Abstract
Observation angular distance error, as the difference between the actual observation angular distance and the reference angular distance, is an important parameter that affects the identification success rate, attitude measurement accuracy, and real-time performance of a terrestrial star tracker. It is the criterion to determine whether stars are identified in star identification but is still unclarified to date. To resolve the problem, the observation angular error model is presented in this work. This model determines the variation range of the observation angular distance error by analyzing the factors of astrometric transformations. Then, the optimal angular distance matching threshold expression for a terrestrial star tracker is presented on the basis of the proposed model for the optimal efficiency in star identification. Numerical simulations and a night sky experiment demonstrate that the differences between the theoretical model, simulation and actual experiment results are less than 0.5′′ and thereby validate the reliability of our conclusions.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
As a high-precision attitude measurement device, terrestrial star trackers are widely used in aircraft, ship, and other Earth’s surface working systems. Such device can determine the real-time three-axis attitude information, such as pitch, yaw, and roll angles, by matching the observation stars obtained from the focal plane of star tracker with the reference stars in star catalog. Among the attitude determination devices, the star tracker is deemed the most accurate [1–4].
Due to the effects of the Earth’s rotation and atmospheric refraction, the attitude measurement of the terrestrial star tracker is more complicated than the traditional star sensors that are used in satellites or space vehicles. Therefore, additional studies are needed to improve the working performance of terrestrial star trackers.
Li et al. [5] proposed an analytical method to estimate the stellar instrument magnitude, characterize associated errors, and make the guide star selection efficient. Yu et al. [6,7] divided the exposure time into segments to improve attitude update rate and estimate the complete motion parameters of the star tracker. In [8,9], the suppression of stray light is implemented to achieve accurate star spot extraction. Katake and Bruccoleri introduced the intensified image detector into the star-tracker field to solve the problem of low sensitivity [10,11]. Imaging performance of a star tracker was analyzed to obtain high positional accuracy in [12–14]. Fast centroid extraction methods have been proposed to break the application limitations of new image sensors and were beneficial to the improvement of attitude updating rate [15,16]. Additional efficient star identification algorithms have been presented to improve the success rate of star identification [17–20].
However, the difference between the actual observation angular distance and the reference angular distance, named observation angular distance error, remains to be unresolved by far. It is the criterion to determine whether stars are identified in star identification, and so that affects the identification success rate, attitude measurement accuracy, and real-time performance of terrestrial star trackers. Fewer studies can be found to determine the observation angular distance error, nor have studies determined the value of angular distance matching threshold in star identification. In [2,17,18,21,22], the size of angular distance matching threshold was set as different values directly.
The determination of observation angular distance error is complex and related not only to the properties of stars, which make up angular distances, but also to various factors, such as angular distance value, revolution and rotation of the Earth, observation time, observation location, observation environment, atmospheric refraction, and optical parameters of the star tracker. The demand to clarify the observation angular distance error of the terrestrial star tracker is substantial.
In this work, an observation angular error model is presented to resolve the issue. The optimal angular matching threshold expression for the terrestrial star tracker is obtained on the basis of the proposed model. Numerical simulation and night sky experiments support the conclusions.
2. Principle of terrestrial star tracker and observation angular distance error
2.1 Principle of terrestrial star tracker
As a high-precision attitude measurement device, terrestrial star trackers are widely used in aircraft, ship, and other Earth’s surface working systems to obtain the real-time attitude information, such as pitch, yaw, and roll angles.
Figure 1(a) shows the structure of the airborne terrestrial star tracker, where, XeYeZe, XnYnZn, XbYbZb, and XsYsZs represent the terrestrial, geographical, carrier, and star tracker coordinate systems, respectively. These coordinate systems are denoted as E-frame, N-frame, B-frame, and S-frame for simplicity.
Assuming that the real-time attitude information of the aircraft (i.e., pitch, yaw, and roll angles) is denoted as $\left [ {\theta ,\psi ,\gamma } \right ]$, the real-time attitude transformation matrix between B- and N-frames can be expressed as follows [23]:
If the reference star vectors $\textbf {V}_i$ are obtained in the N-frame, the attitude transformation matrix between S- and N-frames (i.e., $C_{n}^s$ ) is obtained. Then, by using the installation matrix between B- and S-frames (i.e., $C_s^b$), the real-time attitude matrix $C_n^b$ is given by
Finally, the real-time attitude information of the Earth’s surface working systems is obtained on the basis of Eqs. (1) and (2).2.2 Observation angular distance error
As discussed in the principle of terrestrial star tracker, the corresponding relationship between the observation and the reference stars is the premise for the calculation of the real-time transformation matrix $C_\textrm {ref}^s$.
Suppose that the observation vectors of two stars $i$ and $j$ in the terrestrial star tracker are $\textbf {W}_i$ and $\textbf {W}_j$ , and their corresponding reference vectors in the star catalog are $\textbf {V}_i$ and $\textbf {V}_j$, respectively. Then, the observation angular distance $d_{i,j}^s$ and the reference angular distance $d_{i,j}^\textrm {ref}$ made up of stars $i$ and $j$ are shown respectively as follows:
At observation time $t$, assuming that the real-time observed angular distance of stars $i$ and $j$ is $d_{i,j}^s \left ( t \right )$, the real-time observation angular distance error $\varepsilon _{i,j} \left ( t \right )$ is expressed as
3. Analysis and modeling of the observation angular distance error
In this section, we aim to contribute to the modeling of observation angular distance error. Given that the observation angular distance error results from the inconsistent change of star vectors, the factors influencing the observation star vectors from the ref-frame to the S-frame are analyzed in the first step.
3.1 Factors affecting the observation star vectors
The reference star catalog used by the star tracker is derived from astronomical catalogues [25]. Astronomical catalogue provide the position of a star at a specified epoch and referred to a selected reference system. In this work, the Hipparcos catalogue is selected.
The Hipparcos catalogue stores the position parameters of nearly 120,000 stars in the International Celestial Reference System (ICRS), and detailed positional parameters of the stars in the ICRS at epoch $J2000.0$ can be obtained in [26]. Evidently, a considerable difference exists between the star vectors provided in Hipparcos catalogue and observed from the terrestrial star tracker,as shonw in Fig. 3.
The chain of astrometric transformations linking the star vector from a catalog to the Earth’s surface observation discussed in the International Astronomical Union (IAU) Standards of Fundamental Astronomy (SOFA) [27] is shown in Fig. 4.
Factors that affect the star vector in the chain of astrometric transformations include the space motion, annual parallax, solar gravitational light deflection, annual aberration, Earth nutation-precession, Earth rotation, polar motion, diurnal aberration and parallax, and atmospheric refraction. The coordinate systems based on the different observation positions include the ICRS, astronomical coordinate system (AT-frame), celestial inertial coordinate system (G-frame), and E-, N-, and S-frames.
On the basis of the Astronomical Almanac [28], the real-time star vector in the N-frame at observation time $t$ is transmitted as follows:
In view of the influence of atmospheric refraction [29], the real-time observation reference vector of the star in the S-system (i.e., $\textbf {W}_S \left ( t \right )$) can be obtained as
3.2 Analysis of observation angular distance error
The inconsistent changes of star vectors result in the observation angular distance error in real-time observation of terrestrial star trackers. In this section, we analyze the influence of the aforementioned astrometric transformation factors on the observation angular distance error.
First, the reference angular distance $d_{i,j}^\textrm {ref}$ is determined by setting the ref-frame and reference time ( $T_\textrm {ref}$) of the star catalog as ICRS and $t_0 = J2000.0$, i.e., $d_{i,j}^\textrm {ref} = d_{i,j}^\textrm {ICRS} (t_0 )$. The three-axis rotation transformations, such as Earth nutation-precession and rotation, are neglected because they are not influence the angular distance values.
Then, the factors, namely, space motion, annual parallax, solar gravitational light deflection, annual aberration, diurnal aberration, and atmospheric refraction, are analyzed as follows.
(1) Influences of space motion and annual parallax
The motion in space of stellar objects and the annual parallax are considered because they are functions of time and changes during the earth’s revolution.
Given that the corrections of space motion can be directly made to the right ascension and declination and the reduction for annual parallax refers to the barycentric coordinates of the Earth (Eqs.(7.27) and (7.37) in [28]), the observation angular distance error caused by space motion and the annual parallax should be
(2) Influence of solar gravitational light deflection
When the stellar light passes through the gravitational field of Sun, the star direction will be changed by the Solar gravitational light deflection. According to Einstein’s general relativity theory, the solar gravitational light deflection is estimated as follows:
When two stars $i$ and $j$ make up an angular distance, the observation angular distance error caused by solar gravitational light deflection should be
Given that the barycentric coordinate of the Earth changes in annual periodic as shown in Fig. 6(a) (NASA Jet Propulsion Laboratory (JPL) DE405 [30] or [31]), $r_i \left ( t \right )$ will change with the Earth’s revolution from Sun’s radius ($R_\textrm {Sun}$) to $E$ in annual periodicin of each year. When $r_i \left ( t \right ) = R_\textrm {Sun}$, $\textrm {asin}\left ( {r_i \left ( t \right )/E} \right ) = 0.25^\circ$, and $\Delta \phi _i$ reaches its maximum value.
If the star tracker is used in daytime, stars will satisfy the case $r_i \left ( t \right ) = R_\textrm {Sun}$ once in each year with the Earth’s revolution. Therefore, the observation angular distance error caused by solar gravitational light deflection should be
On the other hand, if terrestrial star trackers only work at night (as star3 and star4 in Fig. 5(b)), the observation angular distance error caused by solar gravitational light deflection for such trackers should be
(3) Influence of annual aberration
On the basis of special relativity, the apparent direction of a star from a moving observer is changed with the Earth’s motion around the Sun, named annual aberration. The annual aberration can be calculated as
Assuming that the star vectors of $i$ and $j$ are ${{\textbf {OS}}_i}$ and ${\textbf {OS}}_\textbf{j}$, respectively, and the vector resulting from annual aberration is $\textbf {A}_\textbf {E}$, the variation of the angular distance can be approximately expressed as:
Since the annual aberration is influenced by the speed direction of Earth, the value of $\left | {\varepsilon _{i,j}^{\textrm {Annual}\_\textrm {abe}} \left ( t \right )} \right |$ changes with the Earth’s revolution. For instance, based on Eq. (16), the angular distance errors of values, $\alpha = 15.0089^\circ$, $14.9996^\circ$, $10.0059^\circ$, and $4.99942^\circ$ changed in the ranges $[0,5.3522'']$, $[0,5.3489'']$, $[0,3.5738'']$, and $[0,1.7855'']$, respectively, as shown in Fig. 8(a).
(4) Influence of diurnal aberration
Given that the terrestrial star tracker is used in the Earth’s surface working systems, in addition to the speed of the carrier, the Earth’s rotation also provides a speed to the star tracker. These two speed components results in an aberration called diurnal aberration, as follows:
$\textbf {v}_\textrm {Observe}$ is usually considerably smaller than $\textbf {v}_\textrm {Earth}$; thus, regardless of $\textbf {v}_\textrm {Observe}$, the observation angular distance error caused by diurnal aberration should be
The value of $\left | {\varepsilon _{i,j}^{\textrm{Diurnal}\_ \textrm{abe}} (t)} \right |$ changes with the Earth’s rotation. Based on Eq. (18), the variation range of $\left | {\varepsilon _{i,j}^{\textrm{Diurnal}\_ \textrm{abe}} (t)} \right |$ for the angular distance values can be obtained, as shown in Fig. 8(b).
(5) Influence of atmospheric refraction
When a stellar light passes through the atmosphere, its direction will be changed due to the influence of the Earth’s atmosphere (Fig. 9(a)). The value of refraction angle can be estimated on the basis of the atmospheric refraction model [29] as follows:
The atmospheric refraction model is influenced by pressure $P_a$, temperature $T_c$, relative humidity $r_h$, and wavelength $w_l$ [27,32], as shown as follows:
As the $\left | {\varepsilon _{i,j}^{\textrm {Atmo}\_ \textrm {ref }} \left ( t \right )} \right |$ increases with the increasing of $P_a$, and decreases with the increasing of $T_c$ and $r_h$, Eq. (21) is rewritten as follows:
However, because of the Earth’s rotation, the value of a star zenith distance will change from $-90^\circ$ to $90^\circ$ every day and exceed the accuracy limitation of Eq. (19) (i.e., $\left [ {0^\circ ,75^\circ } \right ]$ ). It is difficult to obtain the value of $\left | {\varepsilon _{i,j}^{\textrm {Atmo}\_ \textrm {ref }} \left ( t \right )} \right |$ in the larger variation of stars’ zenith distance.
Actually, the star’s zenith distance that can be observed by the terrestrial star tracker is determined by its attitude performance (i.e., the range of pitch, yaw, and roll angles). If the pitch, yaw, and roll angles of the terrestrial star tracker are $\left | \theta \right | \in \left [ {0^\circ ,\theta _{\max } } \right ]$, $\left | \psi \right | \in \left [ {0^\circ ,\psi _{\max } } \right ]$ and $\left | \gamma \right | \in \left [ {0^\circ ,\gamma _{\max } } \right ]$, the zenith distances $z_0$ of a star observed by the terrestrial star tracker will be in the range of
By substituting Eq. (23) into Eq. (22), the observation angular distance error caused by the atmospheric refraction of the terrestrial star tracker can be obtained as follows:
3.3 Modeling of the observation angular distance error
In summary, the observation angular distance error is a complex issue, which is related not only to the properties of stars that make up the angular distances but also to the various factors, such as the angular value, revolution and rotation of the Earth, observation time, observation location, observation environment, atmospheric refraction, and optical parameters of the star tracker. The angular distance with different values or made up of distinct stars will have completely varying rules.
However, since the angular distance of a single pair of stars is identified only when they belong to the angular distance threshold range in star identification, the upper variation range of the angular distance errors is in demand. Therefore, the variation range of the observation angular distance error $\varepsilon _{i,j} \left ( t \right )$ for the terrestrial star tracker can be obtained by setting $d_{i,j}^{AT}=\alpha =\alpha '=\alpha ''=d_{i,j}^\textrm {ref}$, as follows:
Evidently, the changes on $T_\textrm {ref}$ and ref-frame of the star catalog will influence $\varepsilon _{i,j} \left ( t \right )$. On the basis of the difference of ref-frame, the observation angular distance error can be divided into two items, that is,
For example, if the ref-frame of the star catalog is the AT-frame, $\varepsilon _{i,j}^\textrm {ref} \left ( t \right )$ will consist of two factars, i.e., ${\varepsilon _{i,j}^\textrm {space - move} \left ( {t - T_\textrm {ref} } \right )}$ and ${\varepsilon _{i,j}^{\textrm {Annual}\_ \textrm {par }} \left ( t \right )}$; $\varepsilon _{i,j}^\textrm {ref - S} \left ( t \right )$ will consist of four factors, i.e., ${\varepsilon _{i,j}^{\textrm {Gravity}\_ \textrm{Ref}} \left ( t \right )}$, ${\varepsilon _{i,j}^{\textrm {Annual}\_ \textrm {abe}} \left ( t \right )}$, ${\varepsilon _{i,j}^{\textrm{Diurnal}\_ \textrm{abe}} (t)}$, and ${\varepsilon _{i,j}^{\textrm {Atmo}\_ \textrm {ref }} \left ( t \right )}$. Then, the variation range of the observation angular distance error can be calculated on the basis of Eq. (25).
4. Angular distance matching threshold optimization
On the basis of the aforementioned analysis, the observation angular distance error is a complex issue related to various factors. It is unrealistic for a terrestrial star tracker to determine whether the observation angular distance error $\varepsilon _{i,j} \left ( t \right )$ of all possible observed angular distances is consistent with Eq. (6) all the time.
For convenience, the angular distance matching threshold ($TH$) is used to represent the observation angular distances error of all the possible observed angular distances in star tracker at any observation time $t$, as follows:
Given that the observation angular distance error is divided into two items on the basis of the proposed observation angular distance error model Eq. (26), the angular distance matching threshold $TH$ is also rewritten as
4.1 Star catalog angular distance matching threshold
The star catalog angular distance matching threshold $TH_{\textrm {catalogue }}$ is determined by the ref-frame and $T_\textrm {ref}$ of the star catalog. The ref-frame will be one of the five coordinate systems, namely, ICRS, AT-frame, and G-, E-, and N-frames (Fig. 4). $T_\textrm {ref}$ is determined by the stable time of the star catalog, which is usually larger than 1 year compared with the real-time observation time.
AT-frame is the most suitable coordinate system in practical applications, since the vector variation of stars in the G-frame ( larger than $\left [ {\textrm { - 20}\textrm {.4915'', + 20}\textrm {.4915''}} \right ]$ per year), E-frame, and N-frame ( approximately $\left [ {\textrm {- 180}^\circ \textrm {, + 180}^\circ } \right ]$ per day) are too large to obtain a high-accuracy star catalog over the stable time.
Therefore, on the basis of Eqs. (25) and (28), the star catalog angular distance matching threshold can be expressed as follows by setting the ref-frame of the star catalog as AT-frame:
Suppose that the star catalog is stable in 2019, that is, the stable time of the star catalog is 1 year. Figure 10(a) shows the vector error of stars in the AT-frame versus observation time $t$ in 2019 by setting the $T_\textrm {ref}$ as $2019.07.01$. A total of 5,041 stars are selected from the Hipparcos catalogue, with their magnitudes brighter than 6.0. It is found that the vector variation of different stars keep near-linearly increasing with $\left | {t - T_\textrm {ref} } \right |$. The maximum value of star vector error reach to $3.2''$, and the minimum value equals to $0''$.
The cumulative distribution of the number of stars versus the star vector error is shown in Fig. 10(b). If using a value, denoted as $Err_\textrm {star}$, to represent the vector error range of stars during the available stable time, different value of $Err_\textrm {Star}$ will corresponding to different sets of stars in star catalog. For example, when $Err_\textrm {star} = \left [ {0,0.4''} \right ]$, 4,961 stars are present in the range of $Err_\textrm {star}$, and $TH_{\textrm {catalogue}} \approx 2Err_\textrm {star} = 0.8''$.
Therefore, The value of $TH_{\textrm {catalogue}}$ is determined by the selection of star in star catalog. According to the helix uniform distribution method [33,34], assume that there are $N_{\textrm {Axis}}$ uniformly directions of the FOV axis, denoted as $\textbf {b}^{AT}$ existed in the AT-frame. Then, on the basis of $Err_\textrm {star}$ and the FOV of star tracker $\theta _\textrm {FOV}$, the probability that $N_s$ stars exist in the FOV of the star tracker is
If the FOV of a star tracker is increased from $14^\circ \times 14^\circ$ to $22^\circ \times 22^\circ$, the simulation results of $P\left ( {\theta _\textrm {FOV} ,Err_\textrm {star} ,N_s} \right )$ versus $Err_\textrm {star}$ when $N_\textrm {Axis}=100,000$ and $N_s=4$ is shown in Fig. 11(a); the results of $Err_\textrm {star}$ versus $N_s$ when $P\left ( {\theta _\textrm {FOV} ,Err_\textrm {star} ,N_s} \right )=100\%$ is shown in Fig. 11(b).
With the decreasing of $\theta _\textrm {FOV}$ and the increasing of $N_s$, the value of $Err_\textrm {star}$ increases for the case $P\left ( {\theta _\textrm {FOV} ,Err_\textrm {star} ,N_s} \right )=100\%$. Therefore, on the basis of $\theta _\textrm {FOV}$ and $N_s$, the star catalog angular distance matching threshold is determined by
For instance, when $\theta _\textrm {FOV}= 20^\circ$ and $N_s=11$, the optimal $Err_\textrm {star} = \left [ {0,0.375''} \right ]$ (Fig. 11(b)), $TH_{\textrm {catalogue}} \approx 0.75''$, and a total of 4,950 stars are contained in $\overline {\textrm {Star}_{\textrm {Catalogue}} }$.
4.2 Transmission angular distance matching threshold
The transmission angular distance matching threshold is determined by the astrometric transmissions from the ref-frame to the S-frame. According to Eq. (25), when the ref-frame of the star catalog is the AT-frame, the transmission angular distance matching threshold is
Actually, the observed angular distance values of the terrestrial star tracker is determined by the value of $\theta _\textrm {FOV}$ (i.e., $d_{i,j}^\textrm {ref} \in \left ( {0,\theta _\textrm {FOV} } \right ]$). Therefore, on the basis of Eq. (25), when these four factors reach their maximum at the same time, the transmission angular distance matching threshold $TH_{\textrm {ref - S}}$ is obtained as
For instance, by setting the values of parameters as Tabel 1, the $TH_{\textrm {ref - S}}$ versus the value of $\theta _\textrm {FOV}$ when $z_{s,\max }=45^\circ$ is shown in Fig. 12(a), and the $TH_{\textrm {ref - S}}$ versus the value of $z_{s,\max }$ when $\theta _\textrm {FOV}=20^\circ$ is shown in Fig. 12(b). If $\theta _\textrm {FOV}=20^\circ$ and $z_{s,\max }=45^\circ$, the $TH_{\textrm {ref - S}}=58.1974''$ when the star tracker is only used in night; otherwise, $TH_{\textrm {ref - S}}=61.9281''$.
4.3 Optimal angular distance matching threshold
In summary, the total angular distance matching threshold of the terrestrial star tracker is determined by accumulating the star catalog and transmission matching thresholds.
If the value of angular distance matching threshold $TH$ larger than these two matching thresholds resources, i.e., ($TH_{\textrm {catalogue}}$ +$TH_{\textrm {ref - S}}$), the observation angular distance error of the observation stars satisfies Eq. (27) and ensures the success of star identification. However, the value of $TH$ is not the larger the better. It is also related to the identification efficiency. A larger value of $TH$ means that more misrecognized pairs of stars satisfy Eq. (27), and make it more difficult and inefficient to identify the true stars form the misrecognized stars [22].
Therefore, the optimal angular matching threshold $\overline {TH}$ should be the minimum value of $TH$ that ensure the star identification success rate equals to 100%, as shown as follows:
Additionally, in applications, the angular distance errors of the observation stars are also influenced by the measurement capabilities of the star tracker itself and the atmospheric turbulence. Therefore, the optimal angular distance matching threshold $\overline {TH}$ is rewritten as:
5. Simulations and night sky experiment
In the previous sections, we present the observation angular distance error model and derive the optimal angular matching threshold expression for terrestrial star trackers. In this section, numerical simulations and night sky experiment are conducted to validate our conclusions.
5.1 Numerical simulations
A series of simulations is conducted to verify the optimization of angular distance matching threshold. Assuming that the values of simulations parameters are presented as Tabel 1, and the star tracker is only used at night.
Regardless of the positional accuracy ${\varepsilon _p }$ of the star tracker and the atmospheric turbulence error ${\varepsilon _{\textrm {turbulence}} }$, a total of 10,000 simulation vector star images are generated in the test as follows. (1) $\overline {\textrm {Star}_{\textrm {Catalogue}} }$, which consists of 4,950 stars, and $Err_\textrm {star} = 0.375''$ is used as the reference star catalog. (2) A total of 10,000 random unit vectors of the FOV axis in the N-frame $\textbf {b}^N$ is generated with random observed zenith distance $z_s \in \left [ {0^\circ ,45^\circ } \right ]$ and observed azimuth $\alpha _s \in \left [ {0,360^\circ } \right ]$ . (3) The unit vectors of FOV axis in the AT-frame $\textbf {b}^{AT}$ are obtained by transmitting the $\textbf {b}^N$ from the N-frame to the AT-frame on the basis of Eq. (7), where the observation longitude is random in $\left [ { - 180^\circ ,180^\circ } \right ]$, the latitude is random in $\left [ { - 90^\circ ,90^\circ } \right ]$, the height is 960 m, and the local time is 00:00:00 on any day in 2019. (4) All the stars $\textbf {V}_{AT} ^{i}$ from $\overline {\textrm {Star}_{\textrm {Catalogue}} }$, with their angular distances less than $10^\circ$ between $\textbf {V}_{AT} ^{i}$, are selected. (5) The stars $\textbf {V}_{AT} ^{i}$ from the AT-frame to the S-frame are transmitted on the basis of Eqs. (7) and (8) to generate the observation vector $\textbf {W}_{i}$ and formed the simulation vector star images, where $\left \langle {C_n^S } \right \rangle = {\textbf I}$. The examples of two typical simulation vector star images are shown in Fig. 13.
Nine value of the tested angular distance matching threshold $TH$ (i.e., $30''$, $35''$, $40''$, $45''$, $50''$, $55''$, $58''$, $60''$, and $65''$) are used to implement the star identification for the 10,000 simulation vector star images. The star identification for the simulation vector star image is regarded successful only when all the stars are identified. The average number of angular distance pairs that satisfy Eq. (27) for different values of $TH$ is recoded.
Figure 14 shows the simulation results of the star identification success rate and the average number of angular distance pairs versus the angular distance matching threshold $TH$. Theoretically, on the basis of Eq. (35), the optimal angular distance matching threshold of the simulation terrestrial star tracker is $\overline {TH} = \textrm {58}\textrm {.9474''}$, as shown by the dashed line in Fig. 14.
From the figure, the identification success rate has increased from 72% to 100% as $TH$ increased from $30''$ to approximate $59''$. Although the cases when $TH$ is larger than $59''$ also ensure the success of star identification, the average number of angular distance pairs is increased drastically. When the value of $TH$ increased from $59''$ to $65''$, the average number of angular distance pairs is increased by more than $0.71\times 10^4$. Thus, the optimal angular distance matching threshold for the simulation observation results $\overline {TH}$ is approximately equal to $59''$.
The value of theoretical optimal angular distance matching threshold $\overline {TH}$ is approximately equal to the simulation observation results, which means that all the angular distances in the simulation star images agree with the optimal angular distance matching threshold model.
5.2 Night sky experiment
A night sky experiment was conducted at the Xinglong Station (National Astronomical Observatories, China) to verify our theoretical conclusions. In Fig. 15, the star tracker is mounted on a two-axis turntable, the install zenith distance of the star tracker is approximately equals to $1.8^\circ$, and the parameters of the test star tracker is shown in Table 2.
In the night sky experiment, a total of 1,000-frame centroid data were acquired by rotating the turntable within $-40^\circ$ to $40^\circ$. Twelve values of tested angular distance matching threshold (i.e., $25''$, $30''$, $35''$, $40''$, $45''$, $50''$, $55''$, $60''$, $65''$, $70''$, $75''$ and $80''$) are used to implement the star identification for the 1,000-frame centroid data. The $\overline {\textrm {Star}_{\textrm {Catalogue}} }$, which consists of 4,950 stars and $Err_\textrm {star} = 0.375''$, is used as the reference star catalog. The star identification is regarded successful only when the number of recognized stars is larger than six. The results of the identification success rate and the average number of angular distances versus the angular distance matching threshold $TH$ are shown in Fig. 16.
On the basis of Eq. (35), the theoretical $\overline {TH} = \textrm {69}\textrm {.5430''}$, where, the observation longitude is $117.580176^\circ$, the latitude is $40.397073^\circ$, the height is 958 m, $z_{s,\max } = 40^\circ + 1.8^\circ$, $P_a \in \left [ {980,1020} \right ]$, $T_c \in \left [ {14^\circ C,20^\circ C} \right ]$, $r_h \in \left ( {0.2,0.4} \right )$, and $\max \theta _\textrm {FOV} = \sqrt 2 \times 20^\circ$, the residual measurement error of the star tracker after laboratory calibration $\varepsilon _p = 1''$, and $\left | {\varepsilon _{\textrm {turbulence}} } \right |\le 1''$.
As shown in Fig. 16, the identification success rate and the average number of angular distances increase with the distance matching threshold $TH$. As $TH$ increased from $25''$ to $70''$, the identification success rate has increased from 74.6% to 100%, and the average number of angular distances was increased from 0.5 to $2.339\times 10^6$. Therefore, the optimal angular distance matching threshold for the actual observation results is approximately equal to $70''$.
The difference between the theoretical value of $\overline {TH}$ ($\textrm {69}\textrm {.5430''}$) and the actual experiment results ($70''$) is less than $0.5''$. Therefore, the night sky experiment validates the reliability of our conclusions for the observation angular distance error model and angular distance matching threshold optimization.
6. Conclusion
This study focuses on the modeling of the observation angular distance error for terrestrial star trackers. This parameter is important and affects the identification success rate, attitude measurement accuracy, and real-time performance of the terrestrial star tracker but is yet to be clarified to date. To resolve this problem, a stellar observation angular distance error model is derived. The factors, namely, space motion, annual parallax, solar gravitational light deflection, annual aberration, diurnal aberration, and atmospheric refraction, are analyzed to determine the variation range of observation angular distance error. Then, the optimization solution of the angular distance matching threshold is obtained to achieve the optimal efficiency in the star identification. Numerical simulations and night sky experiment validate the reliability of our conclusions for error model and threshold optimization.
Funding
National Natural Science Foundation of China (61725501).
References
1. C. C. Liebe, “Accuracy performance of star trackers-a tutorial,” IEEE Trans. Aerosp. Electron. Syst. 38(2), 587–599 (2002). [CrossRef]
2. G. Zhang, STAR IDENTIFICATION: Methods, Techniques and Algorithms (SPRINGER, 2019).
3. C. C. Liebe, “Star trackers for attitude determination,” IEEE Aerosp. Electron. Syst. Mag. 10(6), 10–16 (1995). [CrossRef]
4. T. Sun, F. Xing, X. Wang, Z. You, and D. Chu, “An accuracy measurement method for star trackers based on direct astronomic observation,” Sci. Rep. 6(1), 22593 (2016). [CrossRef]
5. J. Li, G. Wang, and X. Wei, “Generation of guide star catalogue for star trackers,” IEEE Sens. J. 18(11), 4592–4601 (2018). [CrossRef]
6. W. Yu, J. Jiang, and G. Zhang, “Multi exposure imaging and parameter optimization for intensified star trackers,” Appl. Opt. 55(36), 10187–10197 (2016). [CrossRef]
7. W. Yu, J. Jiang, and G. Zhang, “Star tracking method based on multiexposure imaging for intensified star trackers,” Appl. Opt. 56(21), 5961–5971 (2017). [CrossRef]
8. G. Wang, F. Xing, M. Wei, and Z. You, “Rapid optimization method of the strong stray light elimination for extremely weak light signal detection,” Opt. Express 25(21), 26175–26185 (2017). [CrossRef]
9. T. Sun, F. Xing, J. Bao, S. Ji, and J. Li, “Suppression of stray light based on energy information mining,” Appl. Opt. 57(31), 9239–9245 (2018). [CrossRef]
10. A. Katake and C. Bruccoleri, “Starcam sg100: a high-update rate, high-sensitivity stellar gyroscope for spacecraft,” Proc. SPIE 7536, 753608 (2010). [CrossRef]
11. A. B. Katake, Modeling, image processing and attitude estimation of high speed star sensors Doctoral thesis (Texas A&M University, 2006).
12. T. Sun, F. Xing, Z. You, X. Wang, and B. Li, “Smearing model and restoration of star image under conditions of variable angular velocity and long exposure time,” Opt. Express 22(5), 6009–6024 (2014). [CrossRef]
13. G. Zhang, J. Jiang, and J. Yan, “Dynamic imaging model and parameter optimization for a star tracker,” Opt. Express 24(6), 5961–5983 (2016). [CrossRef]
14. Z. Wang, J. Jiang, and G. Zhang, “Global field-of-view imaging model and parameter optimization for high dynamic star tracker,” Opt. Express 26(25), 33314–33332 (2018). [CrossRef]
15. Z. Wang, J. Jiang, and G. Zhang, “Distributed parallel super-block-based star detection and centroid calculation,” IEEE Sens. J. 18(19), 8096–8107 (2018). [CrossRef]
16. M. S. Wei, F. Xing, and Z. You, “A real-time detection and positioning method for small and weak targets using a 1d morphology-based approach in 2d images,” Light: Sci. Appl. 7(5), 18006 (2018). [CrossRef]
17. G. Wang, J. Li, and X. Wei, “Star identification based on hash map,” IEEE Sens. J. 18(4), 1591–1599 (2018). [CrossRef]
18. D. Mortari, M. A. Samaan, C. Bruccoleri, and J. L. Junkins, “The pyramid star identification technique,” Navigation 51(3), 171–183 (2004). [CrossRef]
19. M. Aghaei and H. A. Moghaddam, “Grid star identification improvement using optimization approaches,” IEEE Trans. Aerosp. Electron. Syst. 52(5), 2080–2090 (2016). [CrossRef]
20. M. D. Shuster and S. D. Oh, “Three-axis attitude determination from vector observations,” J Guid. Control. Dynam 4(1), 70–77 (1981). [CrossRef]
21. Q. Fan and X. Zhong, “A triangle voting algorithm based on double feature constraints for star sensors,” Adv. Space Res. 61(4), 1132–1142 (2018). [CrossRef]
22. H. Zhu, B. Liang, and T. Zhang, “A robust and fast star identification algorithm based on an ordered set of points pattern,” Acta Astronaut. 148, 327–336 (2018). [CrossRef]
23. S. Huang, J. Huang, D. Tang, and F. Chen, Research on uav flight performance test method based on dual antenna gps/ins integrated system, in 2018 IEEE 3rd International Conference on Communication and Information Systems (ICCIS), (2018), pp. 106–116.
24. C. C. Liebe, “Pattern recognition of star constellations for spacecraft applications,” IEEE Aerosp. Electron. Syst. Mag. 8(1), 31–39 (1993). [CrossRef]
25. H. G. Walter and O. Sovers, Astrometry of fundamental catalogues: the evolution from optical to radio reference frames (Springer Science & Business Media, 2000).
26. VizieR, “Hipparcos catalogue,” http://vizier.u-strasbg.fr/.
27. “International astronomical union standards of fundamental astronomy,” http://www.iausofa.org/.
28. S. E. Urban and P. Kenneth Seidelmann, Explanatory supplement to the astronomical almanac (University Science Books, 2013).
29. R. C. Stone, “An accurate method for computing atmospheric refraction,” Publ. Astron. Soc. Pac. 108, 1051–1058 (1996). [CrossRef]
30. G. Sitarski, “Warsaw ephemeris of the solar system: De405/waw,” Acta Astronomica -Warsaw and Cracow- 52, 471–486 (2002).
31. X. Moisson and P. Bretagnon, “Analytical planetary solution vsop2000,” Celest. Mech. & Dyn. Astron. 80(3/4), 205–213 (2001). [CrossRef]
32. R. M. Green, Spherical astronomy (Cambridge University Press, 1985).
33. R. Bauer, “Distribution of points on a sphere with application to star catalogs,” J. Guid. Control. Dyn. 23(1), 130–137 (2000). [CrossRef]
34. W. Wang, X. Wei, J. Li, and G. Zhang, “Guide star catalog generation for short-wave infrared (swir) all-time star sensor,” Rev. Sci. Instrum. 89(7), 075003 (2018). [CrossRef]
35. J. Jiang, K. Xiong, W. Yu, J. Yan, and G. Zhang, “Star centroiding error compensation for intensified star sensors,” Opt. Express 24(26), 29830–29842 (2016). [CrossRef]
36. Y. Li, J. Zhang, W. Hu, and J. Tian, “Laboratory calibration of star sensor with installation error using a nonlinear distortion model,” Appl. Phys. B: Lasers Opt. 115(4), 561–570 (2014). [CrossRef]
37. T. Sun, F. Xing, and Z. You, “Optical system error analysis and calibration method of high-accuracy star trackers,” Sensors 13(4), 4598–4623 (2013). [CrossRef]
38. X. Wei, G. Zhang, Q. Fan, J. Jiang, and J. Li, “Star sensor calibration based on integrated modelling with intrinsic and extrinsic parameters,” Measurement 55, 117–125 (2014). [CrossRef]
39. T. S. McKechnie, “Atmospheric turbulence and the resolution limits of large ground-based telescopes,” J. Opt. Soc. Am. A 9(11), 1937–1954 (1992). [CrossRef]