A star pattern recognition method for multi-field star sensors

Abstract

The present invention discloses a star map recognition method for a multi-field star sensor, comprising the following steps: Step S1, constructing a navigation database, the navigation database comprising a navigation star library for storing basic information of stars, a library for storing star pairs with small angular distances within the field of view, and a library for storing angular distances between fields of view with large angular distances; Step S2, identifying primary stars in multiple fields of view; and Step S3, identifying stars observed in the full field of view. In the present invention, a "one small, two large" model is constructed, and by limiting the selection range of observation stars in the model, the number of models is reduced, thereby reducing the database capacity and improving the recognition speed. Based on attitude solution results, iterative identification of stars observed in the full field of view is carried out, and the accuracy of attitude solution is gradually improved through iteration, thereby improving the recognition rate of stars observed in the full field of view. Only angular distance information is used for recognition, and image fusion between fields of view is not required. The method is not limited by the number of fields of view or the angle between fields of view, and is more suitable for multi-field star sensors with different numbers of fields of view and boresight angles.

Description

Technical Field

[0001] The invention belongs to the technical field of celestial navigation, and in particular relates to a star map recognition method for a multi-field-of-view star sensor. Background Art

[0002] Currently, most spacecraft use star sensors as attitude measurement instruments to achieve high-precision, real-time attitude determination. This is due to their advantages, including arc-second accuracy, the ability to perform attitude measurement without prior information, and high reliability. Star sensors determine the spacecraft's attitude by measuring the positions of stars in an observed star map and comparing them with those in an ephemeris. The process involves four steps: star point imaging, centroid extraction, star map recognition, and attitude solution. Star map recognition is a crucial step in this workflow, as it identifies observed star points as those stored in the ephemeris. Its accuracy directly determines the accuracy of the attitude solution.

[0003] Under high-dynamic conditions, the number of observable star points in a single field of view decreases dramatically, and may even be unable to meet the recognition requirements alone. Multi-field-of-view configurations can lead to an exponential increase in observation information. Therefore, multi-field-of-view star sensors are an important technical approach to further improve the dynamic performance, measurement accuracy, and reliability of star sensors. Typical multi-field-of-view star sensors generally use three or more optical systems to form multiple fields of view. The field of view of a single field of view is small and has a high angular resolution, thus ensuring the accuracy of the star sensor. Because it has multiple fields of view, it can ensure that the star sensor has a sufficient number of observation stars in all fields of view. At the same time, due to the structural characteristics of the multi-field-of-view star sensor, it can improve the attitude measurement accuracy of the roll axis of the star sensor, improving the defect of large differences in the three-axis measurement accuracy of the single-field-of-view star sensor.

[0004] Currently, there are few publicly available star pattern recognition algorithms for multi-field star sensors in China. Jiang Di et al. fused observation images from multiple fields of view using a common set of observed stars as a benchmark. They then constructed characteristic patterns based on the path optimization principle of ant colony traversal to identify observed stars in the fused image. However, this method is only applicable when the observation images overlap and is not suitable for multi-field star sensors with large inter-boresight angles. Gou Bin et al. proposed a multi-stage star pattern recognition algorithm. This algorithm first uses a generalized regression neural network to identify the observation star pattern for each field of view. The recognition results are then verified and roughly calculated using the identification results. Finally, this inter-boresight direction is used to assist in the identification or correction of unidentified or incorrectly identified stars. However, since the individual fields of view of multi-field star sensors are typically small and contain fewer usable stars, this can lead to failure in identifying the star pattern for a single field of view, which in turn leads to algorithm failure and a reduced recognition rate. Xie Junfeng et al. used the sum of the grayscale values ​​of all stars in multiple fields of view and the sum of their angular distances from the primary star as evaluation metrics for star pattern recognition. However, due to the low reliability of star magnitude information, when the magnitude noise in the observed image is large, the recognition rate of the algorithm will decrease. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, the object of the present invention is to provide a star map recognition method for a multi-field star sensor.

