal deformation of optical loads in medium and low orbits, including the following steps:

[0041] S1, Construct an on-orbit camera imaging model based on stars. The camera imaging model includes an internal positioning model and an external positioning model. The camera line-of-sight vector is obtained based on the imaging model.

[0042] S11: Construct the internal positioning model of the low-to-medium orbit camera imaging model, and use the internal positioning model to transform the pixel coordinate system into the camera coordinate system to obtain the camera outgoing vector in the camera coordinate system.

[0043] like Figure 2 The diagram shows the internal positioning model, where O1-ij is the pixel coordinate system, O2-uv is the focal plane coordinate system, O4-xyz represents the ideal space coordinate system, and O5-X represents the ideal space coordinate system. cp Y cp Z cp This is the camera coordinate system. The pixel coordinate system O1-ij directly corresponds to the camera's output image data. We define its origin at the top left corner of the image, with the j-axis pointing downwards along the image column direction and the i-axis pointing to the right along the image row direction. The pixel coordinate system uses pixels as its unit, and its origin is (1, 1). The focal plane coordinate system O2-uv is established based on the pixel coordinate system. Its function is to calculate the physical distance of each pixel relative to the center of the focal plane. The u and v axes of the focal plane coordinate system are parallel to the i and j axes of the pixel coordinate system, respectively, and its origin is at the physical location corresponding to the center of the pixel coordinate system.

[0044] Optical distortion is prevalent in the design, manufacturing, and assembly of optical instruments. Therefore, a virtual coordinate system, the ideal focal plane coordinate system O3-u'v', is introduced for distortion correction. Assuming the camera distortion model is D(u, v), according to distortion theory, the correspondence between coordinate systems O3-u'v' and O2-uv is as follows:

[0045]

[0046] Where (u, v) and (u′, v') represent the coordinates of the image point in the focal plane coordinate system and the coordinates of the actual image point in the ideal focal plane coordinate system, respectively. Figure 3 This is a schematic diagram of the focal plane coordinate system and the ideal focal plane coordinate system.

[0047] Assuming the coordinates of the origin O3 in the O2-uv coordinate system are (u0, v0), which represents the center of the ideal image plane of the camera, i.e., the position of the camera's principal point. If the distortion correction parameters for a certain pixel are expressed as (Δu(u, v), Δv(u, v)), then the distortion correction model for that pixel is expressed as:

[0048]

[0049] The ideal spatial coordinate system O4-xyz is a three-dimensional coordinate system with its origin O4 located at the camera projection center. It is used to transform the two-dimensional coordinates (u', v') in the ideal focal plane into the three-dimensional coordinate system (u', v', f). Here, f is the camera principal distance. The x and y axes are parallel to u' and v', respectively.

[0050] Camera coordinate system O5-X cp Y cp Z cpThe origin is located at the geometric center of the camera reference prism. The transformation matrix R between the camera exit vector in the ideal spatial coordinate system and the camera coordinate system is... IS2C It is related to the reflection matrix of the pointing mirror inside the camera, as well as its azimuth and elevation angles, and needs to be calibrated according to the theodolite station setup.

[0051] The transformation process of the internal positioning model is expressed as follows:

[0052]

[0053] Among them, LOS cam Du(u) represents the camera exit vector corresponding to image point (i, j) in the camera coordinate system. ij v ij ) and Dv(u ij v ij ) represents the distortion correction parameter of the pixel corresponding to image point (i, j); dx and dy are the pixel size.

[0054] S12: Construct an external positioning model for the imaging model of the low-to-medium orbit camera. Use the external positioning model to transform the camera coordinate system into the celestial coordinate system (J2000.0) to obtain the camera exit vector in the celestial coordinate system. The camera exit vector in the celestial coordinate system is the camera line-of-sight vector.

