CN116091804B

CN116091804B - Star suppression method based on adjacent frame configuration matching

Info

Publication number: CN116091804B
Application number: CN202310022949.0A
Authority: CN
Inventors: 郑开源; 汪夏; 邱嘉钰; 徐灿; 李智; 张雅声; 罗迪; 程文华; 方宇强; 郑洁
Original assignee: Peoples Liberation Army Strategic Support Force Aerospace Engineering University
Current assignee: Peoples Liberation Army Strategic Support Force Aerospace Engineering University
Priority date: 2023-01-08
Filing date: 2023-01-08
Publication date: 2023-07-25
Anticipated expiration: 2043-01-08
Also published as: CN116091804A

Abstract

Aiming at the problems of large distortion, dynamic distortion and insufficient priori information of a large-view telescope, the invention provides a star suppression method based on adjacent frame configuration matching, which comprises the following steps: establishing a star configuration variable index according to the stability of the star configuration; acquiring three continuous frames of images in a target star image, extracting a centroid and performing image registration; performing inter-frame primary matching on stars of adjacent frames of each centroid point after image registration by a centroid neighborhood matching method to obtain primary identification stars; based on the primary identification of the star, establishing a main satellite combination structure body, and recording the main satellite pair meeting the star configuration variable index as successful matching; and setting structural body stability evaluation indexes according to the stability of the inter-star configuration, and carrying out star recognition on the main companion combined structural body successfully matched. The invention does not need any prior information and does not need to construct a distortion model, thereby realizing light-weight, simple and convenient rapid identification, effectively inhibiting dynamic distortion and realizing accurate identification of fixed stars.

Description

Star suppression method based on adjacent frame configuration matching

Technical Field

The invention belongs to the technical field of star suppression in the technical field of aerospace, and relates to a star suppression method based on adjacent frame configuration matching.

Background

In recent years, the number of space targets has increased exponentially, while the ground observation equipment is limited, and the observation pressure has increased continuously. Thus, small-bore, large-field telescopes are increasingly being used (Research in Astronomy and Astrophysics,2022,22 (10): 105003). The method has the advantages of low cost, large field of view, strong real-time perception capability and the like, but has the defects of weak light gathering capability and large distortion degree, so that the space target star image has high ambiguity and darkness. In the space target detection, fixed stars are the largest false alarm sources, especially for a wide-field wide-area monitoring system, the fixed stars are densely distributed, the number of the fixed stars is far more than that of the space targets, and the fixed stars have large and bright spot light spots and are easy to mask the space target points. Meanwhile, partial fixed star light spots are similar to the space target in size, shape and gray level distribution, are easy to confuse and cause serious interference to space detection. Therefore, the inhibition of sidereal is an important means for improving the accuracy of space target detection. Star suppression methods based on star map recognition and star suppression methods based on successive frames are the two most common star suppression methods (Journal of Modern Optics,2018,65 (1): 85-97) for spatial object detection.

The star table matching method is mainly used for matching with star image information in a field of view by combining information of an external reference star table, so that the corresponding relation between the observation star in the field of view and stars in the star table is determined (Astronomy & Astronomyics, 2011, 527:A126). After the star is identified, the interference caused by false targets generated by the star can be effectively removed when the space targets are detected. The method has the advantages that accurate information of the position, brightness, shape and the like of the true stars can be obtained, when the observation view field is smaller, the number of stars contained in the view field is smaller, the distortion influence is small, the star image outline is clear, the method has higher identification accuracy, and the false identification condition is less. The disadvantage of the method is that the matching result is very susceptible to preprocessing results. Distortion can cause shifting of star positions and deformation of point spread functions in an image coordinate system (group-based and Airborne Instrumentation for Astronomy VII. International Society for Optics and Photonics,2018, 10702:1070232), and because most existing star map recognition algorithms are not designed for large fields of view, that is, serious distortion problems existing in large field of view observed images are not usually considered, the star map matching method is often poor in recognition effect (Journal of Modern Optics,2018,65 (1): 85-97) for the conditions of incorrect distortion correction process and inaccurate star image centroid extraction.

The star suppressing method based on continuous frames is divided into an original frame difference method and a centroid neighborhood matching method. Since stars can be considered to be at infinity and the position is constant, dense stars can be considered as a background map, rigid motion (laser and infrared, 2015,45 (1): 6) is made between frames as a whole, with this motion characteristic being used to register stars of adjacent frames. The original frame difference method is to directly differential the adjacent frame images after fixed star registration, and offset each other by utilizing the characteristic that the star images of the same fixed star are kept unchanged in the adjacent frames, so as to achieve the effect of fixed star inhibition. However, in the case of dense stars, obvious background fluctuation and serious distortion of the field of view, a great number of edge trails are easily distinguished, and the detection of a space target is greatly interfered (Chinese image and graphic school newspaper, 2013, 18 (7): 799-804). In order to solve the problem, researchers put forward a centroid neighborhood matching method, namely, firstly, extracting the centroids of all star image points, registering adjacent frames by taking the centroids as characteristic points, and then, carrying out neighborhood matching identification on the centroids of all star images of the adjacent frames based on a given threshold value, and if the matching is successful, identifying as a star. The method has the advantages of high fixed star recognition rate and suitability for dense fixed star recognition under large view field and large distortion. The disadvantage is that when there are more interference points, the method has higher omission rate in space object detection due to the lack of related prior information of sidereal, and the method has the advantages of high omission rate (China image and graph school report, 2010,15 (03): 435-442.).

The matching result of the star suppressing method based on star map recognition is extremely susceptible to the preprocessing result. Distortion can cause shifting of star positions in the image coordinate system and distortion of the point spread function (International Society for Optics and Photonics,2018, 10702:1070232). Because most of the existing star map recognition algorithms are not designed for a large field of view, that is, the serious distortion problem of the large field of view observation image is not usually considered, the star map matching method is often poor in recognition effect (Journal of Modern Optics,2018,65 (1): 85-97) for the situations of improper distortion correction processing and inaccurate star image centroid extraction. The large number of stars in a large field of view and the tedious interference points also present great difficulties for matching. Meanwhile, the accuracy of identification is closely related to the star meter capacity, so that star meters with extremely large storage capacity are often required to be carried in order to ensure higher identification rate, and higher requirements are placed on the performance of a processor.

