CN103363987A - Star map identification method of multi-view-field star sensor - Google Patents

Abstract

The invention relates to a star map identification method of a multi-view-field star sensor. The method comprises the following steps: selecting a certain quantity of fixed stars from all view fields; converting star image coordinates of the fixed stars of the other view fields to a first view field image space coordinate; then identifying by utilizing a star map identification method using a double-view-field star sensor; finally, calculating a posture of the multi-view-field star sensor and sending the posture to a navigation computer. Compared with the identification process of a conventional multi-view-field star sensor, the method only needs to carry out the identification process and the posture calculation process for one time; the data updating rate, the instantaneity and the dynamic performance of the star sensor are improved under the same condition.

Description

The method for recognising star map of a kind of many visual fields star sensor

Technical field

The present invention relates to the method for recognising star map of a kind of many visual fields star sensor.

Background technology

Star sensor is to be reference system with celestial coordinates, the high-precision attitude surveying instrument take fixed star as the detection of a target, and it provides high-precision attitude information for all kinds of spacecrafts such as satellite, deep space probes.The ultimate principle of star sensor is by optical system the some fixed stars in the field of View to be imaged on the imageing sensor photosurface, then identify the position of fixed star on celestial sphere by star Pattern Recognition Algorithm, come the attitude information of calculating aircraft according to these positions, the key problem of star sensor is importance in star map recognition, and the key index of weighing a star Pattern Recognition Algorithm is recognition speed and recognition success rate.

The method of importance in star map recognition grows up from single star identification, single star method of identification is direct matching method, this method is that a fixed star star picture in the permanent sensor visual field is directly mated with a star catalogue that is positioned at the predetermined allowance scope, late 1980s, the Bezooi jen of the JPL of the U.S. just begins to consider the global feature of star chart, the main geometric properties that importance in star map recognition is utilized be star to angular distance because the measuring accuracy of magnitude is lower, can be used as the supplemental characteristic utilization.The angular distance matching method is exactly the angular distance d (O to the direct matching method of the utilization in the star sensor visual field a pair of star chart picture that the match is successful _i, O _j) with the angular distance d (S of a pair of star catalogue star _i, S _j) mate, the angular distance matching method does not use separately as a kind of method now, but as the basis of new method for recognising star map.The triangle map method is current the most widely used recognition methods, and the advantage of this algorithm is to realize simply accounting for memory capacity little, has utilized view picture star chart feature, is not prone to the mistake coupling, can be used for the whole day soccer star and scheme identification.The fast star identification algorithm is a kind of improvement of diabolo matching method.Main selective rule and employing binary tree search procedure two aspects that comprise star group and matching characteristic of improving.Principal feature is greatly to have improved recognition speed, has reduced the Computer Storage amount.Mortari proposed to adopt the k-vector method to carry out importance in star map recognition in 1997.Because the mistake that the magnitude uncertainty is brought is identified, and this algorithm only adopts star that angular distance is identified, this algorithm has played fast the just effect of location, has reduced and has searched number of times in order to reduce.Experiment is found: along with star increases information table, matched curve y=f (x) precision reduces.And because fitting precision reduces, can not guarantee that the optimum matching star is to dropping on the subset (δ is angular travel error) between (y-δ) and (y+ δ).Therefore this algorithm has only played fast the just effect of location in regional area.Neural network is a kind of new algorithm that occurs in recent years, most of algorithms all adopt the form of direct coupling, in Guide star database, store in advance the eigenvector of fixed star, in identifying, the eigenvector of the eigenvector of measuring with storage compared, along with eigenvector ground increases, recognition time is also along with increase.Neural network is the advantage of utilizing neural network and expert system, can improve accuracy of identification, and can reduce because the error that the magnitude measuring accuracy causes, but the neural network algorithm training need is than intensive, require very large training set, to finish various modes identification, therefore has the long shortcoming of recognition time; Precision is vulnerable to train the impact of set sizes and training time length; Need potentially to store weights than large buffer memory, so higher to hardware requirement.Because the advantage of above method for recognising star map, these method for recognising star map are widely used in the monoscopic star sensor.

