A method for generating a high-fidelity spatial far-field time-sensitive target optical data set

Abstract

The present invention relates to a method for generating an optical dataset of a far-field time-sensitive target in space, and specifically to a method for generating an optical dataset of a far-field time-sensitive target in space with high fidelity, which solves the technical problem of the current lack of a method for quickly generating an optical dataset of a far-field time-sensitive target with complete label information in a complex space environment. The present invention first establishes a space camera imaging model that converts stars from a celestial coordinate system to a space camera imaging coordinate system, and then performs coordinate transformation and star-to-grayscale conversion on the stars in the field of view of the space camera according to the navigation star catalog, and displays them on a display to obtain an ideal background star map. At the same time, a simulated image of the time-sensitive target is obtained according to the same method, and a light path is traced to obtain an initial simulated image including the position and brightness characteristics of the time-sensitive target. Then, the noise in the imaging is simulated, and the ideal background star map, the noise and the initial simulated image are fused to achieve the generation of a high-fidelity optical image of a far-field time-sensitive target in space.

Description

Technical Field

[0001] The present invention relates to a method for generating a spatial far-field time-sensitive target optical data set, and in particular to a method for generating a high-fidelity spatial far-field time-sensitive target optical data set. Background Art

[0002] Generating optical images of time-sensitive targets in the far-field space is of great significance in fields such as satellite navigation, space debris warning, and deep space exploration. Existing detection of time-sensitive targets in the far-field space is primarily based on traditional methods. While deep learning-based methods significantly outperform traditional methods in many computer vision problems, they rely on large-scale annotated datasets to train network models. However, in the field of optical detection and warning of time-sensitive targets in the far-field space, there is currently no available dataset for such targets, and there is also a lack of methods for rapidly generating far-field optical images and annotating multidimensional information in complex environments.

[0003] Currently, the generation of optical images of time-sensitive targets in the far-field space mainly relies on star map simulation and emulation. Star map simulation is a prerequisite for research on detection algorithms for time-sensitive targets in the far-field space. Since real star maps are difficult to obtain, digital simulation of star maps is necessary. High-precision simulated star maps can ensure the accuracy of detection of time-sensitive targets in the far-field space.

[0004] Extensive research has been conducted on high-fidelity star map simulation technology. In their paper, "Simulation of Sequential Star Maps of Space Targets Based on STK / MATLAB," published in the 2014 issue of Infrared and Laser Engineering, Volume 43, Issue 9, Liu Fucheng et al. proposed a method for generating sequential star maps capable of simulating faint, point-like moving targets against a complex starry sky. Using STK software, they modeled far-field time-sensitive targets, predicted the relative geometric relationship between the observation satellite and the far-field time-sensitive targets in geosynchronous orbit, and systematically analyzed the visibility of high-orbit targets. STK generated predicted orbital data for the space targets, which were simulated using Matlab programming based on the Hipparcos catalog. In their dissertation, "Research on Star Map Recognition Simulation for Star Sensors Based on Scilab / Xcos," published on CNKI in 2020, Fan Meng used the Scilab / Xcos simulation platform and the Hipparcos catalog to analyze mathematical models for star-to-grayscale conversion, grayscale diffusion models for star points, and the sources of star map noise, thereby simulating star maps captured by narrow-field-of-view cameras. However, the aforementioned star map simulation and emulation work did not consider the high-fidelity characteristics of visible light optical images. Wang Yupeng et al. published "A Fast Simulation Method for Space-Based Optical Observation Images of Massive Space Debris" in the journal Progress in Lasers and Optoelectronics, Vol. 59, No. 16, 2022. They proposed a strategy for rapidly solving the visibility of massive non-cooperative targets in space under multiple constraints. This strategy ensures the fidelity of the simulated images and speeds up the imaging simulation. However, it lacks the ability to simulate images containing real-time, time-sensitive target label information.

[0005] In summary, there is currently no public, high-fidelity far-field time-sensitive target optical dataset with complete label information at home and abroad, and there is also a lack of a method to quickly generate far-field time-sensitive target optical dataset with complete label information in complex spatial environments. Summary of the Invention

[0006] The purpose of the present invention is to solve the technical problem of the current lack of a method for quickly generating a far-field time-sensitive target optical dataset with complete label information in a complex spatial environment, and to provide a high-fidelity spatial far-field time-sensitive target optical dataset generation method.

[0007] To achieve the above object, the technical solution adopted by the present invention is:

[0008] A method for generating a high-fidelity spatial far-field time-sensitive target optical data set is characterized in that it includes the following steps:

[0009] Step 1: Set the space camera parameters. Then, based on the space camera parameters and the given space camera's visual axis, establish the rotation relationship between the celestial coordinate system and the space camera coordinate system, as well as the constraints for imaging stars within the space camera's field of view. Then, use the projection transformation method to obtain the space camera imaging model that transforms stars from the celestial coordinate system to the space camera imaging coordinate system.

[0010] Step 2: Based on different navigation star catalogs, the right ascension and declination of all stars in the catalog are obtained in the celestial coordinate system. Then, based on the space camera imaging model, the right ascension and declination of the stars in the celestial coordinate system are converted into the coordinates of the space camera imaging coordinate system to obtain the ideal background star map dataset;

[0011] Step 3: Based on the real time-sensitive target information, the right ascension and declination of the time-sensitive target in the celestial coordinate system are obtained, and according to the method of step 2, a simulated image set of the time-sensitive target is obtained; then, the light path of the time-sensitive target in the simulated image set is traced to obtain a spot intensity map of the time-sensitive target in the imaging coordinate system of the space camera, thereby obtaining an initial simulated image set including the position and brightness characteristics of the time-sensitive target;

[0012] Step 4: simulate spatial stray light and spatial camera noise to obtain a set of spatial stray light simulation images and a spatial camera noise simulation image;

[0013] Step 5: Using a double loop nesting method, the ideal background star map in the ideal background star map dataset is superimposed with each spatial stray light simulation map in the spatial stray light simulation image set to obtain a complex star background map dataset; then all the complex star background maps in the complex star background map dataset are superimposed with the spatial camera noise simulation map to obtain a fused complex star background map dataset;

[0014] Step 6. Brightness enhancement is performed on the time-sensitive targets in all the initial simulated images in the initial simulated image set to obtain a brightness-enhanced image set; then, the brightness-enhanced images in the brightness-enhanced image set are embedded into the fused complex starry sky background image in the fused complex starry sky background image dataset to obtain a high-fidelity spatial far-field time-sensitive target optical dataset.