The original frame difference method in the star suppressing method based on continuous frames is easy to distinguish very much edge trails for the conditions of denser stars, obvious background fluctuation and serious field distortion, and causes great interference to space target detection (Chinese image graphic school newspaper, 2013, 18 (7): 799-804). When more interference points exist in the centroid neighborhood matching method, the centroid neighborhood matching method is very easy to misidentify other interference points comprising the space target as stars due to the lack of related prior information of stars, and therefore the subsequent space target detection has high omission ratio.

Disclosure of Invention

In order to solve the problem of large distortion and blurred star image of the small-caliber large-view-field telescope; the invention provides a fixed star suppression method based on adjacent frame configuration matching, which is applicable to all fixed star images and is mainly applicable to large-view-field strong-distortion dense fixed star images. The method utilizes the stability of the configuration between stars, increases configuration discrimination on the basis of centroid neighborhood matching, and establishes a star configuration variable index and a configuration body stability evaluation index. Effectively inhibit dynamic distortion and achieve the aim of accurately identifying the sidereal. Meanwhile, light-weight, simple and convenient rapid identification is realized, no prior information is needed, and a distortion model is not needed to be constructed. In the semi-physical simulation experiment, the influence of distortion errors and centroid extraction errors are considered at the same time, and finally, the recognition result superior to that of a star table matching method and a centroid neighborhood matching method is obtained.

The aim of the invention is realized by the following technical scheme:

the invention discloses a star inhibiting method based on adjacent frame configuration matching, which is mainly applicable to star images with large visual field and strong distortion and dense density, and comprises the following steps:

Step one, establishing a star configuration variable index according to the stability of the star configuration;

step two, acquiring continuous three-frame images in the target star image, extracting the mass centers of the continuous three-frame images, and then carrying out image registration;

thirdly, performing inter-frame primary matching on stars of adjacent frames of all centroid points after image registration by a centroid neighborhood matching method to obtain primary identification stars;

step four, based on the primary identification star obtained by primary matching between frames, establishing a main satellite combination structural body, and recording a main satellite pair meeting star configuration variable indexes as successful matching;

setting a configuration body stability evaluation index for judging the stability of the main combination configuration body according to the stability of the configuration between stars, and carrying out star recognition on the main combination configuration body successfully matched according to the configuration body stability evaluation index.

In the first step, according to the stability of the configuration between the stars, the step of establishing the star configuration variable index specifically comprises the following steps:

the stability of the inter-star configuration is as follows: the fixed star is at infinity, the relative position is kept unchanged, and the configuration between the fixed star and the fixed star is kept unchanged;

based on the stability of the inter-star configuration, the established star configuration variable indexes are as follows:

Wherein delta _l As a distance index, the ratio of the distances of the same two stars in adjacent frames should satisfy the index; l (L) ₁₂ A distance configuration variable in an image coordinate system for the first star and the second star; l (L) ₁ ' ₂ A distance configuration variable for the next frame; θ ₁₂ The configuration of the angle between the vector of the first star and the vector of the second star and the X axis in the image coordinate systemA variable; θ ₁ ' ₂ The included angle configuration variable is the next frame; delta _θ As the phase angle index, it means that the difference between the phase angles of the same two sidereal vectors in adjacent frames should meet the index.

Further, the distance configuration variables are:

wherein l ₁₂ A distance configuration variable in an image coordinate system for the first star and the second star; (x) ₁ ,y ₁ ) Image coordinates for the first star; (x) ₂ ,y ₂ ) Image coordinates for the second star;

the included angle configuration variables are as follows:

wherein θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; (x) ₁ ,y ₁ ) Image coordinates for the first star; (x) ₂ ,y ₂ ) Is the image coordinates of the second star.

The method further comprises the step of establishing a phase angle and included angle calculation formula by combining the established star configuration variable index and the included angle configuration variable, wherein the phase angle and included angle calculation formula is used for avoiding inconvenience caused by quadrant judgment in arctangent calculation, and the phase angle and included angle calculation formula is as follows:

Wherein θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; θ ₁ ' ₂ The included angle configuration variable is the next frame;an azimuth vector representing the current frame from the first star to the second star; />Representing the azimuth vector of the next frame from the first star to the second star.

In the second step, a continuous three-frame image in the target star image is obtained, and the steps of extracting the mass centers of the continuous three-frame image and then carrying out image registration specifically comprise the following steps:

acquiring continuous three-frame images, and performing basic image processing on the three-frame images;

selecting a star set with brightness larger than a threshold value to extract a centroid, forming any three centroid points in each frame into a triangle, and calculating invariant moment of the three centroid points of the triangle;

and searching a first group of triangles meeting the conditions on the condition that invariant moment is equal or the change is smaller than a threshold value in the three adjacent frames of images, and estimating global motion parameters by using the control points to realize image registration, wherein the triangles are used as the basis for judging star control points.

In the third step, the stars of adjacent frames of all centroid points after image registration are subjected to primary matching between frames by a centroid neighborhood matching method, so as to obtain a primary step of identifying the stars, and the method specifically comprises the following steps:

For three frames of images after image registration, taking an intermediate image as a reference, calculating centroid points of all front and rear frames in the neighborhood of a preset radius range around each centroid point, and marking the centroid points as neighborhood matching points which serve as preliminary matching results of the centroid points in the front and rear frames;

if the centroid point to be identified in the preliminary matching result has a neighborhood matching point in any adjacent frame in the front and rear frames, the centroid point is preliminarily identified as a star, and the follow-up steps are participated;

if the centroid point to be identified in the preliminary matching result does not have a neighborhood matching point in the front frame and the rear frame, the centroid point is not a star and is directly removed.

In the fourth step, based on the primary identification star obtained by primary matching between frames, a primary satellite pair combination structure body is established, and the primary satellite pair meeting the star configuration variable index is recorded as a successful matching step, which specifically comprises the following steps:

traversing all primary identification stars in the current frame, taking the primary identification stars as main stars in sequence, finding out the preset number K around and taking the primary identification stars closest to the primary identification stars as the satellites of the main stars, forming preset number K main satellite pairs, and finding out corresponding main satellite pairs in the previous frame and the next frame aiming at each main satellite pair;

And respectively calculating and comparing the configuration variables of the main satellite pairs in the current frame and the main satellite pairs corresponding to the previous frame and the next frame, and recording the main satellite pairs which simultaneously meet the distance index and the phase angle index in the star configuration variable index as successful matching.