[0006] In order to achieve the above-mentioned purpose and the above-mentioned technical effect, the technical solution adopted by the present invention is:

[0007] A star image recognition method for a multi-field star sensor comprises the following steps:

[0008] Step S1, constructing a navigation database, wherein the navigation database includes a navigation star library for storing basic information of stars, a star pair library for storing small angular distances within a field of view, and a large angular distance library for storing angular distances between fields of view;

[0009] Step S2, multi-field primary star identification;

[0010] Step S3: Identify stars observed in the entire field of view.

[0011] Furthermore, in step S1, the step of constructing a navigation database includes:

[0012] First, a navigation star library is constructed to store basic information about stars.

[0013] Secondly, a library for storing star pairs with small angular distances within the field of view is constructed;

[0014] Finally, a large angular distance library is constructed to store the angular distances between fields of view.

[0015] Furthermore, the steps of constructing a navigation star library for storing basic star information include:

[0016] The stars, binary stars and variable stars whose magnitudes are lower than the observation threshold of the star sensor in the basic star catalog are eliminated, and the numbers, vectors, magnitudes and right ascension and declination information of the remaining stars in the celestial coordinate system are stored to obtain the navigation star library.

[0017] Furthermore, the steps of constructing a library for storing star pairs with small angular distances within the field of view include:

[0018] Traverse and select each navigation star in the navigation star library, calculate the angular distance between it and other stars, select the star with the smallest angular distance to form a star pair, and store the index, number and angular distance between the two stars in the star pair in the star pair library;

[0019] Group the star pair library:

[0020] Assuming the angular distance between the star pairs is γ, the group they fall into is calculated as follows:

[0021]

[0022] Where λ group is the grouping interval, and int() is the rounding function.

[0023] Furthermore, the steps of constructing a large angular distance library for storing angular distances between fields of view include:

[0024] Traverse and select star pairs in the star pair library. Use the geometric center of the star pair as the boresight direction of each field of view of the multi-field star sensor. Select three or more stars that are in different fields of view from the star pair and closest to the geometric center of the star pair, preferably 10. The more stars, the higher the recognition rate, but the longer the recognition time. Calculate the angular distances α and β between the star and the star pair. Quantize the angular distances α and β between the star and the star pair into a two-dimensional grid and store them in a lookup table. The two-dimensional grid (i, j) where each star falls is calculated by the following formula:

[0025]

[0026] Where, angle boresight is the angle between the visual axes, FOV is the field of view, λ grid is the grid size, int() is the rounding function;

[0027] Then, the value of the two-dimensional grid f(i, j) is determined by whether there are stars in the grid:

[0028]

[0029] At this point, each star pair forms a unique two-dimensional grid and its values ​​based on the maximum angular separation information between its fields of view. To facilitate search and storage, the two-dimensional grid is converted into a one-dimensional grid from top to bottom and from left to right, and a lookup table is constructed to store the values ​​of the one-dimensional grid. The i-th row of the lookup table will store the index of all star pairs whose i-th grid value of the one-dimensional grid is 1.

[0030] Furthermore, in step S2, the step of multi-field primary star identification includes:

[0031] First, the brightest observed star in each field of view is selected as the main star to be identified;

[0032] Secondly, the observation stars in the same field of view as each primary star are selected as proximal neighbor stars, and the observation stars in different fields of view as each primary star are selected as distal neighbor stars; the two proximal neighbor stars closest and second closest to the primary star are selected to form two star pairs with the primary star respectively; three or more distal neighbor stars closest to the geometric center of the star pair are selected, preferably 10. The more distal neighbor stars, the higher the recognition rate, but the longer the recognition time, so as to construct multiple "one small and two large" characteristic patterns of the primary star;

[0033] Then, identify the two star pairs formed by each main star;

[0034] Finally, the main star identification results are verified.

[0035] Furthermore, the step of identifying the two star pairs formed by each primary star includes:

[0036] Calculate the angular separation between star pairs, i.e. the smaller angular separation in the "one small and two large" strategy, search the star pair database, and select star pairs in the corresponding group and adjacent groups as candidate star pairs;

[0037] Calculate the angular distance between the distant neighbor and the pair, i.e., the larger angular distance in the "one small, two large" scenario, and obtain the two-dimensional grids that each distant neighbor falls into. Find the rows in the lookup table corresponding to these two-dimensional grids. If the corresponding row in the lookup table records the index of a candidate pair, add 1 to the vote count of this candidate pair. After all distant neighbors have voted, select the candidate pair with the most votes and a number of votes greater than a certain threshold as the final matching result. The threshold is ≥3. A larger threshold results in higher recognition accuracy, but fewer matching stars.

