A weighted neighborhood consistency star catalog matching method for high dynamic star map reliability
By fusing the quality information of detected stars and the support of neighborhood structures, the accuracy and anti-interference issues of star catalog matching under high dynamic conditions are solved, and reliable weighted neighborhood consistency star catalog matching of high dynamic star maps is achieved.
Patent Information
- Application Number
- CN202611585932.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-14
- Publication Date
- 2026-11-10
AI Technical Summary
Under high dynamic conditions, the angular motion of spacecraft causes star point tailing, weak star response attenuation, and centroid position shift. Existing neighborhood structure matching methods do not fully consider the reliability differences of detected stars, resulting in false star points and repetitive responses. Furthermore, the lack of one-to-one constraints leads to insufficient accuracy and anti-interference capability in star table matching.
By acquiring the quality information of stars detected in the star map, calculating the fusion reliability, and combining attitude priors and neighborhood structure support, a weighted fusion method is used to screen candidate matching pairs and apply one-to-one constraints to improve the accuracy and anti-interference ability of the matching results.
The impact of false detection stars, missed detection stars, and star tails on matching results is reduced, improving the accuracy and anti-interference ability of star catalog matching and ensuring the accuracy of one-to-one allocation.
Smart Images

Figure CN123065485A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of star sensor image processing and spacecraft autonomous navigation technology, and specifically relates to a high dynamic star map reliability weighted neighborhood consistency star catalog matching method. Background Technology
[0002] With the continuous development of autonomous navigation technology for spacecraft, star sensors are widely used for spacecraft attitude determination due to their high measurement accuracy and the fact that errors do not accumulate over time. Star sensors typically detect star points and extract their positions from captured star images, and establish a correspondence between the detected stars and stars in the navigation star catalog, thereby obtaining star identifiers for subsequent attitude calculations.
[0003] In star map recognition with attitude priors, stars within the current field of view are typically projected onto the imaging plane based on attitude priors, camera imaging parameters, and a navigation star catalog. A fixed distance threshold is then used to filter candidate matches between detected stars and projected stars, and the matching result is determined based on candidate distances and neighborhood geometric relationships. Some existing methods construct a graph structure from the stars, using distances, directions, and relative displacement relationships between stars for graph matching; others employ data-driven models such as graph neural networks or graph attention networks to learn star structure features to complete star classification or candidate matching.
[0004] However, under high dynamic conditions, spacecraft angular motion can easily cause star trails, weak star response attenuation, and centroid position shifts. Image noise, hot pixels, and tail breaks can also produce false stars or repetitive responses, resulting in significant differences in the authenticity and positioning reliability of different detected stars. The star detection process can usually output detection confidence and thermal map response information, but existing distance matching methods mainly rely on the absolute position of individual stars, failing to fully utilize the aforementioned quality information and making them susceptible to attitude prior errors and star positioning errors.
[0005] While existing neighborhood structure matching methods can utilize the geometric relationships between multiple star points, they typically do not adequately consider the reliability differences among different detection stars. This leads to low-reliability spurious star points, weak star responses, or tailed repetitive responses potentially participating in structure comparisons with high-reliability star points on an equal footing. Furthermore, multiple candidate matching correspondences may share the same detection star or the same projected star; without a unified one-to-one constraint, the same star point can be repeatedly occupied, and erroneous correspondences can be retained. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a high dynamic star map reliability weighted neighborhood consistency star catalog matching method. This method introduces the detection of star fusion reliability into the comparison of multi-star neighborhood structures, and combines the absolute distance information provided by attitude priors, and applies one-to-one constraints to the matching results, which can improve the accuracy of star catalog matching under complex imaging conditions.