In the fifth step, the step of performing star recognition on the successfully matched main partner-combined structural body according to the structural body stability evaluation index specifically comprises the following steps:

counting the number SP of main satellite pairs successfully matched with the previous frame in the current frame _i Calculate SP _i A duty cycle in the total number of primary satellite pairs K;

counting the number SN of main satellite pairs successfully matched with the next frame in the current frame _i Calculating SN _i A duty cycle in the total number of primary satellite pairs K;

if the duty ratio meets the set structural body stability evaluation index lambda, judging that the main star is a star, and if the duty ratio does not meet the set structural body stability evaluation index lambda, judging that the main star is not the star.

Further, the judgment basis for the duty ratio to meet the set structural body stability evaluation index lambda comprises:

performing configuration stability judgment on the target to-be-identified star positioned outside the preset range of the edge of the view field by using a first judgment basis of star identification, and performing configuration stability judgment on the target to-be-identified star positioned in the preset range of the edge of the view field by using a second judgment basis of star identification; or alternatively, the first and second heat exchangers may be,

Firstly, carrying out configuration stability judgment on all target stars to be identified by using a first judgment basis of star identification, and then carrying out configuration stability judgment on the target stars to be identified which are positioned in a preset range of the edge of a visual field by using a second judgment basis of star identification;

the configuration is considered stable by meeting the first judgment basis, and the star to be identified is judged to be a star; if not, judging that the star is not the star; the configuration is considered stable by meeting the second judgment basis, and the star to be identified currently is judged to be a star; if not, judging that the star is not the star;

the first judgment basis of star identification is as follows:

the second judgment basis of star identification is as follows:

wherein (P_X) _i ,P_Y _i ) Is pixel coordinates;

the method further comprises the step of establishing a performance evaluation index of a star recognition algorithm, wherein the performance evaluation index is used for performing performance evaluation on a result of performing star recognition on the successfully matched main companion combined structural body according to the structural body stability evaluation index in the step five, and the performance evaluation index of the star recognition algorithm comprises the following steps:

identification rate:

false recognition rate:

wherein eta represents the proportion of successfully identified stars to the total stars in the field of view, M _star Representing the number of identified stars, N _star Representing the total number of stars in the field of view; mu represents the position in the identified region The mass center point, the duty ratio of the interference point, M _noise Representing the number of interfering points that are misrecognized as stars, and M represents the total number of centroid points that are identified as stars.

The invention discloses a star suppression method based on adjacent frame configuration matching, which establishes star configuration variable indexes and configuration body stability evaluation indexes, analyzes configuration bodies under the influence of projection transformation and distortion by taking an observation scene of the invention as a background, establishes index establishment rules, and finally converts the judgment of a single centroid point into the judgment of a star group structure body; extracting a distortion model based on measured data to obtain a simulation data support simulation experiment; the influence of distortion errors and centroid extraction errors is considered; extracting the mass center of each star image point as a characteristic point, and further realizing the registration of star images; a centroid neighborhood matching method based on centroid points is adopted, so that the influence of a specific star image is avoided; when telescope priori information is known, preliminary configuration variable indexes can be calculated according to specific parameters of the telescope, when telescope priori information is unknown, two looser configuration variable indexes are estimated first, the satellite number is increased appropriately, and meanwhile the requirement for judging the stability of the configuration body is improved. The method ensures that the configuration difference caused by the interframe geometric transformation and telescope system error can be contained, and simultaneously ensures lower false recognition probability; and evaluating by combining the matching conditions of three continuous frames, so that the probability of mismatching is reduced.

For different centroid positioning errors, the method has better recognition rate and lowest false recognition rate. Particularly, under the condition of small centroid positioning error, the method has obvious advantages in the identification rate and the false identification rate.

The accuracy and precision of centroid extraction largely determine the performance of the present invention.

Because the invention is obtained by adding the configuration discrimination requirement on the basis of the direct centroid neighborhood matching method, the identification rate of the sidereal is probably slightly smaller than that of the direct centroid neighborhood matching method, but the invention has the advantage of lower false identification rate compared with the direct centroid neighborhood matching method, and the false identification rate can effectively inhibit the occurrence of false alarm during the detection of a space target.

The invention does not need to load any priori information and construct a distortion model, greatly simplifies the star identification process, and is suitable for identifying large-view-field strong-distortion dense star images.

Drawings

The invention is described in further detail below with reference to the drawings and examples.

Fig. 1 is a flowchart of a star suppression method based on adjacent frame configuration matching.

Fig. 2a is a schematic illustration of a single configuration provided by the invention.

Fig. 2b is a schematic illustration of a single configuration provided by the invention.

Fig. 3 is a schematic diagram of celestial sphere projective transformation distortion provided by the invention.

Fig. 4 is a schematic diagram of the inter-frame preliminary identification in the present invention.

FIG. 5 is a schematic diagram of a configuration matching scheme in the present invention.

Fig. 6a is a schematic representation of a previous frame configuration match.

Fig. 6b is a current frame configuration matching pictorial intent.

Fig. 6c is a schematic representation of the matching of the configuration of the next frame.

Fig. 7 is a schematic diagram of a distortion fitting process based on star map identification.

Fig. 8 is a schematic diagram of the selection of distortion control points.

Fig. 9a is a schematic diagram of the result of the distortion fitting to the X-coordinate.

Fig. 9b is a schematic diagram of the result of the distortion fitting to the Y-coordinate.

Fig. 10 is a centroid positioning error schematic.

Fig. 11 is a schematic diagram of an intermediate frame simulation image.

FIG. 12 is a graphical illustration of dynamic distortion trends affected by natural environments.

FIG. 13a is a graph of recognition rate and misrecognition rate for centroid localization errors of + -0.1 pixel.

Fig. 13b is a schematic diagram of recognition rate and misrecognition rate for centroid localization errors of ±4 pixels.

