CN102840860B - A kind of method for recognising star map based on Hybrid Particle Swarm - Google Patents

Abstract

The invention discloses a method for identifying a star map based on a hybrid particle swarm algorithm, which includes two steps of establishing a navigation star database and identifying a star map based on a hybrid particle swarm algorithm. Group algorithm for fast path optimization. The star map recognition method based on the hybrid particle swarm algorithm proposed by the present invention solves the shortcomings of low star map recognition rate and poor robustness to noise under the condition of large field of view and high sensitivity star sensor; realizes fast matching and recognition of navigation stars , which improves the recognition rate and is more robust to noise.

Description

A Star Map Recognition Method Based on Hybrid Particle Swarm Algorithm

technical field

The invention relates to a star map recognition method based on a hybrid particle swarm algorithm, which belongs to the field of spacecraft navigation, guidance and control.

Background technique

The star sensor is the main working part of the astronomical navigation system based on the star map matching and recognition technology. The star map matching and recognition method is the core of the star sensor technology, which can quickly and accurately obtain the attitude information of the carrier. Therefore, it is of great theoretical and practical significance to study a star map recognition algorithm with fast recognition speed and high recognition success rate.

At present, there are many relatively mature star map matching recognition methods, such as triangle algorithm, convex polygon matching recognition algorithm, grid algorithm, recognition method based on Delaunay triangulation, star map recognition algorithm based on neural network, star map recognition algorithm based on genetic algorithm. Graph recognition method and star map recognition method based on ant colony algorithm, etc. These star map recognition algorithms have achieved relatively good results, but they have their own advantages and disadvantages in recognition speed, recognition success rate, database size, and robustness to noise; especially for large field of view and high sensitivity star sensors. Under these conditions, due to the large distortion of the optical system, many sensitive star points, and difficulty in extracting star point centroids, the recognition speed and success rate are significantly reduced.

Contents of the invention

The purpose of the invention is to solve the problems of slow star map recognition speed and low recognition success rate of the star sensor under the condition of large field of view. The invention proposes a star map recognition method based on a hybrid particle swarm algorithm, which significantly improves the success rate of star map recognition and the robustness to noise.

A kind of star map recognition method based on hybrid particle swarm algorithm proposed by the present invention comprises the following steps:

Step 1: Establish a navigation star library to realize the construction of the star map recognition algorithm navigation star library;

Step 2: star map matching and identification, using the hybrid particle swarm optimization algorithm to construct a feature pattern for the navigation star, and then using the binary search method to match and identify the navigation star.

The advantages of the present invention are:

(1) The star map recognition method disclosed in the present invention uses a hybrid particle swarm algorithm to optimize the path of star points, indexes according to the optimal path length value, and uses the angular distance between navigation stars in the optimal path for matching and identification, improving The recognition rate of the method is improved, and the matching time is reduced;

(2) The star map recognition method disclosed in the present invention proposes a method for adaptively determining the circle radius r, so that the optimization scale of the recognition method is within a reasonable range and the real-time performance of the method is guaranteed.

(3) The star map recognition algorithm disclosed in the present invention adds a navigation star verification link, which improves the accuracy of the recognition result and reduces redundant matching at the same time.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Fig. 2 is when the circle radius r of the present invention gets 4 °, the distribution of the number of navigation stars when the number of star points in the circle is different;

Fig. 3 is the navigation star that the present invention is less than 6 to r=4 ° of star points in the circle, when getting radius r=5 °, the star point number distribution in the circle;

Fig. 4 is that the present invention is more than 25 navigation stars to r=4 ° of star points in the circle, when getting radius r=2.5 °, the number of star points in the circle is distributed;

Fig. 5 is the star point path optimization diagram in the circle based on the hybrid particle swarm algorithm proposed by the present invention;

Fig. 6 is the process flow of the star map recognition method based on the hybrid particle swarm algorithm proposed by the present invention;

Fig. 7 is a relationship diagram between the recognition rate of the method proposed by the present invention and the star point position error.

detailed description

The present invention will be further described in detail with reference to the accompanying drawings and embodiments.

