A method for selecting navigation stars based on the magnitude characteristics of star sensor cameras

Abstract

This invention discloses a navigation star selection method based on the magnitude features of a star sensor camera, comprising: S1: obtaining a continuous scene star sensor imaging star map; S2: extracting background noise from the star map in S1 and performing star map denoising; S3: binarizing the denoised star map; S4: solving for 8-connected components using the binarized star map to create set A; S5: sorting set A. This method first establishes the relationship between the imaging characteristics of the star sensor camera and the magnitudes of stars, then uses this relationship to select a parent star catalog for magnitude filtering, and finally distributes the star catalog evenly according to the star sensor camera characteristics to form the final navigation star catalog. Compared with traditional navigation star catalog construction methods, this invention increases the correlation with the star sensor camera, and can be applied to different sensors. The navigation star catalog constructed by this invention has high completeness, low redundancy, and good uniformity.

Description

Technical Field

[0001] This invention belongs to the field of navigation star catalog construction technology, and particularly relates to a navigation star selection method based on the star magnitude characteristics of a star sensor camera. Background Technology

[0002] Currently, most high-resolution remote sensing satellites choose star sensors as the primary control for attitude determination, mainly due to their reliable attitude determination results, high accuracy, and strong autonomy. As image products demand increasingly higher geometric positioning accuracy, hardware development is rapidly advancing towards multi-field-of-view and large-field-of-view star sensors, while CCD angular resolution is continuously improving. Software development focuses on enhancing the efficiency and accuracy of star image recognition algorithms, which is a major research direction for scholars both domestically and internationally.

[0003] Currently, the efficiency of star map recognition algorithms can be improved from two directions: one is to select a suitable matching strategy and find the best storage structure; the other is to build a navigation star catalog with low redundancy and high reliability.

[0004] This invention takes a second approach, seeking a strategy to minimize the number of navigation stars without losing key information. Chinese patent application number CN201210344509.9, entitled "A Method for Screening Navigation Stars for a Star Sensor," also discloses a method for screening navigation stars. This invention establishes a relationship between the light spot of the star sensor's CCD imaging and the magnitude of the stars; the two have a strong correlation. The advantages of this method are: ① It can more accurately screen navigation stars based on magnitude, avoiding the omission of detectable small stars. ② The relationship between the light spot and magnitude can be used in subsequent star map recognition, accelerating the efficiency of the star map recognition algorithm and improving its effectiveness. The cited document only utilizes the vague concept of "the limiting magnitude of the star sensor." It should be noted that, given current hardware capabilities, there is always a deviation between the design value and the theoretical value. Screening magnitudes based on the design value cannot achieve the accuracy level of this invention, nor can it support subsequent star map recognition algorithms.

[0005] The number of stars in a star catalog is closely related to their magnitude. Stars are classified according to their luminosity, with each magnitude differing by a factor of 2.512. Lower magnitudes indicate brighter stars. Furthermore, the number of stars in the catalog increases dramatically with magnitude. The empirical formula for the total number of stars across the entire celestial sphere as a function of magnitude is:

[0006] N = 6.57e 1.08Mv

[0007] In the formula, N is the total number of stars distributed across the entire celestial sphere; Mv is the magnitude.

[0008] The table below shows the number of stars corresponding to different magnitudes:

[0009] Table 1. Number of stars corresponding to magnitudes in the sky.

[0010]

[0011] Current mainstream methods for constructing navigation stars include self-organizing selection algorithms and orthogonal grid methods. Their main purpose is to build a uniformly distributed navigation star library. The idea is to project candidate navigation stars from a unit sphere onto a plane using an equal-area mapping method, divide the space into non-overlapping orthogonal grids with equal areas, and then select the star with the lowest magnitude from each small grid as the navigation star. These methods require magnitude screening before processing, but this screening process only considers given instrument parameters, i.e., determining the maximum detection performance of the star sensor and selecting an upper magnitude limit. Current research often leaves the selection of the lower magnitude limit vague, frequently using empirical values ​​such as 0 magnitude stars (Resource series satellites, Gaofen series satellites) or 2 magnitude stars (GLAS series satellites). Furthermore, the detection performance of star sensor cameras is mostly obtained through laboratory measurements, and the actual situation may be more complex. The selection of the upper magnitude limit should ideally be based on the actual working performance of the star sensor.

[0012] Regarding how to select the star magnitude threshold, the following research has been conducted:

[0013] A navigation star database was constructed using a magnitude-weighted method, increasing the weight coefficient of brighter stars. 4846 stars were selected for subsequent identification in order of magnitude from low to high. However, this method may result in the loss of a lot of stellar information. Most other studies use empirical thresholds for magnitude selection, which can lead to redundancy in the navigation star catalog. This is more pronounced with star cameras that have higher detection capabilities.