[0038] If both star pairs of the primary star are successfully identified and one star pair receives more votes, the matching result of that star pair will be selected for the subsequent process; if the two star pairs receive the same number of votes and there is a common star in the matching results, the matching results of both star pairs will be used for the subsequent process; otherwise, both star pairs are considered to have failed to be identified, and the second brightest star in the field of view will be replaced as the primary star, and the primary star identification will be performed again, and so on.

[0039] Furthermore, the steps of verifying the primary star identification result include:

[0040] The star sensor attitude is calculated using the primary star identification result, and this attitude is used to identify the remaining observation stars in the same field of view as the primary star, excluding the identified stars. If any observation star is successfully identified, the primary star identification result is considered correct. Otherwise, the primary star identification result is considered incorrect and the field of view is changed.

[0041] Furthermore, in step S3, the step of identifying the full-field observation stars includes:

[0042] First, the main star identification results of each field of view are combined to roughly calculate the attitude of the star sensor;

[0043] Secondly, all observed stars in the full field of view are identified based on the attitude solution results;

[0044] Finally, based on the recognition results of the full-field observation stars, the star sensor attitude is solved again, and the full-field observation stars are iteratively identified until the recognition results of two adjacent iterative cycles are the same.

[0045] Furthermore, based on the attitude solution results, the steps of identifying all observed stars in the full field of view include:

[0046] The observation vector of each observation star in the star sensor coordinate system is calculated, and the rotation matrix from the star sensor to the celestial coordinate system is obtained using the star sensor attitude, and then the vector of each observation star in the celestial coordinate system is calculated; based on the vector of the observation star in the celestial coordinate system, the angular distance between each observation star and the navigation star in the star library is calculated, and the navigation star with the smallest angular distance from the observation star and less than a certain value is selected as the observation star identification result, completing the identification of the observation star with the full field of view.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. This invention constructs a "one small, two large" model. By limiting the selection range of observed stars in the model, the number of models is reduced, thereby reducing the database capacity and improving the recognition speed. The angular distance within the field of view and between fields of view are simultaneously used in the recognition process, fully integrating the information within the field of view and between fields of view to improve the recognition rate.

[0049] 2. The present invention iteratively identifies full-field-of-view observation stars based on attitude solution results, gradually improving the accuracy of attitude solution through iteration, thereby improving the recognition rate of full-field-of-view observation stars;

[0050] 3. The present invention uses only angular distance information for recognition, rather than less reliable magnitude information, which improves robustness. It also eliminates the need for image fusion between fields of view. Its use is not limited by the number of fields of view or the angle between them. It boasts high recognition speed, good recognition rate, and robustness, making it more suitable for multi-field star sensors with different numbers of fields of view and different boresight angles. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of the "one small, two large" model for main star identification;

[0052] Figure 2 A flowchart for multi-field primary star identification;

[0053] Figure 3 This is the flow chart for star identification in full-field observation. DETAILED DESCRIPTION

[0054] The present invention is described in detail below so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more precise definition of the protection scope of the present invention.

[0055] The following is a brief summary of one or more aspects to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceivable aspects and is neither intended to identify key or critical elements of all aspects nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that will be provided later.

[0056] like Figure 1-3 As shown, a star image recognition method for a multi-field star sensor includes the following steps:

[0057] Step S1: Building a navigation database

[0058] First, a navigation star library is constructed to store basic star information. Stars, binary stars, and variable stars with magnitudes below the star sensor's observation threshold are removed from the basic star catalog. The remaining stars' numbers, vectors in the celestial coordinate system, magnitudes, and right ascension and declination information are stored.

[0059] Next, a library is constructed to store star pairs with small angular separations within the field of view. Each navigation star in the library is selected, and the angular separations between it and other stars are calculated. The star with the smallest angular separation is selected to form a pair, or the star closest to it is selected to form a pair. The index, number, and angular separation between the two stars in the pair are stored in the library. To facilitate searching the library, the pairs are grouped. If the angular separation between the star pairs is γ, the group they fall into is calculated as follows:

[0060]

[0061] Where λ group is the grouping interval, and int() is the rounding function.