The present invention is a star map identification method based on hybrid particle swarm optimization algorithm, the method flow shown in Figure 1, including the following steps:

Step 1: Establish the navigation star library, and realize the construction of the navigation star library by the star map recognition method

For the star map identification method based on the hybrid particle swarm optimization algorithm, the construction method of the navigation star database is as follows:

Step 1: Select a navigation star and build a basic navigation star list

Select the navigation star from the basic star catalog, and eliminate the double stars whose angular distance between stars is less than a certain value (generally 3-5 pixels), and obtain the basic navigation star catalog for star map recognition. The basic navigation star catalog includes each The number, magnitude, right ascension and declination information of a navigation star.

Step 2: Construct the feature data set of each navigation star

For each navigation star in the basic navigation star list, a circle is drawn with the navigation star as the center and radius r, and all the navigation stars in the circle form the feature data set of the center navigation star.

Star map recognition is a feature matching method, which needs to construct a unique feature pattern for each navigation star, and then use the constructed features to realize the matching and recognition of the navigation star. The star map recognition method based on the hybrid particle swarm optimization algorithm needs to give the circle radius r when constructing the characteristic pattern, and the size of the circle radius r affects the number of navigation stars in the circle, and then determines the difficulty of the hybrid particle swarm optimization algorithm path optimization degree. In order to improve the accuracy and real-time performance of the star map identification method, the present invention proposes an adaptive adjustment method to determine the circle radius r, so as to ensure that the optimization scale of the hybrid particle swarm optimization algorithm is within a reasonable range, and the specific steps include:

(1) Take a navigation star as the center, take the circle radius r=4°, and count the number of star points in the circle. When the number of star points is less than 6, enter step (2). When the number of star points is greater than 25 , enter step (3), when the number of star points is greater than or equal to 6 and less than or equal to 25, determine the circle radius r=4°;

In order to ensure the accuracy and real-time performance of the star map identification method, the present invention adjusts the circle radius r of the above two situations;

(2) When the number of star points in the circle is less than 6, increase the circle radius r to 5°;

(3) When the number of star points in the circle is greater than 25, reduce the circle radius r to 2.5°.

Step 3: Select a navigation star in the basic navigation star catalog, obtain the angular distance between two stars in the feature data set, and then start from the navigation star at the center of the feature data set, use the hybrid particle swarm algorithm to quickly Path optimization, obtain the optimal path of the feature data set, obtain the length of the optimal path, and select the navigation star closest to the center of the optimal path as the direction of the optimal path, and obtain the first three navigation stars of the optimal path .

The fast path optimization steps of the hybrid particle swarm optimization algorithm are as follows:

(1) Assuming that there are m navigation stars in the feature data set, first encode the navigation stars in the feature data set, and arrange the order in which the particles traverse all the navigation stars with natural numbers as the solution of the particles; then initialize the hybrid particle swarm optimization algorithm Parameters, set the number of particles to n, and the maximum number of iterations to Nmax; then randomly generate n initial solutions (initial paths), and use the sum of the path lengths of the particles to traverse all navigation stars as the fitness function of the particles, and then according to The fitness value is calculated from the current position of the particle, and the individual extremum of each particle and the global extremum of all particles are set.

(2) Intersect the current solution of the i-th particle with the individual extremum. The specific intersecting method is: for the current solution of the particle, randomly select an intersection area (a piece of data in the solution), and then delete the solution and the individual extremum The same elements in the intersection area, and then insert the elements in the individual extreme value intersection area into the current solution of the particle to obtain a new solution for the i-th particle.

(3) Intersect the new solution obtained in step (2) with the global extremum. The specific intersecting method is: for the current solution of the particle, randomly select an intersecting area, delete the elements in the solution that are the same as the global extremum intersecting area, and then Insert the elements in the global extremum intersection area into the current solution of the particle to get the new solution of the i-th particle.