[0007] The technical solution provided by this invention is as follows:
[0008] A high-dynamic star map reliability weighted neighborhood consistency star catalog matching method includes:
[0009] Obtain the detection star quality information of the current frame of the star map, calculate the fusion reliability of each detection star based on the detection star quality information, and project the navigation stars in the current field of view onto the imaging plane to form projection stars, thereby obtaining a local projection star set;
[0010] Based on the absolute image plane distance between the detected star and the projected star, candidate matching pairs between the detected star and the projected star are established within the first search radius;
[0011] The detection star neighborhood and the projection star neighborhood of the candidate matching pair are determined respectively. Based on the relative displacement of the neighboring detection stars in the detection star neighborhood and the relative displacement of the neighboring projection stars in the projection star neighborhood, the neighborhood structure support of the candidate matching pair is determined.
[0012] The neighborhood structure support of the candidate matching pair, the fusion reliability of the detected stars in the candidate matching pair, and the absolute distance similarity of the candidate matching pair are weighted and fused to obtain the joint score of the candidate matching pair.
[0013] Candidate matching pairs with high joint scores are selected based on the constraints and used as the final matching results;
[0014] The constraint is that neither the detection star nor the projection star is used repeatedly.
[0015] Preferably, the detection satellite quality information includes: the detection confidence of the detection satellite, the thermal map response of the detection satellite, and the graph context reliability of the detection satellite.
[0016] Preferably, the formula for calculating the fusion reliability of the detection stars is:
[0017] ;
[0018] in, Indicates detection star The reliability of fusion For detecting stars The normalized value of the detection confidence level. For detecting stars The normalized value of the heatmap response, For detecting stars The normalized value of the graph context reliability, , , These are the fusion weights for detection confidence, heatmap response, and graph context reliability, respectively. ; This is the amplitude limiting function.
[0019] Preferably, the method for establishing candidate matching pairs is as follows:
[0020] Calculate the detection star With projected stars absolute image plane distance :
[0021] ;
[0022] If satisfied , will detect stars With projected stars The combination of these is retained as a candidate matching pair;
[0023] in, For detecting stars Two-dimensional image plane coordinates, For projection stars Two-dimensional image plane coordinates, This represents the first search radius.
[0024] Preferably, the method for determining the neighborhood structure support of candidate matching pairs includes:
[0025] Computing neighborhood detection stars For candidate matching pairs Optimal Vector Consistency :
[0026] ;
[0027] In the formula, This represents the number of projected stars in the local projected star set. The relative displacement consistency kernel scale. Detection stars for candidate matching pairs Pointing to its neighboring detection stars The relative displacement, Projected stars for candidate matching pairs Pointing to its neighboring projection star The relative displacement;
[0028] Candidate matching pairs are calculated based on the optimal vector consistency. neighborhood structure support :
[0029] ;
[0030] In the formula, Weighting the structure, The detection neighborhood of the candidate matching pair.
[0031] Preferably, the structural weighting is calculated using the following formula:
[0032] ;
[0033] in, The neighborhood basic weights, and satisfying , For neighboring detection stars The reliability of the fusion.
[0034] Preferably, candidate matching pairs absolute distance similarity Calculated using the following formula:
[0035] ;
[0036] in, The absolute distance kernel scale, For detecting stars With projected stars The absolute image-plane distance.
[0037] Preferably, candidate matching pairs Combined score The calculation formula is:
[0038] ;
[0039] In the formula, , and These are the fusion weights for neighborhood structure support, detection star fusion reliability, and distance similarity, respectively, and they satisfy: .
[0040] Preferably, the high dynamic star map reliability weighted neighborhood consistency star catalog matching method further includes:
[0041] Candidate matching pairs that simultaneously meet the following conditions Add to the set of candidate matching pairs to be assigned:
[0042] , ;
[0043] in, The second search radius, ; This is the threshold for the candidate joint score;
[0044] Based on the constraints, candidate matching pairs with high joint scores are selected from the set of candidate matching pairs to be assigned, and these are used as the final matching results.
[0045] Preferably, the method for selecting candidate matching pairs from the set of candidate matching pairs to be assigned is as follows:
[0046] The candidate matching pairs in the set of candidate matching pairs to be assigned are sorted from high to low according to the joint candidate score to form a candidate matching pair sequence;
[0047] Candidate matching pairs are read sequentially according to the sorting order and used as the matching results;
[0048] If the detection star of the currently read candidate matching pair has been occupied by a previous matching result, the currently read candidate matching pair is discarded; until all candidate matching pairs in the candidate matching pair sequence have been read, the final matching result is obtained.