Yet the star sensor of monoscopic is subject to the restriction of self structure, and its roll angle precision is low, generally than crab angle and the approximately low magnitude of the angle of pitch.In order to improve the precision of star sensor roll angle, current is Star-Sensor Design a plurality of visual fields, adopt the method for data fusion to utilize the data of a plurality of visual fields to improve the precision of star sensor, many visual fields star sensor is because effective expansion of visual field, bring abundanter observation information, can further improve measuring accuracy and the functional reliability of star sensor.Current many visual fields star sensor adopts following processing procedure: at first levy and adopt respectively the method for recognising star map of monoscopic star sensor to identify to different visual fields, next adopts the Attitude Calculation of monoscopic star sensor to calculate each visual field attitude, angle relation between again pointing to according to each visual field optical axis is (in order to improve many visual fields precision of star sensor, angle is 90 degree between general each visual field), the attitude of other visual fields is transformed into attitude under the coordinate of some visual fields, adopts at last two vector attitudes to determine that method calculates many visual fields star sensor.Although the method part can be improved the attitude accuracy of star sensor, it is still processed many visual fields star sensor as the monoscopic star sensor, the key index of the current star sensors such as the sensitivity of impossible raising star sensor, dynamic property.In addition, adopt current many visual fields star sensor owing to need to all visual field fixed stars be identified respectively, therefore increased the identification number of times, the star sensor processor is had higher requirement, and because the identification number of times increases, reduced the data updating rate of star sensor.

Summary of the invention

Based on above weak point, the invention provides the method for recognising star map of a kind of many visual fields star sensor, this method has improved data updating rate, real-time and the dynamic property of star sensor.

The technology used in the present invention means are as follows: the method for recognising star map of a kind of many visual fields star sensor, and step is as follows:

Step 1: receive the fixed star star of two visual fields of star sensor as coordinate;

Step 2: utilize formula (1) calculate respectively the first visual field and the second visual field fixed star star as coordinate respectively at the image space coordinate of the first visual field and the second visual field;

$W_i=\left[\begin{array}{c}{X}_i-X_0\\ Y_i-Y_0\\ f\end{array}\right]\frac{1}{\sqrt{{(X_i-X_0)}^2+{(Y_i-Y_0)}^2+f^2}}---\left(1\right)$

Wherein:

(X _i, Y _i)---the fixed star star is as coordinate in the visual field;

(X ₀, Y ₀)---the coordinate of star sensor lens center under the picture plane;

F---the star sensor focal length of lens;

Step 3: utilize formula (2) to calculate the coordinate of image space coordinate under the first field image space coordinates of the second visual field institute any stars star picture;

$W_{12}=\left(\begin{array}{ccc}0& -1& 0\\ 0& 0& -1\\ 1& 0& 0\end{array}\right)\·W_2---\left(2\right)$

Wherein,

W ₁₂---the coordinate of image space coordinate under the first field image space coordinates of the second visual field fixed star star picture,

W ₂---the coordinate of image space coordinate under the second field image space coordinates of the second visual field fixed star star picture;

Step 4: adopt the bubble sort method that institute's any stars star picture of the first visual field and the second visual field is sorted according to the order of stellar magnitude from bright to dark;

Step 5: select three the brightest fixed star star pictures, be respectively S ₁, S ₂And S ₃Three fixed stars, the order of three fixed stars from bright to dark is: S ₁, S ₂And S ₃, difference fixed star S ₁With fixed star S ₂Star to angular distance D _S1S2, fixed star S ₁With fixed star S ₃Star to angular distance D _S1S3And fixed star S ₂With fixed star S ₃Star to angular distance D _S2S3

Step 6: calculate S ₁, S ₂And S ₃Three stars of three fixed stars to angular distance and SD ₁₂₃=D ₁₂+ D ₁₃+ D ₂₃