(4) The new solution obtained in step (3) is mutated. In order to minimize the sum of the path lengths, the roulette selection method is used to select the corresponding path in the path with a higher probability (the navigation star with a higher probability is easier to be selected). Two navigation stars with a large angular distance between adjacent navigation stars, and then insert other navigation stars between them to obtain a new solution after mutation.

The specific method is: obtain the angular distance between adjacent navigation stars in the new solution obtained in step (3), which are respectively recorded as l(k), k=1,2,...,m, where l(1) represents the The angular distance between the first navigation star and the second navigation star, and so on, l(m-1) represents the angular distance between the m-1th navigation star and the m-th navigation star in the solution, and l(m) represents the angular distance in the solution The angular distance between the mth and the first navigation star, then the probability of each navigation star being selected is:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}$

According to the probability obtained by the above formula, use the roulette selection method to select the navigation star i ₁ . When the roulette selection method is used, each navigation star in the solution is similar to a small sector in the roulette, and the area of the sector is proportional to the probability that the navigation star is selected. The probability of being selected is also greater.

After the navigation star i ₁ is selected, the star point closer to the star point i ₁ is selected as the next traversal point of the particle with a higher probability. Let d(i ₁ , j) represent the distance between the navigation star i ₁ and the navigation star j in the solution, then the distance of the farthest navigation star from the navigation star i ₁ can be calculated as d _max , where k=1,...,m and k≠i ₁ , in order to prevent the next navigation star after traversing the navigation star i ₁ from being i ₁ itself, let d(i ₁ ,i ₁ )=d _max , then each navigation star is The probability of being selected as the next traversal navigation star is:

$p_j=\frac{d_{\max }-d(i_1,j)}{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}(d_{\max }-d(i_1,k))},$ j=1,...,m and j≠i ₁

According to the probability obtained by the above formula, use the roulette selection method to select the star point j ₁ , that is, the star point with a high probability is more likely to be selected, and then arrange the navigation star j ₁ after the navigation star i ₁ , and keep the rest unchanged to obtain the variation Later new solution.

(5) To update the individual extremum, the specific method is: first calculate the fitness value of the new solution after the i-th particle has mutated, and compare it with the fitness value of the individual extremum of the particle to obtain the i-th particle’s fitness Then the simulated annealing mechanism is used to determine whether to accept the new solution, and if accepted, the individual extremum of the particle is updated.

Determine whether the individual extreme values of all particles have been updated, if they have been updated, go to step (6), otherwise return to step (2).

(6) Update the global extremum, determine the global extremum and the global extremum path according to the individual extremum of all particles, and judge whether the number of iterations is greater than the maximum number of iterations Nmax, if greater, the path optimization ends, and the global extremum The value path is the optimal path, otherwise, return to step (2).

Step 4: Obtain the information of the first three navigation stars in the optimal path, that is, obtain the right ascension, declination information and magnitude corresponding to the first navigation star, the second navigation star and the third navigation star respectively, Angular distance information between the first navigation star and the second navigation star, between the second navigation star and the third navigation star, and between the first navigation star and the third navigation star.

The angular distance information between the first navigation star and the second navigation star, between the second navigation star and the third navigation star, the first navigation star, the second navigation star and the third navigation star respectively correspond to The right ascension and declination information are used as the basic identification information of the navigation star.

The angular distance between the first navigation star and the third navigation star and the magnitudes of the first navigation star, the second navigation star and the third navigation star are used as verification information for star map identification.

Step 5: Repeat Step 3 and Step 4 to obtain the characteristic information of each navigation star in the navigation star catalog. The feature information includes the optimal path length of the navigation star, the basic identification information of the navigation star and the verification information of star map identification . For the navigation star identification information with a radius of r=5°, the navigation star identification information is arranged in ascending order according to the optimal path length, which constitutes the first part of the navigation star database; for the navigation star identification information with a radius r=4°, it is arranged in ascending order according to the optimal path length, and constitutes the navigation In the second part of the star library, the identification information of the navigation stars with a radius of r=2.5° is arranged in ascending order according to the optimal path length, constituting the third part of the navigation star library, and the last three parts constitute a complete navigation star library.

