CN117782065A - Star map identification method based on directed singular value and hierarchical credibility verification - Google Patents
Star map identification method based on directed singular value and hierarchical credibility verification Download PDFInfo
- Publication number
- CN117782065A CN117782065A CN202311843142.XA CN202311843142A CN117782065A CN 117782065 A CN117782065 A CN 117782065A CN 202311843142 A CN202311843142 A CN 202311843142A CN 117782065 A CN117782065 A CN 117782065A
- Authority
- CN
- China
- Prior art keywords
- star
- main
- stars
- matching
- credibility
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000012795 verification Methods 0.000 title claims abstract description 26
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 69
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000012163 sequencing technique Methods 0.000 claims description 16
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 17
- 238000003909 pattern recognition Methods 0.000 abstract description 9
- 238000010276 construction Methods 0.000 abstract description 8
- 230000007547 defect Effects 0.000 abstract description 2
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012896 Statistical algorithm Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Navigation (AREA)
Abstract
The invention relates to the field of star map recognition, in particular to a star map recognition method based on directed singular value and hierarchical credibility verification, which is used for solving the defects that a pattern recognition algorithm is easy to influence in a pattern construction process and further fails in star map recognition because all observation star points in an observation star map to be recognized participate in the pattern construction process, and redundant matching and error matching probability in a matching process of a subgraph isomorphic algorithm are high. The star map identification method based on directed singular value and hierarchical credibility verification fully considers the robustness of the matching features to interference factors such as star point position noise, star point brightness noise and the like in the process of constructing the observation feature triangle and the feature extraction.
Description
Technical Field
The invention relates to the field of star map identification, in particular to a star map identification method based on directed singular value and hierarchical credibility verification.
Background
In the existing attitude sensor, gyroscopes, magnetometers, earth sensors and the like are limited by factors such as error accumulation, measurement accuracy, application range and the like, and are not suitable for being independently applied to deep space detectors with long distance and long task period. The star sensor has the advantages of subthreshold second level, no drift error, strong autonomy and reference star spread over the space, thereby becoming the first choice for deep space detection. Wherein, star map identification is the key to finish astronomical navigation and positioning.
The star map recognition methods currently applied in engineering can be broadly categorized into two main categories: the pattern recognition algorithm is isomorphic with the subgraph algorithm. The pattern recognition algorithm distributes a unique pattern or mark for each navigation star, the pattern features are related to the positions of adjacent stars around the navigation star and are used for representing the topological relation of the adjacent stars, so that the star map recognition problem is converted into the pattern recognition problem, and the star map matching can be completed by searching the navigation star pattern which is most in line with the similarity judgment standard with the observation pattern. The subgraph isomorphism algorithm is easy to realize, but has low recognition rate in noise environment, long running time and larger database capacity.
The pattern recognition algorithm is large in deviation between the observed star map and the standard star table for constructing the pattern database because the camera instrument star and the like are inconsistent with the star and the like in the star table, and for the pattern recognition algorithm recognition process, all the observed star points in the star map participate in the pattern construction process, so that the pattern construction process is easily influenced and the star map recognition is failed. The subgraph isomorphism algorithm can complete the star map matching process by means of a plurality of observation satellites, so that the inconsistency of the camera instrument satellites and the like with the constancy of the star and the like in the star table has relatively small influence on the subgraph construction and matching process, but the redundancy matching and error matching probability in the subgraph isomorphism algorithm matching process is higher.
Disclosure of Invention
The invention aims to solve the defects that the pattern recognition algorithm is easy to influence in the pattern construction process and further causes star map recognition failure because all observation star points in the observation star map to be recognized participate in the pattern construction process, and the probability of redundant matching and error matching in the matching process of the subgraph isomorphic algorithm is high, and provides a star map recognition method based on directed singular value and hierarchical credibility verification.
The star map identification method based on directional singular value and hierarchical credibility verification is characterized by comprising the following steps:
step 1, selecting star points of N before brightness sequencing in a field of view of a star sensor as N navigation stars respectively, constructing a characteristic triangle of each navigation star, and calculating a directed singular value p corresponding to each characteristic triangle 1 、p 2 The method comprises the steps of carrying out a first treatment on the surface of the N is an integer of 10 or more and 15 or less;
step 2, selecting star points of a to-be-identified observation star map before gray level sequencing as main stars, constructing an observation feature triangle of a kth main star, and calculating directed singular values corresponding to the observation feature trianglek=1, 2, 3..a; a is a positive integer of 1.2 to 1.5 times N;
step 3, the directional singular values obtained in the step 2 are processed Respectively with the directional singular values p obtained in the step 1 1 、p 2 Comparing, selecting the difference between the corresponding oriented singular values to be in advanceSetting navigation satellites in an error range as candidate matching satellites;
step 4, screening and verifying the candidate matching star by a credibility judgment method, determining the unique candidate identification of the kth main star, removing the wrong matching, and outputting a final star map identification result; the credibility judging method comprises star voting, credibility attribute judgment and verification.