[0062] Finally, a large angular distance library is constructed to store angular distances between fields of view. Star pairs are traversed and selected from the library. The geometric center of each star pair is used as the boresight direction for each field of view of the multi-field star sensor. Three or more stars are selected that are in different fields of view from the star pair and closest to the geometric center of the star pair. A larger number of stars increases the recognition rate, but also increases the recognition time. Ten stars are preferred. The angular distances α and β between the star and the star pair are calculated. To facilitate the storage and search of large angular distances, the angular distances α and β between the star and the star pair are quantized into a two-dimensional grid and stored using a lookup table. The two-dimensional grid (i, j) where each star falls is calculated using the following formula:

[0063]

[0064] In the above formula, angle boresight is the angle between the visual axes, FOV is the field of view, λ grid is the grid size, and int() is the rounding function. Subsequently, the value of the two-dimensional grid f(i, j) is determined by whether there is a star in the grid:

[0065]

[0066] At this point, each star pair forms a unique two-dimensional grid and its values ​​based on the maximum angular separation between their fields of view. To facilitate search and storage, the two-dimensional grid is converted into a one-dimensional grid from top to bottom and left to right. A lookup table is then constructed to store the values ​​of the one-dimensional grid. The i-th row of the lookup table stores the index of all star pairs whose i-th grid cell has a value of 1.

[0067] Step S2: Multi-field primary star identification

[0068] First, the brightest observed star in each field of view is selected as the main star to be identified;

[0069] Secondly, select the observation stars in the same field of view as each main star as the proximal neighbor stars, and the observation stars in different fields of view as each main star as the distal neighbor stars. Select the two proximal neighbor stars that are closest and second closest to the main star to form two star pairs with the main star respectively. Select 3 or more stars closest to the geometric center of the star pair respectively. The more stars, the higher the recognition rate, but the longer the recognition time. 10 stars are preferred to construct multiple "one small, two large" feature patterns of the main star. Calculate the angular distance between the star pairs and the angular distance from the distal neighbor star to the two stars in the star pair, that is, multiple "one small, two large" angular distances for identification;

[0070] Then, the two star pairs formed by each primary star are identified. Based on the angular distance between the star pairs, that is, the smaller angular distance in the "one small two large" scenario, the star pair library is searched to determine the candidate star pairs. Based on the angular distance from the distant neighbor star to the star pair, that is, the larger angular distance in the "one small two large" scenario, the lookup table is searched to select the star pair with the largest number of matches to the distant neighbor star that is greater than a certain value as the final matching result.

[0071] Specifically, follow the steps below to identify the two star pairs separately:

[0072] 1. Calculate the angular separation between star pairs, i.e. the smaller angular separation in the "one small, two large" scenario, search the star pair database, and select the star pairs in the corresponding group and adjacent groups as candidate star pairs;

[0073] 2. Calculate the angular distance from the distant neighbor to the pair, i.e., the larger angular distance in the "one small, two large" scenario, to determine the two-dimensional grid cells that each distant neighbor falls into. Find the rows in the lookup table corresponding to these grid cells. If the corresponding row in the lookup table records the index of a candidate pair, add 1 to the vote count for that candidate pair. After all distant neighbors have voted, the candidate pair with the most votes, and a vote count greater than a certain threshold, is selected as the final match. The threshold is ≥3. A higher threshold results in higher recognition accuracy, but fewer matching stars.

[0074] If both pairs of the primary star are successfully identified and one pair receives more votes, the matching result of that pair is selected for subsequent processes. If the two pairs receive the same number of votes and there is a common star in the matching results, both pairs are used for subsequent processes. Otherwise, both pairs are considered to have failed identification, and the second brightest star in the field of view is replaced as the primary star, and primary star identification is repeated, and so on.

[0075] Finally, the primary star identification results are verified. The star sensor attitude is calculated using the primary star identification results, and this attitude is used to identify all other observed stars within the same field of view as the primary star, excluding the identified star. If any observed star is successfully identified, the primary star identification result is considered correct; otherwise, the primary star identification result is considered incorrect, and the field of view is changed.