[0014] It should be noted that the above content falls within the inventor's technical knowledge and does not necessarily constitute prior art. Summary of the Invention

[0015] To address the aforementioned problems, the present invention aims to provide a navigation star selection method based on the magnitude characteristics of a star sensor camera. This method establishes a relationship between magnitude and star sensor camera, and the selected navigation star library features high completeness, low redundancy, and good uniformity.

[0016] To achieve the above objectives, this invention proposes a navigation star selection method based on the magnitude characteristics of a star sensor camera, comprising:

[0017] S1: Determine the size of the star sensor camera array and the field of view to obtain a continuous star sensor imaging star map.

[0018] S2: Extract the background noise from the star map in S1 and perform star map denoising.

[0019] S3: Binarize the denoised star map to obtain the binarized star map.

[0020] S4: Use the binarized star map to solve for the 8-connected connected components, extract the adaptive star window, and determine the size (length × width), centroid coordinates (x, y), and gray values ​​of each pixel (Gray(x1, y1), Gray(x2, y2)...Gray(xn, yn)) of each star window to form set A.

[0021] S5: Sort set A by sorting the set A by first sorting the star window size and then by summing the pixel grayscale values.

[0022] S6: A star map recognition algorithm that is not sensitive to star magnitude is used for star map recognition. Sample training is performed. All star maps adopt the all-sky autonomous recognition mode. If the recognition result of the same star image point is unique in different star maps, it is considered as a successful recognition. The star magnitude (Mag), centroid coordinates (x, y), and right ascension and declination (Ra, Dec) of the recognized stars constitute a set B.

[0023] S7: Perform the same operation as S6 on the multi-track star map to construct the set B of successful matches.

[0024] S8: After multiple training sessions, a one-to-one correspondence is established between set A and the identified set B to determine the star magnitude threshold.

[0025] S9: Download the Hipparcos star catalog as the parent catalog, filter the parent catalog according to the magnitude threshold, make the catalog sparse, and homogenize the catalog according to the star sensor camera parameters.

[0026] S10: Full-day traversal, determine the number of stars within the field of view of the star sensor camera. If the number is less than 3, increase the magnitude threshold and continue traversing until the number of stars within the field of view of all star sensor cameras is greater than 3.

[0027] The navigation star selection method based on star magnitude characteristics of a star sensor camera proposed in this invention can bring the following beneficial effects:

[0028] (1) By constructing the relationship between star image points and magnitude, the magnitude threshold can be accurately determined, which greatly reduces the number of navigation stars;

[0029] (2) By denoising and binarizing the star map, the coordinates of the star center and the adaptive star window information of the star can be extracted quickly and accurately. In the future satellite operation, the adaptive star window information can be transmitted by the star sensor to reduce the data transmission pressure.

[0030] (3) The navigation star catalog constructed based on this method already has a comparison between star image points and magnitudes. Therefore, in the subsequent star identification process, a high-matching magnitude can be used as a matching condition, which greatly speeds up the matching process. Attached Figure Description

[0031] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0032] Figure 1 This is a flowchart of a navigation star selection method based on the star magnitude characteristics of a star sensor camera;

[0033] Figure 2 This is a schematic diagram of star window extraction before noise reduction;

[0034] Figure 3 This is a diagram of a star window.

[0035] Figure 4 This is a diagram of star points.

[0036] Figure 5 This is a diagram of set A and set B. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below with reference to the embodiments and accompanying drawings.

[0038] like Figure 1 The image shows a navigation star selection method based on the magnitude characteristics of a star sensor camera, which includes the following steps:

[0039] S1: Determine the size of the star sensor camera array and the field of view to obtain a continuous star sensor imaging star map.

[0040] S2: Extract background noise from the star map and perform star map denoising. The specific process of star map denoising is as follows:

[0041] Dynamic noise template denoising is performed on the continuous star map of the downlink star sensor in S1. The basic formula is as follows:

[0042]

[0043] Where R0(i,j) is the minimum gray value of a star within a certain number of star maps from when it enters the star map to when it leaves the star map neighborhood, l is the weight of the static background noise, which is usually set to 5 or 6; R1(i,j) is the noise template value.

[0044] Observe the denoised star map and compare it with the original star map before denoising. Set the gray value of areas that are obviously noise but not zero to 0 (this usually occurs at the edge of the image).

[0045] S3: Binarize the denoised star map to obtain the binarized star map.

[0046] The threshold for binarization is calculated using the maximum inter-class variance method, which yields:

[0047]

[0048] Where thresh1 is the threshold obtained by using the Otsu's method.