Further, the step 1 specifically includes:
step 1.1, adopting a uniform scanning mode of an all-celestial sphere, and selecting star points of N before brightness sequencing in a field of view of a star sensor as N navigation satellites respectively; n is an integer of 10 or more and 15 or less;
step 1.2, taking each navigation star as a center, taking all stars in an R-R interval as neighbor stars, and forming a characteristic triangle by two neighbor stars before brightness sequencing and the navigation star; wherein R and R are respectively the buffer radius of the neighbor star and the mode radius of the neighbor star;
step 1.3, calculating the directed singular value p corresponding to the characteristic triangle 1 、p 2 ；
Step 1.4, establishing a navigation database, and recording characteristic triangles corresponding to all navigation stars and directed singular values p corresponding to the characteristic triangles 1 、p 2 。
Further, the step 1.3 specifically includes:
define the navigation star as S 1 The two adjacent stars of the first two of the brightness sequences are S in turn 2 、S 3 ；
Step 1.3.1, determining a direction coefficient c;
the direction coefficient c defining the feature triangle is as follows: if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle between the two is anticlockwise, and the direction coefficient c is 1; if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle between the two is clockwise, and the direction coefficient c is-1;
step 1.3.2, calculating a feature matrix:
defining the feature matrix G of the feature triangle as
Wherein d ij Represent S i And S is equal to j Angular distance between, i=1, 2,3, j=1, 2,3, angular distance d ij The value of (2) can be calculated by the following formula:
d ij ＝arc cos(cos(σ i )*cos(σ j )*cos(α i -α j )+sin(σ i )*sin(σ j ))
wherein alpha is i And sigma (sigma) i Respectively represent S i Corresponding right ascension and declination values in the star table; alpha j And sigma (sigma) j Respectively represent star S j Corresponding right ascension and declination values in the star table;
step 1.3.3, directional singular value decomposition of a feature matrix G;
singular value g=u Σv of feature matrix G obtained by SVD conversion T ；
Wherein U is an orthogonal matrix, and the column vector is a left singular vector; sigma is a diagonal matrix with element e on the diagonal 1 、e 2 、e 3 Is a singular value; v (V) T The orthogonal matrix is transposed, and the row vector is a right singular vector;
the directional singular value vector P defining the feature triangle is:
P＝[p 1 p 2 p 3 ]＝c*[e 1 e 2 e 3 ]；
p in directed singular value vector P 1 、p 2 Is the directed singular value of the feature triangle.
Further, the step 2 specifically includes:
selecting star points of a to-be-identified observation star map before gray level sequencing as main stars, taking a kth main star as a center, and forming an observation characteristic triangle by two adjacent stars of a second to-be-identified gray level sequencing in all adjacent stars in an R-R interval and the kth main star; calculating directed singular values corresponding to observation feature triangles of kth main starR and R are the neighbor star buffer radius and the neighbor star pattern radius, respectively.
Further, the step 3 specifically includes:
step 3.1, the directional singular values obtained in the step 2 are used forAnd directed singular value p 1 Comparing, and rapidly selecting the directed singular value p by using a dichotomy 1 At->All navigation stars in the interval form a navigation star set, and epsilon represents the allowable error of the directional singular value;
step 3.2, the directional singular values obtained in the step 2 are used forAnd step 3.1 navigation of the directed singular values p in the star set 2 Performing traversal comparison, and selecting directed singular values p in a navigation star set 2 At->B navigation satellites in the interval are respectively used as m candidate matching satellites C of the kth main satellite in the observation star map to be identified km The method comprises the steps of carrying out a first treatment on the surface of the b is an integer not less than 1, m=1, 2.
Further, the epsilon has a value of 3 to 5 pixels corresponding to the angle of the field of view.
Further, the step 4 specifically includes:
step 4.1, voting by star points;
calculating the numerical value vn of the counter of all candidate matched stars of the kth main star respectively by utilizing the consistency of the observed angular distance and the angular distance between the corresponding initial matching results, and selecting the candidate matched star with the most numerical value vn of the counter as the unique candidate identification of the kth main star;
step 4.2, judging the credibility attribute of the unique candidate identification of the kth main star;
if the number vn of the counter is more than or equal to 4 and the number of unique candidate identification is more than or equal to 5, judging that the credibility attribute is high credibility, and executing the step 4.3.1;
if the unique candidate identification of the numerical value vn of the counter is more than or equal to 3 is equal to 4, or the unique candidate identification of the numerical value vn of the counter is more than or equal to 2 is equal to 3, judging the credibility attribute as medium credibility, and executing the step 4.3.2;
if the number of unique candidate identifications with the value vn of the counter equal to 1 is more than or equal to 2, or the number of unique candidate identifications with the value vn of the counter smaller than or equal to 2 is less than 3, judging that the credibility attribute is low credibility, and executing the step 4.3.3;
step 4.3.1, outputting a unique candidate identification of the kth main star as a final star map identification result of the kth main star;
step 4.3.2, selecting unique candidate identifications with the numerical value vn of the counter being more than or equal to 2 or 3, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the errors between the two angular distances are smaller than a threshold epsilon', and if yes, outputting the unique candidate identifications as final star map identification results of the corresponding main stars; otherwise, the algorithm identification fails, and the observation star map to be identified of the next frame is identified;
step 4.3.3, selecting unique candidate identifications with the value vn of the counter equal to 1 or 2, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the star pair meets the error between the two angular distances to be a threshold epsilon', and if yes, outputting a final star map identification result of the unique candidate identification as the corresponding main star; otherwise, the algorithm fails to identify, and the observation star map to be identified in the next frame is identified.