[0049] The beneficial effects of this invention are:
[0050] The high-dynamic star map reliability weighted neighborhood consistency star catalog matching method provided by this invention introduces the fusion reliability of detection stars into the verification of multi-star neighborhood structure, which can reduce the impact of false detection stars, missed detection stars, weak star responses and star point tails on the matching results. At the same time, by combining neighborhood relative displacement, detection star fusion reliability and absolute distance information, it improves the discrimination ability of candidate matching pairs, and avoids the repeated occupation of detection stars or projection stars by one-to-one allocation, thereby improving the accuracy and anti-interference ability of star catalog matching under complex imaging conditions. Attached Figure Description
[0051] Figure 1 This is a flowchart of the overall process for the reliability-weighted neighborhood structure consistency star table matching method described in this invention.
[0052] Figure 2 This is a schematic diagram illustrating the calculation process for detecting the reliability of star fusion as described in this invention.
[0053] Figure 3 This is a graph showing the relative displacement consistency and neighborhood structure support calculation between the detected star neighborhood and the projected star neighborhood as described in this invention.
[0054] Figure 4 This is a flowchart illustrating the one-to-one allocation of detection stars and projection stars according to the present invention.
[0055] Figure 5 A comparison chart showing the matching performance of narrow radius distance matching baselines and the method of this invention under different interference conditions; wherein, Figure 5In Figure (a), the curve of the matching F1 value variation under position noise conditions is shown. Figure 5 (b) shows the curve of the matching F1 value under the condition of spurious detector interference. Figure 5 (c) shows the curve of the matching F1 value change under the condition of missed satellite interference. Figure 5 In the middle (d), the curve of the matching F1 value change under the star tail condition is shown.
[0056] Figure 6 This is a diagram showing the recognition results of the method of the present invention on a simulated star map generated based on a navigation star catalog; wherein, Figure 6 (a) is the original simulated star map. Figure 6 (b) shows the identification result of the method of the present invention.
[0057] Figure 7 A comparison diagram of the recognition results of the narrow radius distance matching baseline and the method of the present invention on the same representative simulated star map; wherein... Figure 7 In the middle (a), the identification results of the narrow radius distance matching baseline are shown. Figure 7 (b) shows the identification result of the method of the present invention.
[0058] Figure 8 This is a comparison chart showing the recognition results of the narrow radius distance matching baseline under star trailing noise conditions and the recognition results of the method of the present invention on the same simulated star map; wherein... Figure 8 In the middle (a), the identification results of the narrow radius distance matching baseline are shown. Figure 8 (b) shows the identification result of the method of the present invention. Detailed Implementation
[0059] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.
[0060] like Figure 1 As shown, this invention provides a high-dynamic star map reliability weighted neighborhood consistency star catalog matching method. This method is applicable to star map matching scenarios with known attitude priors. It calculates the fusion reliability of detected stars by detecting star quality information, evaluates the candidate matching correspondence by utilizing the neighborhood relative displacement consistency between the detected star set and the locally projected star set, and completes one-to-one star catalog matching by combining the absolute image plane distance. The specific implementation process is as follows.
[0061] Step S1: Obtain relevant information of the current frame of the star map, including: detection star quality information, current frame attitude prior, camera imaging parameters and navigation star catalog; calculate the detection star fusion reliability based on the detection star quality information, and establish a local projected star set based on the attitude prior, camera imaging parameters and navigation star catalog.
[0062] Suppose that the current frame has obtained a total of The first detection satellite, the first One detection star (detection star) The two-dimensional image plane coordinates are The quality information of a detected star is a quality metric generated for the corresponding detected star during the star detection process. It can include one or more of the following: detection confidence, heatmap response, local signal-to-noise ratio, centroid fitting quality, tail integrity, and graph context reliability.