Step 7: adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD ₁₂₃-E, and less than or equal to SD ₁₂₃Three fixed star information of+E, this set is SD _A∈ [SD ₁₂₃-E, SD ₁₂₃+ E], wherein E is identification error;

Step 8: S set D _AIn only have the information of one group of three fixed star, identify successfully this moment; S set D _AIn the information of three fixed stars of many groups, execution in step 9 are arranged;

Step 9: choose again a fixed star, be made as S _i(i ∈ [4, n]), wherein n is two visual field fixed star summations;

Step 10: calculate respectively fixed star S _iWith fixed star S ₁, S ₂Star to angular distance D _1iAnd D _2i

Step 11: calculate this moment three stars to angular distance and SD _12i=D ₁₂+ D _1i+ D _2i

Step 12: adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD _12i-E, and less than or equal to SD _12iThree fixed star information of+E, this set is SD _i∈ [SD _12i-E, SD _12i+ E];

Step 13: get S set D _AWith S set D _iCommon factor SD=SD _A∩ SD _i

Step 14: only have the information of one group of three fixed star among the common factor SD, identify successfully this moment; Common factor SD has the information of three fixed stars of many groups, gets the 5th fixed star again, repeats repeating step 9-step 13, until all common factors only have the information of one group of three fixed star.

The features and advantages of the invention:

First: compare with many visual fields of tradition star sensor identifying, the method only need to be carried out identifying and Attitude Calculation process, under equal conditions can improve data updating rate and the real-time of star sensor.

Second: when carrying out many visual fields star sensor identifying owing to employing the method, reduced calculated amount, therefore reduced the requirement of star sensor data processing section hardware platform, thereby reduced the cost of star sensor hardware platform.

The the 3rd: in identifying, owing to only need to from each visual field, select 2-3 fixed star the brightest to identify, according to the star sensor principle of work, if be lowered into the quantity of picture fixed star in the visual field, time shutter can be reduced, make star sensor within the less time shutter, can make the fixed star star picture of fixed star sufficient amount in the visual field, therefore not only can increase the dynamic property data of star sensor, and can improve the data updating rate of star sensor.

The 4th: traditional recognition method is in order to increase the fixed star quantity in the star sensor visual field, need to increase the time shutter, and because the increase of time shutter, must increase " hangover " of fixed star star picture, be only to need " on a small quantity " fixed star in the visual field, must reduce the time shutter and adopt the method to identify, " hangover " of fixed star imaging shortened, improved and obtained the fixed star star as coordinate, thereby improved the attitude accuracy of star sensor.

Description of drawings

Fig. 1 visual field sensor general structure schematic diagram;

Fig. 2 plants double-view field star sensor method for recognising star map workflow diagram;

Fig. 3 visual field star sensor attitude error off-line curve map;

Fig. 4 plants the enforcement figure of double-view field star sensor recognition methods.

Embodiment

Embodiment 1

Many visual fields star sensor (method for recognising star map of many visual fields star sensor being described, lower same here as an example of the double-view field star sensor example) mainly is divided into three parts on system consists of: the first view field imaging components of system as directed, the second view field imaging part and data processing section.Main working process is as follows: fixed star is by the camera lens of the first visual field, be imaged in the first imageing sensor of the first visual field, the first imageing sensor drives the image that receives, send in the first image pretreatment unit, the first image pretreatment unit utilizes gravity model appoach to extract institute's persevering star as coordinate from the image that the first visual field is taken, and sends to data processing section by interface.In like manner, the second image pretreatment unit utilizes gravity model appoach to extract institute's persevering star as coordinate from the image that the second visual field is taken, and sends to data processing section by interface.Data processing section receives institute's any stars star of the first visual field and the second visual field as coordinate, adopt ranking method that fixed star star picture is sorted according to the order from bright to dark, then adopt many visual fields of the present invention star sensor method for recognising star map to identify institute's any stars star picture of the first visual field and the second visual field, adopt at last QUEST algorithm computing system structure as shown in Figure 1.