Further, the step 4.1 specifically includes:
a counter is allocated to each candidate matching star, and the value vn of all the counters is set to 0 before voting begins;
calculate the kth 1 The main star and the kth 2 Angular distance between main satellitesk 1 ＝1，2，3...a，k 2 =1, 2, 3..a, and k 1 ≠k 2 ；
Candidate matching starMatching star->Angular distance between-> Is the kth 1 M of the main star 1 Candidate matching star, ++>Is the kth 2 M of the main star 2 Candidate matching stars, m 1 、m 2 Is an integer of 1 or more and b or less;
judging angular distanceDistance from angle->If the error between the two is smaller than the threshold epsilon', the matching is considered to be successful, and the candidate matching star is used for +.>For the kth 1 Voting once by the main star, and selecting matching star +.>For the kth 2 One by oneAnd voting is carried out once, the value vn of the counter is increased by 1, otherwise, the matching is considered to be failed, and voting is not carried out.
Further, the value of epsilon' is the view field angle corresponding to 3 to 5 pixels.
Compared with the prior art, the invention has the beneficial effects that:
(1) The star map recognition method based on directional singular value and hierarchical credibility verification fully considers the robustness of the matching features to interference factors such as star point position noise, star point brightness noise and the like in the process of constructing the observation feature triangle and feature extraction, so that the reliability in the sub-map matching process is improved as much as possible by the initial matching scheme based on the directional singular value, and further, in order to improve algorithm matching operation efficiency, a database is searched by adopting a dichotomy method, and the initial recognition result is screened and verified by adopting a credibility judgment method, so that the correct recognition result is kept as much as possible, and the requirement of a star map recognition algorithm on high recognition success rate in the whole celestial sphere range is met.
(2) According to the star map identification method based on directed singular value and hierarchical credibility verification, a credibility judgment method is adopted to obtain the unique identification result of the observation star, and the problem that the subsequent star map identification fails due to redundant matching and incorrect matching in candidate matching stars is avoided.
(3) Compared with a grid algorithm and an SVD algorithm, the star map identification method based on directed singular value and hierarchical credibility verification has higher robustness to star point position noise, star point brightness noise and pseudo-star noise.
Drawings
FIG. 1 is a flow chart of an embodiment of a star map identification method based on directed singular value and hierarchical confidence verification in the present invention;
FIG. 2 is a schematic view of a characteristic triangle of each navigation star in step 1 according to an embodiment of the present invention;
FIG. 3 is a graph showing the variation of the recognition rate and the error rate of the SVD algorithm and the grid algorithm under different star point position noises;
FIG. 4 is a graph showing the variation of the recognition rate and the error rate of the SVD algorithm and the grid algorithm under different star point brightness noises;
FIG. 5 is a graph showing the variation of the recognition rate and the error rate of the SVD algorithm and the grid algorithm under different numbers of pseudo-satellites according to the embodiment of the invention.
Detailed Description
The invention is further described below with reference to the drawings and exemplary embodiments.