Step 2: Carry out star map recognition based on the hybrid particle swarm optimization algorithm;

For a star map, after star map preprocessing and star extraction, the position coordinates and gray information of each star point in the star map are obtained, and then the star map is matched and identified, as shown in Figure 6, based on the hybrid particle swarm optimization algorithm The star map identification method specifically includes the following steps:

Step 1: Determine the candidate main star for identification. The selected main star should have the characteristics of high brightness and the angular distance from the edge of the field of view is greater than the radius r of the circle. The selection method is as follows: Calculate all the angular distances from the edge of the field of view in the star map. The average gray value of the star point with radius r, and then select the star point whose gray value is greater than the average gray value as the candidate identification main star, and obtain the candidate identification main star set.

The user sets the maximum number of recognition failures as N, sets the recognition times of the method as n, and sets the initial value of n to 0.

Step 2: Perform star point fast path optimization. In the set of candidate identification main stars, first select the star point that is n+1th closest to the center of the field of view as the center of the circle (that is, for the first recognition, select the nearest star point, the first In the second recognition, select the next closest star point, and so on), and then draw a circle with radius r. The determination method of radius r is the same as that described in the second step of step 1. All the star points in the circle constitute the feature The data set is shown in Figure 5, where the white star point indicates that the gray value of the star point is greater than the average gray value.

Calculate the angular distance between two star points in the feature data set, then use the star point at the center of the feature data set as the starting point, use the hybrid particle swarm optimization algorithm to perform fast path optimization, and obtain the optimal path of the star point feature data set, Obtain the length of the optimal path, select the star point closest to the center of the circle in the optimal path as the forward direction of the optimal path, and obtain the first three star points in the optimal path.

Wherein, using the hybrid particle swarm optimization algorithm for fast path optimization is the same as the fast path optimization method of the hybrid particle swarm optimization algorithm described in the third step of step one, and the feature data set of star points is replaced by the data feature set of navigation stars.

Obtain the information of the first three star points in the optimal path, that is, obtain the magnitudes corresponding to the first star point, the second star point, and the third star point respectively, and the first star point and the second star point The angular distance d ₁₂ between, the angular distance d ₂₃ between the second star point and the third star point, the angular distance d ₁₃ between the first star point and the third star point.

The third step: Carry out star map matching and identification, according to the radius r of the star point, determine the part of the star point corresponding to the radius r in the navigation star database, and use the binary search method to find the navigation according to the optimal path length obtained in the second step The navigation stars in the star database that are close to the optimal path length of the star point are used as the candidate identification stars, and then d ₁₂ and d ₂₃ obtained in the second step are matched with the basic identification information of the candidate identification stars to obtain matching results. The matching result is no guide star match, only one guide star match or multiple guide star matches.

Step 4: Verification of the navigation star recognition result, verify the matching result obtained in the third step, the specific verification method is as follows:

(1) If only one navigation star is matched, use the angular distance d ₁₃ obtained in the second step to match with the star map identification and verification information of the navigation star. If it matches, the star map identification is successful. If there is no match , then the star map matching fails this time, and the number of star map recognition failures n increases by 1;

(2) If there is no navigation star matching, the star map matching fails this time, and the number of star map recognition failures n increases by 1;

(3) If there are multiple navigation stars matching, use the angular distance d ₁₃ obtained in the second step to match with the star map identification and verification information of multiple navigation stars to obtain further matching results. If there is no navigation star matching, the star map matching fails this time, and the number of star map recognition failures n is increased by 1; if only one navigation star matches, the star map recognition is successful; if there are still multiple navigation star matches, use In the second step of the optimal path, the star magnitudes corresponding to the first star point, the second star point and the third star point are matched with the verification information of the star map identification. If there is no navigation star matching or there are still multiple navigation stars If the star matches, it is considered that the star map has failed to match, and the number of star map recognition failures n is increased by 1, otherwise the star map is recognized successfully.