[0055] like Figure 4 The diagram shown is a schematic of the external positioning model, where O6-X c Y c Z c Indicates the camera coordinate system, O7-X s Y s Z s Indicates the satellite's coordinate system, O8-X o Y o Z o It is the satellite orbital coordinate system, O9-X J Y J Z J It is a celestial coordinate system. The camera is mounted on a satellite platform, and the camera turntable's rotation axis has mounting angles α, β, and γ relative to the three coordinate axes of the satellite's body coordinate system. Therefore, the camera coordinate system is O6-X. c Y c Z c Transform to satellite body coordinate system O7-X s Y s Z s The installation matrix is ​​represented as R c2s (α, β, γ). Satellite body coordinate system O7-X s Y s Z s Transform to orbital coordinate system O8-X oY o Z o We need to know the satellite's three-axis attitude (pitch, roll, and yaw), and its coordinate transformation matrix is ​​represented as R. s2o (pitch, roll, yaw). The X-axis of the satellite orbital coordinate system points along the tangent to the satellite's motion, and the Z-axis points to the Earth's center. The coordinate axes follow the right-hand screw law. Orbital coordinate system and celestial coordinate system O9-X J Y J Z J The transformation relationship between (J2000.0) depends on the satellite's orbital parameters, namely the satellite's instantaneous position vector and velocity vector, and its transformation matrix is ​​expressed as R. o2J .

[0056] Based on the above description, the emission vector of a pixel in the detector array can be expressed in celestial coordinates as follows:

[0057] LOS J =R o2J ·R s2o (pitch, roll, yaw)·R c2s (α, β, γ)·LOS cam

[0058] Among them, LOS J This represents the camera exit vector corresponding to image point (i, j) in the celestial coordinate system, which is the camera line-of-sight vector at that moment.

[0059] S2, Analyze the reasons for the error caused by sunlight on the camera's line-of-sight vector and construct feature parameters;

[0060] Based on the camera imaging process, line-of-sight pointing errors are divided into interior orientation parameter errors and exterior orientation parameter errors. Interior orientation parameter errors refer to calibration errors in parameters such as the camera principal point, principal distance, and pixel distortion caused by changes in the optical system distortion model after the camera begins operation in orbit. In addition, interior orientation parameter errors also include errors in extracting the image centroid position and measurement errors in the camera's internal pointing mirror parameters. Exterior orientation parameter errors mainly include camera installation errors and errors in measuring satellite attitude and orbital parameters. Furthermore, errors caused by changes in the space thermal environment, satellite platform vibration, and time asynchrony also exist.

[0061] Measurement errors during line-of-sight determination are random errors. These errors typically have zero mean and a fixed variance, changing little over time and remaining essentially constant. An effective correction method is to use more precise measuring instruments to reduce their impact on the line of sight.

[0062] Thermal deformation error and camera mounting error are systematic errors, which can cause the corresponding measured values ​​to vary within a certain range. Among them, due to the influence of spatial thermal deformation, the optical distortion model and camera mounting matrix change drastically, which has a significant impact on the accuracy of the line of sight.

[0063] Sunlight is considered the primary cause of thermal deformation in the payload, having the greatest impact on the temperature environment and optical distortion model. Parameters directly reflecting the influence of sunlight include the satellite's real-time position, attitude, and corresponding camera attitude. However, satellite orbits are constantly changing, with different orbital positions for each orbital cycle. Achieving a complete illumination analysis across the entire orbit is virtually impossible. Furthermore, different satellites have different orbits. Using satellite position parameters as the analysis object lacks the universality required for low-Earth orbit optical payloads. Analyzing satellite attitude and the corresponding camera attitude at any given moment is even more complex, requiring analysis of camera attitude based on information such as the installation matrix under varying satellite attitude conditions. Essentially, the analysis of these parameters aims to investigate the impact of incident light from the camera on optical distortion.