Detailed Description

Example 1

As shown in fig. 1, a first embodiment of the present invention provides a star suppressing method based on adjacent frame configuration matching, which is mainly applicable to star images with large field of view and dense strong distortion, and the method includes the following steps:

In the embodiment of the invention, the centroid point is the point after the centroid of the star point is extracted.

wherein delta _l Is a distance index and represents that the same two fixed stars are in phaseThe ratio of distances in adjacent frames should meet the index; l (L) ₁₂ A distance configuration variable in an image coordinate system for the first star and the second star; l (L) ₁ ' ₂ A distance configuration variable for the next frame; θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; θ ₁ ' ₂ The included angle configuration variable is the next frame; delta _θ As a phase angle index, the phase angle difference between the pointing vectors of the same two stars in adjacent frames should meet the index;

based on the stability of the inter-frame configuration, the configuration is in a basically stable state between adjacent frames, and the configuration variable is not changed greatly, so the invention firstly predicts two looser configuration variable indexes, ensures that the configuration difference caused by the geometric transformation between frames and the telescope system error can be contained, but increases the probability of false recognition at the same time, therefore, the satellite number is properly increased, and meanwhile, the requirement of judging the stability of the configuration body is improved.

Further, the distance configuration variables are:

further, the included angle configuration variables are:

wherein θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; θ'. ₁₂ The included angle configuration variable is the next frame;an azimuth vector representing the current frame from the first star to the second star; />Representing the azimuth vector of the next frame from the first star to the second star.

In the first step, the star configuration variable index principle is established as follows:

and (3) using a stable configuration relation between stars as a matching basis, and carrying out star identification based on a method of matching adjacent frames. The constant configuration description variables to be selected are the distance between two centroid points and the phase angle of the azimuth vector between two stars in the image coordinate system, as shown in fig. 2a, and a single star pair connection forms a single configuration.

In FIG. 2a l ₁₂ Representing the distance, θ, of the first Star Star1 and the second Star Star2 in the image coordinate system ₁₂ The vector representing the direction Star1 to Star2 is at an angle to the X-axis in the image coordinate system, i.e. phase angle. The configuration variable establishment procedure is as follows.

Let Star1 and Star2 have the right and left declination of (. Alpha.) in the equatorial coordinate system of the station core ₁ ,β ₁ ) And (alpha) ₂ ,β ₂ ) Telescope visual axis direction coordinate (alpha) ₀ ,β ₀ ) Telescope roll angle phi ₀ . Can be used forAnd obtaining rectangular coordinates of the star under the camera coordinate system.

(X _C ,Y _C ,Z _C ) ^T ＝R·(X _I ,Y _I ,Z _I ) ^T (1)

Wherein (X) _I ,Y _I ,Z _I ) ^T Rectangular coordinate system representing object under earth inertial coordinate system, (X) _C ,Y _C ,Z _C ) ^T Representing rectangular coordinates of the object in the camera coordinate system. (X) _I ,Y _I ,Z _I ) ^T Obtained from the following conversion formula.

R represents a coordinate transformation matrix, and the expression is as follows.

Obtaining the barycenter point coordinate under the image coordinate system as

Where x and y represent the coordinates of the centroid point in the image coordinate system, substituted (α) ₁ ,β ₁ ) Obtain the image coordinates (x) ₁ ,y ₁ ) Substituted into (alpha) ₂ ,β ₂ ) Obtaining the image coordinates (x) ₂ ,y ₂ ) F represents the camera focal length. The two configuration variables are calculated as follows.

The sorting judgment basis is as follows:

wherein l' ₁₂ And θ' ₁₂ Representing the configuration variable, delta, of the next frame _l As a distance index, the ratio of the distances of the same two stars in adjacent frames should satisfy the index; delta _θ As the phase angle index, it means that the difference between the phase angles of the same two sidereal vectors in adjacent frames should meet the index. Separately calculating θ ₁₂ And θ' ₁₂ In order to avoid inconvenience caused by quadrant judgment in arc tangent calculation, the following changes are carried out on formulas (6) and (7) to construct a phase angle and included angle calculation formula. The angle difference, namely the included angle, is directly obtained by using a phase angle included angle calculation formula, quadrant judgment is not needed, and arccos is used for calculating to be a positive value all the time. The phase angle included angle calculation formula is as follows:

wherein, the liquid crystal display device comprises a liquid crystal display device,an azimuth vector representing the current frame pointing from Star1 to Star2,,, is->Representing the azimuth vector for the next frame pointed to by Star1 to Star 2.

Different indexes should be selected for different telescopes, different working modes, different distortion degrees and the like. The projection of the star configuration in the celestial coordinate system to the reference plane is affected by the projective transformation, and distortion to some extent occurs with the transformation of the telescope viewing angle. As shown in fig. 3, the star position is fixed on the celestial sphere, the telescope viewing angle changes as the earth rotates, the light area represents the telescope's field of view, The middle dark region represents the region where two target stars are located, V ₁ V ₂ And V' ₁ V' ₂ Representing a reference plane, w ₁ w ₂ And w' ₁ w' ₂ Representing the area of two target stars on the reference plane. Then it is easy to know that w ₁ w ₂ And w' ₁ w' ₂ Unequal, we refer to distortion due to projective transformation. Therefore, the degree of influence of projective transformation and distortion on the configuration variable is analyzed below. Since the formula derivation is extremely complex, the device according to the invention and the observation case only take the form of a value analysis.

In the application scene of the invention, the telescope is in a fixed fixation mode, and rotates only along with the rotation of the earth, the shooting interval time is 5s (including exposure time), the focal length is 220+/-3 mm, and the roll angle is 0deg. Stars at the center and corner points are selected as the calculation objects of the configuration respectively for explanation. Table 1 shows specific parameter value information.

TABLE 1

Substituting the formula (1) - (7) to obtain l ₁₂ /l' ₁₂ ＝0.99995，|θ ₁₂ -θ' ₁₂ |= 0.0012465deg. It follows that only projective transformation has a negligible effect on the configuration between stars.

The invention describes telescope distortion by using a common third-order polynomial of static geometric distortion as follows:

wherein, (x) _d ,y _d ) Representing the coordinates of the centroid point after distortion in the image coordinate system, (x) _u ,y _u ) Representing the coordinates of the centroid point in the image coordinate system without distortion. a, a ₀ ～a ₉ And b ₀ ～b ₉ Is the corresponding distortion coefficient. From equation (9), it can be initially known that the distortion isThe metamodel is a continuously varying surface, so that in the same distortion model, the same configuration can be considered approximately unchanged within a certain error range under the condition that the translation amount of adjacent frames is not large. The error is analyzed as an important reference for index establishment.

The centroid point in the center of the field of view is much less affected by distortion than the edge centroid point and therefore still uses the values of the table above for analysis. And substituting the conversion result of the formula (4) into the formula (9) for calculation, substituting the result into the formulas (5) to (7), and calculating by using a value substitution method to obtain the distance configuration variable as follows.