[0063] like Figure 2 As shown, in this embodiment, the normalized detection confidence level is used. Heatmap response And graph context reliability As information on the quality of the detection satellite. Detection satellite Fusion reliability Calculate according to the following formula:
[0064] (1)
[0065] In the formula, , and These are the fusion weights for detection confidence, heatmap response, and graph context reliability, respectively. , and All are non-negative values and satisfy: Among them, the detection star The detection confidence level is directly output by the star detector for that detection star; the thermal map response is determined based on the thermal map output by the star detector, specifically by taking the detection star. The thermal map response value at the centroid coordinates, or the maximum response value within a preset local window centered at those centroid coordinates; the reliability of the graph context is determined by the graph context evaluation module when setting the star detector, based on the detected stars. The context features of the detected stars and their neighboring stars are output. For the quality information of detected stars whose values are not between 0 and 1, normalization is performed before fusion. When the star detector does not output the graph context reliability, its corresponding fusion weights are used. Set it to 0, and renormalize the remaining fusion weights.
[0066] For the amplitude limiting function, Indicates the numerical value Limited to the range of 0 to 1: when When 0 is taken, Take 1 in the first case, and take 1 in the other cases. When a certain satellite quality information is missing, the corresponding fusion weight is set to 0, and the remaining fusion weights are renormalized.
[0067] After normalization, multiplication by the corresponding fusion weight, weighted summation according to the fusion weight, and amplitude limiting, the detection star fusion reliability is formed with a value ranging from 0 to 1. The higher the detection star fusion reliability, the higher the probability that the corresponding detection star was formed from a real star and that its position measurement results are reliable. Graph context reliability is only used as an optional piece of evidence for detection star quality in the fusion process and does not directly output navigation star numbers or catalog target identifiers.
[0068] The attitude prior is provided by the star image tracking result of the previous frame, inertial device output, integrated navigation result, or other attitude estimation result. Navigation stars within the current field of view are selected based on the attitude prior, camera intrinsics, image size, field of view, magnitude threshold, and navigation star table. The selected navigation stars are then projected onto the imaging plane, with each projected star having a corresponding navigation star identifier. The projected navigation stars are the projected stars themselves, and the projected stars on the current imaging plane form a local projected star set. Let the local projected star set contain... The projection star, the first Projected stars (projected stars) The two-dimensional image plane coordinates are Each detected star has a corresponding navigation star identifier. The attitude prior can be provided by the star map tracking results of the previous frame, inertial device output, integrated navigation results, or other attitude estimation results. Therefore, step S1 outputs the fusion reliability and local projected star set of each detected star, which are used for subsequent candidate scoring and candidate matching pair construction.
[0069] Step S2, calculate the detection star With projected stars absolute image plane distance between :
[0070] (2)
[0071] Only when At that time, the detection star With projected stars The combination of detection stars and projection stars is retained as candidate matching pairs. The projected stars that will form candidate matching pairs Let be denoted as candidate projection stars, where ... This is the first search radius. Combinations outside the first search radius are not included in subsequent neighborhood structure calculations to reduce the computational cost caused by invalid candidates. Step S2 outputs the initial set of candidate matching pairs within the first search radius. This is used for step S3 to perform neighborhood structure consistency calculation.
[0072] Step S3, Figure 3 As shown, based on the image plane distance between the detection stars, each detection star is selected The nearest neighbors, or the detection stars within a preset neighborhood radius, are selected to form a detection star system. neighborhood Detecting stars The detection stars within the neighborhood of a candidate matching pair are denoted as neighborhood detection stars; Candidate matching pairs Projected stars in Hereinafter referred to as candidate projection stars Establish candidate projection stars in the local projection star set. The neighborhood of the local projected stars. In this embodiment, the local projected star set excluding candidate projected stars is selected. Projected stars other than themselves are considered members of that neighborhood, i.e., neighborhood projected stars; The range of values for the quantity is and In other embodiments, neighboring members can be selected according to the number of nearest neighbors of the projected star or a preset projection neighborhood radius.