Referring to fig. 1, a star map identification method based on directed singular value and hierarchical reliability verification includes the steps of:
step 1, referring to fig. 2, selecting N star points of front N of brightness sequence in a field of view of a star sensor as N navigation satellites respectively, constructing a characteristic triangle of each navigation satellite, and calculating a directed singular value p corresponding to each characteristic triangle 1 、p 2 ；
Step 1.1, adopting a uniform scanning mode of the whole celestial sphere, and selecting N star points before brightness sequencing in a field of view of a star sensor as N navigation satellites respectively; n is an integer of 10 or more and 15 or less;
step 1.2, taking each navigation star as a center, taking all stars in an R-R interval as neighbor stars, and forming a characteristic triangle by two neighbor stars before brightness sequencing and the navigation star;
wherein R and R are respectively a neighbor star buffer radius and a neighbor star mode radius, the outer circle radius R is limited by the field of view of the star sensor, and the introduction of the inner circle radius R avoids the interference of double stars in the construction process of the characteristic triangle; r is the view angle corresponding to 10 pixels, and R is half of the view angle;
step 1.3, calculating the directed singular value p corresponding to the characteristic triangle 1 、p 2 ；
Define the navigation star as S 1 The two adjacent stars of the first two of the brightness sequences are S in turn 2 、S 3 ；
Step 1.3.1, determining a direction coefficient c;
the direction coefficient c defining the feature triangle is as follows: if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle (0-180) is anticlockwise, and the direction coefficient c is 1; if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle (0-180) is clockwise, and the direction coefficient c is-1;
step 1.3.2, calculating a feature matrix:
defining the feature matrix G of the feature triangle as
Wherein d ij Represent S i And S is equal to j Angular distance between, i=1, 2,3, j=1, 2,3, angular distance d ij The value of (2) can be calculated by the following formula:
d ij ＝arc cos(cos(σ i )*cos(σ j )*cos(α i -α j )+sin(σ i )*sin(σ j ))
wherein alpha is i And sigma (sigma) i Respectively represent S i Corresponding right ascension and declination values in the star table; alpha j And sigma (sigma) j Respectively represent star S j Corresponding right ascension and declination values in the star table;
step 1.3.3, directional singular value decomposition of a feature matrix G;
obtaining singular values g=u Σv of the feature matrix G by SVD (Singular Value Decomposition) transformation T ；
Wherein U is an orthogonal matrix, and the column vector is a left singular vector; sigma is a diagonal matrix with element e on the diagonal 1 、e 2 、e 3 Is a singular value; v (V) T The orthogonal matrix is transposed, and the row vector is a right singular vector;
the directional singular value vector P defining the feature triangle is:
P＝[p 1 p 2 p 3 ]＝c*[e 1 e 2 e 3 ]；
c is the direction coefficient of step 1.3.1;
step 1.3.4, feature dimension reduction;
p in directed singular value vector P 1 、p 2 Directed singular values that are feature triangles;
step 1.4, establishing a navigation database, and recording characteristic triangles corresponding to all navigation stars and directed singular values p corresponding to the characteristic triangles 1 、p 2 ；
Step 2, selecting a star points of a to-be-identified observation star map before gray level sequencing as k-th main stars, wherein k=1, 2 and 3. a is a positive integer of 1.2 to 1.5 times N;
taking the kth main star as a center, and forming an observation characteristic triangle by the two adjacent stars before gray level sequencing in all adjacent stars in the R-R interval and the kth main star;
calculating directed singular values corresponding to observation feature triangles of kth main star(same as step 1.3);
step 3, the directional singular values obtained in the step 2 are processed Respectively with the directional singular values p obtained in the step 1 1 、p 2 Comparing, and selecting navigation satellites with the difference of corresponding oriented singular values within a preset error range as candidate matching satellites;
step 3.1, the directional singular values obtained in the step 2 are used forDirected singular value p in a navigation database 1 Comparing, and rapidly selecting the directed singular value p in the navigation database by using a dichotomy 1 At->All navigation stars in the interval form a navigation star set, and epsilon represents a directionThe allowable error of the singular value is the view field angle corresponding to 3 to 5 pixels;
step 3.2, the directional singular values obtained in the step 2 are used forAnd step 3.1 navigation of the directed singular values p in the star set 2 Performing traversal comparison, and selecting directed singular values p in a navigation star set 2 At->B navigation satellites in the interval are respectively used as m candidate matching satellites C of the kth main satellite in the observation star map to be identified km The method comprises the steps of carrying out a first treatment on the surface of the b is an integer not less than 1, m=1, 2..b;
after the directional singular value feature matching is completed, candidate matching stars corresponding to some main stars in the imaging plane are determined; however, among these candidate matching stars, there may be some redundant matches, even false matches, which are fatal to subsequent pose calculations; therefore, the invention adopts a credibility judgment method to obtain the unique identification result of the observation star;
step 4, screening and verifying the candidate matching star by a credibility judgment method, determining the unique candidate identification of the kth main star, removing the wrong matching, and outputting a final star map identification result; the credibility judging method comprises star voting, credibility attribute judgment and verification;
step 4.1, voting by star points;
calculating the numerical value vn of the counter of all candidate matched stars of the kth main star respectively by utilizing the consistency of the observed angular distance and the angular distance between the corresponding initial matching results, and selecting the candidate matched star with the most numerical value vn of the counter as the unique candidate identification of the kth main star;