Using a separation constant method and adding a in a denominator ₂ The coefficients of (2) are divided into about 1.

O (p) in the molecule represents a coefficient of less than 1×10 ^-6 Is a term of (2). As can be seen from the coefficients of the above equation, the distortion has a small influence on the configuration variable for a general distortion model, and is negligible. The phase angle configuration variable can be obtained by the same method, and the calculation formula is complex and is not given here.

From the above analysis, it is possible to consider the effect of projective transformation on the star configuration for adjacent frames with shorter shooting intervals, and the configuration between stars can be considered approximately unchanged under the influence of static distortion only. For the situation that the distortion model is known a priori, the initial value of the configuration variable index can be calculated by referring to the formula (11), and then the configuration variable index is further adjusted on the basis.

acquiring continuous three-frame images, and performing basic image processing on the three-frame images; the basic image processing includes: filtering, background modeling, image segmentation and connected domain extraction.

The neighborhood radius of centroid neighborhood matches may be set as loosely as possible to reduce miss recognition. Allowing a single point to be identified in the current frame to occur where there are multiple matching points in adjacent frames.

For example, the present invention adopts centroid neighborhood matching based on centroid points. For the three frames of images after registration, taking the intermediate image as the reference, calculating the centroid points of all the front and rear frames in the neighborhood of a certain radius range around each centroid point, and taking the centroid points as the preliminary matching result of the centroid points in the front and rear frames, as shown in fig. 4, the centroid points in the neighborhood represented by the dashed line boxes in the front frame and the rear frame are identified as the preliminary matching stars of the centroid points to be identified in the current frame.

For example: for a large-view-field dense star field, a large number of interference sources exist, the star is identified by configuration matching of one star pair, and misidentification and missing identification are easy to cause, so that the invention designs a structure body of 'main and auxiliary combination', a centroid point to be identified in a current frame is taken as a main star, a certain number of auxiliary stars are combined around the main star, and a plurality of single structures form a divergent structure body, as shown in fig. 2 b. Traversing all primary identification stars in the current frame, taking the primary identification stars as main stars in sequence, finding out K surrounding primary identification stars which are closest to the primary identification stars as satellites to form K main satellite pairs, and finding out corresponding main satellite pairs in the front and rear frames according to each main satellite pair. And respectively calculating and comparing configuration variables of the main satellite pairs in the current frame and the corresponding main satellite pairs in the previous frame and the next frame, and recording the main satellite pairs meeting the distance index and the phase angle index as successful matching. And finally, respectively counting how many groups of main satellite pairs are successfully matched when the current frame is matched with the previous frame and the current frame is matched with the next frame in configuration.

the first judgment basis of star identification is as follows:

the second judgment basis of star identification is as follows:

wherein (P_X) _i ,P_Y _i ) Is pixel coordinates;

for example, in combination with the previous analysis, the configuration between stars in the image coordinate system is stable between frames, so the invention uses the stability of the configuration as a basis for discrimination to discriminate the principal star, and if the configuration x is stable between frames, the principal star can be regarded as a star. Meanwhile, as the distortion characteristic shows, the smaller the coverage range, the more stable the configuration is between frames, so that a plurality of centroid points closest to the centroid point to be identified are selected as satellites.

Recording the current point to be identified as P _i The neighborhood matching point set corresponding to the previous frame is P _i P＝{P _i p ₁ ,P _i p ₂ ,..}, the neighborhood matching point set corresponding to the next frame is P _i N＝{P _i n ₁ ,P _i n ₂ ,...}. Record P _i Corresponding K satellites are C _j (j=1..k), record satellite C _j The neighborhood matching point set corresponding to the previous frame is C _j P＝{C _j p ₁ ,C _j p ₂ ,..}, the neighborhood matching point set corresponding to the next frame is C _j N＝{C _j n ₁ ,C _j n ₂ ,...}。

Sequentially calculating each group of 'main satellite pairs' P of the current frame _i -C _i Configuration variable of (2) and corresponding 'primary satellite pair' P in adjacent frames _i P-C _j P and P _i N-C _j Configuration variables of N. For the case that a main star has a plurality of neighborhood matching points in a certain adjacent frame, the points are respectively used as the main star in the adjacent frame to guide a plurality of structural bodies together: { P _i p ₁ -C _j P,P _i p ₂ -C _j P..} and { P } _i n ₁ -C _j N,P _i n ₂ -C _j And N., judging the point to be identified as a star if at least one configuration body meets the stability evaluation index.

There are multiple neighborhood matches in adjacent frames for satelliteIn the case of points, when the configuration variables of the adjacent frames are calculated, all the points participate in calculation, which can be expressed as: { P _i p _g -C _j p ₁ ,P _i p _g -C _j p ₂ ,...}(P _i p _g ∈P _i P) and { P _i n _g -C _j n ₁ ,P _i p _g -C _j n ₂ ,...}(P _i n _g ∈P _i And N), if the configuration variable of at least one neighborhood matching point and the current main satellite meets the index, the 'main satellite pair' matching is successful. And judging that the main satellite or satellite with the neighborhood matching point does not exist as the matching failure for the main satellite or satellite with the neighborhood matching point. As shown in fig. 5, which is a schematic diagram of configuration matching of adjacent frames, light gray and black arrows represent configuration connection lines of different primary identification satellites corresponding to the main satellites and primary identification satellites of all satellites, and solid line and dotted line lines represent configurations calculated for different primary identification satellite groups in a configuration body guided by the current primary identification satellites.