[0015] Furthermore, step 1 is specifically as follows:

[0016] Step 1.1, setting the space camera parameters, including the field of view of the space camera, the focal length of the space camera, the number of pixels of the space camera, and the pixel size;

[0017] Step 1.2: Based on the visual axis of the space camera, obtain the rotation relationship between the celestial coordinate system and the space camera coordinate system as shown in the following formula:

[0018]

[0019]

[0020] in, The star is in the space camera coordinate system O ′- XYZ The direction vector in X 、 Y 、 Z The stars are in the space camera coordinate system O ′- XYZ The direction vector in X axis, Y axis, Z Components on the axis; is the star in the celestial coordinate system O - UVW The unit direction vector in , U 、 V 、 W The stars are in the celestial coordinate system O - UVW The unit direction vector in U axis, V axis, W Components on the axis; R is the rotation matrix; is the right ascension of the visual axis of the space camera in the celestial coordinate system, is the declination of the visual axis of the space camera in the celestial coordinate system;

[0021] Step 1.3: Based on the field of view of the space camera, obtain the constraint conditions for the star imaging within the field of view of the space camera as shown in the following formula:

[0022]

[0023] in, For the i The right ascension of a star in the celestial coordinate system, For the i The declination of a star in the celestial coordinate system, i is an integer, and 1≤ i ≤H, H is the number of stars; r is the field of view radius of the space camera, , represents the field of view of the space camera, Indicates that the space camera is in the space camera imaging coordinate system x Axis field of view, Indicates that the space camera is in the space camera imaging coordinate system y Axis field of view;

[0024] Step 1.4: Based on the right ascension and declination of the star in the celestial coordinate system, obtain the formula for calculating the unit direction vector of the star in the celestial coordinate system:

[0025]

[0026] in, For the i The unit direction vector of a star in the celestial coordinate system, 、 、 Respectively i The unit direction vector of a star in the celestial coordinate system is U axis, V axis, W Components on the axis;

[0027] Step 1.5: Based on the rotation relationship between the celestial coordinate system and the space camera coordinate system obtained in step 1.2, the constraints for imaging stars within the field of view of the space camera obtained in step 1.3, and the calculation formula for the unit direction vector of stars in the celestial coordinate system obtained in step 1.4, establish a star point spatial distribution model that converts stars from the celestial coordinate system to the space camera coordinate system:

[0028]

[0029]

[0030] in, For the i The unit direction vector of a star in the space camera coordinate system, 、 、 Respectively iThe unit direction vector of a star in the space camera coordinate system is X axis, Y axis, Z Components on the axis;

[0031] Step 1.6: Use projection transformation to establish the projection relationship between the space camera coordinate system and the space camera imaging coordinate system to obtain the perspective projection transformation model:

[0032]

[0033]

[0034] in, 、 Respectively i The stars in the space camera imaging coordinate system x Axis coordinates, y axis coordinates; f is the focal length of the space camera, L is the number of pixels of the space camera, pixel_size is the pixel size;

[0035] Step 1.7: Substitute the star point spatial distribution model established in step 1.5 into the perspective projection transformation model obtained in step 1.6 to obtain the space camera imaging model:

[0036]

[0037]

[0038] in, 、 Represent the imaging coordinate system of the space camera x axis, y Axis pixel number, based on the number of pixels of the space camera L get.

[0039] Furthermore, step 2 is specifically as follows:

[0040] Step 2.1, define the sky area scanning grid;

[0041] Step 2.2, obtain the right ascension and declination of one of the navigation catalog stars in the celestial coordinate system according to the sky scanning grid;

[0042] Step 2.3: Based on the space camera imaging model, obtain the coordinates of the star in the space camera imaging coordinate system according to the right ascension and declination of the star in the celestial coordinate system;

[0043] Step 2.4: According to the following formula, translate the coordinates of the star in the imaging coordinate system of the space camera to obtain the coordinates of the star within the field of view of the space camera on the display:

[0044]

[0045] in, 、 Respectively i The horizontal and vertical coordinates of each star on the display;

[0046] Step 2.5: Define the point spread function, simulate the expansion effect based on the point spread function, convert the star magnitude value into grayscale distribution, and obtain the expanded star image:

[0047]

[0048] in, Indicates the i The star image after the expansion of the stars, 、 Respectively represent the horizontal and vertical coordinates of any point on the display; A 1 is the total energy of all stars, σ 1 is the energy concentration, A 1 and σ 1 are obtained based on the magnitude of the star;

[0049] Step 2.6: Synthesize the expanded star image to obtain the ideal background star map for the navigation star catalog;

[0050] Step 2.7: Repeat steps 2.2 to 2.6 until the ideal background star maps of all navigation star catalogs are obtained, thereby obtaining an ideal background star map dataset.

[0051] Furthermore, step 2.3 is specifically as follows:

[0052] Step 2.3.1. Based on the star point spatial distribution model, convert the right ascension and declination of the star in the celestial coordinate system into the unit direction vector of the star in the space camera coordinate system;

[0053] Step 2.3.2: Based on the unit direction vector of the star in the space camera coordinate system, use the following formula to select the stars that can be imaged in the space camera imaging coordinate system:

[0054]

[0055] in, , represents the spatial direction vector of the visual axis of the space camera in the space camera coordinate system, 、 、 They are the spatial direction vectors of the visual axis of the space camera in the space camera coordinate system. X axis, Y axis, Z Components on the axis; , indicating the i The unit direction vector of each star in the space camera coordinate system;

[0056] Step 2.3.3: Based on the perspective projection transformation model, the unit direction vector of the star that can be imaged in the space camera imaging coordinate system in the space camera coordinate system is converted into the coordinate in the space camera imaging coordinate system.