specifically, a counter is allocated to each candidate matching star, and before voting begins, the values vn of all the counters are set to 0;
calculate the kth 1 The main star and the kth 2 Angular distance between main satellitesk 1 ＝1，2，3...a，k 2 =1, 2, 3..a, and k 1 ≠k 2 ；
Candidate matching starMatching star->Angular distance between-> Is the kth 1 M of the main star 1 Candidate matching star, ++>Is the kth 2 M of the main star 2 Candidate matching stars, m 1 、m 2 Is an integer of not less than 1 and not more than b;
judging angular distanceDistance from angle->If the error between the two is smaller than the threshold epsilon', the matching is considered to be successful, and the candidate matching star is used for +.>For the kth 1 Voting once by the main star, and selecting matching star +.>For the kth 2 The primary star votes once, the value vn of the counter is increased by 1, otherwise, the matching is considered to be failed, and the voting is not carried out; epsilon' is the field angle corresponding to 3 to 5 pixels;
step 4.2, judging the credibility attribute of the unique candidate identification of the kth main star;
if the number vn of the counter is more than or equal to 4 and the number of unique candidate identification is more than or equal to 5, judging that the credibility attribute is high credibility, and executing the step 4.3.1;
if the unique candidate identification of the numerical value vn of the counter is more than or equal to 3 is equal to 4, or the unique candidate identification of the numerical value vn of the counter is more than or equal to 2 is equal to 3, judging the credibility attribute as medium credibility, and executing the step 4.3.2;
if the number of unique candidate identifications with the value vn of the counter equal to 1 is more than or equal to 2, or the number of unique candidate identifications with the value vn of the counter smaller than or equal to 2 is less than 3, judging that the credibility attribute is low credibility, and executing the step 4.3.3;
step 4.3, respectively adopting corresponding verification according to the credibility attribute to determine a final star chart recognition result;
step 4.3.1, outputting a unique candidate identification of the kth main star as a final star map identification result of the kth main star;
step 4.3.2, selecting unique candidate identifications with the numerical value vn of the counter being more than or equal to 2 or 3, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the errors between the two angular distances are smaller than a threshold epsilon', and if yes, outputting the unique candidate identifications as final star map identification results of the corresponding main stars; otherwise, the algorithm identification fails (namely, any star pair exists to meet the error between two angular distances and is larger than or equal to a threshold epsilon', the algorithm identification fails), and the observation star map to be identified in the next frame is identified;
step 4.3.3, selecting unique candidate identifications with the value vn of the counter equal to 1 or 2, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the star pair meets the error between the two angular distances to be a threshold epsilon', and if yes, outputting a final star map identification result of the unique candidate identification as the corresponding main star; otherwise, the algorithm fails to identify, and the observation star map to be identified in the next frame is identified.
In order to evaluate the performance of the algorithm of the invention, a simulated star chart is generated by adopting an SAO star table for simulation verification, and the algorithm of the invention is compared with a SVD (Singular Value Decomposition) algorithm and a grid algorithm, wherein parameters of a star sensor platform and key parameters of the algorithm in the simulation are shown in a table 1.
TABLE 1
| Star sensor image plane size | 1024 pixels×1024 pixels |
| Star sensor pixel size | 0.015 mm by 0.015 mm |
| Focal length | 54.646 mm |
| Field of view size | 16 degree by 16 degree |
| Limit star etc | 6.0 Star etc |
| Neighbor star pattern radius | R=5 degrees |
| Neighbor star buffer radius | r=0.3 degrees |
The simulation experiment mainly tests the influence of the number of pseudo-stars in the field of view of the star sensor on the algorithm performance on the star point position noise and the star point brightness noise. In noise testing, one noise source is kept at a typical value, while the other noise source is linearly transformed, while the algorithm parameters are constant throughout the test. In each case 10000 Monte Carlo tests are adopted to identify the correct rate and error rate conditions of the statistical algorithm.
Because of the influence of image sensor noise, optical lens distortion, sub-pixel positioning algorithm and other interference factors, the imaging plane position coordinate where the real calculated star is located has certain deviation from the corresponding ideal position, namely star point position noise is introduced. In the test, gaussian noise is added at the theoretical position of a star point to be used as the projection coordinate of the star on the plane of the detector to simulate random noise, and in addition, the measurement data of the brightness of the star is limited to a certain extent by the existing photoelectric detection system due to the fact that the spectrum characteristic and brightness change of the star, the characteristics of an image sensor and an optical system and other factors are limited, namely, the star target actually detected by a camera is not completely consistent with the corresponding star in the star table on the star and the like. The brightness noise is represented in the real star map as follows: some dark stars, such as stars higher than the limit stars of the camera, appear in the real-time star map in reality, while other bright stars, such as stars lower than the limit stars of the star sensor, do not appear in the real-time star map, so Gaussian random noise is added to the brightness value of the star point in the simulation experiment to serve as the brightness value of the star point in the simulated star map. In addition, due to the influence of the planet, the artificial satellite or the space debris, certain non-fixed targets are recognized as fixed stars by mistake, and the star pattern recognition is interfered, so that the star pattern recognition is failed.