[0076] Step S3: Identify full-field observation stars

[0077] First, the star sensor attitude is roughly calculated by combining the main star identification results of each field of view, and the QUEST algorithm is preferably used to solve the star sensor attitude.

[0078] Secondly, based on the attitude solution results, all observed stars in the full field of view are identified. The main steps are:

[0079] Calculate the observation vector of each observation star in the star sensor coordinate system, use the star sensor attitude to obtain the rotation matrix from the star sensor to the celestial coordinate system, and then calculate the vector of each observation star in the celestial coordinate system; based on the vector of the observation star in the celestial coordinate system, calculate the angular distance between each observation star and the navigation star in the star library, and select the navigation star with the smallest angular distance from the observation star and less than a certain value as the observation star identification result, completing the identification of the observation star in the full field of view;

[0080] Finally, based on the recognition results of the full-field observation stars, the star sensor attitude is solved again, and the full-field observation stars are iteratively identified until the recognition results of two adjacent iterative cycles are the same.

[0081] Example 1

[0082] like Figure 1-3 As shown, a star image recognition method for a multi-field star sensor includes the following steps:

[0083] Step S1, constructing a navigation database, wherein the navigation database includes a navigation star library for storing basic information of stars, a star pair library for storing small angular distances within a field of view, and a large angular distance library for storing angular distances between fields of view;

[0084] In step S1, the steps of constructing a navigation database include:

[0085] First, build a navigation star library to store basic star information, including the following steps:

[0086] Eliminate stars, binary stars, and variable stars whose magnitudes are lower than the observation threshold of the star sensor in the basic star catalog, and store the numbers, vectors, magnitudes, and right ascension and declination information of the remaining stars in the celestial coordinate system to obtain a navigation star library;

[0087] Secondly, a library for storing small angular distance star pairs within the field of view is constructed, which includes the following steps:

[0088] Traverse and select each navigation star in the navigation star library, calculate the angular distance between it and other stars, select the star with the smallest angular distance to form a star pair, and store the index, number and angular distance between the two stars in the star pair in the star pair library;

[0089] Group the star pair library:

[0090] Assuming the angular distance between the star pairs is γ, the group they fall into is calculated as follows:

[0091]

[0092] Where λ group is the grouping interval, int() is the rounding function;

[0093] Finally, a large angular distance library for storing the angular distances between fields of view is constructed, which includes the following steps:

[0094] The star pairs in the star pair library are traversed and selected. The geometric center of the star pair is used as the boresight direction of each field of view of the multi-field star sensor. The 10 stars that are in different fields of view from the star pair and closest to the geometric center of the star pair are selected. The angular distances α and β between the star and the star pair are calculated. The angular distances α and β between the star and the star pair are quantized into a two-dimensional grid and stored in a lookup table. The two-dimensional grid (i, j) where each star falls is calculated by the following formula:

[0095]

[0096] Where, angle boresight is the angle between the visual axes, FOV is the field of view, λ grid is the grid size, int() is the rounding function;

[0097] Then, the value of the two-dimensional grid f(i, j) is determined by whether there are stars in the grid:

[0098]

[0099] At this point, each star pair forms a unique two-dimensional grid and its values ​​based on the maximum angular separation information between its fields of view. To facilitate search and storage, the two-dimensional grid is converted into a one-dimensional grid from top to bottom and from left to right, and a lookup table is constructed to store the values ​​of the one-dimensional grid. The i-th row of the lookup table will store the index of all star pairs whose i-th grid value of the one-dimensional grid is 1.

[0100] Step S2, multi-field primary star identification, includes the following steps:

[0101] First, the brightest observed star in each field of view is selected as the main star to be identified;

[0102] Secondly, we selected the observation stars in the same field of view as each primary star as the proximal neighbor stars, and the observation stars in different fields of view as the distal neighbor stars. We selected the two proximal neighbor stars closest to the primary star and the second closest to the primary star to form two star pairs with the primary star respectively. We also selected the 10 distal neighbor stars closest to the geometric center of each star pair to construct multiple "one small, two large" characteristic patterns of the primary star.