[0057] Furthermore, in step 2.5, when 0<star magnitude≤3, A 1=0.8, ;

[0058] 3<When the magnitude of the star is ≤6, A 1=0.5, .

[0059] Furthermore, in step 4, the simulated spatial stray light is simulated linearly distributed spatial stray light or simulated Gaussian distributed spatial stray light.

[0060] Furthermore, in step 4, the linearly distributed spatial stray light is simulated by the following formula:

[0061]

[0062] in, is a simulation diagram of linearly distributed spatial stray light, is the column vector of spatial stray light on the display, is the row vector of spatial stray light on the display; is the bias coefficient; is the gain coefficient of the linear distribution background, set k = (255 -30) / (img_width*a), where img_width is the size of the spatial stray light simulation image and a is the variation coefficient of the background grayscale value. By adjusting the variation coefficient a of the background grayscale value, the linear distribution of spatial stray light simulation images with different grayscale values is obtained, thereby obtaining a set of spatial stray light simulation images.

[0063] In step 4, the Gaussian-distributed spatial stray light is simulated by the following formula:

[0064]

[0065] in, is a simulation diagram of Gaussian distributed spatial stray light, A2 is the amplitude, σ 2 is the standard deviation of the Gaussian distribution; 、 They are the horizontal and vertical coordinates of the Gaussian function diffusion center on the display; by adjusting the diffusion center The value and amplitude of A 2. It can form Gaussian distribution spatial stray light simulation images at different illumination angles, thereby obtaining a set of spatial stray light simulation images.

[0066] Furthermore, in step 4, the spatial camera noise is simulated by the following formula:

[0067]

[0068] in, This is a noise simulation diagram of a space camera. It means that the generated object has a mean of 0 and a variance of The size of the normal distribution is M × N Simulated image of space camera noise; is the standard deviation of the space camera noise; M 、 N are the number of pixel rows and columns of the space camera respectively, and M = N .

[0069] Furthermore, in step 5, the specific method of the double loop nesting method is:

[0070] Step a1: superimposing a spatial stray light simulation image in the spatial stray light simulation image set with all the ideal background star images in the ideal background star image data set;

[0071] Step a2: Process the next spatial stray light simulation image according to the method of step a1 until all spatial stray light simulation images in the spatial stray light simulation image dataset are traversed.

[0072] Furthermore, in step 6, the specific method of embedding the brightness enhanced image in the brightness enhanced image set into the fused complex starry sky background image in the fused complex starry sky background image dataset is:

[0073] The fused complex starry sky background image in the fused complex starry sky background image dataset is read, the brightness enhanced image in the brightness enhanced image dataset is matched through modulo operation, and the complex starry sky background image and the matched brightness enhanced image are superimposed one by one.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1. The present invention provides a method for generating a high-fidelity space far-field time-sensitive target optical data set. The method is based on a three-stage framework of physical modeling and multi-layer synthesis. First, a space camera imaging model is established to convert stars from the celestial coordinate system to the space camera imaging coordinate system according to the space camera parameters and the given space camera's line of sight. The celestial coordinate system is mapped to the space camera imaging coordinate system through rotation transformation and projection transformation according to the navigation star catalog and the real time-sensitive target information, respectively, to obtain an ideal background star map and a simulated image of the time-sensitive target. Then, the simulated image of the time-sensitive target is subjected to ray path tracing to obtain an initial simulated image including the position and brightness characteristics of the time-sensitive target. At the same time, spatial stray light and space camera noise are simulated and fused with the ideal background star map to construct a fused complex starry sky background map. Finally, the initial simulated image is brightness enhanced and then fused with the fused complex starry sky background map to generate a high-fidelity space far-field time-sensitive target optical image. This method fully realizes the detection simulation imaging process of the space optical sensor for time-sensitive targets in a deep space environment, and can provide high-fidelity test data for subsequent target detection algorithms.

[0076] 2. The present invention provides a method for generating a high-fidelity far-field, time-sensitive target optical dataset. The space camera imaging model divides the star imaging chain in the space camera into the star's rotation transformation from the celestial coordinate system to the space camera coordinate system, and the perspective projection transformation from the space camera coordinate system to the space camera imaging coordinate system. This can simplify the modeling and calibration process, improve model accuracy, and enhance the model's flexibility and compatibility.

[0077] 3. This invention provides a method for generating a high-fidelity far-field time-sensitive target optical data set. The method defines a point spread function that can simulate the expansion effect caused by focal distortion, thereby adjusting the size and width of stars according to their magnitude to better extract the star's center of mass.

[0078] 4. The present invention provides a method for generating a high-fidelity spatial far-field time-sensitive target optical data set. By simulating space camera noise through Gaussian distribution, the method can simulate background noise, dark current noise, photon noise, etc. in actual shooting, conforming to the physical characteristics of space camera noise, simplifying modeling, and facilitating parameter estimation of space camera noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0080] Figure 2 Schematic diagram of the rotation transformation from the celestial coordinate system to the space camera coordinate system in step 1.2 of an embodiment of the present invention;

[0081] Figure 3 Schematic diagram of the spatial camera projection transformation in step 1.6 of an embodiment of the present invention;

[0082] Figure 4 The ideal background star map obtained in step 2.5 of the embodiment of the present invention;

[0083] Figure 5 The initial simulated image of the time-sensitive target obtained in step 3 of the embodiment of the present invention, wherein (a) is the initial simulated image of a single time-sensitive target, and (b) is the initial simulated image of a dual time-sensitive target;

[0084] Figure 6 The spatial stray light simulation diagram obtained in step 4 of the embodiment of the present invention, wherein (a) is a linearly distributed spatial stray light simulation diagram, and (b) is a Gaussian distributed spatial stray light simulation diagram;

[0085] Figure 7 The fused complex starry sky background image obtained in step 5 of the embodiment of the present invention, (a) is a fused complex starry sky background image with linear distribution, and (b) is a fused complex starry sky background image with Gaussian distribution;