Star point brightness noise standard deviation 0.3Mv, star point position noise standard deviation is increased from 0pixel to
In the case of 2 pixels, the change curves of the recognition rate and the error rate of the SVD algorithm and the grid algorithm are shown in figure 3, and as can be seen from figure 3, when the standard deviation of the star point position noise increases to 2 pixels, the recognition rate of the algorithm reaches more than 97.0%, and still keeps at a higher level. As comparison, when the standard deviation of the star point position noise is increased from 0pixel to 2pixel, the recognition rates of the SVD algorithm and the grid algorithm are respectively reduced from 95.0% to 81.1% and 97.2% to 90.0%. In terms of error rate, the algorithm of the invention does not have error identification, and the error rate of the grid algorithm is increased from 1.8% to 4.5% under the condition that the standard deviation of the star point position noise is increased from 0pixel to 2 pixel. In conclusion, compared with the grid algorithm and the SVD algorithm, the algorithm provided by the invention has higher robustness to star point position noise.
As shown in FIG. 4, when the standard deviation of the star point position noise is 0.5pixel and the standard deviation of the star point brightness noise is increased from 0Mv to 0.6Mv, the change curves of the recognition rate and the error rate of the SVD algorithm and the grid algorithm are shown, and when the standard deviation of the star point brightness noise is increased to 0.6Mv, the recognition rate of the algorithm reaches more than 96.2%, and the high level is still maintained. As comparison, when the standard deviation of the star point brightness noise is increased from 0Mv to 0.6Mv, the recognition rates of the SVD algorithm and the grid algorithm are respectively reduced from 95.0% to 75.7%, and 99.3% to 90.4%. In terms of error rate, the algorithm of the invention does not have error recognition, and the recognition error rate of the grid algorithm is increased from 1.9% to 6.2% under the condition that the standard deviation of star point brightness noise is increased from 0Mv to 0.6 Mv. In conclusion, compared with the grid algorithm and the SVD algorithm, the algorithm provided by the invention has higher robustness to star point brightness noise. When the standard deviation of the star point brightness noise is increased from 0Mv to 0.6Mv, the recognition rate of the algorithm in the all-day star map recognition task is higher, and the algorithm is proved to have strong robustness on the star point brightness noise.
The change curves of the identification rate and the error rate of the three star map identification algorithms are shown in figure 5 when the number of the added pseudo-star is increased from 0 to 3, and the identification rate of the algorithm reaches more than 93.2% and still remains at a higher level when the number of the added pseudo-star is increased to 3. As comparison, when the number of the added pseudo-star is increased from 0 to 3, the recognition rate of the SVD algorithm and the grid algorithm is respectively reduced from 94.4% to 68.6% and 96.1% to 93.8%, and the algorithm has no error recognition condition in the aspect of error rate. In summary, compared with the grid algorithm and the SVD algorithm, the algorithm of the invention has higher robustness to the pseudo-star noise.
The algorithm of the invention comprises two steps, namely initial matching based on directed singular value characteristics and final matching based on credibility judgment. The former mainly relates to directed singular value feature matching of a feature triangle, the latter uses star voting to judge and identify the credibility, and by constructing the feature triangle and introducing reliable directed singular value features in the process, on one hand, the algorithm searching and matching efficiency is accelerated, and on the other hand, the feature dimension is increased to reduce the occurrence probability of redundant matching. The candidate matching results under different conditions are verified in a grading manner on the basis of the matching results, so that no error matching occurs in the output results, correct matching is reserved as far as possible, and a final matching star pair is obtained.
Claims (9)
1. A star map identification method based on directed singular value and hierarchical credibility verification is characterized by comprising the following steps:
step 1, selecting star points of N before brightness sequencing in a field of view of a star sensor as N navigation stars respectively, constructing a characteristic triangle of each navigation star, and calculating a directed singular value p corresponding to each characteristic triangle 1 、p 2 The method comprises the steps of carrying out a first treatment on the surface of the N is an integer of 10 or more and 15 or less;
step 2, selecting star points of a to-be-identified observation star map before gray level sequencing as main stars, constructing an observation feature triangle of a kth main star, and calculating directed singular values corresponding to the observation feature trianglek=1, 2,3 … a; a is a positive integer of 1.2 to 1.5 times N;
step 3, the directional singular values obtained in the step 2 are processedRespectively with the directional singular values p obtained in the step 1 1 、p 2 Comparing, and selecting navigation satellites with the difference of corresponding oriented singular values within a preset error range as candidate matching satellites;
step 4, screening and verifying the candidate matching star by a credibility judgment method, determining the unique candidate identification of the kth main star, removing the wrong matching, and outputting a final star map identification result; the credibility judging method comprises star voting, credibility attribute judgment and verification.