Step 5: When the number of failures n is less than the maximum number of recognition failures N, return to Step 2, otherwise, the star map matching fails, and the star map recognition ends.

Example:

When selecting the navigation star and constructing the navigation star catalog, the maximum magnitude of the navigation star is selected as 6.5, and the double stars whose inter-star angular distance is less than a certain value (take 3 pixels) are eliminated, and finally the basic navigation star catalog obtained has a total of 8996 navigation stars star. When determining the circle radius r, take the circle radius r=4°, the number of navigation stars when the number of star points in the circle takes different values approximately obeys the normal distribution, there are 404 navigation stars with the number of star points in the circle less than 6, and more than 25 There are 628 navigation stars in total, as shown in Figure 2. When the number of navigation stars in the circle is less than 6, increase the circle radius r to 5°, and the distribution of navigation stars with different numbers of star points in the circle is shown in Figure 3; when the number of navigation stars in the circle is greater than 25, reduce the circle radius r is 2.5°, and the distribution of navigation stars with different numbers of star points in the circle is shown in Figure 4.

The star map recognition method based on the hybrid particle swarm optimization algorithm first constructs a navigation star library for star map recognition according to step 1. The constructed navigation star library consists of an optimal path length table, an angular distance identification star table, and a recognition verification star table. The optimal path length table stores the length value of the optimal path, and the angular distance identification table includes the angular distance between the first navigation star and the second navigation star, the second navigation star and the third navigation star in the optimal path. The angular distance between the stars and the right ascension and declination information corresponding to the three navigation stars, the identification verification star list includes the angular distance between the first navigation star and the third navigation star in the optimal path and the forward The magnitudes of the three navigation stars and the storage contents of the three tables in the navigation star library are shown in Table 1, Table 2 and Table 3 respectively. Then, according to the second step, the matching recognition of the navigation star is realized, and the user sets the maximum value of the number of recognition failures (generally a positive integer within 4). The recognition process of the method is shown in Figure 6.

Table 1 Optimal path length table

Optimal path value (°) ... 11.030192 11.453998 11.607923 12.864802 ...

Table 2 Star map angular distance identification table

Angular distance 1 (°) Angular distance 2 (°) Right ascension 1 (°) Declination 1 (°) Right ascension 2 (°) Declination 2 (°) Right ascension 3 (°) Declination 3 (°) ... ... ... ... ... ... ... ... 2.682591 3.831108 48.137904 -57.321636 45.903517 -59.737742 53.214933 -61.016722 2.761853 5.601960 47.819796 -16.025144 47.771804 -13.263681 49.671696 -18.559589 1.184397 1.059622 213.017000 69.432614 210.460967 68.678717 207.745563 68.314947 2.760843 1.592563 144.260825 16.437953 141.385337 16.585803 140.314033 15.371067 ... ... ... ... ... ... ... ...

Table 3 Identification Verification Form

Angular distance (°) star magnitude 1 star magnitude 2 star 3 ... ... ... ... 4.516492 5.7 5.2 6.3 3.090185 6.3 6.5 5.8 2.203262 5.4 6.4 6.4 3.942676 5.9 6.3 6.5 ... ... ... ...

Applying the star map recognition method based on the hybrid particle swarm algorithm proposed by the invention can significantly improve the recognition success rate and robustness to noise under the conditions of a large field of view and a high-sensitivity star sensor. Using the star map recognition method based on the hybrid particle swarm optimization algorithm of the present invention, the traditional triangle algorithm and the improved neural network star map recognition algorithm, respectively identify 1000 randomly generated star maps, wherein the star map recognition method based on the hybrid particle swarm optimization algorithm The maximum number of identification failures selected by the method is 1, that is, only one path optimization is performed. When performing identification, the three methods are based on the following conditions: the field of view of the star sensor is 20°×20°, the size of the star sensor array is 1024×1024, and the pixel size is 15 μm×15 μm. During the simulation, the mean value of the star point position error is 0, and the variance is 0 to 2.5 pixels. The three star map recognition methods are simulated 1000 times under different position error conditions. The recognition rate of the method is shown in Figure 7.