[0086] Figure 8 These are the high-fidelity spatial far-field time-sensitive target optical images obtained in step 6 of an embodiment of the present invention, wherein (a) is a high-fidelity spatial far-field time-sensitive target optical image of a single time-sensitive target, and (b) is a high-fidelity spatial far-field time-sensitive target optical image of a dual time-sensitive target. DETAILED DESCRIPTION

[0087] The following is a detailed description of the method for generating a high-fidelity spatial far-field time-sensitive target optical dataset proposed by the present invention, in conjunction with the accompanying drawings and specific embodiments. It should be understood by those skilled in the art that these embodiments are merely intended to illustrate the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0088] A method for generating high-fidelity spatial far-field time-sensitive target optical data sets, such as Figure 1 As shown, the following steps are included:

[0089] Step 1: Set the space camera parameters. Then, based on the space camera parameters and the given space camera's visual axis, establish the rotation relationship between the celestial coordinate system and the space camera coordinate system, as well as the constraints for imaging stars within the space camera's field of view. Then, use the projection transformation method to obtain the space camera imaging model that transforms stars from the celestial coordinate system to the space camera imaging coordinate system. Specifically:

[0090] Step 1.1: Set the space camera parameters, including the field of view, focal length, number of pixels, magnitude sensitivity limit, and pixel size of the space camera. The specific parameters are shown in Table 1.

[0091] Table 1

[0092]

[0093] Step 1.2: Based on the visual axis of the space camera, obtain the rotation relationship between the celestial coordinate system and the space camera coordinate system as shown in the following formula:

[0094]

[0095]

[0096] in, The star is in the space camera coordinate system O ′- XYZ The direction vector in X 、 Y 、 Z The stars are in the space camera coordinate system O ′- XYZ The direction vector in X axis, Y axis, Z Components on the axis; is the star in the celestial coordinate system O - UVW The unit direction vector in , U 、 V 、 W The stars are in the celestial coordinate system O - UVW The unit direction vector in U axis, V axis, W Components on the axis; R is the rotation matrix; is the right ascension of the visual axis of the space camera in the celestial coordinate system, is the declination of the visual axis of the space camera in the celestial coordinate system.

[0097] Let the coordinates of the star in the celestial coordinate system be expressed in right ascension and declination, and denote them as , the direction vector in the space camera coordinate system is , the coordinates in the imaging plane coordinate system of the space camera are The starting point for right ascension is the vernal equinox, while the starting point for declination is the celestial equator. The northern range is 0° to 90°, while the southern range is 0° to -90°. In the celestial coordinate system, the direction vector of a star is derived as follows:

[0098] The space camera is fixed on the spacecraft. The distance between the center of the space camera and the center of the spacecraft is negligible compared to the distance between the star and the earth. Therefore, the center of the space camera and the center of the spacecraft are considered to coincide. Figure 2 As shown, O - UVW represents the celestial coordinate system, where OIt is the center of the celestial sphere. O ′- XYZ represents the space camera coordinate system, where O ′ is the center of the space camera. Since the distance between the star and the earth is far, the radius of the earth can be ignored, so the coordinate center of the space camera is O ′ and the center of the earth (i.e. the center of the celestial sphere O ) can be considered as approximately overlapping.

[0099] Celestial coordinate system O - UVW After three rotations, it becomes the space camera coordinate system O ′- XYZ , the rotation relationship between the two coordinate systems can be expressed as:

[0100]

[0101] Celestial coordinate system O - UVW , first around W Axis rotation , and obtain the coordinate system O - U 1 V 1 W 1, then around U 1-axis rotation ,make W 1 and Z The axes coincide and the coordinate system is obtained O - U 2 V 2 W 2. Finally, go around W 2-axis rotation , and obtain the coordinate system O - XYZ , then after three rotations the rotation matrix Expressed as:

[0102]

[0103] like Figure 2 As shown, in the celestial coordinate system, The axis points to the vernal equinox, The axis points to the celestial north pole, 、 and The three axes form a right-handed coordinate system. In the space camera coordinate system, The axis is the visual axis of the space camera, and the direction coordinate of the visual axis of the space camera in the celestial coordinate system is marked as , then the visual axis of the space camera is Axis angle , Axis Projection in the plane and The angle between the axes is .when Axis and If the axes do not coincide and are not in reverse direction, Axis and The axes together form a plane .because and Parallel and There is an intersection , then we can conclude plane. Because plane, and Flat, so flat , and the two planes have an intersection line .set up Axis and intersection line The angle is , The angle between the axis and .because axis, The axis is related to the installation position of the space camera. Different installation positions, The value of is different. The space camera is mounted on the spacecraft. If its installation position is determined, is a fixed value. For the convenience of calculation, According to the rotation relationship, we can conclude that: , , then the rotation matrix It can be expressed as:

[0104] .

[0105] In step 1.2, the celestial coordinate system is mapped to the space camera coordinate system using the third Euler rotation transformation.

[0106] Step 1.3: Based on the field of view of the space camera, obtain the constraint conditions for the star imaging within the field of view of the space camera as shown in the following formula:

[0107]

[0108] in, For the i The right ascension of a star in the celestial coordinate system, For the i The declination of a star in the celestial coordinate system, i is an integer, and 1≤ i ≤H, H is the number of stars; r is the field of view radius of the space camera, , represents the field of view of the space camera, Indicates that the space camera is in the space camera imaging coordinate system x Axis field of view, Indicates that the space camera is in the space camera imaging coordinate system y Axis field of view.

[0109] Step 1.4: Based on the right ascension and declination of the star in the celestial coordinate system, obtain the formula for calculating the unit direction vector of the star in the celestial coordinate system:

[0110]

[0111] in, For the i The unit direction vector of a star in the celestial coordinate system, 、 、 Respectively i The unit direction vector of a star in the celestial coordinate system is U axis, V axis, W Components on the axis.