According to many visual fields star sensor Attitude Calculation principle, point to should pairwise orthogonal for optical axis between the star sensor visual field, and star sensor attitude accuracy of exporting in many visual fields of guarantee is the highest like this.Therefore the angle between double-view field star sensor optical axis points to is 90 degree.

Double-view field star sensor importance in star map recognition principle (polarity take star sensor is consistent with example with the polarity of the first visual field here, and is lower same): at first receive the fixed star star of two visual fields of star sensor as coordinate, utilize

$W_i=\left[\begin{array}{c}{X}_i-X_0\\ Y_i-Y_0\\ f\end{array}\right]\frac{1}{\sqrt{{(X_i-X_0)}^2+{(Y_i-Y_0)}^2+f^2}}---\left(1\right)$

Wherein,

(X _i, Y _i)---the fixed star star is as coordinate in the visual field;

(X ₀, Y ₀)---the coordinate of star sensor lens center under the picture plane;

F---the star sensor focal length of lens.

Calculate the first visual field and the second visual field fixed star star as coordinate respectively at the image space coordinate of the first visual field and the second visual field, utilize

$W_{12}=\left(\begin{array}{ccc}0& -1& 0\\ 0& 0& -1\\ 1& 0& 0\end{array}\right)\·W_2---\left(2\right)$

Calculate the coordinate of image space coordinate under the first field image space coordinates of the second visual field institute any stars star picture.

Wherein,

W ₁₂---the coordinate of image space coordinate under the first field image space coordinates of the second visual field fixed star star picture;

W ₂---the coordinate of image space coordinate under the second field image space coordinates of the second visual field fixed star star picture.

Adopt the bubble sort method that institute's any stars star picture of the first visual field and the second visual field is sorted according to the order of stellar magnitude from bright to dark, select three the brightest fixed star star pictures (to be respectively S ₁, S ₂And S ₃Three fixed stars, the order of three fixed stars from bright to dark is: S ₁, S ₂And S ₃), difference fixed star S ₁With fixed star S ₂Star to angular distance D _S1S2, fixed star S ₁With fixed star S ₃Star to angular distance D _S1S3And fixed star S ₂With fixed star S ₃Star to angular distance D _S2S3, calculate three stars to angular distance and SD ₁₂₃=D ₁₂+ D ₁₃+ D ₂₃, suppose that identification error is E, adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD ₁₂₃-E, and less than or equal to SD ₁₂₃Three fixed star information of+E, this set is SD _A∈ [SD ₁₂₃-E, SD ₁₂₃+ E], this moment S set D _AThe information of three fixed stars of many groups is arranged, choose again a fixed star, be made as S _i(i ∈ [4, n]), wherein n is two visual field fixed star summations, calculates respectively fixed star S _iWith fixed star S ₁, S ₂Star to angular distance D _1iAnd D _2i, calculate this moment three stars to angular distance and SD _12i=D ₁₂+ D _1i+ D _2i, adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD _12i-E, and less than or equal to SD _12iThree fixed star information of+E, this set is SD _i∈ [SD _12i-E, SD _12i+ E], get S set D _AWith S set D _iCommon factor SD=SD _A∩ SD _i, only have the information of one group of three fixed star among the common factor SD, identify successfully this moment; If common factor SD has the information of three fixed stars of many groups, choose again a fixed star, above-mentioned steps is until all common factors only have the information of one group of three fixed star.

Embodiment 2

Two visual field main performance index of star sensor:

Visual field: 6 ° * 6 °

Face battle array: 512 * 512

Survey magnitude: 4Mv

Data updating rate: 15Hz

Star sensor data processing section the key technical indexes:

Processor: TMS320VC33

Frequency of operation: 50MHz

SRAM memory size: 512K * 32

FLASH memory size: 512K * 32

We have chosen certain model double-view field star sensor and visual field star simulator more than, the model SSM-1 of simulator, before the experiment, double-view field star sensor and visual field multi-star simulator more than be placed into carry out semi-physical system in the darkroom and verify simulation results show algorithm complexity, accuracy, robustness etc.The double-view field star sensor is placed into experiment porch, the power supply that connects star sensor, star sensor is connected the RS422 serial ports to connect with host computer, before many fixed stars simulator being put into the camera lens of star sensor, many fixed stars simulator is according to orbit parameter and star sensor parameter, show in real time star chart, the star chart that shows is in real time taken in two visual fields, and from the star chart of taking, extract the fixed star star as coordinate, the fixed star star that extracts is sent to the star sensor data processing section respectively as coordinate, the fixed star star that data processing section receives two visual fields adopts the recognition methods of double-view field star sensor to identify after as coordinate, to the importance in star map recognition time of star sensor, discrimination, the leading indicators such as Attitude Calculation error are tested.

(1) the importance in star map recognition time

The random star chart of selecting the double-view field star sensor to take, according to many visual fields star sensor method for recognising star map principle, statistics adds when carrying out once many visual fields star sensor method for recognising star map, subtract, take advantage of, except and the number of times such as extracting operation, by statistics, 4680 sinusoidal computings have been carried out when carrying out once many visual fields star sensor method for recognising star map, 3120 cos operation, 4680 multiplyings, 9984 sub-addition computings, 1872 subtractions, 312 square root calculations are because processor working frequency is 50MHz, therefore recognition time is 85 milliseconds, satisfies the requirement of engineering real-time.

(2) discrimination

The double-view field star sensor is taken 6000 width of cloth star charts continuously, if identifying, star sensor successfully exports 1, if star sensor identification just exports 0, host computer is added up these results after receiving recognition result, by statistics, total star chart of input star sensor is 6000 frames, and identifying successfully is 5953 frames, discrimination is 99.22% so, satisfies the engineering use and is no less than 98% requirement.

(3) attitude error statistics

Utilize the attitude after the QUEST method is calculated identification, and send to host computer by RS422 during the attitude fructufy after the double-view field star sensor identification, upper computer software shows the attitude resultant error after receiving the attitude result at once, and preserves in real time the attitude resultant error.The continuous working of double-view field star sensor was cut off three visual field star sensor power supplys after 30 minutes, and off-line shows attitude error (mistake as shown in Figure 3! Do not find Reference source.), and add up the precision of three visual field star sensor attitudes, by statistics, star sensor, crab angle, the precision of the angle of pitch and roll angle is respectively 1.7909 " (3 σ), 1.3829 " (3 σ), 1.1727 " (3 σ).

Embodiment 3

As shown in Figure 4, it is a kind of specific embodiment of double-view field star sensor laboratory testing method, wherein the processor selection of the data processing section of star sensor TMS320VC33, the frequency of operation of processor: 50MHz, operation recognizer and preserve the SRAM memory size in star storehouse: 512K * 32, preserve the FLASH memory size in star Pattern Recognition Algorithm and star storehouse: 512K * 32, in order to compress storage space, optimize the asterisk of rear three fixed stars and star and to the data structure of the characteristic quantity database of angular distance be:

Suppose d ₁Represent the brightest fixed star of navigational triangle constellation and navigational triangle constellation time bright fixed star star to angular distance, d ₂Represent the brightest fixed star of navigational triangle constellation and the darkest fixed star star of navigational triangle constellation to angular distance, d ₃Expression navigational triangle constellation time bright fixed star and the darkest fixed star star of navigational triangle constellation are to angular distance.According to the figure place of storer, displacement storage d ₁, d ₂And d ₃, because if the storer of TMS320VC33 processor is 32, d then ₃Move to left 20, d ₂Moving to left 10 has:

Wherein

Expression moves to left.

Obviously, such storage mode is so that constellation star is relatively independent unified again to the relation of an angular distance.The most directly show and saved storage space, but prior being under identical algorithm design conditions, improved the retrieval rate of navigation constellation characteristic quantity greatly, improved the data updating rate of system.