As can be seen from Figure 7, when the variance of the star point position error is less than 2 pixels, the star map recognition method of the present invention has the highest recognition success rate, followed by the improved neural network algorithm recognition rate, and the recognition rate of the traditional triangle algorithm. lowest. When the star point position error variance is greater than 2 pixels, the recognition rate based on the present invention is initially lower than the improved neural network recognition method, but better than the traditional triangle algorithm.

For the recognition speed of the method of the present invention, on the IntelCorei3 processor 2.39GHz PC, VS2010 is used to test, the method of the present invention requires an average time of 0.15s from the construction of the feature pattern to the matching recognition of the star map. Since the feature data in the navigation star library is the optimal path value obtained by the hybrid particle swarm optimization algorithm path optimization, it is stored in ascending order, so the binary search method can be used for searching and matching. Its retrieval speed is faster than the traditional triangle algorithm, which is equivalent to the improved neural network star map recognition algorithm. star pattern recognition method.

Claims (1)

1. A star pattern recognition method based on hybrid particle swarm optimization algorithm, is characterized in that, comprises the following steps: Step 1: Establish a navigation star library Specifically include the following steps: Step 1: Select a navigation star and build a basic navigation star list Select the navigation star from the basic star catalog, and eliminate the double stars whose angular distance between stars is less than 3 to 5 pixels, and obtain the basic navigation star catalog for star map recognition. The basic navigation star catalog includes the number, star etc., right ascension and declination information; Step 2: Construct the feature data set of each navigation star For each navigation star in the basic navigation star list, draw a circle with each navigation star as the center and radius r, and form all the navigation stars in the circle into the feature data set of the center navigation star; Step 3: Select a navigation star in the basic navigation star catalog, obtain the angular distance between two stars in the feature data set, and then start from the navigation star at the center of the feature data set, use the hybrid particle swarm algorithm to quickly Path optimization, obtain the optimal path of the feature data set, obtain the length of the optimal path, and select the navigation star closest to the center of the optimal path as the direction of the optimal path, and obtain the first three navigation stars in the optimal path star; Step 4: Obtain the information of the first three navigation stars in the optimal path, that is, obtain the right ascension, declination information and magnitude corresponding to the first navigation star, the second navigation star and the third navigation star respectively, Angular distance information between the first navigation star and the second navigation star, between the second navigation star and the third navigation star, and between the first navigation star and the third navigation star; The angular distance information between the first navigation star and the second navigation star, between the second navigation star and the third navigation star, the first navigation star, the second navigation star and the third navigation star respectively correspond to Right ascension and declination information, as the basic identification information of navigation stars; The angular distance between the first navigation star and the third navigation star and the magnitudes of the first navigation star, the second navigation star and the third navigation star are used as verification information for star map identification; Step 5: Repeat Step 3 and Step 4 to obtain the characteristic information of each navigation star in the navigation star catalog. The feature information includes the optimal path length of the navigation star, the basic identification information of the navigation star and the verification of star map identification Information, the characteristic information of each navigation star is arranged in ascending order of the optimal path length, and constitutes the navigation star library; Step 2: Star map recognition based on hybrid particle swarm optimization algorithm For a star map, after star map preprocessing and star extraction, the position coordinates and grayscale information of each star point in the star map are obtained, and then the star map is matched and recognized. The specific steps of the star map recognition method based on the hybrid particle swarm optimization algorithm as follows: Step 1: Determine the candidate main star for identification. The selected main star should have the characteristics of high brightness and the angular distance from the edge of the field of view is greater than the radius r of the circle. The selection method is as follows: Calculate all the angular distances from the edge of the field of view in the star map. The average gray value of the star point of radius r, then select the star point whose gray value is greater than the average gray value as the candidate identification main star, and obtain the candidate identification main star set; The user sets the maximum value of the number of recognition failures as N, and sets the number of recognitions as n, and the initial value of n is set to 0; Step 2: Perform star point fast path optimization. In the set of candidate identification main stars, first select the star point n+1th closest to the center of the field of view as the center of the circle, and then draw a circle with radius r, and divide all the stars in the circle Points form a feature data set; Calculate the angular distance between two star points in the feature data set, then use the star point at the center of the feature data set as the starting point, use the hybrid particle swarm optimization algorithm to perform fast path optimization, and obtain the optimal path of the star point feature data set, Obtain the length of the optimal path, select the star point closest to the center of the circle in the optimal path as the forward direction of the optimal path, and obtain the first three star points in the optimal path; Obtain the information of the first three star points in the optimal path, that is, obtain the magnitudes corresponding to the first star point, the second star point, and the third star point respectively, and the first star point and the second star point The angular distance d ₁₂ between, the angular distance d ₂₃ between the second star point and the third star point, the angular distance d ₁₃ between the first star point and the third star point; Step 3: Carry out star map matching and recognition. According to the radius r ₁ of the star point, determine the part of the star point corresponding to the radius r in the navigation star library. According to the optimal path length obtained in the second step, use the binary search method to find The navigation stars in the navigation star database that are close to the optimal path length of the star point are taken as candidate identification stars, and then the d ₁₂ and d ₂₃ obtained in the second step are used to match the basic identification information of the candidate identification stars to obtain Matching result; the matching result is no navigation star match, only one navigation star match, or multiple navigation star matches; Step 4: Verification of the navigation star recognition result, verify the matching result obtained in the third step, the specific verification method is as follows: (1) If only one navigation star is matched, use the angular distance d ₁₃ obtained in the second step to match with the star map identification and verification information of the navigation star. If it matches, the star map identification is successful. If there is no match , then the star map matching fails this time, and the number of star map recognition failures n ₁ plus 1; (2) If there is no navigation star matching, the star map matching fails this time, and the number of star map recognition failures n ₁ plus 1; (3) If there are many navigation stars matching, then use the angular distance d ₁₃ obtained in the second step to match the star map identification and verification information of multiple navigation stars to obtain further matching results; if there is no navigation star matching, then If the star map matching fails this time, the number of star map recognition failures n ₁ plus 1; if only one navigation star matches, the star map recognition is successful; if there are still multiple navigation stars matching, use the second step in the optimal path The star magnitudes corresponding to the first star point, the second star point and the third star point are matched with the verification information of the star map identification. If there is no guide star match or there are still multiple guide star matches, it is considered that the star If the map fails to match, the number of star map recognition failures n ₁ plus 1, otherwise the star map recognition is successful; Step 5: When the number of failures n ₁ is less than the maximum number of identification failures N, return to step 2, otherwise, the star map matching fails, and the star map identification ends; In the second step of step one and in the first step of step two, the selection method of the radius r of the circle is specifically: (1) Take a navigation star as the center, take the circle radius r=4°, count the number of star points in the circle, when the number of star points is less than 6, enter step (2), when the number of star points is greater than 25 , enter step (3), when the number of star points is greater than or equal to 6 and less than or equal to 25, determine the circle radius r=4 °; (2) When the number of star points in the circle is less than 6, increase the circle radius r to 5°; (3) When the number of star points in the circle is greater than 25, reduce the circle radius r to 2.5°; In the third step of step one and in the second step of step two, the fast path optimization steps of the hybrid particle swarm optimization algorithm are as follows: (1) Assuming that there are m navigation stars in the feature data set, first encode the navigation stars in the feature data set, and arrange the order in which the particles traverse all the navigation stars with natural numbers as the solution of the particles; then initialize the hybrid particle swarm optimization algorithm Parameters, set the number of particles as n ₂ , and the maximum number of iterations as Nmax; then randomly generate n initial solutions, and use the sum of the path lengths of the particles to traverse all navigation stars as the