[0112] Step 1.5: Based on the rotation relationship between the celestial coordinate system and the space camera coordinate system obtained in step 1.2, the constraints for imaging stars within the field of view of the space camera obtained in step 1.3, and the calculation formula for the unit direction vector of stars in the celestial coordinate system obtained in step 1.4, establish a star point spatial distribution model that converts stars from the celestial coordinate system to the space camera coordinate system:

[0113]

[0114]

[0115] in, For the i The unit direction vector of a star in the space camera coordinate system, 、 、 Respectively i The unit direction vector of a star in the space camera coordinate system is X axis, Y axis, Z Components on the axis.

[0116] Step 1.6, such as Figure 3 As shown in the figure, using projection transformation and based on the geometric principle of similar triangles, the projection relationship between the space camera coordinate system and the space camera imaging coordinate system is established, and the perspective projection transformation model is obtained:

[0117]

[0118]

[0119] in, 、 Respectively i The stars in the space camera imaging coordinate system x Axis coordinates, y axis coordinates; f is the focal length of the space camera, L is the number of pixels of the space camera, pixel_size is the pixel size.

[0120] Step 1.7: Substitute the star point spatial distribution model established in step 1.5 into the perspective projection transformation model obtained in step 1.6 to obtain the space camera imaging model:

[0121]

[0122]

[0123] in, 、 Represent the imaging coordinate system of the space camera x axis, y Axis pixel number, based on the number of pixels of the space camera L get.

[0124] In this embodiment, step 1 divides the imaging link of the star in the space camera into the rotation transformation of the star from the celestial coordinate system to the space camera coordinate system, and the perspective projection transformation from the space camera coordinate system to the space camera imaging coordinate system. This can simplify the modeling and calibration process, improve the model accuracy, and enhance the flexibility and compatibility of the model.

[0125] Step 2: Based on different navigation star catalogs, obtain the right ascension and declination of all stars in the celestial coordinate system. Then, based on the space camera imaging model, convert the right ascension and declination of the stars in the celestial coordinate system to the coordinates of the space camera imaging coordinate system to obtain the ideal background star map dataset. Specifically:

[0126] Step 2.1: Define the sky scanning grid as follows: the sky scanning range is 0.1° to 359.9° in right ascension and -89.9° to 89.9° in declination, with a sky scanning step size of 1°.

[0127] To reduce the amount of calculation, according to the grid where the visual axis of the space camera is located, in step 2.2, only the navigation star catalogs of the 8 adjacent grids near the grid are searched.

[0128] Step 2.2: Obtain the right ascension and declination of one of the stars in the navigation catalog in the celestial coordinate system based on the sky scanning grid.

[0129] Step 2.3: Based on the space camera imaging model, obtain the coordinates of the star in the space camera imaging coordinate system according to the right ascension and declination of the star in the celestial coordinate system. Specifically:

[0130] Step 2.3.1: Based on the star point spatial distribution model, convert the right ascension and declination of the star in the celestial coordinate system into the unit direction vector in the space camera coordinate system.

[0131] Step 2.3.2: Based on the unit direction vector of the star in the space camera coordinate system, use the following formula to select the stars that can be imaged in the space camera imaging coordinate system:

[0132]

[0133] in, , represents the spatial direction vector of the visual axis of the space camera in the space camera coordinate system, 、 、 They are the spatial direction vectors of the visual axis of the space camera in the space camera coordinate system. X axis, Y axis, Z Components on the axis; , indicating the i The unit direction vector of a star in the space camera coordinate system.

[0134] Step 2.3.3: Based on the perspective projection transformation model, the unit direction vector of the star that can be imaged in the space camera imaging coordinate system in the space camera coordinate system is converted into the coordinate in the space camera imaging coordinate system.

[0135] Step 2.4: According to the following formula, translate the coordinates of the star in the imaging coordinate system of the space camera to obtain the coordinates of the star within the field of view of the space camera on the display:

[0136]

[0137] in, 、 Respectively j The horizontal and vertical coordinates of the stars that can be imaged in the imaging coordinate system of the space camera on the display are: j is an integer, and 1≤ j ≤J, J is the number of stars that can be imaged in the imaging coordinate system of the space camera; 、 Respectively jThe stars that can be imaged in the space camera imaging coordinate system are x Axis coordinates, y Axis coordinates.

[0138] In this step, the coordinate origin of the space camera imaging coordinate system is located at the center of the imaging plane, while the coordinate origin of the display is in the upper left corner. If the ideal background star map is to be displayed on the display, the coordinates of the stars in the space camera imaging coordinate system need to be translated so that the final ideal background star map can be displayed normally on the display.

[0139] Step 2.5: Define the point spread function, simulate the expansion effect based on the point spread function, convert the star magnitude value into grayscale distribution, and obtain the expanded star image:

[0140]

[0141] in, Indicates the j An expanded star image of a star that can be imaged in the imaging coordinate system of a space camera, 、 Respectively represent the horizontal and vertical coordinates of any point on the display; A 1 is the total energy of all stars that can be imaged in the imaging coordinate system of the space camera, σ 1 is the energy concentration, A 1 and σ 1 are obtained based on the magnitude of the star:

[0142] 0<When the magnitude of the star is ≤3, A 1=0.8, ;

[0143] 3<When the magnitude of the star is ≤6, A 1=0.5, .

[0144] Because the grayscale distribution of star images follows the point spread function of the optical system, it can be approximated by a two-dimensional Gaussian distribution function. The main function of the point spread function is to adjust the size and width of the star according to its magnitude, so as to better extract the center of mass in subsequent verification. For each star, a point spread function is applied to simulate the expansion effect caused by focal distortion.

[0145] Step 2.6: Synthesize the expanded star image to obtain the ideal background star map for the navigation star catalog, such as Figure 4 shown.

[0146] Step 2.7: Repeat steps 2.2 to 2.6 until the ideal background star maps of all navigation star catalogs are obtained, thereby obtaining an ideal background star map dataset.