2. The star map identification method based on directional singular value and hierarchical credibility verification according to claim 1, wherein the step 1 is specifically:
step 1.1, adopting a uniform scanning mode of an all-celestial sphere, and selecting star points of N before brightness sequencing in a field of view of a star sensor as N navigation satellites respectively; n is an integer of 10 or more and 15 or less;
step 1.2, taking each navigation star as a center, taking all stars in an R-R interval as neighbor stars, and forming a characteristic triangle by two neighbor stars before brightness sequencing and the navigation star; wherein R and R are respectively the buffer radius of the neighbor star and the mode radius of the neighbor star;
step 1.3, calculating the directed singular value p corresponding to the characteristic triangle 1 、p 2 ；
Step 1.4, establishing a navigation database, and recording characteristic triangles corresponding to all navigation stars and directed singular values p corresponding to the characteristic triangles 1 、p 2 。
3. The star map identification method based on directional singular value and hierarchical credibility verification according to claim 2, wherein the step 1.3 is specifically:
define the navigation star as S 1 The two adjacent stars of the first two of the brightness sequences are S in turn 2 、S 3 ；
Step 1.3.1, determining a direction coefficient c;
defining characteristic trianglesThe direction coefficient c is as follows: if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle between the two is anticlockwise, and the direction coefficient c is 1; if the vector S 1 S 2 Rotated to vector S 1 S 3 The included angle between the two is clockwise, and the direction coefficient c is-1;
step 1.3.2, calculating a feature matrix:
defining the feature matrix G of the feature triangle as
Wherein d ij Represent S i And S is equal to j Angular distance between, i=1, 2,3, j=1, 2,3, angular distance d ij The value of (2) can be calculated by the following formula:
d ij ＝arccos(cos(σ i )*cos(σ j )*cos(α i -α j )+sin(σ i )*sin(σ j ))
wherein alpha is i And sigma (sigma) i Respectively represent S i Corresponding right ascension and declination values in the star table; alpha j And sigma (sigma) j Respectively represent star S j Corresponding right ascension and declination values in the star table;
step 1.3.3, directional singular value decomposition of a feature matrix G;
singular value g=u Σv of feature matrix G obtained by SVD conversion T ；
Wherein U is an orthogonal matrix, and the column vector is a left singular vector; sigma is a diagonal matrix with element e on the diagonal 1 、e 2 、e 3 Is a singular value; v (V) T The orthogonal matrix is transposed, and the row vector is a right singular vector;
the directional singular value vector P defining the feature triangle is:
P＝[p 1 p 2 p 3 ]＝c*[e 1 e 2 e 3 ]；
p in directed singular value vector P 1 、p 2 Is characterized byDirectional singular values of triangles.
4. The star map identification method based on directional singular value and hierarchical credibility verification according to claim 1, wherein the step 2 is specifically:
selecting star points of a to-be-identified observation star map before gray level sequencing as main stars, taking a kth main star as a center, and forming an observation characteristic triangle by two adjacent stars of a second to-be-identified gray level sequencing in all adjacent stars in an R-R interval and the kth main star; calculating directed singular values corresponding to observation feature triangles of kth main starR and R are the neighbor star buffer radius and the neighbor star pattern radius, respectively.
5. The star map identification method based on directional singular value and hierarchical credibility verification according to claim 1, wherein the step 3 is specifically:
step 3.1, the directional singular values obtained in the step 2 are used forAnd directed singular value p 1 Comparing, and rapidly selecting the directed singular value p by using a dichotomy 1 At->All navigation stars in the interval form a navigation star set, and epsilon represents the allowable error of the directional singular value;
step 3.2, the directional singular values obtained in the step 2 are used forAnd step 3.1 navigation of the directed singular values p in the star set 2 Performing traversal comparison, and selecting directed singular values p in a navigation star set 2 At->B navigation satellites in the interval are respectively used as m candidate matching satellites C of the kth main satellite in the observation star map to be identified km The method comprises the steps of carrying out a first treatment on the surface of the b is an integer not less than 1, m=1, 2.
6. The star map identification method based on directed singular value and hierarchical confidence verification of claim 5, wherein the star map identification method is characterized by: the epsilon is a view field angle corresponding to 3 to 5 pixels.
7. The star map identification method based on directional singular value and hierarchical credibility verification according to any one of claims 1 to 6, wherein the step 4 is specifically:
step 4.1, voting by star points;
calculating the numerical value vn of the counter of all candidate matched stars of the kth main star respectively by utilizing the consistency of the observed angular distance and the angular distance between the corresponding initial matching results, and selecting the candidate matched star with the most numerical value vn of the counter as the unique candidate identification of the kth main star;
step 4.2, judging the credibility attribute of the unique candidate identification of the kth main star;
if the number vn of the counter is more than or equal to 4 and the number of unique candidate identification is more than or equal to 5, judging that the credibility attribute is high credibility, and executing the step 4.3.1;
if the unique candidate identification of the numerical value vn of the counter is more than or equal to 3 is equal to 4, or the unique candidate identification of the numerical value vn of the counter is more than or equal to 2 is equal to 3, judging the credibility attribute as medium credibility, and executing the step 4.3.2;
if the number of unique candidate identifications with the value vn of the counter equal to 1 is more than or equal to 2, or the number of unique candidate identifications with the value vn of the counter smaller than or equal to 2 is less than 3, judging that the credibility attribute is low credibility, and executing the step 4.3.3;
step 4.3.1, outputting a unique candidate identification of the kth main star as a final star map identification result of the kth main star;
step 4.3.2, selecting unique candidate identifications with the numerical value vn of the counter being more than or equal to 2 or 3, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the errors between the two angular distances are smaller than a threshold epsilon', and if yes, outputting the unique candidate identifications as final star map identification results of the corresponding main stars; otherwise, the algorithm identification fails, and the observation star map to be identified of the next frame is identified;
step 4.3.3, selecting unique candidate identifications with the value vn of the counter equal to 1 or 2, forming a star pair by each unique candidate identification and all unique candidate identifications which do not correspond to the same main star, calculating the angular distance of the star pair and the angular distance between the main stars corresponding to the two unique candidate identifications in the star pair, judging whether the star pair meets the error between the two angular distances to be a threshold epsilon', and if yes, outputting a final star map identification result of the unique candidate identification as the corresponding main star; otherwise, the algorithm fails to identify, and the observation star map to be identified in the next frame is identified.