[0073] like Figure 3 As shown, for candidate matching pairs By detection star Pointing to its neighboring detection stars The relative displacement is:
[0074] (3)
[0075] From candidate projection stars Pointing to the neighboring projected star in the neighborhood of the candidate projected star The relative displacement is:
[0076] (4)
[0077] In the formula, , Excluding candidate projection stars in the local projection star set Any projection star other than the one mentioned above.
[0078] For each neighboring detector star in the detection star's neighborhood. Compare the relative displacements of the detected stars respectively Relative displacement with each projected star The vector difference between the two is calculated, and the relative displacement of the projected star with the smallest difference from the relative displacement vector of the detected star is selected to obtain the neighborhood detected star. For candidate matching pairs Optimal Vector Consistency :
[0079] (5)
[0080] In the formula, It serves as the relative displacement consistency kernel scale, used to control the sensitivity of optimal vector consistency to relative displacement errors; The larger the value, the higher the tolerance for relative displacement error. The smaller the value, the stricter the consistency determination. The relative displacement consistency kernel scale is a positive number, which can be determined based on the detection star's centroid positioning error, the navigation star's projection error, and the image resolution; in this embodiment, it is taken as 1.5 pixels. The smaller the error between the relative displacement of the detection star and the relative displacement of a certain projected star, the more stringent the consistency determination. The closer to 1, the better. The above calculation does not require pre-determining neighboring detection stars. The corresponding navigation star, but in the candidate matching pair The search is conducted within the local relative structure centered on the star to find the most consistent star-surface relative displacement.
[0081] To reduce the impact of spurious and unstable detectors on structure determination, based on neighboring detectors... Fusion reliability Determine its structural weighting :
[0082] (6)
[0083] In the formula, These are the neighborhood base weights, which are greater than 0 and less than 1. and The calculation method is the same.
[0084] Candidate Matching Pairs neighborhood structure support Calculate according to the following formula:
[0085] (7)
[0086] Therefore, highly reliable neighboring detectors contribute more to structural support, while the impact of less reliable spurious responses or unstable detectors is suppressed. When the detector neighborhood is empty, the structural support can be set to a preset default value; by setting... This avoids situations where the denominator is 0 when the neighborhood is not empty. Step S3 outputs the neighborhood structure support of each candidate matching pair.
[0087] Step S4, detect the star With candidate projected stars absolute image plane distance Convert to distance similarity that decreases with increasing distance. :
[0088] (8)
[0089] In the formula, The absolute distance kernel scale is used to control the sensitivity of distance similarity to the absolute image plane distance between the detected star and the candidate projected star; The larger the value, the higher the tolerance for absolute image plane distance deviation. The smaller the value, the more stringent the requirements for the absolute image plane distance. The absolute distance kernel scale is a positive number and can be determined based on the attitude prior error, the detection star's centroid positioning error, the navigation star's projection error, and the image resolution. In this embodiment, it is taken as 3 pixels.
[0090] Subsequently, the neighborhood structure support was... Fusion reliability of candidate detection stars and distance similarity The candidate joint score of the candidate matching pairs is obtained by fusion. :
[0091] (9)
[0092] In the formula, , and These are the fusion weights for neighborhood structure support, detection star fusion reliability, and distance similarity, respectively. , and All are non-negative values and satisfy: .
[0093] Among them, neighborhood structure support is used to evaluate the relative geometric relationship between multiple stars, detection star fusion reliability is used to evaluate the quality of the candidate detection stars themselves, and distance similarity is used to preserve the local position constraints provided by the attitude prior. These three types of information jointly participate in the candidate scoring, which can reduce the possibility of erroneous matching when matching is based solely on single-point image-plane distance or solely on relative structure. Step S4 outputs the joint candidate score for each candidate matching pair.
[0094] Step S5, as follows Figure 4 As shown, as a preferred option, the following conditions are first retained: and The candidate matching pairs are used to form a set of candidate matching pairs to be assigned. .in, The second search radius, and , This is the candidate joint score threshold. Determine the set of candidate matches to be assigned. Is it empty? If If empty, output "No valid match in the current frame" as the matching result, and end the matching process for the current frame; if If the value is not empty, then proceed to the one-to-one allocation process of candidate matching pairs.