The real-time star chart that utilizes star sensor to take is finished star as the normalization of coordinate extraction and starlight vector, magnitude structure observation triangle constellation according to observation star in the visual field, and calculate observation triangle Constellation optimization characteristic quantity DisABC ', if DisABC ' ∈ is [DisABC-E, DisABC+E] (wherein E is identification error) think that then the match is successful, if the observation triangle thinks then that through repeatedly still not finding identical navigation constellation to optimize characteristic quantity after the perturbation it fails to match.

Claims (1)

1. the method for recognising star map of visual field star sensor more than a kind is characterized in that, method is as follows:

Step 1: receive the fixed star star of two visual fields of star sensor as coordinate;

Step 2: utilize formula (1) calculate respectively the first visual field and the second visual field fixed star star as coordinate respectively at the image space coordinate of the first visual field and the second visual field;

$W_i=\left[\begin{array}{c}{X}_i-X_0\\ Y_i-Y_0\\ f\end{array}\right]\frac{1}{\sqrt{{(X_i-X_0)}^2+{(Y_i-Y_0)}^2+f^2}}---\left(1\right)$

Wherein,

(X _i, Y _i)---the fixed star star is as coordinate in the visual field;

(X ₀, Y ₀)---the coordinate of star sensor lens center under the picture plane;

F---the star sensor focal length of lens;

Step 3: utilize formula (2) to calculate the coordinate of image space coordinate under the first field image space coordinates of the second visual field institute any stars star picture;

$W_{12}=\left(\begin{array}{ccc}0& -1& 0\\ 0& 0& -1\\ 1& 0& 0\end{array}\right)\·W_2---\left(2\right)$

Wherein,

W ₁₂---the coordinate of image space coordinate under the first field image space coordinates of the second visual field fixed star star picture,

W ₂---the coordinate of image space coordinate under the second field image space coordinates of the second visual field fixed star star picture;

Step 4: adopt the bubble sort method that institute's any stars star picture of the first visual field and the second visual field is sorted according to the order of stellar magnitude from bright to dark;

Step 5: select three the brightest fixed star star pictures, be respectively S ₁, S ₂And S ₃Three fixed stars, the order of three fixed stars from bright to dark is: S ₁, S ₂And S ₃, difference fixed star S ₁With fixed star S ₂Star to angular distance D _S1S2, fixed star S ₁With fixed star S ₃Star to angular distance D _S1S3And fixed star S ₂With fixed star S ₃Star to angular distance D _S2S3

Step 6: calculate S ₁, S ₂And S ₃Three stars of three fixed stars to angular distance and SD ₁₂₃=D ₁₂+ D ₁₃+ D ₂₃

Step 7: adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD ₁₂₃-E, and less than or equal to SD ₁₂₃Three fixed star information of+E, this set is SD _A∈ [SD ₁₂₃-E, SD ₁₂₃+ E], wherein E is identification error;

Step 8: S set D _AIn only have the information of one group of three fixed star, identify successfully this moment; S set D _AIn the information of three fixed stars of many groups, execution in step 9 are arranged;

Step 9: choose again a fixed star, be made as S _i(i ∈ [4, n]), wherein n is two visual field fixed star summations;

Step 10: calculate respectively fixed star S _iWith fixed star S ₁, S ₂Star to angular distance D _1iAnd D _2i

Step 11: calculate this moment three stars to angular distance and SD _12i=D ₁₂+ D _1i+ D _2i

Step 12: adopt dichotomy searches out all three fixed stars from star catalogue star to angular distance and more than or equal to SD _12i-E, and less than or equal to SD _12iThree fixed star information of+E, this set is SD _i∈ [SD _12i-E, SD _12i+ E];

Step 13: get S set D _AWith S set D _iCommon factor SD=SD _A∩ SD _i

Step 14: only have the information of one group of three fixed star among the common factor SD, identify successfully this moment; Common factor SD has the information of three fixed stars of many groups, gets the 5th fixed star again, repeats repeating step 9-step 13, until all common factors only have the information of one group of three fixed star.

Images