[0064] To make the proposed method universal for low-orbit optical payloads and to simplify the calculation and correction process, two parameters reflecting the influence of sunlight on the payload are proposed for analysis: the angle θ between the solar position vector and the satellite position vector in the celestial coordinate system. Sun-Sat In the celestial coordinate system, the angle θ between the solar position vector and the camera's line-of-sight vector is... Sun-LOS The former reflects the effect of changes in ambient temperature caused by sunlight on the thermal deformation of the load. The latter characterizes the effect of light on the optical distortion model.

[0065] S21: Calculate the angle θ between the solar vector and the satellite vector in the celestial coordinate system (J2000.0). Sun-Sat included angle θ Sun-Sat These are characteristic parameters representing the influence of space thermal environment temperature changes caused by sunlight on load thermal deformation.

[0066] like Figure 5 As shown, θ is the angle between the solar vector and the satellite vector. Sun-Sat A schematic diagram.

[0067] First, calculate the solar position vector in the celestial coordinate system, that is, the vector pointing from the Earth's center to the Sun's center. First, convert the current time t to a Julian century number T starting from j2000.0. Then, based on the Julian century number, calculate the solar geometrical longitude L at the current time. sun Sun's aperiphery angle M sun The intermediate equation of the sun, C, is obtained. sun The expression.

[0068] C sun =(1.914600-0.004817·T-0.000014·T 2 )·sin(M sun )+(0.019993-0.000101·T)·sin(2·M sun )+0.000290·sin(3·M sun )

[0069] Θ sun =L sun +C sun

[0070] V sun =M sun +C sun

[0071] Where, Θ sun V sun These are the true ecliptic longitude and true anomalous angle of the Sun, respectively. For Θ sun After correcting for nutation and skewing aberration, the apparent ecliptic longitude λ of the sun in the ecliptic coordinate system is obtained. sun :

[0072] λ sun =Θ sun -0.00569-0.00478·sin(125.04-1934.136·T)

[0073] Calculate the true obliquity of the ecliptic ε sun :

[0074] ε sun =ε0+Δε

[0075]

[0076] Where ε0 is the obliquity of the ecliptic and the angle between the two sides, and Δε is the nutation correction for the angle.

[0077] The apparent right ascension and declination of the Sun at time t (δ sun , σ sun This can be expressed as:

[0078]

[0079] Right ascension and declination (δ, σ) and vector (X) in celestial coordinate system J Y J Z J The conversion relationship between them is as follows:

[0080]

[0081] Combining the above formulas, we can calculate the three-dimensional vector of the Sun in the celestial coordinate system:

[0082] P sun =(X sun Y sun Z sun ).

[0083] Assume the instantaneous position vector of the satellite at the corresponding moment is P. sat =(X sat Y sat Z sat If the angle θ between the solar vector and the satellite vector is 0, then... Sun-Sat The expression is:

[0084]

[0085] Where, θ Sun-Sat The range is (-180° to 180°).

[0086] S22: Calculate the angle θ between the solar vector and the camera line-of-sight vector in the celestial coordinate system (J2000.0). Sun-LOS included angle θ Sun-LOS These are the characteristic parameters that affect the camera's optical distortion model due to sunlight.

[0087] like Figure 6 As shown, θ is the angle between the solar vector and the camera's line-of-sight vector. Sun-LOS Schematic diagram.

[0088] In the celestial coordinate system (J2000.0), the solar vector P sun =(X sun Y sun Z sun ) and camera line-of-sight vector (LOS) J The included angle θ Sun-LOS Represented as:

[0089]

[0090] Where, θ Sun-LOS The angle range is (0°~180°).

[0091] S3. Construct a camera thermal deformation error model and calculate the thermal deformation error matrix R under different characteristic parameters. STD ;

[0092] In coordinate transformation, the error caused by a small angle (<1000μrad) can usually be described by the following approximation:

[0093] R(n+Δn)R(ξ+Δξ)=R(n)R(Δn)R(ξ)R(Δξ)≈R(Δn)R(Δξ)R(n)R(ξ)

[0094] ≈R(Δη+Δξ)R(η)R(ξ)

[0095] Based on the above description, the problem of inaccurate load line of sight caused by thermal deformation can be equivalently represented by the thermal deformation error matrix R. STD Problem. Assume that the thermal deformation error matrix under space thermal conditions is only related to the two characteristic parameters θ proposed in S2. Sun-Sat and θ Sun-LOS The thermal deformation error matrix is ​​the same under a fixed space thermal environment. Based on the above assumptions, it is only necessary to construct the mapping relationship between the characteristic parameters and the error matrix to achieve the correction of thermal deformation error. Therefore, the problem of correcting thermal deformation error for medium and low orbit loads is transformed into the problem of estimating the thermal deformation error matrix under a specific environment.

[0096] The theoretical value of the right ascension and declination of stars determined from the star catalog is (δ). star , σ star ), calculate the true value (LOS) of the camera's line-of-sight vector. J-star :

[0097]

[0098] Thermal deformation error matrix R STD The formula for calculation is:

[0099] LOS J-star =R STD (φ,ω)·LOS J-cal

[0100] Among them, LOS J-cal The calculated value of the camera's line-of-sight vector is obtained based on the imaging model in S1.

[0101] S4. Improve the BP neural network by using a Newton-Raphson optimizer to adjust the network structure of the BP neural network and using the rime optimization algorithm to initialize the network parameters of the BP neural network.

[0102] S41: Use the Newton-Raphson optimizer to adjust the learning rate, convergence error, and number of hidden layer nodes in the BP neural network;

[0103] The Newton-Raphson optimizer is a novel metaheuristic algorithm that uses the Newton-Raphson search rule and a trap avoidance operator to complete the entire search process and explore the optimal result. The Newton-Raphson search method improves exploration capability and convergence speed to achieve a better search space location. The trap avoidance operator helps avoid local optima traps.

[0104] S42: Use the rime optimization algorithm to initialize the weights and biases of the BP neural network. The weight values ​​include the weight values ​​from the input layer to the hidden layer and between each hidden layer. The bias values ​​include the bias values ​​of each hidden layer.

[0105] The rime ice optimization algorithm is a highly efficient optimization algorithm based on the physical phenomenon of icing. This algorithm simulates the soft and hard fogging growth processes of rime ice, constructing a soft fogging search strategy and a hard fogging penetration mechanism to realize the exploration and utilization behaviors in the optimization method. Simultaneously, it employs an active greedy selection mechanism and updates the population during the optimal solution selection phase to minimize the algorithm's tendency to get trapped in local optima.

[0106] S5: Combine the characteristic parameters in S2 and the thermal deformation error matrix R in S3 STD As input data, the input data is trained using the improved BP neural network in S4 to predict thermal deformation error and complete the correction of camera line-of-sight error.

[0107] S51: Combine the two characteristic parameters θ Sun-Sat and θ Sun-LOS The thermal deformation error matrix R is used as the independent variable. STD Create a dataset as the dependent variable;

[0108] S52: Divide the dataset into a training set and a test set in a 7:3 ratio. Input the training set into the improved BP neural network for training to obtain a network that represents the mapping relationship between feature parameters and thermal deformation error matrix.

[0109] S53: Input the independent variable portion of the test set into the trained network to predict the dependent variable thermal deformation error matrix R. STD ;

[0110] S54: Using the thermal deformation error matrix R STD The formula for correcting the camera's line-of-sight vector is:

[0111] LOS J-correct =R STD ·LOS J-cal ;

[0112] Among them, LOS J-cal LOS is the calculated value of the camera's line-of-sight vector. J-correct This is the correction value for the camera's line-of-sight vector.

[0113] In this embodiment, to verify the effectiveness of the proposed method for correcting the line-of-sight error of spatial thermal deformation of optical payloads, 90 sets of on-orbit data were used for verification.