Counting the number of main satellite pairs successfully matched with the previous frame in the current frame respectively, and marking the number as SP _i . Counting the number of main satellite pairs successfully matched with the next frame in the current frame, and marking the number as SN _i . And judging the stability of the configuration by combining the configuration body stability evaluation index, and further judging whether the main star in the current frame is a star or not. Assuming that the structural body stability evaluation index is lambda epsilon [0,1]The proportion of the main satellite pair to the total main satellite pair which is successfully matched in the current frame is shown. Considering that the matching result of the current frame and the previous frame or the next frame is different, the following three criteria are set for configuration stabilization:

and if one or more of the three evaluation indexes are met, the configuration is considered to be stable, namely, the star to be identified currently is judged to be a star. However, the evaluation basis has the weakness that if an interference point happens to exist at the astronomical coordinates of the same adjacent frame, the method can misidentify the star to be identified (actually the interference point) of the current frame as a star, because the star is identical to the surrounding star in configuration, fig. 6a, 6b and 6c are configuration matching diagrams of three continuous frames, fig. 6a is a configuration matching diagram of the previous frame, fig. 6b is a configuration matching diagram of the current frame, and fig. 6c is a configuration matching diagram of the next frame. In the figure, gray points are interference points, white points are fixed stars, gray circles represent stars to be identified, white circles represent surrounding selected fixed stars, and straight lines represent star pairs successfully matched. Fig. 6a, 6b and 6c illustrate the identification of an interference point in the current frame, where there happens to be an interference point at the same location in the next frame, which is in the same configuration as the satellite, and where there is no centroid point in the previous frame, which is not identified. Therefore, in order to reduce the probability of occurrence of such mismatching, the following design is required for evaluation in combination with matching situations of three consecutive frames:

or

Wherein formula (13) indicates that there are successfully matched star pairs for the currently identified star in all three consecutive frames. Equation (14) is used to process centroid points located at the edges of the field of view, where (P_X) _i ,P_Y _i ) Is the pixel coordinates. Because of the field of view interlacing, there is a non-coincident region of adjacent frames, and the stars of this partial region do not have corresponding centroid points in their adjacent frames. Therefore, the centroid point of the partial region in the current frame can be matched only with a certain frame adjacent thereto, and therefore, the expression (12) holds. By combining the influence of the motion speed, the distortion condition and the centroid extraction error of the visual axis in the application scene, the edge width of the visual field is defined as 20 pixels, and the formula (14) is obtained.

2. Method verification

1. Simulation data verification

In consideration of the fact that the real number of stars cannot be accurately known in the actually measured image, the performance of the algorithm cannot be accurately evaluated, the simulation data are utilized to test the performance of the algorithm. The method comprises the following specific steps: and the method comprises three parts of distortion model extraction based on measured data, experimental data generation, experiment and result analysis.

1. Distortion model extraction based on measured data

Since the distortion of an image photographed in practice is not only related to the lens distortion of the telescope itself but also to natural environmental factors, the simulation using only a conventional geometric distortion model cannot meet the requirements. Therefore, a distorted star map with distortion is simulated by using a distortion model obtained by fitting in the actual measurement image, and the gap between simulation data and actual measurement data is reduced. The fitting process is shown in fig. 7.

1) Star map pretreatment: firstly, basic background suppression, non-uniformity correction, filtering denoising, image segmentation and the like are carried out on a telescope real-time star map. Sub-pixel centroid localization is then performed to obtain the image coordinates of each centroid point, which can be done by a secotractor (Astronomy and Astrophysics Supplement Series,1996,117 (2)).

2) Searching the corresponding relation between the actual coordinates and the theoretical coordinates: the method for obtaining the distortion model based on the actual measured star map is generally based on the established basic polynomial model, and the residual error between the actual measured coordinate and the theoretical coordinate is fitted to obtain the coefficient of the model. The key point of obtaining the residual is to find out the corresponding relation between the actual coordinates and the theoretical coordinates of each centroid point on the image. The common method is to perform astronomical positioning on all centroid points in an image according to telescope pointing information or a navigation star matching positioning method. Then a larger angular distance threshold is set (according to the actual observation in the present invention, the angular distance threshold is set to 50 as), each centroid point in the image is matched (Chinese Astronomy and Astrophysics,2016,40 (1)) with a star in the star table by cross matching (Astronomy & Astrophysics,2011,527), and the matching is considered successful when there is and only one known star in the neighborhood matches with the measured centroid point.

3) Selecting control points: among the centroid points that match successfully, the control point is selected for fitting. The principle of control point selection is that the positioning is accurate and the distribution is uniform. The whole view field is divided into 144 equal-sized areas with 12 multiplied by 12, and the centroid point with the largest star and the complete star image is searched in the successfully matched centroid point of each area to be used as a control point, as shown in fig. 8.

The control point is a reference star and is used for calculating the subsequent global motion parameters and calculating the translation and rotation of the graphics between adjacent frames, thereby realizing the registration of the images.

4) Residual distance calculation: and converting astronomical coordinates of stars in the star table into an image coordinate system through coordinate conversion. And (3) making a difference with the corresponding measured centroid point coordinates, and taking the result as a residual error. The degree of distortion is enlarged by 10 times as follows to highlight the superiority of the present invention in processing a large distortion image.

Wherein x is _d And y _d Representing the actual coordinates of centroid points in an image, x _u And y _u And (5) representing the star theoretical coordinates obtained after coordinate conversion. Δx and Δy represent image coordinate residuals. Taking the residual error as an objective function value, and establishing a distortion model as follows:

since the degree of lens distortion is related to the telescope field size and not to the camera resolution, the input pixel coordinates are normalized with respect to the image resolution size as follows:

Obtaining

5) Distortion parameter fitting: firstly, eliminating the outliers by using the Laida criterion, and secondly, fitting the above 20 coefficients by using a least square method to obtain fitting results shown in fig. 9a and 9 b.

The crosses in the figure represent the outliers that are screened out with greater impact on the fit results. The obtained fitting model is

2. Simulation data generation

Based on the distortion model obtained above, the simulation data generation steps are as follows:

1) And screening star meters according to the set view field range, wherein the star meters are selected from common UCAC4 star meters, and stars with the star number larger than 12 are screened out according to the detection capability of the telescope.

2) And converting the screened fixed stars into an image pixel coordinate system to obtain the theoretical coordinate of each fixed star.

3) According to equation (19), distortion is added to each star, and a random error of 1% or less is added to each coefficient in the distortion model.

4) The centroid positioning error caused by the star image shape variation caused by distortion is considered. Since the present invention focuses on the processing of each centroid point after centroid extraction and does not involve the processing of a specific star image spot, in order to simplify the simulation process, the error of centroid positioning is simulated by adding random errors in a certain range to each centroid point position, thereby replacing the simulation of star image spots, as shown in fig. 10.

In the suppressing process of the star, it should be ensured that other objects than the star are not misrecognized as the star. In the experimental simulation of the invention, the algorithm performance is checked by setting the interference point. 5000 interference points consisting of unremoved noise points and space targets are randomly added, 3 frames of images are formed by the symbiosis according to the steps, and the visual axis direction and the number of stars contained in each frame are shown in table 1. Given the intermediate frame simulation image as shown in fig. 11, white dots in fig. 11 represent the fixed star actual positioning positions, and gray dots represent the added interference points. Fig. 12 shows the distortion trend of the actual observation image, and the overall rule is that the distortion of the boundary of the field of view is large and the distortion of the center is small. Unlike static geometric distortion, because of the influence of natural environment, four distortion centers exist in the field of view, and no obvious rule exists, so that great difficulty is brought to distortion correction work.