[0095] Figure 4 In the embodiment shown, the set of candidate matching pairs to be assigned is... Candidate matching pairs are ranked according to their joint candidate scores. Sort from highest to lowest to form a candidate matching pair sequence. Initialize the set of matched detection stars. Matched projection star set and the final matching set All are empty sets. Read the sequences sequentially according to their arrangement. Current candidate matching pairs in If the detection star And the projected star If the current candidate matching pair satisfies the one-to-one allocation condition, then accept the candidate matching pair and... Add to the final matching set , will detect stars Add to the set of matched detection stars And project the star Add to the set of matched projected stars Otherwise, discard the current candidate matching pair and do not add it to the final matching set. After processing the current candidate matching pair, determine whether the candidate matching pair is a sequence. The last candidate match in the sequence; if not, read the sequence in the order it is sorted. The next candidate matching pair is selected, and the above judgment and allocation process is repeated; if it is, the pair sequence ends. The traversal of the sequence forms a candidate matching pair sequence. If candidate matching pairs have the same candidate joint score, they can be sorted locally deterministically in sequence according to absolute image plane distance, detection star index, and projection star index.
[0096] In other embodiments, as Figure 4 The alternative to the sorted allocation method shown can also be to use candidate joint scores as candidate gains, employing a global maximum weight one-to-one allocation method with unmatched virtual endpoints. By setting unmatched virtual endpoints, detectors with insufficient quality or scores are allowed to remain unmatched, avoiding the forced acceptance of low-scoring candidates to obtain a full match. Regardless of whether sorted allocation or global maximum weight allocation is used, the constraints of each detector matching at most one projected star and each projected star matching at most one detector star must be satisfied.
[0097] Complete sequence After processing all candidate matching pairs, based on the final matching set The navigation star identifiers corresponding to each projected star are used to output the final matching set. The system has already matched the one-to-one correspondence between the detection satellites and the navigation satellites (matching results). It can also output the corresponding detection satellite fusion reliability, neighborhood structure support, candidate joint score, absolute image plane distance, and assignment status for subsequent attitude calculation, matching quality control, or fault diagnosis.
[0098] Example
[0099] In this embodiment, the number of nearest neighbors of the detected star is... Choose 6, first search radius Take 7 pixels, second search radius Take 4 pixels, relative displacement consistency kernel scale Take 1.5 pixels, absolute distance from kernel scale Take 3 pixels; fusion weights , and The candidate joint score thresholds are set to 0.55, 0.25, and 0.20 respectively. Set the value to 0.18. The above parameters can be adjusted based on camera resolution, star density, attitude prior error, and star detection accuracy.
[0100] In a group of Figure 5 In the parameter settings corresponding to the simulation experiment shown, the fusion weights of detection confidence, thermogram response, and graph context reliability are... , and The neighborhood base weights are set to 0.45, 0.30, and 0.25 respectively. The value is set to 0.15. The specific values mentioned above are used to illustrate feasible parameter combinations and are not intended to limit the invention; in other embodiments, adjustments can be made based on the detector output calibration results and imaging conditions.
[0101] To verify the adaptability of the above implementation method under conditions of position noise, spurious detectable stars, missed detectable stars, and star trails, a comparative experiment was conducted using a controllable simulated star map generated based on the navigation star catalog. The comparison method was a narrow radius distance matching baseline, which only uses the absolute image plane distance within the second search radius for candidate screening and one-to-one allocation, without using the reliability of detectable star fusion and neighborhood structure information. The experiment used the harmonic mean of matching precision and matching recall, F1, as the evaluation index. Each experimental condition was independently repeated 100 times, with a random seed of 20,260,726. Each experiment generated 30 locally projected stars, and 20 of them were randomly selected to generate corresponding effective detectable stars. When generating the coordinates of the detectable stars, the translation vector (2.2, -1.5) pixels was subtracted from the corresponding projected star coordinates, and then position noise, spurious detectable stars, missed detectable stars, or star trails were superimposed according to the corresponding scene.