fitness function of the particles; then according to the current The position calculates its fitness value, sets the individual extremum of each particle, and the global extremum of all particles; (2) Intersect the current solution of the i-th particle with the individual extremum. The specific intersecting method is: for the current solution of the particle, randomly select an intersecting area, and then delete the elements in the solution that are the same as the individual extremum intersecting area, and then Insert the elements in the individual extremum crossing area into the current solution of the particle to obtain the new solution of the i-th particle; (3) Intersect the new solution obtained in step (2) with the global extremum. The specific intersecting method is: for the current solution of the particle, randomly select an intersecting area, delete the elements in the solution that are the same as the global extremum intersecting area, and then Insert the elements in the global extremum crossing region into the current solution of the particle to obtain the new solution of the i-th particle; (4) The new solution obtained in step (3) is mutated. In order to minimize the sum of the path lengths, the roulette selection method is used to select the two adjacent navigation stars with a larger angular distance in the path with a greater probability. guide stars, and then insert other guide stars between them to obtain a new solution after mutation; The specific method is: obtain the angular distance between adjacent navigation stars in the new solution obtained in step (3), which are respectively recorded as l(k), k=1,2,...,m, where l(1) represents the The angular distance between the first navigation star and the second navigation star, and so on, l(m-1) represents the angular distance between the m-1th navigation star and the m-th navigation star in the solution, and l(m) represents the angular distance in the solution The angular distance between the mth and the first navigation star, then the probability of each navigation star being selected is: ${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}$ According to the probability obtained by the above formula, use the roulette selection method to select the navigation star i ₁ ; when the roulette selection method is used, each navigation star in the solution is similar to a small sector in the roulette, and the area of the sector is Proportional to the probability of the navigation star being selected, the larger the sector area, the greater the probability of the navigation star being selected as the next traversing navigation star; After selecting the navigation star i ₁ , select the star point closer to the star point i ₁ with a higher probability as the next traversal point of the particle; let d(i ₁ , j) represent the relationship between the navigation star i ₁ and the navigation star in the solution. j, the distance between the navigation star and the farthest navigation star i ₁ can be calculated as d _max , where d _max =m _k axd(i ₁ ,k), k=1,...,m and k≠i ₁ , in order to prevent the next navigation star after traversing navigation star i ₁ from being i ₁ itself, let d(i ₁ ,i ₁ )=d _max , then the probability of each navigation star being selected as the next traversal navigation star is: $p_j=\frac{d_{max}-d(i_1,j)}{\underset{k=1}{\overset{m}{\&Sigma;}}(d_{max}-d(i_1,k\left)\right)},j=1,...,m$ And j≠i ₁ According to the probability obtained by the above formula, use the roulette selection method to select the star point j ₁ , that is, the star point with a high probability is more likely to be selected, and then arrange the navigation star j ₁ after the navigation star i ₁ , and keep the rest unchanged to obtain the variation new solution after (5) To update the individual extremum, the specific method is: first calculate the fitness value of the new solution after the i-th particle mutates, and compare it with the fitness value of the individual extremum of the particle to obtain the fitness value of the i-th particle Then use the simulated annealing mechanism to determine whether to accept the new solution, if accepted, update the individual extremum of the particle; Determine whether the individual extreme values of all particles have been updated, if they have been updated, go to step (6), otherwise return to step (2); (6) Update the global extremum, determine the global extremum and the global extremum path according to the individual extremum of all particles, and judge whether the number of iterations is greater than the maximum number of iterations Nmax, if it is greater, the path optimization ends, and the global extremum The value path is the optimal path, otherwise, return to step (2); In the fifth step of step one, the method for constructing the navigation star library is as follows: for the navigation star identification information of radius r ₁ =5°, arrange in ascending order according to the optimal path length to form the first part of the navigation star library; for radius r ₁ = The navigation star identification information of 4° is arranged in ascending order according to the optimal path length, and constitutes the second part of the navigation star library; for the navigation star identification information of radius r ₁ =2.5°, it is arranged in ascending order of the optimal path length, and constitutes the second part of the navigation star library. Three parts, the last three parts constitute a complete navigation star library.