TABLE 1

3. Experiment and result analysis

Since the simulated image of the invention does not provide brightness information, three stars near the center of the field of view (less distorted) are selected directly from the known stars set on the image and registered as control points. On the basis of improving the satellite number, the index setting of corresponding configuration variables is as strict as possible based on a lower configuration body stability evaluation standard. Two algorithm performance evaluation indexes are defined respectively:

Identification rate:

false recognition rate:

wherein eta represents the proportion of successfully identified stars to the total stars in the field of view, M _star Representing the number of identified stars, N _star Representing the total number of stars in the field of view, μ representing the duty cycle of the interference point, M, in the identified centroid point _noise Representing the number of interfering points that are misrecognized as stars, and M represents the total number of centroid points that are identified as stars.

For centroid positioning errors of different degrees, respectivelyTwo experiments were performed. The first group, centroid positioning error, is set as low as possible, here with an error radius of 0.1pixel. Referring to the calculation result of the configuration variable, determining the configuration variable index as follows: (1) distance index delta _l =0.99; (2) phase angle index delta _θ =0.1 deg. The satellite number k=60 is set, and the structural body stability evaluation index λ=1/6. And the second group is provided with larger centroid positioning errors, and for the observation scene of the invention, most star images are more than 10 pixels in size, so that the centroid positioning error radius can be set to be 4 pixels, the stability of the configuration is mainly influenced by distortion and centroid positioning errors, the constraint of various configuration variable indexes is relaxed, the satellite number is increased, and the stability evaluation index requirement of the configuration body is improved. The modification index is as follows: (1) distance index delta _l =0.95; (2) phase angle index delta _θ =0.85 deg. The satellite number k=100, and the structural body stability evaluation index λ=1/5 are set.

The distance of one pixel on the image corresponds to the angular distance 9as in the field of view, according to the telescope-related parameter conversion. Four methods were compared, respectively: direct inter-frame field matching, direct star-table matching and distortion corrected star-table matching and the method of the invention. Firstly, setting the star table cross matching angular distance threshold as 50as, namely setting the centroid neighborhood matching threshold as 50/9 pixel distance, wherein the centroid positioning error radius 0.1 pixel result is shown in table 2 and the centroid positioning error radius 4pixel result is shown in table 3.

TABLE 2

TABLE 3 Table 3

The star recognition rate and the false recognition rate have close relation with the size of the neighborhood matching threshold, and the larger threshold can improve the recognition rate to a certain extent, but improve the false recognition rate. As shown in fig. 13a and 13b, the relationship between the recognition rate and the false recognition rate and the matching threshold (the matching threshold is uniformly converted and then expressed by angular distance) is given by four methods.

Compared with the results, the star inhibiting method provided by the invention has the following advantages:

1. the direct star-table matching method has the lowest recognition rate and the highest false recognition rate, so that star-table matching cannot be directly performed without correction. The star table matching method after distortion correction has lower false recognition rate, but the recognition rate is lower. The direct centroid neighborhood matching method has simple and loose judging conditions, so that the method has the highest recognition rate and the highest false recognition rate. For different centroid positioning errors, the method has better recognition rate and lowest false recognition rate. Particularly, under the condition of small centroid positioning error, the method has obvious advantages in the identification rate and the false identification rate.

2. The centroid neighborhood matching method, the direct star list matching method and the corrected star list matching method have small influence on the identification result by the centroid positioning error, and the method is greatly influenced by the centroid positioning error, so that the accuracy and precision of centroid extraction determine the performance of the method to a great extent.

3. Because the invention is obtained by adding the configuration discrimination requirement on the basis of the direct centroid neighborhood matching method, the identification rate of the sidereal is probably slightly smaller than that of the direct centroid neighborhood matching method, but the invention has the advantage of lower false identification rate compared with the direct centroid neighborhood matching method, and the false identification rate can effectively inhibit the occurrence of false alarm during the detection of a space target.

4. The invention does not need to load any priori information and construct a distortion model, greatly simplifies the star identification process, and is suitable for identifying large-view-field strong-distortion dense star images;

for further explanation of the technical solution of the present invention, two examples are now provided for illustration.

Example-algorithm Performance analysis based on semi-physical simulation

And generating simulation data based on the obtained distortion model.

Building configuration indexes, calculating specific configuration variable indexes through the formulas (1) - (11), and determining the configuration variable indexes as follows: (1) distance index delta _l =0.99; (2) phase angle index delta _θ ＝0.1deg。

The centroid positioning error radius is set to be 0.1pixel, three continuous frames of images are acquired, and basic image processing (filtering, background modeling, image segmentation and connected domain extraction) is performed. Selecting a star set with larger brightness to extract the mass center, forming a triangle by any 3 mass center points in each frame, and calculating the invariant moment of 3 points. And (3) searching a first group of triangles meeting the conditions on the condition that invariant moment is equal (or change is small) in the adjacent 3 frames, and estimating global motion parameters by using the control points to realize the registration of star images by taking the triangles as the basis for judging star control points.

And taking the intermediate image as a reference, calculating the centroid points of all the front and rear frames in the neighborhood of a certain radius range around each centroid point, and taking the centroid points as the preliminary matching result of the centroid points in the front and rear frames. If the point to be identified has a neighborhood matching point in any adjacent frame, the point is preliminarily identified as a fixed star and participates in the subsequent step, otherwise, the point is not fixed star and is directly removed.

Traversing all primary identification stars in the current frame, taking the primary identification stars as main stars in sequence, finding out 60 surrounding primary identification stars which are closest to the primary identification stars as satellites to form 60 main satellite pairs, and finding out corresponding main satellite pairs in the front and rear frames according to each main satellite pair. And respectively calculating and comparing configuration variables of the main satellite pairs in the current frame and the corresponding main satellite pairs in the previous frame and the next frame, and recording the main satellite pairs meeting the distance index and the phase angle index as successful matching.