[0102] like Figure 5As shown, Figure 5 In (a), under the condition of adding 14 pseudo-detection stars in each frame, the standard deviation of position noise is increased from 0 to 3.0 pixels; Figure 5 In (b), with the position noise standard deviation fixed at 0.45 pixels, the number of pseudo-detection stars per frame is increased from 0 to 30; Figure 5 In (c), with the position noise standard deviation fixed at 0.45 pixels and 14 pseudo-detection stars added per frame, the number of missed detection stars per frame is increased from 0 to 12. Figure 5 In section (d), with the position noise standard deviation fixed at 0.45 pixels and 14 pseudo-detection stars added per frame, the star trail length is increased from 0 to 6 pixels. Each curve represents the narrow radius distance matching baseline and the average matching F1 value of the method of this invention; the shaded area represents the 95% confidence interval of the average value.
[0103] When the standard deviation of the position noise is 0 and 3.0 pixels, the average F1 value of the narrow radius distance matching baseline is 88.39% and 51.44%, respectively, while the average F1 value of the method of the present invention is 97.53% and 53.93%, respectively. When the number of pseudo-detected stars per frame increases from 0 to 30, the average F1 value of the narrow radius distance matching baseline decreases from 98.33% to 80.45%, while the average F1 value of the method of the present invention decreases from 99.59% to 93.70%. When the number of missed detection stars per frame increases from 0 to 12, the average F1 value of the narrow radius distance matching baseline decreases from 88.60% to 78.67%, while the average F1 value of the method of the present invention decreases from 97.16% to 86.43%. When the star trail length increases from 0 to 6 pixels, the average F1 value of the narrow radius distance matching baseline decreases from 89.42% to 55.99%, while the average F1 value of the method of the present invention decreases from 96.72% to 58.98%. The above results show that, under the four types of interference conditions set in this embodiment, the average matching F1 value of the method of the present invention is higher than that of the narrow radius distance matching baseline.
[0104] Furthermore, a representative simulated star map containing background noise, weak star responses, and star trails is generated from the navigation star catalog, and the method of this invention is used for identification. For example... Figure 6 As shown, Figure 6 (a) is the original simulated star map. Figure 6 Image (b) shows the identification results of the method of the present invention; the circles in the identification results are used to mark the correctly matched detection stars, and the numbers near the circles are the corresponding navigation star identifiers. In this frame of simulated star map, the method of the present invention obtained 12 correct matches and no incorrect matches were generated.
[0105] like Figure 7As shown, on a representative simulated star map of the same frame, the narrow radius distance matching baseline is compared with the method of the present invention, where circles represent correct matches and crosses represent incorrect matches. The narrow radius distance matching baseline yielded 12 correct matches and 4 incorrect matches, while the method of the present invention yielded 12 correct matches and no incorrect matches, indicating that the reliability of the detected star fusion and the support of the neighborhood structure can suppress erroneous correspondences generated solely based on absolute image plane distance.
[0106] like Figure 8 As shown, the recognition results of the two methods are further compared on another frame of simulated star trailing star map. The narrow radius distance matching baseline yielded 17 correct matches and 1 incorrect match, while the method of the present invention yielded 18 correct matches and no incorrect matches, indicating that the method of the present invention can still maintain good matching accuracy when star trailing causes centroid shift and local structural disturbance.
[0107] comprehensive Figures 5 to 8 The simulation results shown indicate that, within the parameters and interference range given in this embodiment, the matching F1 value of the method of the present invention is higher than that of the narrow radius distance matching baseline under conditions of position noise, spurious detection stars, missed detection stars, and star tails, and it reduces erroneous matching on representative simulated star maps.
[0108] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.