[0114] The formula for calculating the total absolute error of the camera's line of sight is as follows:

[0115]

[0116] The average total absolute error of the camera line of sight after correction is 0.0011 rad, which is 80.28% higher than the original 0.0056 rad. Experimental results show that the proposed algorithm is effective in correcting camera line of sight errors caused by thermal deformation.

[0117] The above specific embodiments are merely explanations of the present invention and are not intended to limit the present invention. After reading this specification, those skilled in the art can make modifications to these embodiments without contributing any inventive step, but as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. A method for correcting line-of-sight errors caused by thermal deformation of optical loads in medium and low orbits, characterized in that, Includes the following steps: S1, Construct a star-based on-orbit camera imaging model, and obtain the camera line-of-sight vector based on the imaging model, specifically: S11: Construct the internal positioning model of the low-to-medium orbit camera imaging model, and use the internal positioning model to transform the pixel coordinate system into the camera coordinate system to obtain the camera outgoing vector in the camera coordinate system. S12: Construct an external positioning model for the imaging model of the low-to-medium orbit camera. Use the external positioning model to transform the camera coordinate system into the celestial coordinate system to obtain the camera exit vector in the celestial coordinate system. The camera exit vector in the celestial coordinate system is the camera line-of-sight vector. S2, analyze the reasons for the error caused by sunlight on the camera's line-of-sight vector and construct feature parameters, specifically: S21: Calculate the angle θ between the solar vector and the satellite vector in the celestial coordinate system. Sun-Sat The included angle θ Sun-Sat These are characteristic parameters representing the influence of space thermal environment temperature changes caused by sunlight on load thermal deformation. S22: Calculate the angle θ between the solar vector and the camera line-of-sight vector in the celestial coordinate system. Sun-LOS The included angle θ Sun-LOS These are the characteristic parameters that affect the camera's optical distortion model due to sunlight exposure; S3. Construct a camera thermal deformation error model and calculate the thermal deformation error matrix R under different characteristic parameters. STD The specific method for constructing the camera thermal deformation error model is as follows: S31, Assume the thermal deformation error matrix R under the space thermal environment. STD Only with two feature parameters θ Sun-Sat and θ Sun-LOS Related to, and under a fixed space thermal environment, the thermal deformation error matrix R STD same; S32, Construct the characteristic parameters and thermal deformation error matrix R STD The mapping relationship between them; S4. Improve the BP neural network by using a Newton-Raphson optimizer to adjust the network structure and using the rime optimization algorithm to initialize the network parameters. The specific improvement method is as follows: S41: Use the Newton-Raphson optimizer to adjust the learning rate, convergence error, and number of hidden layer nodes in the BP neural network; S42: Use the rime optimization algorithm to initialize the weights and biases of the BP neural network. The weight values ​​include the weight values ​​from the input layer to the hidden layer and between each hidden layer. The bias values ​​include the bias values ​​of each hidden layer. S5: Combine the characteristic parameters in S2 and the thermal deformation error matrix R in S3. STD As input data, the input data is trained using the improved BP neural network in S4 to predict thermal deformation error and correct camera line-of-sight error, specifically as follows: S51: Combine the two characteristic parameters θ Sun-Sat and θ Sun-LOS The thermal deformation error matrix R is used as the independent variable. STD Create a dataset as the dependent variable; S52: Divide the dataset into a training set and a test set in a 7:3 ratio. Input the training set into the improved BP neural network for training to obtain a network that represents the mapping relationship between feature parameters and thermal deformation error matrix. S53: Input the independent variable portion of the test set into the trained network to predict the dependent variable thermal deformation error matrix R. STD ; S54: Using the thermal deformation error matrix R STD The camera line-of-sight vector is corrected using the formula (1): THE J-correct =R STD ·THE J-cal (1); Among them, LOS J-cal LOS is the calculated value of the camera's line-of-sight vector. J-correct This is the correction value for the camera's line-of-sight vector.

Images