[0103] Then, the two star pairs formed by each primary star are identified separately, including the following steps:

[0104] Calculate the angular separation between star pairs, i.e. the smaller angular separation in the "one small and two large" strategy, search the star pair database, and select star pairs in the corresponding group and adjacent groups as candidate star pairs;

[0105] Calculate the angular distance between the distant neighbor and the pair, i.e., the larger angular distance in the "one small, two large" match, and obtain the two-dimensional grids that each distant neighbor falls into. Find the rows in the lookup table corresponding to these two-dimensional grids. If the corresponding row in the lookup table records the index of a candidate pair, add 1 to the vote count of this candidate pair. After all distant neighbors have voted, select the candidate pair with the most votes (greater than 3 votes) as the final match result.

[0106] If both pairs of the primary star are successfully identified and one pair receives more votes, the matching result of that pair is selected for subsequent processes. If the two pairs receive the same number of votes and there is a common star in the matching results, both pairs are used for subsequent processes. Otherwise, both pairs are considered to have failed identification, and the second brightest star in the field of view is replaced as the primary star, and primary star identification is repeated, and so on.

[0107] Finally, the primary star identification result verification is carried out, which includes the following steps:

[0108] The star sensor attitude is calculated using the primary star identification result, and this attitude is used to identify the remaining observation stars in the same field of view as the primary star, excluding the identified stars. If any observation star is successfully identified, the primary star identification result is considered correct. Otherwise, the primary star identification result is considered incorrect and the field of view is changed.

[0109] Step S3, full-field observation star identification, includes the following steps:

[0110] First, the main star identification results of each field of view are combined to roughly calculate the attitude of the star sensor;

[0111] Secondly, based on the attitude solution results, all observed stars in the full field of view are identified, including the following steps:

[0112] Calculate the observation vector of each observation star in the star sensor coordinate system, use the star sensor attitude to obtain the rotation matrix from the star sensor to the celestial coordinate system, and then calculate the vector of each observation star in the celestial coordinate system; based on the vector of the observation star in the celestial coordinate system, calculate the angular distance between each observation star and the navigation star in the star library, and select the navigation star with the smallest angular distance from the observation star and less than a certain value as the observation star identification result, completing the identification of the observation star in the full field of view;

[0113] Finally, based on the recognition results of the full-field observation stars, the star sensor attitude is solved again, and the full-field observation stars are iteratively identified until the recognition results of two adjacent iterative cycles are the same.

[0114] Parts or structures not specifically described in the present invention may adopt existing technologies or existing products and will not be described in detail here.

[0115] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A star map recognition method for a multi-field star sensor, characterized in that: The following steps are involved: Step S1, constructing a navigation database, wherein the navigation database includes a navigation star library for storing basic information of stars, a star pair library for storing small angular distances within a field of view, and a large angular distance library for storing angular distances between fields of view; Step S2, multi-field primary star identification; Step S3, identifying full-field observation stars; In step S1, the steps of constructing a navigation database include: First, a navigation star library is constructed to store basic information about stars. Secondly, a library for storing star pairs with small angular distances within the field of view is constructed; Finally, a large angular distance library is constructed to store the angular distances between fields of view; In step S2, the steps of multi-field primary star identification include: First, the brightest observed star in each field of view is selected as the main star to be identified; Secondly, we selected the observation stars in the same field of view as each primary star as the proximal neighbor stars, and the observation stars in different fields of view as the distal neighbor stars. We selected the two proximal neighbor stars closest and second closest to the primary star to form two pairs with the primary star respectively. We also selected three or more distal neighbor stars closest to the geometric center of each pair to construct multiple "one small, two large" characteristic patterns of the primary star. Then, identify the two star pairs formed by each main star; Finally, the main star identification results are verified; The steps for identifying the two pairs of stars formed by each primary star include: Calculate the angular separation between star pairs, i.e. the smaller angular separation in the "one small and two large" scenario, search the star pair database, and select star pairs in the corresponding group and adjacent groups as candidate star pairs; Calculate the angular distance between the distant neighbor and the pair, i.e., the larger angular distance in the "one small, two large" scenario, and obtain the two-dimensional grids that each distant neighbor falls into. Find the rows in the lookup table corresponding to these two-dimensional grids. If the corresponding row in the lookup table records the index of a candidate pair, add 1 to the vote count for this candidate pair. After all distant neighbors have voted, select the candidate pair with the most votes, and a vote count greater than a certain threshold, as the final match result. If both pairs of the primary star are successfully identified and one pair receives more votes, the matching result of that pair is selected for subsequent processes. If the two pairs receive the same number of votes and there is a common star in the matching results, both pairs are used for subsequent processes. Otherwise, both pairs are considered to have failed identification, and the second brightest star in the field of view is replaced as the primary star, and primary star identification is repeated, and so on. The steps for verifying the primary star identification results include: The star sensor attitude is calculated using the primary star identification result, and this attitude is used to identify the remaining observation stars in the same field of view as the primary star, excluding the identified stars. If any observation star is successfully identified, the primary star identification result is considered correct. Otherwise, the primary star identification result is considered incorrect and the field of view is changed.

2. The star image recognition method of a multi-field star sensor according to claim 1, characterized in that: The steps to build a navigation star library for storing basic star information include: The stars, binary stars and variable stars whose magnitudes are lower than the observation threshold of the star sensor in the basic star catalog are eliminated, and the numbers, vectors, magnitudes and right ascension and declination information of the remaining stars in the celestial coordinate system are stored to obtain the navigation star library.

3. The star map recognition method for a multi-field star sensor according to claim 1, characterized in that: The steps to build a library for storing small angular distance star pairs in the field of view include: Traverse and select each navigation star in the navigation star library, calculate the angular distance between it and other stars, select the star with the smallest angular distance to form a star pair, and store the index, number and angular distance between the two stars in the star pair in the star pair library; Group the star pair library: Assuming the angular distance between the star pairs is γ, the group they fall into is calculated as follows: Where λ group is the grouping interval, and int() is the rounding function.

4. The star map recognition method for a multi-field star sensor according to claim 1, characterized in that: The steps for building a large angular distance library for storing the angular distance between fields of view include: The star pairs in the star pair library are traversed and selected. The geometric center of the star pair is used as the boresight direction of each field of view of the multi-field star sensor. Three or more stars that are in different fields of view from the star pair and closest to the geometric center of the star pair are selected. The angular distances α and β between the star and the star pair are calculated. The angular distances α and β between the star and the star pair are quantized into a two-dimensional grid and stored in a lookup table. The two-dimensional grid (i, j) where each star falls is calculated by the following formula: Where, angle boresight is the angle between the visual axes, FOV is the field of view, λ grid is the grid size, int() is the rounding function; Then, the value of the two-dimensional grid f(i, j) is determined by whether there are stars in the grid: At this point, each star pair forms a unique two-dimensional grid and its values ​​based on the maximum angular separation information between its fields of view. To facilitate search and storage, the two-dimensional grid is converted into a one-dimensional grid from top to bottom and from left to right, and a lookup table is constructed to store the values ​​of the one-dimensional grid. The i-th row of the lookup table will store the index of all star pairs whose i-th grid value of the one-dimensional grid is 1.

5. The star image recognition method of a multi-field star sensor according to claim 1, characterized in that: In step S3, the steps of identifying full-field observation stars include: First, the main star identification results of each field of view are combined to roughly calculate the attitude of the star sensor; Secondly, all observed stars in the full field of view are identified based on the attitude solution results; Finally, based on the recognition results of the full-field observation stars, the star sensor attitude is solved again, and the full-field observation stars are iteratively identified until the recognition results of two adjacent iterative cycles are the same.

6. The star image recognition method for a multi-field star sensor according to claim 5, characterized in that: Based on the attitude solution results, the steps to identify all observed stars in the full field of view include: The observation vector of each observation star in the star sensor coordinate system is calculated, and the rotation matrix from the star sensor to the celestial coordinate system is obtained using the star sensor attitude, and then the vector of each observation star in the celestial coordinate system is calculated; based on the vector of the observation star in the celestial coordinate system, the angular distance between each observation star and the navigation star in the star library is calculated, and the navigation star with the smallest angular distance from the observation star and less than a certain value is selected as the observation star identification result, completing the identification of the observation star with the full field of view.

Images