And finally, performing fixed star identification according to the inter-frame stability of the configuration. Counting the number of main satellite pairs successfully matched with the previous frame in the current frame respectively, and marking the number as SP _i . System for managing a plurality of dataCounting the number of 'main satellite pairs' successfully matched with the next frame in the current frame, and marking as SN _i . Their respective duty cycles in the total "primary satellite pair" number are calculated. And setting a structural body stability evaluation index lambda=1/6, and judging the structural stability. If the formula (14) is satisfied, the configuration is considered stable, and the star to be identified is judged to be a star. If not, judging that the star is not the star. If the star to be identified is located within the field of view edge 20pixel, then the configuration stability determination is made using equation (15). If the formula (15) is satisfied, the configuration is considered stable, and the star to be identified currently is judged to be a star. If not, judging that the star is not the star.

And generating simulation data based on the obtained distortion model in the second example.

Building configuration indexes, calculating specific configuration variable indexes through the formulas (1) - (11), and determining the configuration variable indexes as follows: (1) distance index delta _l =0.95; (2) phase angle index delta _θ ＝0.85deg。

The centroid positioning error radius is set to be 4 pixels, three continuous frames of images are acquired, and basic image processing (filtering, background modeling, image segmentation and connected domain extraction) is performed. Selecting a star set with larger brightness to extract the mass center, forming a triangle by any 3 star image points in each frame, and calculating the invariant moment of 3 points. And (3) searching a first group of triangles meeting the conditions on the condition that the invariant moment is equal or the change is small in the adjacent 3 frames, and estimating global motion parameters by using the control points to realize the registration of star images by taking the triangles meeting the conditions as the basis for judging the star control points.

Traversing all primary identification stars in the current frame, taking the primary identification stars as main stars in turn, finding out 100 surrounding primary identification stars which are closest to the primary identification stars as satellites to form 100 main satellite pairs, and finding out corresponding main satellite pairs in the front and rear frames aiming at each main satellite pair. And respectively calculating and comparing configuration variables of the main satellite pairs in the current frame and the main satellite pairs corresponding to the previous frame and the next frame, and recording the main satellite pairs meeting the distance index and the phase angle index as successful matching.

And finally, performing fixed star identification according to the inter-frame stability of the configuration. Counting the number of main satellite pairs successfully matched with the previous frame in the current frame respectively, and marking the number as SP _i . Counting the number of main satellite pairs successfully matched with the next frame in the current frame, and marking the number as SN _i . Their respective duty cycles in the total number of primary satellite pairs are calculated. And setting a structural body stability evaluation index lambda=1/6, and judging the structural stability. If the formula (14) is satisfied, the configuration is considered stable, and the star to be identified is judged to be a star. If not, judging that the star is not the star. If the star to be identified is located within the field of view edge 20pixel, then the configuration stability determination is made using equation (15). If the formula (15) is satisfied, the configuration is considered stable, and the star to be identified currently is judged to be a star. If not, judging that the star is not the star.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. The star suppressing method based on adjacent frame configuration matching is mainly suitable for star images with large field of view and strong distortion, and is characterized by comprising the following steps:

Setting a configuration body stability evaluation index for judging the stability of the main combination configuration body according to the stability of the configuration between stars, and carrying out star recognition on the main combination configuration body successfully matched according to the configuration body stability evaluation index;

in the first step, according to the stability of the configuration between the stars, the step of establishing the star configuration variable index specifically comprises the following steps: the stability of the inter-star configuration is as follows: the fixed star is at infinity, the relative position is kept unchanged, and the configuration between the fixed star and the fixed star is kept unchanged; based on the stability of the inter-star configuration, the established star configuration variable indexes are as follows:

wherein delta _l As a distance index, the ratio of the distances of the same two stars in adjacent frames should satisfy the index; l (L) ₁₂ A distance configuration variable in an image coordinate system for the first star and the second star; l's' ₁₂ A distance configuration variable for the next frame; θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; θ ₁ ' ₂ The included angle configuration variable is the next frame; delta _θ As the phase angle index, it means that the difference between the phase angles of the same two sidereal vectors in adjacent frames should meet the index.

2. The method of claim 1, wherein,

the distance configuration variables are:

the included angle configuration variables are as follows:

3. The method of any of claims 1-2, further comprising the step of establishing a phase angle and angle calculation formula for avoiding inconveniences associated with quadrant judgment in arctangent calculation by combining the established star configuration variable index and the angle configuration variable, wherein the phase angle and angle calculation formula is:

wherein θ ₁₂ Forming a variable of an included angle between a vector of the first star and the second star and an X axis in an image coordinate system; θ ₁ ' ₂ The included angle configuration variable is the next frame;an azimuth vector representing the current frame from the first star to the second star;representing the azimuth vector of the next frame from the first star to the second star.

4. The method of claim 1, wherein in the second step, three consecutive images of the target sidereal image are acquired, and the steps of extracting centroids of the three consecutive images and then performing image registration specifically include:

5. The method of claim 1, wherein in the third step, the step of initially matching stars of adjacent frames of each centroid point after image registration by using a centroid neighborhood matching method to obtain initial identification stars specifically comprises:

6. The method of claim 1, wherein in the fourth step, based on the primary identification of stars obtained by the primary matching between frames, a primary satellite pair combination structure is established, and the primary satellite pair meeting the star configuration variable index is recorded as successful matching, specifically comprising:

7. The method of claim 6, wherein in the fifth step, the step of performing star recognition on the successfully matched primary partner-binding conformation according to the conformation stability evaluation index specifically comprises:

8. The method according to claim 7, wherein the criterion for the duty ratio to satisfy the set structural body stability evaluation index λ includes:

the first judgment basis of star identification is as follows:

the second judgment basis of star identification is as follows:

wherein (P_X) _i ,P_Y _i ) Is the pixel coordinates.

9. The method of claim 1, further comprising the step of establishing a performance evaluation index of a star recognition algorithm for performing performance evaluation on a result of star recognition on the successfully matched primary and secondary combined structural bodies according to the structural body stability evaluation index in the fifth step, wherein the performance evaluation index of the star recognition algorithm comprises:

identification rate:

false recognition rate:

wherein eta represents the proportion of successfully identified stars to the total stars in the field of view, M _star Representing the number of identified stars, N _star Representing the total number of stars in the field of view; mu represents the duty cycle of the interference point in the identified centroid point, M _noise Representing the number of interfering points that are misrecognized as stars, and M represents the total number of centroid points that are identified as stars.