8. The star map identification method based on directional singular value and hierarchical credibility verification according to claim 7, wherein the step 4.1 is specifically:
a counter is allocated to each candidate matching star, and the value vn of all the counters is set to 0 before voting begins;
calculate the kth 1 The main star and the kth 2 Angular distance between main satellitesk 1 ＝1,2,3…a，k 2 =1, 2,3 … a, and k 1 ≠k 2 ；
Candidate matching starMatching star->Angular distance between-> Is the kth 1 M of the main star 1 Candidate matching star, ++>Is the kth 2 M of the main star 2 Candidate matching stars, m 1 、m 2 Is an integer of 1 or more and b or less;
judging angular distanceDistance from angle->If the error between the two is smaller than the threshold epsilon', the matching is considered to be successful, and the candidate matching star is used for +.>For the kth 1 Voting once by the main star, and selecting matching star +.>For the kth 2 The primary star votes once, the value vn of the counter is increased by 1, otherwise, the matching is considered to be failed, and the voting is not carried out.
9. The star map identification method based on directed singular value and hierarchical confidence verification of claim 8, wherein the star map identification method is characterized by: the value of epsilon' is the view field angle corresponding to 3 to 5 pixels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311843142.XA CN117782065A (en) | 2023-12-28 | 2023-12-28 | Star map identification method based on directed singular value and hierarchical credibility verification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311843142.XA CN117782065A (en) | 2023-12-28 | 2023-12-28 | Star map identification method based on directed singular value and hierarchical credibility verification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117782065A true CN117782065A (en) | 2024-03-29 |
Family
ID=90387141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311843142.XA Pending CN117782065A (en) | 2023-12-28 | 2023-12-28 | Star map identification method based on directed singular value and hierarchical credibility verification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117782065A (en) |
-
2023
- 2023-12-28 CN CN202311843142.XA patent/CN117782065A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705457B (en) | Remote sensing image building change detection method | |
| Kolomenkin et al. | Geometric voting algorithm for star trackers | |
| CN111145227B (en) | Iterative integral registration method for space multi-view point cloud of underground tunnel | |
| CN109813303B (en) | Star map identification method independent of calibration parameters based on angular pattern cluster voting | |
| CN110132263B (en) | Star map identification method based on representation learning | |
| CN105548976A (en) | Shipborne radar offshore precision identification method | |
| CN103673995A (en) | Calibration method of on-orbit optical distortion parameters of linear array push-broom camera | |
| Clouse et al. | Small field-of-view star identification using Bayesian decision theory | |
| CN109631912A (en) | A kind of deep space spherical object passive ranging method | |
| CN106056625A (en) | Airborne infrared moving target detection method based on geographical homologous point registration | |
| CN111814597A (en) | Urban function partitioning method coupling multi-label classification network and YOLO | |
| Samaan et al. | Nondimensional star identification for uncalibrated star cameras | |
| CN107907126B (en) | Star map identification method, device and system based on triangle matching | |
| CN111174776B (en) | Radial and dynamic circumferential mode-based star map identification method | |
| CN112179335B (en) | Star map identification method based on simplest general subgraph | |
| CN117782065A (en) | Star map identification method based on directed singular value and hierarchical credibility verification | |
| CN112729277A (en) | Star sensor star map identification method based on dynamic included angle matching | |
| CN116091804B (en) | Star suppression method based on adjacent frame configuration matching | |
| CN115578448B (en) | Astronomical positioning method and system based on CCD observation data batch processing | |
| CN111156991A (en) | Space debris real-time astronomical positioning method based on automatic pointing error determination | |
| CN114255051B (en) | Authenticity inspection method of orthometric product based on stereo mapping satellite | |
| CN115290100A (en) | Partitioned fast star map identification method based on attitude information | |
| CN113483751A (en) | Radial triangle mapping matrix-based star map identification method | |
| CN117788849A (en) | All-celestial sphere star map identification method based on composite calibration star verification | |
| CN114543789B (en) | Star map identification method and system based on one-dimensional convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |