CN1796939A - Method for recognising star map based on LPT transforms - Google Patents
Method for recognising star map based on LPT transforms Download PDFInfo
- Publication number
- CN1796939A CN1796939A CN 200410102588 CN200410102588A CN1796939A CN 1796939 A CN1796939 A CN 1796939A CN 200410102588 CN200410102588 CN 200410102588 CN 200410102588 A CN200410102588 A CN 200410102588A CN 1796939 A CN1796939 A CN 1796939A
- Authority
- CN
- China
- Prior art keywords
- star
- lpt
- nautical
- observation
- conversion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Navigation (AREA)
Abstract
The invention belongs to the field of spaceflight technique, relating to the improvement of the star map identification method. The steps of the invention: 1. making LPT transform on all navigation stars in a navigation star table and storing the obtained character modes after transform; 2.making LTP transform on all observing stars in an observing star map; 3. searching navigation stars matching with observing stars; 4. verifying. The invention can realize fast matching between observing star and navigation star mode; occupies less memory space; and has very strong robustness to star point position noise, star noise and other noises.
Description
Technical field
The invention belongs to the spationautics field, relate to improvement method for recognising star map.
Technical background
Star sensor is a kind of high precision that extensively adopts in the current aerospacecraft, the attitude measurement parts of high reliability, its principle of work is: CCD (perhaps CMOS) imageing sensor is taken the starry sky image in the current field range, image is through signal Processing, extract position (and brightness) information of celestial body in the observation visual field, and in navigational star table, find the correspondence coupling of observation star by star Pattern Recognition Algorithm, utilize these coupling stars to calculating the three-axis attitude of star sensor at last.Star Pattern Recognition Algorithm is a core in the star sensor.
Existing method for recognising star map can roughly be divided into two big classes: based on star to the method for angular distance (as angle of polygon apart from algorithm, maximum match group algorithm, triangle algorithm etc.) with based on the method (as grid algorithm etc.) of " star pattern ".Traditional method for recognising star map general using angular distance is as recognition feature.These method principles are simple, be easy to realize, but discrimination are not high, and recognition time is longer, can not reach the requirement of real-time.In addition, some method also needs to rely on the attitude information of priori, can not satisfy the requirement of independence.As the representative in second class methods, the grid algorithm is a kind of more outstanding method, and it has high recognition, and the capacity of recognition speed and navigational route database is also very little faster.But also there is a problem in it: under the situation of location star (need in the structural model process look for to close on star) identification error, and pattern vector that can generation error and cause recognition failures.Because the discrimination of location star is not high, therefore also influenced the discrimination of algorithm to a certain extent.
Summary of the invention
The objective of the invention is: propose a kind of method for recognising star map based on the Log-Polar conversion.Can extract the feature mode of star by conversion, and utilize this pattern to mate identification, in the hope of all obtaining excellent performance aspect minimizing memory capacity and the raising antinoise interference performance.
Technical scheme of the present invention is: a kind of method for recognising star map based on the LPT conversion, it is characterized in that,
1, all nautical stars in the navigational star table are carried out the LPT conversion, its step is as follows:
1.1, select the neighborhood R of nautical star, select any one nautical star s, direction vector with nautical star s is the optical axis sensing of star sensor, with nautical star s radius is that other nautical stars in the R neighborhood project on the imaging plane, as original image, nautical star s projects to the initial point of original image, with the neighborhood of this original image as this nautical star with the star map image that obtains like this, with the companion star of other nautical stars in this neighborhood as this nautical star, R=4 °～10 °;
1.2, all companion stars of this nautical star s are done the LPT conversion, the plane right-angle coordinate that is about to the companion star place converts the polar coordinate system of logarithm radius to, be located at that bianry image is expressed as f (x under the plane right-angle coordinate, y), be f ' (r then by the graphical representation under polar coordinate system after the LPT conversion, θ), the LPT transform definition is:
θ=tan
-1(y/x), r is the ordinate of the logarithm radius under the polar coordinate system in the formula, and θ is the horizontal ordinate (constitute owing to star map image can be regarded as by the two-dimensional discrete point, so the LPT conversion of star chart only need be carried out conversion to the pixel at companion star place) under the polar coordinate system;
1.3, determine the proper vector of nautical star s, if the image that the original image at this nautical star s place promptly projects on the imaging surface to be become size is M * N, image size after the LPT conversion is m * n, promptly resolving power is respectively 360 °/m and R/n on θ and r direction, and the bianry image after the LPT conversion is represented with the sparse matrix A of a m * n:
i＝1，...，m，j＝1，...，n
Image after the conversion to θ axial projection, is obtained the vector of a 1 * m: lpt (s)=(a
1, a
2... a
i... a
m), a
i(i=1 ..., m) be defined as follows:
If 1. for j ∈ arbitrarily (1 ..., n), A (i, j)=0, a then
i=0;
If 2. exist j ∈ (1 ..., n), make A (i, j)=1, then ai equals to make A (lpt (s) that obtains according to above definition is the proper vector of the nautical star s that conversion obtains through LPT for i, the j value of minimum j)=1;
1.4, according to the method for top 1.1.1 to 1.1.3 all nautical stars in the navigational star table are carried out the LPT conversion, obtain the proper vector lpt (s) of all nautical stars and store;
2, all the observation stars in the observation star chart are carried out the LPT conversion, the transform method of its transform method and nautical star is identical,
3, seek and observe the nautical star of star coupling, the feature mode lpt (t) of certain observation star and the feature mode lpt (s) of the nautical star in the navigational star table are compared one by one, find out nautical star with this observation star coupling, in the navigational star table, the feature mode lpt (t) of observation star t is defined as with the similarity of the feature mode lpt (s) of nautical star s:
In the formula, cs (lpt (s), v) represent lpt (s) circulation left side or the v position that moves to right, same is defined as the number that corresponding nonzero digit is complementary in two vectors, and the same value is big more, shows that the number that corresponding nonzero digit is complementary is many more, two pattern vectors are close more, sim (lpt (s), lpt (t) is big more, shows that the similarity degree of lpt (t) and lpt (s) is high more;
If satisfy following formula: sim (lpt (s), lpt (t))＞ξ, then observe star t and nautical star s coupling, wherein, ξ is a similarity threshold, the value of ξ and the number n that observes star neighborhood companion star
NeighborRelevant, suppose that allowing companion star's number (nonzero digit does not match) of mistake coupling is n
Wrong, then desirable ξ=n
Neighbor-n
Wrong, when finding 2 observation stars and nautical star coupling, stop to seek;
4, checking, the observation star that utilization is found above and the coupling of nautical star are calculated the attitude of star sensor, and then, produce the reference star chart of width of cloth simulation according to this attitude, compare with reference to star chart and observation star chart with this, if all can find corresponding observation star in the less radius of neighbourhood in the star chart in observation with reference to the star in the star chart, promptly think consistent with reference to star chart and observation star chart, then identification is correctly; Otherwise, repeat 1.3 and 1.4 step, till identification is correct.
Advantage of the present invention is:
1, the LPT conversion in the image recognition is incorporated into importance in star map recognition, utilizes the LPT conversion to extract the feature of star chart, the mode of employing string encoding and string matching realizes observing the quick coupling of star and nautical star pattern.
2, the storage space of nautical star library is less, is better than other recognizer.
3, asterism position noise and magnitude noise all had very strong robustness.
Description of drawings
Fig. 1 is a LPT conversion principle schematic.
Fig. 2 is the synoptic diagram that star chart is carried out the LPT conversion.
Fig. 3 is the synoptic diagram of the string matching in the importance in star map recognition in the inventive method.
Fig. 4 is the employed recognizer process flow diagram of the inventive method.
Fig. 5 is the crab angle of the inventive method at star sensor, and the angle of pitch and roll angle are respectively 27.50 °, the identification situation of the width of cloth star chart that simulation generates when-14.18 ° and 17.20 °.
Fig. 6 is the influence of asterism position noise to identification.
Fig. 7 is the influence of magnitude noise to identification.
Embodiment
Below the present invention is described in further details.At first, brief description is done in conversion to Log-Polar.Schwartz once proposed, and had the mapping of a kind of Log-Polar between human retina surface and visual cortex, for the important effect of the Target Recognition with size, displacement, rotational invariance.Log-Polar conversion (Log-Polar Transform, hereinafter to be referred as LPT) be a kind of a kind of conversion that plane right-angle coordinate is transformed into the polar coordinate system of logarithm radius, by mapping, size, displacement and the rotation of target is converted to single change in displacement, the problem that makes is simplified greatly.The LPT conversion has a wide range of applications in a lot of fields such as moving target identification, character recognition.
Referring to Fig. 1, be located under the plane right-angle coordinate bianry image be expressed as f (x, y), by the graphical representation under polar coordinate system after the LPT conversion be f ' (r, θ), the LPT conversion can be defined as:
θ＝tan
-1(y/x)
By the LPT conversion, the rotational transform in the original image shows as the axial circulation translation of θ (circular shift) under the polar coordinate system, and the transformation of scale in the original image shows as the axial translation of r under the polar coordinate system.Therefore, the LPT conversion can be used for extracting invariable rotary feature and constant rate feature in images match.
The observation star chart can be regarded as the star chart in a certain zone on the celestial sphere through rotation, and importance in star map recognition can be understood as the process of utilizing the rotational invariance feature to carry out images match identification to a certain extent.
Figure 2 shows that the synoptic diagram of the LPT conversion of star chart.Fig. 2 (a) is the star chart that star constituted of nautical star s vicinity and the result of LPT conversion (is true origin with s) thereof in the navigational star table; Fig. 2 (b) is the result that true origin is carried out the LPT conversion for the observation star chart with observation star t.If observation star and nautical star coupling then observes star chart should overlap with the star chart that nautical star in 2 (a) constitutes after t rotates to an angle, being reflected on the image after the LPT conversion is circulation translation on the θ direction of principal axis.It with nautical star (perhaps observing star) result that obtains after the LPT conversion at center feature mode as nautical star (perhaps observing star).
The step of method for recognising star map of the present invention is as follows:
All nautical stars in the navigational star table are carried out the LPT conversion, and its step is as follows:
1.1, select the neighborhood R of nautical star, select any one nautical star s, direction vector with nautical star s is the optical axis sensing of star sensor, with nautical star s radius is that other nautical stars in the R neighborhood project on the imaging plane, as original image, nautical star s projects to the initial point of original image, with the neighborhood of this original image as this nautical star with the star map image that obtains like this, with the companion star of other nautical stars in this neighborhood as this nautical star, R=4 °～10 °.
1.2, all companion stars of this nautical star s are done the LPT conversion, the plane right-angle coordinate that is about to the companion star place converts the polar coordinate system of logarithm radius to, be located at that bianry image is expressed as f (x under the plane right-angle coordinate, y), be f ' (r then by the graphical representation under polar coordinate system after the LPT conversion, θ), the LPT transform definition is:
θ=tan
-1(y/x), r is a logarithm under the polar coordinate system in the formula
The ordinate of radius, θ are the horizontal ordinate (constitute owing to star map image can be regarded as by the two-dimensional discrete point, so the LPT conversion of star chart only need be carried out conversion to the pixel at companion star place) under the polar coordinate system;
1.3, determine the proper vector of nautical star s, if the original image at this nautical star s place (project on the imaging surface become image) size is M * N, image size after the LPT conversion is m * n, promptly resolving power is respectively 360 °/m and R/n on θ and r direction, and the bianry image after the LPT conversion is represented with the sparse matrix A of a m * n:
Image after the conversion to θ axial projection, is obtained the vector of a 1 * m: lpt (s)=(a
1, a
2... a
i... a
m), a
i(i=1 ..., m) be defined as follows:
If 1. for j ∈ arbitrarily (1 ..., n), A (i, j)=0, a then
i=0;
If 2. exist j ∈ (1 ..., n), make A (i, j)=1, a then
i(lpt (s) that obtains according to above definition is the proper vector of the nautical star s that conversion obtains through LPT for i, the j value of minimum j)=1 to equal to make A.
1.4, according to top 1.1 to 1.3 method all nautical stars in the navigational star table are carried out the LPT conversion, obtain the proper vector lpt (s) of all nautical stars and store.
2, all the observation stars in the observation star chart are carried out the LPT conversion, the transform method of its transform method and nautical star is identical.
3, seek and observe the nautical star of star coupling, the feature mode lpt (t) of certain observation star and the feature mode lpt (s) of the nautical star in the navigational star table are compared one by one, find out nautical star with this observation star coupling, in the navigational star table, the feature mode lpt (t) of observation star t is defined as with the similarity of the feature mode lpt (s) of nautical star s:
In the formula,
Cs (lpt (s), v) represent lpt (s) circulation left side or the v position that moves to right, same is defined as the number that corresponding nonzero digit is complementary in two vectors, the same value is big more, show that the number that corresponding nonzero digit is complementary is many more, two pattern vectors are close more, sim (lpt (s), lpt (t)) big more, show that the similarity degree of lpt (t) and lpt (s) is high more.For example, for the proper vector of two nautical stars and the observation star of m=20
lpt(s)＝(0 23 0 0 54 0 10 0 0 0 21 0 0 0 0 0 0 19 0 0)
lpt(t)＝(10 0 46 0 21 00 12 0 0 0 19 0 0 0 20 0 0 54 0)
Its similarity is sim (lpt (s), lpt (t))=same (cs (lpt (s), 6), lpt (t))=4, the proper vector that the proper vector of expression nautical star circulates left after 6 with the observation star has maximum likeness in form degree, and promptly the number that is complementary of nonzero digit is 4 to the maximum.
If satisfy following formula: sim (lpt (s), lpt (t))＞ξ, then observe star t and nautical star s coupling, wherein, ξ is a similarity threshold, the value of ξ and the number n that observes star neighborhood companion star
NeighborRelevant, suppose that allowing companion star's number (nonzero digit does not match) of mistake coupling is n
Wrong, then desirable ξ=n
Neighbor-n
Wrong,
When finding 2 observation stars and nautical star coupling, stop to seek.
4, checking, the main effect of checking link is:
1. the correctness of checking identification filters out wrong coupling simultaneously.
2. find its coupling star, the precision that helps Tracking Recognition like this and improve Attitude Calculation for observation star as much as possible in the observation star chart.
3. can provide rough attitude information.
The basic thought of checking link is: if there is the identification of two stars correct, utilize these two stars calculate the attitude of star sensor should be accurately.And then the star chart (being called with reference to star chart) of the width of cloth simulation that produces according to this attitude also should consistent with original observation star chart (position of asterism be corresponding consistent).If all can find corresponding observation star in the less radius of neighbourhood in the star chart in observation with reference to the star in the star chart, promptly think consistent with reference to star chart and observation star chart, think that promptly identification is correct, the algorithm end.Otherwise, repeat 1.3 and 1.4 step, select other lower observation star of ordering successively and discern, till identification is correct.The whole algorithm process flow diagram as shown in Figure 4.
Because the feature mode of nautical star and observation star all is the vector of 1 * m, wherein except the nonzero digit of minority, numerical digit is 0 mostly, if directly be used for identification, not only can increase the capacity in pattern vector storehouse, but also can increase the time of pattern search coupling.Therefore, can carry out recompile, represent the feature mode of star with the form of character string to the pattern vector that obtains through the LPT conversion, not only can conserve storage, the speed of search is improved.Methods for Coding is:
A, when all nautical stars in the navigational star table are carried out the LPT conversion, obtain after the proper vector lpt (s) of nautical star s, proper vector to nautical star s is carried out recompile, the proper vector of representing star with the form of character string, coding method is: at first, to vectorial end, make nonzero digit become the first place of 1 * m vector zero ring shift lefts all before first nonzero digit among the lpt (s); Then, recompile obtains the model string str (s) of nautical star, and odd bits is the nonzero digit among the lpt (s), and even bit is the number of zero between the adjacent nonzero digit among the lpt (s); All nautical stars in the navigational star table are carried out the LPT conversion, obtain the model string str (s) of all nautical stars and store.As follows for sequence number in the navigational star table is the pattern vector (m=100) that 5 nautical star obtains after the LPT conversion, the character string of str (s) after for coding;
lpt(s)＝(0 0 0 0 0 0 0 0 0 0 0 0 0 0 35 0 0 0 0
0 0 53 0 0 0 0 0 0 0 0 0 44 0 0 0 0 0 52 51 0 0
0 0 0 0 0 54 48 0 0 0 0 0 0 0 0 0 0 31 0 49 0 0
0 0 0 0 0 0 0 0 0 0 0 0 53 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 29 0 0 0 0 0 0 0)
str(s)＝(35 65 39 44 5 52 0 51 7 54 0 48 10 31 1 49 14 53 16 29 21)
B, when observation all observation stars in the star chart are carried out the LPT conversion, obtain observing after the proper vector lpt (t) of star, proper vector lpt (t) to observation star t carries out recompile, the proper vector of representing star with the form of character string, coding method is: at first, to vectorial end, make nonzero digit become the first place of 1 * m vector zero ring shift lefts all before first nonzero digit among the lpt (t); Then, recompile obtains observing the model string str (t) of star, and its odd bits is the nonzero digit among the lpt (t), and even bit is the number of zero between the adjacent nonzero digit among the lpt (t).
Consider that rotation is presented as the characteristic of circulation translation after the LPT conversion, the coupling of pattern the character string behind the coding of nautical star can be prolonged 1/2, and the coded string of observation star remains unchanged for convenience.Can only need translation to get final product so that in matching process, need not circulate translation more like this.Sequence number is that the model string of 5 nautical star is through becoming after prolonging in the navigational star table as implied above
str′(s)＝(36 6 53 9 44 5 52 0 51 7 54 0 48 10 31 1 49 14 53 16 29 21 36 6 539 44 5 52 0 51 7 54 0)
C, string matching, after LPT conversion and string encoding, can regard the pattern of star as word, and a dictionary is regarded in the set of the model string of all nautical stars as in the whole navigational star table, the process of importance in star map recognition is exactly the process of searching the most close with it word in the dictionary the inside.
If model string is str
1 * m, in dictionary, search word and come down at dictionary dict={p
1, p
2..., p
NPattern p of middle searching
iMake it and str
1 * mBe complementary.A text tex formed in words all in the dictionary
1 * k, in dictionary, search word and can describe, be i.e. searching modes character string str with the string matching problem
1 * mAt text tex
1 * kThe middle position that occurs.Wherein, str (i) ∈ R, i=1,2 ..., m, tex (i) ∈ R, i=1,2 ..., k, R represents character set.Generally speaking, k＞＞m＞0.
Adopt the time complexity m (k-m-1) of method of exhaustion string matching, (search end of text (EOT)) promptly in the worst case and need carry out the inferior comparison of m (k-m-1).KMP (Knuth-Morris-Pratt) algorithm is a kind of fast algorithm commonly used in the string matching.Different with method of exhaustion, the KMP algorithm is not simply to return from text under the situation of finding certain character match failure, but makes full use of the comparison information of front.The KMP algorithm can make the complexity of string matching reduce greatly, only need carry out m+k time in the worst case relatively.
Be applied to the string matching algorithm of importance in star map recognition and distinguish to some extent, therefore can not directly the KMP algorithm is used for discerning with string matching algorithm in general sense.These differences are mainly reflected in:
1. both used character set differences, the meaning difference of character representation.The every represented meaning of common string matching is equal to, and the character string odd bits character of importance in star map recognition culminant star pattern is different with the meaning of even bit character representation, the odd bits character represent the companion star through after the LPT conversion on the r direction of principal axis coordinate figure, the axial interval of θ represented in the even bit character.For importance in star map recognition, string matching is actually the coupling of odd bits character string.As shown in Figure 3, establish observation star character string and nautical star character string with (c
1, d
1) be that starting point is mated c
2i+1And d
2i+1Coupling must satisfy simultaneously:
c
2i+1＝d
2i+1
And c
2+ c
4+ ... + c
2i=d
2+ d
4+ ... + d
2i
Obviously, here in the character string in the coupling of character and the general character string coupling definition of character be different.
2. string matching is actually a kind of fuzzy string matching in the importance in star map recognition.Here " bluring " mainly contains the implication of two aspects: at first, because the pattern of observation star is also imperfect, particularly under the situation of field of view edge, the pattern of observation star has only 1/4 to 1/2 of its corresponding nautical star pattern, the influence of adding the asterism positioning error and disturbing star, observation star generate pattern character string can not be corresponding fully one by one with its nautical star model string that is complementary; Secondly, because the existence of asterism positioning error can not come the coupling of character is defined with the principle of accurate string matching.Therefore, in order to increase the robustness of string matching, need redefine following formula:
When carrying out said star chart coupling, recommend to adopt fuzzy character string matching method, its concrete grammar is: on the basis of KMP algorithm, the model string of establishing the observation star is:
str(t)＝c
1+c
2+…+c
i，
The model string of nautical star is:
str(s)＝d
1+d
2+…+d
i，
Character match is defined as:
|c
2i+1-d
2i+1|＝1
And | (c
2+ c
4+ ... + c
2i)-(d
2+ d
4+ ... + d
2i) |=1
The present invention provides a kind of fuzzy character string identification method that is suitable for the identification of star chart feature string on the basis of KMP algorithm, it redefines the character match principle.General fuzzy string matching algorithm need be considered deletion, insertion and three kinds of situations of replacement of character, and can adopt a kind of simple relatively method in importance in star map recognition.If observation star model string can corresponding characters matched (odd bits) number in the nautical star model string and can not corresponding characters matched number be respectively n
MatchAnd n
Dismatch, n
MatchAnd n
DismatchIn matching process, bring in constant renewal in.By following the tracks of the state of these two count values, with the flow process of control identification search.If observation star model string can the corresponding odd bits character number that mates be n in the nautical star model string
Match, can not corresponding characters matched number be n
Dismatch, then the principle of string matching is:
Correct coupling: if n
MatchGreater than threshold xi, think then and found correct coupling that algorithm successfully returns; ξ and the number n of forming the neighborhood companion star who observes the star pattern
NeighborRelevant.Get ξ=n
Neighbor-2, promptly allow to occur the identification of two character mistakes.
Erroneous matching: when unmatched character is accumulated to more than 2, i.e. n
Dismatch＞2, then to think and can not find correct coupling in this position, failure is returned, and the state that returns is by the failure link decision of KMP algorithm construction.Can jump out from erroneous matching as soon as possible in this way, thus the speed of quickening identification.
The hunting zone: nautical star neighborhood companion star's to be identified number should be within the specific limits, in theory it should greater than with observation star neighborhood companion star's number and can not exceed too many (the neighborhood scope of field of view edge observation star be corresponding nautical star neighborhood scope 1/4 to 1/2).If constituting nautical star model string companion star number is m
NeighborIf, m
Neighbor＜n
Neighbor-2 or m
Neighbor＞2n
Neighbor, promptly ignore this nautical star, then next nautical star of coupling.Can be limited to a small range to the nautical star of search like this, thereby the speed of identification is improved.
Neighborhood companion star's number n
NeighborMany more, the probability that the observation star is correctly validated is high more.Simultaneously, easier being caught in of star that brightness is higher has higher confidence level.For this reason, be every observation star definition Q=M/n
Neighbor(M is a magnitude) will observe star according to the ordering of Q value, have preferential selected coupling of star of minimum Q value.
For the observation star chart, select observation star is carried out LPT change and string encoding, and discern with the method for above-mentioned string matching.If have two stars to find corresponding coupling, then can change the checking link over to.
Be illustrated in figure 5 as the star sensor crab angle, the angle of pitch and roll angle are respectively 27.50 °, the identification situation of the width of cloth star chart that simulation generates when-14.18 ° and 17.20 °.Table 1 is depicted as the information of asterism in the observation star chart.Sequence number is that 6 and 7 observation star has minimum Q value, therefore, earlier these two stars is carried out the LPT conversion line character string encoding of going forward side by side, through conversion and the model string after encoding be
str(6)＝(26 0 21 27 48 6 44 0 48 7 52 3 41 24 48 25)
str(7)＝(48 5 55 0 60 17 47 8 24 3 23 2 48 4 52 13)
Method with string matching is mated itself and the nautical star model string that stores respectively, experiment finds that sequence number in observation star 6 and the navigational star table is that the model string of 2869 nautical star has maximum similarity, its similarity is 7, sequence number is that the model string of 2870 nautical star has maximum similarity in observation star 7 and the navigational star table, and its similarity is 7.Utilize these two coupling stars to verifying, all asterisms in the reference star chart of generation and all find correspondence at the corresponding neighborhood of observation star chart, promptly consistent with reference to star chart and observation star chart, therefore, think and discern correctly.
The asterism information that star chart comprised that table 1 width of cloth generates at random
| Sequence number | The X coordinate | The Y coordinate | Companion star's number | Magnitude (brightness) | |
| 1 | 1023.67. | 992.13 | 3 | 5.9 | 1.97 |
| 2 | 980.18 | 322.73 | 4 | 5.2 | 1.30 |
| 3 | 997.85 | 455.86 | 5 | 5.7 | 1.14 |
| 4 | 774.53 | 294.55 | 5 | 5.6 | 1.12 |
| 5 | 755.40 | 325.60 | 6 | 5.5 | 0.92 |
| 6 | 748.28 | 692.57 | 8 | 5.8 | 0.73 |
| 7 | 583.56 | 332.94 | 8 | 3.6 | 0.45 |
| 8 | 608.24 | 793.53 | 5 | 4.8 | 0.96 |
| 9 | 579.57 | 833.79 | 4 | 3.9 | 0.98 |
| 10 | 387.02 | 305.71 | 5 | 5.7 | 1.14 |
| 11 | 228.68 | 497.37 | 5 | 5.9 | 1.18 |
| 12 | 157.01 | 26.29 | 2 | 5.7 | 2.75 |
| 13 | 146.22 | 6.67 | 2 | 4.2 | 2.10 |
(in the table 1 record be respectively in the observation star chart 13 observation stars star map image in the information of companion star's number, magnitude and Q value in the position, pattern radius neighborhood)
In order to assess the performance of the inventive method, adopted the star chart of simulation to carry out emulation experiment.The parameters of emulation experiment is as shown in table 2.The value of other identification parameter is in the experiment: pattern radius R=6 °, LPT transformation parameter m=100 in the identifying, n=60.Emulation experiment is mainly investigated the influence of identification with regard to asterism position noise and magnitude (brightness) noise.
Table 2 emulation experiment parameter
| Parameter | Numerical value |
| Pixel resolution | 1024×1024 |
| Pixel dimension | 12μm×12μm |
| The principal point position | (512，512) |
| The optical system focal length | 58.4536mm |
| FOV | 12°×12° |
| Can responsive maximum magnitude | 6Mv |
The position noise of asterism is mainly derived from the error of asterism center segmented positioning method.In the experiment, the star map image for simulation generates adds Gaussian distributed on true asterism position, and average is 0, the noise that standard deviation from 0 to 2 pixel changes.Its recognition result as shown in Figure 6.Statistics shows that this method has stronger antijamming capability to position noise, when σ=2 pixels, still can reach the discrimination more than 98%.
The magnitude noise has reflected the sensitivity of imageing sensor to brightness.In the experiment, in the star image simulation process, magnitude is added that the average of Gaussian distributed is 0, standard deviation is from 0 to the 1Mv noise that changes, and is reflected in to be the noise of asterism gradation of image amplitude in the star chart.Equally, get 1000 width of cloth star charts from the whole day ball at random and discern statistics, Fig. 7 is a statistics.As can be seen from the figure, from statistics as can be seen, the magnitude noise is little to the influence of discrimination, and discrimination still can reach about 99% when noise standard deviation is 1Mv, therefore luminance errors can be ignored to the influence of identification.
The inventive method carries out discerning for 1000 times average recognition time on Pentium 800M Hz PC be 15.7ms, time overhead mainly concentrates on the match search stage of character string, observe in the visual field under the many situations of star, obviously increase the working time of identification.In the LPT conversion, the average length of each model string is about 28, calculates with byte of each character, and 3360 nautical stars need 94K bytes of memory space altogether.As seen, be very little based on the method for recognising star map of LPT conversion to the demand of storage space.
Claims (2)
1, a kind of method for recognising star map based on LPT (Log-Polar Transform) conversion is characterized in that,
1.1, all nautical stars in the navigational star table are carried out the LPT conversion, its step is as follows:
1.1.1, select the neighborhood R of nautical star, select any one nautical star s, direction vector with nautical star s is the optical axis sensing of star sensor, with nautical star s radius is that other nautical stars in the R neighborhood project on the imaging plane, as original image, nautical star s projects to the initial point of original image, with the neighborhood of this original image as this nautical star with the star map image that obtains like this, with the companion star of other nautical stars in this neighborhood as this nautical star, R=4 °～10 °;
1.1.2, all companion stars of this nautical star s are done the LPT conversion, the plane right-angle coordinate that is about to the companion star place converts the polar coordinate system of logarithm radius to, be located at that bianry image is expressed as f (x under the plane right-angle coordinate, y), be f ' (r then by the graphical representation under polar coordinate system after the LPT conversion, θ), the LPT transform definition is:
θ=tan
-1(y/x), r is the ordinate of the logarithm radius under the polar coordinate system in the formula, and θ is the horizontal ordinate (constitute owing to star map image can be regarded as by the two-dimensional discrete point, so the LPT conversion of star chart only need be carried out conversion to the pixel at companion star place) under the polar coordinate system;
1.1.3, determine the proper vector of nautical star s, if the image that the original image at this nautical star s place promptly projects on the imaging surface to be become size is M * N, image size after the LPT conversion is m * n, promptly resolving power is respectively 360 °/m and R/n on θ and r direction, and the bianry image after the LPT conversion is represented with the sparse matrix A of a m * n:
Image after the conversion to θ axial projection, is obtained the vector of a 1 * m:
Lpt (s)=(a
1, a
2... a
i... a
m), a
i(i=1 ..., m), be defined as follows:
If 1. for j ∈ arbitrarily (1 ..., n), A (i, j)=0, a then
i=0;
If 2. exist j ∈ (1 ..., n), make A (i, j)=1, a then
i(lpt (s) that obtains according to above definition is the proper vector of the nautical star s that conversion obtains through LPT for i, the j value of minimum j)=1 to equal to make A;
1.1.4, according to the method for top 1.1.1 to 1.1.3 all nautical stars in the navigational star table are carried out the LPT conversion, obtain the proper vector lpt (s) of all nautical stars and store;
1.2, observation all observation stars in the star chart are carried out the LPT conversion, the transform method of its transform method and nautical star is identical,
1.3, seek the nautical star that mates with the observation star, the feature mode lpt (t) of certain observation star and the feature mode lpt (s) of the nautical star in the navigational star table are compared one by one, find out nautical star with this observation star coupling, in the navigational star table, the feature mode lpt (t) of observation star t is defined as with the similarity of the feature mode lpt (s) of nautical star s:
In the formula, cs (lpt (s), v) represent lpt (s) circulation left side or the v position that moves to right, same is defined as the number that corresponding nonzero digit is complementary in two vectors, and the same value is big more, shows that the number that corresponding nonzero digit is complementary is many more, two pattern vectors are close more, sim (lpt (s), lpt (t)) is big more, shows that the similarity degree of lpt (t) and lpt (s) is high more;
If satisfy following formula: sim (lpt (s), lpt (t))＞ξ, then observe star t and nautical star s coupling, wherein, ξ is a similarity threshold, the value of ξ and the number n that observes star neighborhood companion star
NeighborRelevant, suppose that allowing companion star's number of mistake coupling is n
Wrong, then desirable ξ=n
Neighbor-n
Wrong,
When finding 2 observation stars and nautical star coupling, stop to seek;
1.4, the checking, the observation star that utilization is found above and the coupling of nautical star are calculated the attitude of star sensor, and then, produce the reference star chart of width of cloth simulation according to this attitude, compare with reference to star chart and observation star chart with this, if all can find corresponding observation star in the less radius of neighbourhood in the star chart in observation with reference to the star in the star chart, promptly think consistent with reference to star chart and observation star chart, then identification is correctly; Otherwise, repeat 1.3 and 1.4 step, till identification is correct.
2, method for recognising star map according to claim 1 is characterized in that,
2.1, when all nautical stars in the navigational star table are carried out the LPT conversion, obtain after the proper vector lpt (s) of nautical star s, proper vector to nautical star s is carried out recompile, the proper vector of representing star with the form of character string, coding method is: at first, to vectorial end, make nonzero digit become the first place of 1 * m vector zero ring shift lefts all before first nonzero digit among the lpt (s); Then, recompile obtains the model string str (s) of nautical star, and odd bits is the nonzero digit among the lpt (s), and even bit is the number of zero between the adjacent nonzero digit among the lpt (s); All nautical stars in the navigational star table are carried out the LPT conversion, obtain the model string str (s) of all nautical stars and be stored in navigation;
2.2, when observation all observation stars in the star chart are carried out the LPT conversion, obtain observing after the proper vector lpt (t) of star, proper vector lpt (t) to observation star t carries out recompile, the proper vector of representing star with the form of character string, coding method is: at first, to vectorial end, make nonzero digit become the first place of 1 * m vector zero ring shift lefts all before first nonzero digit among the lpt (t); Then, recompile obtains observing the model string str (t) of star, and its odd bits is the nonzero digit among the lpt (t), and even bit is the number of zero between the adjacent nonzero digit among the lpt (t);
2.3, when said star chart coupling, adopt fuzzy character string matching method, its concrete grammar is: on the basis of KMP algorithm, the model string of establishing the observation star is str (t)=c
1+ c
2+ ... + c
i, the model string of nautical star is str (s)=d
1+ d
2+ ... + d
i, with character match redefine into:
|c
2i+1-d
2i+1|＝1
And | (c
2+ c
4+ ... + c
2i)-(d
2+ d
4+ ... + d
2i) |=1
If observation star model string can the corresponding odd bits character number that mates be n in the nautical star model string
Match, can not corresponding characters matched number be n
Dismatch, then the principle of string matching is:
2.3.1, correct coupling: if n
MatchGreater than threshold xi, think then and found correct coupling that algorithm successfully returns;
2.3.2, erroneous matching: when unmatched character is accumulated to more than 2, i.e. n
Dismatch＞2, think then and can not find correct coupling in this position that failure is returned;
2.3.3, the hunting zone: establish that to constitute nautical star model string companion star number be m
Neighbor, observation star model string companion star number is n
NeighborIf, m
NeighboR＜n
Neighbor-2 or m
Neighbor＞2n
Neighbor, promptly ignore this nautical star, then next nautical star of coupling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004101025888A CN100357705C (en) | 2004-12-28 | 2004-12-28 | Method for recognising star map based on LPT transforms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004101025888A CN100357705C (en) | 2004-12-28 | 2004-12-28 | Method for recognising star map based on LPT transforms |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1796939A true CN1796939A (en) | 2006-07-05 |
| CN100357705C CN100357705C (en) | 2007-12-26 |
Family
ID=36818157
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004101025888A Expired - Fee Related CN100357705C (en) | 2004-12-28 | 2004-12-28 | Method for recognising star map based on LPT transforms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100357705C (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101308024B (en) * | 2008-07-03 | 2010-10-13 | 上海交通大学 | Orbit movement target parameter estimation system based on transient relative model |
| CN103162711A (en) * | 2013-02-28 | 2013-06-19 | 北京航空航天大学 | Method and device for high-dynamic star sensor image intensifier error compensation |
| CN102114919B (en) * | 2009-12-31 | 2013-11-06 | 北京控制工程研究所 | Asteroid imaging simulator at deep space exploration transition stage |
| CN104776845A (en) * | 2015-04-29 | 2015-07-15 | 西安电子科技大学 | Autonomous star recognition method based on combination mode |
| CN106289240A (en) * | 2016-08-18 | 2017-01-04 | 中国人民解放军国防科学技术大学 | A kind of two step coupling method for recognising star map based on primary |
| CN106441280A (en) * | 2012-09-17 | 2017-02-22 | 常州工学院 | Method for screening navigation stars for star sensor |
| CN106971189A (en) * | 2017-03-31 | 2017-07-21 | 西北工业大学 | A kind of noisy method for recognising star map of low resolution |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102410844B (en) * | 2011-08-12 | 2013-12-11 | 北京航空航天大学 | Method and device for correcting non-uniformity of image of high-dynamic star sensor |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS61272612A (en) * | 1985-05-29 | 1986-12-02 | Hitachi Ltd | Star identification system of star sensor |
| FR2759474B1 (en) * | 1997-02-13 | 1999-04-30 | Centre Nat Etd Spatiales | METHOD FOR RECOGNIZING IN A PRE-ESTABLISHED CATALOG OF STARS DETECTED BY A STELLAR SENSOR |
-
2004
- 2004-12-28 CN CNB2004101025888A patent/CN100357705C/en not_active Expired - Fee Related
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101308024B (en) * | 2008-07-03 | 2010-10-13 | 上海交通大学 | Orbit movement target parameter estimation system based on transient relative model |
| CN102114919B (en) * | 2009-12-31 | 2013-11-06 | 北京控制工程研究所 | Asteroid imaging simulator at deep space exploration transition stage |
| CN106441280A (en) * | 2012-09-17 | 2017-02-22 | 常州工学院 | Method for screening navigation stars for star sensor |
| CN106441280B (en) * | 2012-09-17 | 2019-05-14 | 常州工学院 | The method of screening nautical star for star sensor |
| CN103162711A (en) * | 2013-02-28 | 2013-06-19 | 北京航空航天大学 | Method and device for high-dynamic star sensor image intensifier error compensation |
| CN103162711B (en) * | 2013-02-28 | 2015-05-27 | 北京航空航天大学 | Method and device for high-dynamic star sensor image intensifier error compensation |
| CN104776845A (en) * | 2015-04-29 | 2015-07-15 | 西安电子科技大学 | Autonomous star recognition method based on combination mode |
| CN104776845B (en) * | 2015-04-29 | 2017-05-03 | 西安电子科技大学 | Autonomous star recognition method based on combination mode |
| CN106289240A (en) * | 2016-08-18 | 2017-01-04 | 中国人民解放军国防科学技术大学 | A kind of two step coupling method for recognising star map based on primary |
| CN106971189A (en) * | 2017-03-31 | 2017-07-21 | 西北工业大学 | A kind of noisy method for recognising star map of low resolution |
| CN106971189B (en) * | 2017-03-31 | 2019-09-24 | 西北工业大学 | A kind of noisy method for recognising star map of low resolution |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100357705C (en) | 2007-12-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109345588B (en) | Tag-based six-degree-of-freedom attitude estimation method | |
| CN103136525B (en) | A kind of special-shaped Extended target high-precision locating method utilizing Generalized Hough Transform | |
| CN103295239B (en) | A kind of autoegistration method of the laser point cloud data based on datum plane image | |
| CN112070770B (en) | High-precision three-dimensional map and two-dimensional grid map synchronous construction method | |
| CN108537844B (en) | Visual SLAM loop detection method fusing geometric information | |
| CN112418129B (en) | Point cloud data processing method and device, electronic equipment and storage medium | |
| Tran et al. | On-device scalable image-based localization via prioritized cascade search and fast one-many ransac | |
| CN101782969A (en) | Reliable image characteristic matching method based on physical positioning information | |
| CN111666434A (en) | Streetscape picture retrieval method based on depth global features | |
| CN108983769B (en) | Instant positioning and map construction optimization method and device | |
| EP4246458A1 (en) | System for three-dimensional geometric guided student-teacher feature matching (3dg-stfm) | |
| CN1796939A (en) | Method for recognising star map based on LPT transforms | |
| Sun et al. | Multi-stage refinement feature matching using adaptive ORB features for robotic vision navigation | |
| CN1896681A (en) | Fast tracting method of star sensor | |
| CN110910349A (en) | Wind turbine state acquisition method based on aerial photography vision | |
| CN112102475B (en) | Space target three-dimensional sparse reconstruction method based on image sequence trajectory tracking | |
| Lusk et al. | Graffmatch: Global matching of 3d lines and planes for wide baseline lidar registration | |
| CN104964684A (en) | High-dynamic rapid star tracking method | |
| CN109840525B (en) | Extraction and matching search method of circumferential binary features | |
| CN116817887A (en) | Semantic visual SLAM map construction method, electronic equipment and storage medium | |
| CN115239902A (en) | Method, device and equipment for establishing surrounding map of mobile equipment and storage medium | |
| CN115187614A (en) | Real-time simultaneous positioning and mapping method based on STDC semantic segmentation network | |
| CN103871068A (en) | High-precision calibration method based on genetic algorithm | |
| WO2020197495A1 (en) | Method and system for feature matching | |
| Munoz et al. | Improving Place Recognition Using Dynamic Object Detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20071226 Termination date: 20111228 |