[0049] S4: Use the binarized star map to solve for the 8-connected connected components, extract the adaptive star window, and determine the size (length × width), centroid coordinates (x, y), and gray values ​​of each pixel (Gray(x1, y1), Gray(x2, y2)...Gray(xn, yn)) of each star window to form set A.

[0050] Set A is represented as:

[0051] A = {Area1,Area2,…,Area} n}

[0052] Each sub-element in set A contains a coordinate range, the center coordinates of the image points within that range, and the grayscale value of each pixel; the center coordinates of the image points are obtained using a Gaussian two-dimensional surface fitting method, and its basic form is as follows:

[0053]

[0054] In the formula, G represents the amplitude of the Gaussian two-dimensional surface distribution, σ x , σ y denoted as x and y, respectively, are the standard deviations in the x and y directions.

[0055] S5: Sort set A by sorting the star window size first, then by the sum of the pixel gray values. That is, the larger the star window range, the earlier it appears. If the star window range is the same, the higher the sum of the pixel gray values, the earlier it appears.

[0056] S6: A star map recognition algorithm that is not sensitive to star magnitude is used for star map recognition. Sample training is performed. All star maps are in an all-sky autonomous recognition mode. Since the star map is a continuous frame, there are several identical image points in the continuous star map. If the recognition result of the same star image point is unique in different star maps, it is considered as a successful recognition. The star magnitude (Mag), centroid coordinates (x, y), and right ascension and declination (Ra, Dec) of the recognized stars constitute a set B.

[0057] The specific process of star map recognition is as follows: First, the star magnitude threshold is increased to basically include all stars in the star table except for binary stars, and a matching algorithm based on the primary star is adopted.

[0058] Track the same star in multiple images. If the star is identified with the same ID each time, it is considered to be successfully identified. This increases the success rate of star map identification and ensures the accuracy of the training set. Add the correctly identified navigation star to set B.

[0059] S6 specifically includes:

[0060] ① The star map recognition algorithm used must be insensitive to the magnitude of stars to ensure a high recognition success rate;

[0061] ② If a star image point is identified, the following conditions are met: The number of times the image point appears in subsequent star charts (n) and the number of times it is identified (m) are counted. The identification is considered successful if:

[0062]

[0063] S7: Perform the same operation as S6 on the multi-track star map to construct the set B of successful matches.

[0064] S8: After multiple training sessions, a one-to-one correspondence is established between set A and the identified set B. The magnitude threshold is determined. Set A and set B are correlated using the center coordinates of the image points. The specific process is as follows: Using the centroid of the star window as a connecting element, a connection is established between set A and set B, thereby obtaining the relationship between magnitude and star window. The minimum and maximum magnitude thresholds are statistically analyzed and set as the screening thresholds when constructing navigation stars.

[0065] S9: The constructed navigation star catalog is homogenized, the threshold required for homogenization is calculated, redundant information is further filtered out, the parent star catalog is filtered according to the magnitude threshold, the star catalog is made sparse, and the star catalog is homogenized according to the star sensor camera parameters.

[0066] When performing star catalog sparseness analysis, the following principles apply:

[0067] In S4, the maximum connected component range is N*N, the star sensor field of view is Fov°, and the number of horizontal and vertical pixels is Num respectively. The threshold thresh2 is then calculated using the following formula:

[0068]

[0069] When performing star catalog homogenization, each star needs to be traversed. If there are other stars within the thresh2 sky region, the brighter star is retained.

[0070] S10: Full-day traversal, determine the number of stars within the field of view of the star sensor camera. If the number is less than 3, increase the magnitude threshold and continue traversing until the number of stars within the field of view of all star sensor cameras is greater than 3.

[0071] The main performance indicators of the star sensor selected in this embodiment are:

[0072] Field of view: 8.9° × 8.9°

[0073] Array size: 2048×2048

[0074] In this embodiment, the Gaofen-7 dual-field-of-view star sensor is selected as the experimental object. The data is the star map transmitted by Gaofen-7 during its on-orbit testing. Taking 10 orbits of data as an example, the star windows of this example are extracted to construct a star window dataset and a star identification dataset.

[0075] (1) Stellar window dataset

[0076] Then, 10 star chart data were selected, and their star windows were extracted respectively. For example, the range of the star window determined based on the denoised star chart is illustrated. Figure 2 As shown, an example of a star window before denoising is given, such as... Figure 3 As shown;

[0077] Dataset elements (in) Figure 4 For example, if we want to store the star's centroid coordinates (39, 385) and star window size (8, 8), we need to store the grayscale values ​​of each pixel: (85, 86, 88, 89, 89, 84, 90, 81, 89, 93, 91, 91, 87, 88, 86, 89, 95, 113, 104, 96, 92, 99, 92, 90, 93, 104). 164, 165, 146, 113, 98, 92, 86, 94, 111, 175, 250, 162, 109, 94, 80, 86, 99, 115, 162, 150, 98, 87, 86, 84, 88, 95, 104, 97, 96, 82, 86, 87, 84, 87, 92, 89, 90, 84)

[0078] (2) Star recognition dataset

[0079] Star identification was performed using S7 and S8, and a star identification dataset was constructed. Some examples (where the magnitudes are magnified by 100 times) are shown in the table below:

[0080] Table 2. Star Identification Results

[0081]

[0082] (3) Training situation

[0083] The experimental hardware environment used in this embodiment was: 11th Gen Intel(R) Core(TM) i7-11800H @ 2.30GHz CPU; 32GB RAM. Computer system: Windows 11. Since the number of star image frames varies in each orbital star image data, with an average of approximately 60 star images per image, the training time was 6 minutes. However, this training can significantly improve the efficiency of star image recognition in subsequent processing, providing a permanent solution.

[0084] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and variations in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for selecting navigation stars based on the magnitude characteristics of a star sensor camera, characterized in that, include: S1: Determine the size of the star sensor camera array and the field of view to obtain a continuous star sensor imaging star map; S2: Extract the background noise from the star map in S1 and perform star map denoising; S3: Binarize the denoised star map to obtain the binarized star map; S4: Use the binarized star map to solve for the connected components, extract the adaptive star windows, and determine the size of each star window (length × width), centroid coordinates (x, y), and gray values ​​of each pixel (Gray(x1, y1), Gray(x2, y2) ... Gray(xn, yn)), forming set A; S5: Sort set A by sorting the star window size first, then by accumulating the pixel grayscale values; S6: A star map recognition algorithm that is not sensitive to star magnitude is used for star map recognition. Sample training is performed. All star maps adopt the all-sky autonomous recognition mode. If the recognition result of the same star image point is unique in different star maps, it is considered as a successful recognition. The star magnitude Mag, centroid coordinates (x, y), and right ascension and declination (Ra, Dec) of the recognized stars constitute a set B. S7: Perform the same operation as S6 on the multi-track star map to construct the set B of successful matches; S8: After multiple training sessions, a one-to-one correspondence is established between set A and the identified set B to determine the star magnitude threshold; S9: Download the Hipparcos star catalog as the parent catalog, filter the parent catalog according to the magnitude threshold, make the catalog sparse, and homogenize the catalog according to the star sensor camera parameters. S10: Full-day traversal, determine the number of stars within the field of view of the star sensor camera. If the number is less than 3, increase the magnitude threshold and continue traversing until the number of stars within the field of view of all star sensor cameras is greater than 3. The set A in S4 is represented as follows: ； Each sub-element in set A contains a coordinate range, the center coordinates of the image points within that range, and the grayscale value of each pixel. The center coordinates of the image points are fitted using a Gaussian two-dimensional surface method, and its basic form is as follows: ； In the formula, The amplitude representing the Gaussian two-dimensional surface distribution, , They are respectively , Standard deviation in direction; S6 specifically includes: ① The star map recognition algorithm used must be insensitive to the magnitude of stars to ensure a high recognition success rate; ② If a star image point is identified, the following conditions are met: The number of times the image point appears in subsequent star charts (n) and the number of times it is identified (m) are counted. The identification is considered successful if: 。 2. The navigation star selection method based on star magnitude characteristics of a star sensor camera according to claim 1, characterized in that, In step S2, the star map denoising process uses a dynamic noise template to denoise the star map. After denoising, it is necessary to check whether there are abnormal pixel values ​​that are obviously not star spots, and combine the original star map before star map denoising to set the gray value of areas that are obviously noise but not 0 to 0.

3. The navigation star selection method based on star magnitude characteristics of a star sensor camera according to claim 1, characterized in that, The binarization threshold calculation in S3 uses the maximum inter-class variance method, which yields: ； In the formula, The threshold is obtained by using the Otsu's method.

4. The navigation star selection method based on star magnitude characteristics of a star sensor camera according to claim 1, characterized in that, In S9, when performing star catalog sparsity calculations, the following principles apply: In S4, the maximum connected component range is N*N, the star sensor field of view is Fov°, and the number of horizontal and vertical pixels is Num respectively. Therefore, the threshold... Solve using the following formula: ； When performing star catalog homogenization, it is necessary to traverse each star. If in If there are other stars within the sky region, then the brighter star will be retained.

Images