[0147] Step 3: Based on the real time-sensitive target information, the right ascension and declination of the time-sensitive target in the celestial coordinate system are obtained, and according to the method of step 2, a simulated image set of the time-sensitive target is constructed; then the light path of the time-sensitive target in the simulated image set is traced to obtain the spot intensity map of the time-sensitive target in the space camera imaging coordinate system, thereby obtaining an initial simulated image set including the position and brightness characteristics of the time-sensitive target, providing high-precision input data for subsequent fusion. The initial simulated image is as follows: Figure 5 As shown in the figure, (a) is the initial simulated image of a single time-sensitive target, and (b) is the initial simulated image of a dual time-sensitive target. The true time-sensitive target information includes the true time-sensitive target position, true time-sensitive target size, and true time-sensitive target brightness. The specific method of ray path tracing is as follows: Based on the principle of reversibility of light paths, a ray is emitted from the viewpoint of the space camera, and the reflection path of the ray at the intersection with the time-sensitive target in the simulated image set is recursively calculated to obtain the spot intensity map of the time-sensitive target in the space camera imaging coordinate system.

[0148] Step 4: simulate spatial stray light and spatial camera noise to obtain a set of spatial stray light simulation images and a spatial camera noise simulation image. Simulating spatial stray light is to simulate linearly distributed spatial stray light or Gaussian distributed spatial stray light, and the following can be obtained: Figure 6 The spatial stray light simulation diagram shown in the figure, where (a) is the linear distribution spatial stray light simulation diagram, and (b) is the Gaussian distribution spatial stray light simulation diagram. The linear distribution spatial stray light is simulated by the following formula:

[0149]

[0150] in, is a simulation diagram of linearly distributed spatial stray light, is the column vector of spatial stray light on the display, is the row vector of spatial stray light on the display; is the bias coefficient; is the gain coefficient of the linear distribution background, set k = (255 -30) / (img_width*a), img_width is the size of the spatial stray light simulation image, a is the variation coefficient of the background grayscale value. By adjusting the variation coefficient a of the background grayscale value, the linear distribution of spatial stray light simulation images with different grayscale values is obtained, thereby obtaining a set of spatial stray light simulation images.

[0151] The Gaussian distribution of spatial stray light is simulated by the following formula:

[0152]

[0153] in, is a simulation diagram of Gaussian distributed spatial stray light, A 2 is the amplitude, σ 2 is the standard deviation of the Gaussian distribution; 、 They are the horizontal and vertical coordinates of the Gaussian function diffusion center on the display. The value and amplitude of A 2. A Gaussian distribution of spatial stray light simulation images at different illumination angles can be formed, thereby obtaining a set of spatial stray light simulation images.

[0154] The spatial camera noise is simulated by the following formula:

[0155]

[0156] in, This is a noise simulation diagram of a space camera. It means that the generated object has a mean of 0 and a variance of The size of the normal distribution is M × N Simulated image of space camera noise; is the standard deviation of the space camera noise; M 、 N are the number of pixel rows and columns of the space camera respectively, and M = N .

[0157] Star image background noise comes from two main sources: one is the starry sky background noise, including stray light; the other is the noise introduced by the imaging device itself, which mainly includes shot noise, transfer noise, output noise, and dark current noise. As for the noise introduced by the imaging device itself, with the advancement of manufacturing process technology, transfer noise and output noise have been reduced to very low levels and can be ignored in simulations. Dark current and shot noise are mainly related to temperature and exposure time and can be approximately considered white noise. Therefore, in this embodiment, Gaussian-distributed random numbers are used to represent the total noise of the space camera.

[0158] Step 5: Using a double loop nesting method, the ideal background star map in the ideal background star map dataset is superimposed with each spatial stray light simulation image in the spatial stray light simulation image set to obtain a complex star background map dataset. Then, the complex star background images in the complex star background map dataset are superimposed with the spatial camera noise simulation images to obtain a fused complex star background map dataset. The fused complex star background map is as follows: Figure 7As shown in the figure, (a) is a linearly distributed fused complex starry sky background image, and (b) is a Gaussianly distributed fused complex starry sky background image. The double loop nesting method is specifically as follows: a spatial stray light simulation image in the spatial stray light simulation image dataset is superimposed with all the ideal background star images in the ideal background star image dataset, and then the next spatial stray light simulation image is processed according to the above method until all spatial stray light simulation images in the spatial stray light simulation image dataset are traversed.

[0159] In this embodiment, step 5 adopts a double loop nesting method, which can achieve full combination and superposition of multiple types of spatial stray light simulation images and ideal background star images, thereby realizing the synthesis of complex starry sky background.

[0160] Step 6: Perform brightness enhancement on the time-sensitive targets in all the initial simulation images in the initial simulation image set to obtain a brightness enhanced image set. Then, embed the brightness enhanced images in the brightness enhanced image set into the fused complex starry sky background image in the fused complex starry sky background image dataset to obtain a high-fidelity space far-field time-sensitive target optical dataset. The high-fidelity space far-field time-sensitive target optical image is as follows: Figure 8 As shown in the figure, (a) is a high-fidelity spatial far-field time-sensitive target optical image of a single time-sensitive target, and (b) is a high-fidelity spatial far-field time-sensitive target optical image of a dual time-sensitive target. The specific method of embedding the brightness enhanced image in the brightness enhanced image set into the fused complex starry sky background image in the fused complex starry sky background image dataset is as follows:

[0161] Read the fused complex starry sky background image in the fused complex starry sky background image dataset, match the brightness enhanced image in the brightness enhanced image dataset through modulo operation, and superimpose the complex starry sky background image and the matched brightness enhanced image in pairs. The matching method through modulo operation is:

[0162] If the number of fused complex starry sky background images in the fused complex starry sky background image dataset is less than or equal to the number of brightness-enhanced images in the brightness-enhanced image dataset, then the complex starry sky background images and the matching brightness-enhanced images are superimposed in pairs in order;

[0163] If the number of fused complex sky background images in the fused complex sky background image dataset exceeds the number of brightness-enhanced images in the brightness-enhanced image dataset, the matching index is "wrapped back" to the first brightness-enhanced image in the brightness-enhanced image dataset through a modulo operation. Specifically, if the number of brightness-enhanced images is P and the number of complex sky background images is Q (Q > P), then the brightness-enhanced image index that matches the pth background image is p mod Q. If P = 5, then the sixth fused complex sky background image matches the first brightness-enhanced image, the seventh fused complex sky background image matches the second brightness-enhanced image, and so on. Because the simulated target is time-sensitive and can appear in all starry sky backgrounds, cyclically reading complex sky background images allows for their reuse.

[0164] In this embodiment, step 6 increases target saliency through luminance gain, and after linear superposition, the pixel values are truncated to the [0, 1] dynamic range. To maintain quantization accuracy, the images in the high-fidelity spatial far-field time-sensitive target optical dataset are converted to 16-bit integers. The final high-fidelity image is output according to the original background file name rules, resulting in the low signal-to-noise ratio target detection scene required for target detection algorithm testing.

[0165] The present invention provides a method for generating a high-fidelity optical dataset for a far-field, time-sensitive target in space. First, based on the given camera's boresight and camera parameters, a space camera imaging model is constructed to transform stars from a celestial coordinate system to the camera's imaging coordinate system. The camera imaging model then performs coordinate transformation and magnitude-to-grayscale conversion on each star within the camera's field of view, based on the right ascension, declination, and magnitude in the navigation star catalog and actual time-sensitive target information. The resulting images are then displayed on a monitor as two-dimensional images, generating an ideal background star map and a simulated image of the time-sensitive target. During the imaging process, due to the influence of stray light in space and the camera itself, a large amount of noise will appear in the image, reducing the image quality. Therefore, factors such as stellar background noise and brightness non-uniformity are also simulated to achieve the generation of a high-fidelity optical image of a far-field, time-sensitive target in space.

Claims

1. A method for generating a high-fidelity spatial far-field time-sensitive target optical data set, characterized in that: The following steps are involved: Step 1: Set the space camera parameters. Then, based on the space camera parameters and the given space camera's visual axis, establish the rotation relationship between the celestial coordinate system and the space camera coordinate system, as well as the constraints for imaging stars within the space camera's field of view. Then, use the projection transformation method to obtain the space camera imaging model that transforms stars from the celestial coordinate system to the space camera imaging coordinate system. Step 2: Based on different navigation star catalogs, obtain the right ascension and declination of all stars in the catalog in the celestial coordinate system. Then, based on the space camera imaging model, convert the right ascension and declination of the stars in the celestial coordinate system into the coordinates of the space camera imaging coordinate system to obtain the ideal background star map dataset; specifically: Step 2.1, define the sky area scanning grid; Step 2.2, obtain the right ascension and declination of one of the navigation catalog stars in the celestial coordinate system according to the sky scanning grid; Step 2.3: Based on the space camera imaging model, obtain the coordinates of the star in the space camera imaging coordinate system according to the right ascension and declination of the star in the celestial coordinate system; Step 2.4: According to the following formula, translate the coordinates of the star in the imaging coordinate system of the space camera to obtain the coordinates of the star within the field of view of the space camera on the display: Among them, x i ′、y i ′ are the horizontal and vertical coordinates of the i-th star on the display; Step 2.5: Define the point spread function, simulate the expansion effect based on the point spread function, convert the star magnitude value into grayscale distribution, and obtain the expanded star image: Among them, μ i (x′, y′) represents the expanded star image of the i-th star, x′ and y′ represent the horizontal and vertical coordinates of any point on the display, respectively; A1 is the total energy of all stars, and σ1 is the energy concentration. Both A1 and σ1 are obtained based on the magnitude of the star; Step 2.6: Synthesize the expanded star image to obtain the ideal background star map for the navigation star catalog; Step 2.7: Repeat steps 2.2 to 2.6 until the ideal background star maps of all navigation star catalogs are obtained, thus obtaining the ideal background star map dataset. Step 3: Based on the real time-sensitive target information, the right ascension and declination of the time-sensitive target in the celestial coordinate system are obtained, and according to the method of step 2, a simulated image set of the time-sensitive target is obtained; then, the light path of the time-sensitive target in the simulated image set is traced to obtain a spot intensity map of the time-sensitive target in the imaging coordinate system of the space camera, thereby obtaining an initial simulated image set including the position and brightness characteristics of the time-sensitive target; Step 4: simulate spatial stray light and spatial camera noise to obtain a set of spatial stray light simulation images and a spatial camera noise simulation image; Step 5: Using a double loop nesting method, the ideal background star map in the ideal background star map dataset is superimposed with each spatial stray light simulation map in the spatial stray light simulation image set to obtain a complex star background map dataset; then all the complex star background maps in the complex star background map dataset are superimposed with the spatial camera noise simulation map to obtain a fused complex star background map dataset; Step 6. Brightness enhancement is performed on the time-sensitive targets in all the initial simulated images in the initial simulated image set to obtain a brightness-enhanced image set; then, the brightness-enhanced images in the brightness-enhanced image set are embedded into the fused complex starry sky background image in the fused complex starry sky background image dataset to obtain a high-fidelity spatial far-field time-sensitive target optical dataset.

2. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 1, characterized in that: Step 1 is as follows: Step 1.1, setting the space camera parameters, including the field of view of the space camera, the focal length of the space camera, the number of pixels of the space camera, and the pixel size; Step 1.2: Based on the visual axis of the space camera, obtain the rotation relationship between the celestial coordinate system and the space camera coordinate system as shown in the following formula: [X,Y,Z] T =R[U,V,W] T Where [X,Y,Z] T is the direction vector of the star in the space camera coordinate system O′-XYZ, where X, Y, and Z are the components of the direction vector of the star in the space camera coordinate system O′-XYZ on the X, Y, and Z axes respectively; [U, V, W] T is the unit direction vector of the star in the celestial coordinate system O-UVW, U, V, and W are the components of the unit direction vector of the star in the celestial coordinate system O-UVW on the U axis, V axis, and W axis respectively; R is the rotation matrix; α0 is the right ascension of the visual axis of the space camera in the celestial coordinate system, and δ0 is the declination of the visual axis of the space camera in the celestial coordinate system; Step 1.3: Based on the field of view of the space camera, obtain the constraint conditions for the star imaging within the field of view of the space camera as shown in the following formula: Among them, α i is the right ascension of the ith star in the celestial coordinate system, δ i is the declination of the i-th star in the celestial coordinate system, i is an integer, and 1≤i≤H, H is the number of stars; r is the field of view radius of the space camera, FOV represents the field of view of the space camera, FOV x Represents the field of view of the space camera in the x-axis direction in the space camera imaging coordinate system, FOV y Represents the field of view of the space camera in the y-axis direction in the space camera imaging coordinate system; Step 1.4: Based on the right ascension and declination of the star in the celestial coordinate system, obtain the formula for calculating the unit direction vector of the star in the celestial coordinate system: Among them, [U i ,V i ,W i ] T is the unit direction vector of the i-th star in the celestial coordinate system, U i 、V i 、W i are the components of the unit direction vector of the i-th star in the celestial coordinate system on the U axis, V axis, and W axis respectively; Step 1.5: Based on the rotation relationship between the celestial coordinate system and the space camera coordinate system obtained in step 1.2, the constraints for imaging stars within the field of view of the space camera obtained in step 1.3, and the calculation formula for the unit direction vector of stars in the celestial coordinate system obtained in step 1.4, establish a star point spatial distribution model that converts stars from the celestial coordinate system to the space camera coordinate system: Among them, [X i ,Y i ,Z i ] T is the unit direction vector of the i-th star in the space camera coordinate system, X i 、Y i 、Z i are the components of the unit direction vector of the i-th star in the space camera coordinate system on the X-axis, Y-axis, and Z-axis respectively; Step 1.6: Use projection transformation to establish the projection relationship between the space camera coordinate system and the space camera imaging coordinate system to obtain the perspective projection transformation model: Among them, x i 、y i are the x-axis coordinate and y-axis coordinate of the i-th star in the imaging coordinate system of the space camera; f is the focal length of the space camera, L is the number of pixels of the space camera, and pixel_size is the pixel size; Step 1.7: Substitute the star point spatial distribution model established in step 1.5 into the perspective projection transformation model obtained in step 1.6 to obtain the space camera imaging model: Among them, L x , L y They represent the number of pixels on the x-axis and y-axis of the imaging coordinate system of the space camera, respectively, and are obtained according to the number of pixels L of the space camera.

3. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 2, characterized in that: Step 2.3 is as follows: Step 2.3.

1. Based on the star point spatial distribution model, convert the right ascension and declination of the star in the celestial coordinate system into the unit direction vector of the star in the space camera coordinate system; Step 2.3.2: Based on the unit direction vector of the star in the space camera coordinate system, use the following formula to select the stars that can be imaged in the space camera imaging coordinate system: Where v0 = [X0, Y0, Z0] T , represents the spatial direction vector of the viewing axis of the space camera in the space camera coordinate system, X0, Y0, and Z0 are the components of the spatial direction vector of the viewing axis of the space camera in the space camera coordinate system on the X axis, Y axis, and Z axis respectively; v i =[X i ,Y i ,Z i ] T , represents the unit direction vector of the i-th star in the space camera coordinate system; Step 2.3.3: Based on the perspective projection transformation model, the unit direction vector of the star that can be imaged in the space camera imaging coordinate system in the space camera coordinate system is converted into the coordinate in the space camera imaging coordinate system.

4. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 3, characterized in that: In step 2.5, when 0 < star magnitude ≤ 3, A1 = 0.8, 3<When the magnitude of the star is ≤6, A1=0.5, 5. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 1, characterized in that: In step 4, the simulated spatial stray light is simulated linearly distributed spatial stray light or simulated Gaussian distributed spatial stray light.

6. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 5, characterized in that: In step 4, the linearly distributed spatial stray light is simulated by the following formula: I1=A×(k·B+h) Where I1 is the linearly distributed spatial stray light simulation image, A is the column vector of spatial stray light on the display, and B is the row vector of spatial stray light on the display; h is the bias coefficient; k is the gain coefficient of the linearly distributed background, set k = (255-30) / (img_width*a), img_width is the size of the spatial stray light simulation image, and a is the coefficient of change of the background grayscale value; In step 4, the Gaussian-distributed spatial stray light is simulated by the following formula: Among them, I2 is the spatial stray light simulation diagram of the Gaussian distribution, A2 is the amplitude, σ2 is the standard deviation of the Gaussian distribution; x0 and y0 are the horizontal and vertical coordinates of the diffusion center of the Gaussian function on the display, respectively.

7. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 5, characterized in that: In step 4, the spatial camera noise is simulated by the following formula: N(x′,y′)=normrand(0,σ3 2 ,M,N) Among them, N(x′,y′) is the noise simulation map of the space camera, normrand(0,σ3 2 ,M,N) means that the generated data has a mean of 0 and a variance of σ3 2 Simulated image of space camera noise with a normal distribution of size M×N; σ3 is the standard deviation of the space camera noise; M and N are the number of pixel rows and columns of the space camera, respectively, and M=N.

8. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 1, characterized in that: In step 5, the specific method of the double loop nesting method is: Step a1: superimposing a spatial stray light simulation image in the spatial stray light simulation image set with all the ideal background star images in the ideal background star image data set; Step a2: Process the next spatial stray light simulation image according to the method of step a1 until all spatial stray light simulation images in the spatial stray light simulation image dataset are traversed.

9. The method for generating a high-fidelity spatial far-field time-sensitive target optical data set according to claim 1, characterized in that: In step 6, the specific method of embedding the brightness enhanced image in the brightness enhanced image set into the fused complex starry sky background image in the fused complex starry sky background image dataset is: The fused complex starry sky background image in the fused complex starry sky background image dataset is read, the brightness enhanced image in the brightness enhanced image dataset is matched through modulo operation, and the complex starry sky background image and the matched brightness enhanced image are superimposed one by one.

Images