Claims
1. A method for matching a weighted neighborhood consistency star catalog with high dynamic star map reliability, characterized in that, include: Obtain the detection star quality information of the current frame of the star map, and calculate the fusion reliability of each detection star based on the detection star quality information; Projecting the navigation stars within the current field of view onto the imaging plane forms a projected star, resulting in a local set of projected stars; Based on the absolute image plane distance between the detected star and the projected star, candidate matching pairs between the detected star and the projected star are established within the first search radius; The detection star neighborhood and the projection star neighborhood of the candidate matching pair are determined respectively. Based on the relative displacement of the neighboring detection stars in the detection star neighborhood and the relative displacement of the neighboring projection stars in the projection star neighborhood, the neighborhood structure support of the candidate matching pair is determined. The neighborhood structure support of the candidate matching pair, the fusion reliability of the detected stars in the candidate matching pair, and the absolute distance similarity of the candidate matching pair are weighted and fused to obtain the joint score of the candidate matching pair. Candidate matching pairs with high joint scores are selected based on the constraints and used as the final matching results; The constraint is that neither the detection star nor the projection star is used repeatedly.
2. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 1, characterized in that, The satellite quality information includes: the detection confidence of the satellite, the thermal map response of the satellite, and the graph context reliability of the satellite.
3. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 2, characterized in that, The formula for calculating the fusion reliability of the detection stars is as follows: ; in, Indicates detection star The reliability of fusion For detecting stars The normalized value of the detection confidence level. For detecting stars The normalized value of the heatmap response, For detecting stars The normalized value of the graph context reliability, , , These are the fusion weights for detection confidence, heatmap response, and graph context reliability, respectively. ; This is the amplitude limiting function.
4. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 3, characterized in that, The method for establishing candidate matching pairs is as follows: Calculate the detection star With projected stars absolute image plane distance : ; If satisfied , will detect stars With projected stars The combination of these is retained as a candidate matching pair; in, For detecting stars Two-dimensional image plane coordinates, For projection stars Two-dimensional image plane coordinates, This represents the first search radius.
5. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 3 or 4, characterized in that, Methods for determining the neighborhood structure support of candidate matching pairs include: Computing neighborhood detection stars For candidate matching pairs Optimal Vector Consistency : ; In the formula, This represents the number of projected stars in the local projected star set. The relative displacement consistency kernel scale. Detection stars for candidate matching pairs Pointing to its neighboring detection stars The relative displacement, Projected stars for candidate matching pairs Pointing to its neighboring projection star The relative displacement; Candidate matching pairs are calculated based on the optimal vector consistency. neighborhood structure support : ; In the formula, Weighting the structure, The detection neighborhood of the candidate matching pair.
6. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 5, characterized in that, The structural weighting is calculated using the following formula: ; in, The neighborhood basic weights, and satisfying , For neighboring detection stars The reliability of the fusion.
7. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 6, characterized in that, Candidate Matching Pairs absolute distance similarity Calculated using the following formula: ; in, The absolute distance kernel scale, For detecting stars With projected stars The absolute image-plane distance.
8. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 7, characterized in that, Candidate Matching Pairs Combined score The calculation formula is: ; In the formula, , and These are the fusion weights for neighborhood structure support, detection star fusion reliability, and distance similarity, respectively, and they satisfy: .
9. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 8, characterized in that, Also includes: Candidate matching pairs that simultaneously meet the following conditions Add to the set of candidate matching pairs to be assigned: , ; in, The second search radius, ; This is the threshold for the candidate joint score; Based on the constraints, candidate matching pairs with high joint scores are selected from the set of candidate matching pairs to be assigned, and these are used as the final matching results.
10. The high dynamic star map reliability weighted neighborhood consistency star catalog matching method according to claim 9, characterized in that, The method for selecting candidate matching pairs from the set of candidate matching pairs to be assigned is as follows: The candidate matching pairs in the set of candidate matching pairs to be assigned are sorted from high to low according to the joint candidate score to form a candidate matching pair sequence; Candidate matching pairs are read sequentially according to the sorting order and used as the matching results; If the detection star of the currently read candidate matching pair has been occupied by a previous matching result, then the currently read candidate matching pair is discarded. The final matching result is obtained after all candidate matching pairs in the candidate matching pair sequence have been read.
