During the operation of the star sensor in orbit, due to the transient effects caused by space radiation, a large amount of transient noise is easily generated on the detector target surface. Famous spatial radiation belts include the South Atlantic Anomaly (SAA) region, where traditional star sensors may experience ineffective attitude and switch to LIS mode due to a large amount of transient noise interference when passing through the SAA region. When the star sensor enters LIS mode, it is difficult to accurately capture and enter tracking mode under the interference of a large amount of transient noise, which is also a problem that is difficult to avoid for most star sensor models during orbital operation. At present, most of the star recognition algorithms studied by domestic and foreign experts focus on feature selection and the improvement of star recognition ability in conventional backgrounds. A few star recognition algorithms for transient noise [116] are limited by the amount of noise and still cannot achieve good recognition results. To solve this problem, it is necessary to propose a star sensor star point recognition algorithm suitable for large amounts of transient noise interference, which can complete star recognition again under the interference of transient noise and output accurate attitude information.
Dr. Lu Kaili (Institute of Optics and Electronics, Chinese Academy of Sciences) proposed a fast star recognition algorithm (FSI) for star sensors. Unlike traditional corner distance matching algorithms, this algorithm mainly utilizes the statistical approach of star repetition to achieve the positioning and recognition of the main star. This algorithm first targets the characteristics of transient noise and completes rough noise suppression through preprocessing; Then, a custom star attribute table that can be quickly queried was constructed, and the angular distance between the main star and neighboring stars was calculated. The K-vector method was used to search for the star pair list corresponding to the angular distance in the star attribute table; Utilize a fast repeated star search technique to achieve preliminary positioning of the main star star; Finally, the correct information of the main star within the field of view is determined through a dual screening method of field of view and angular distance.
When working in a natural space environment, the reliability of star sensors can be reduced due to the influence of high-energy radiation. The electromagnetic radiation generated by the impact of high-energy radiation mainly includes X-rays and γ Radiation. Among them, the situation of high-energy particle radiation is quite special, which can be basically divided into the Earth’s radiation zone
There are three major categories of solar cosmic rays and galactic cosmic rays.
The Earth’s radiation belt, also known as the Van Allen radiation belt, poses the greatest radiation threat in Earth’s orbit. It is the proton, electron, and small amount captured by the Earth’s magnetic field α Composed of spatial radiation particles such as particles. According to the changes in high latitude, it can be divided into two types: inner radiation zone and outer radiation zone.
(1) The orbital height of the inner radiation zone is approximately 600km to 10000km. Considering the anomalous distribution of the geomagnetic field, the lower boundary of the inner radiation zone is located about 200km above the South Atlantic, which is a well-known area of geomagnetic anomalies in the South Atlantic. The corresponding internal radiation band is mainly composed of protons, electrons, and a small amount of α Particle composition, which is not sensitive to solar activity, and the energy distribution of protons is mainly 0.1-400 MeV, while the energy distribution of electrons is 0.04-7 MeV.
(2) The range of orbital heights corresponding to the outer radiation zone is relatively wide, ranging from 10000km to 60000 km, and even expanding to 130000 km. The center position of the orbit is about 20000 km to 25000 km. The majority of the outer radiation zone is composed of electrons, and the energy of electrons is mostly 0.04-4 MeV, while the corresponding proton energy is relatively low. External radiation is easily affected by solar activity, and when the local magnetic field is disturbed, the intensity and position information corresponding to the external radiation zone will undergo significant changes.
The Milky Way cosmic rays are located outside the solar system and are composed of high-energy particles from all directions. The vast majority of high-energy particles are composed of protons (88%), followed by relatively few α Particles, while other particles with higher charges have a content approximately two orders of magnitude less than protons. Its energy range is approximately 0-1010 GeV, and it is distributed in free space more than 50 km above the ground, with a flux of approximately 4 cm-2 · s-1.
The high-energy solar radiation mainly includes the solar wind and X-rays generated during solar flare eruptions γ Rays and high-energy particle streams. The high-energy particle flow generated during solar flares can be referred to as solar cosmic rays. Due to the fact that the main component is high-energy protons and there are very few particles with a charge greater than 3, such flares are also known as solar proton events. The energy of solar cosmic rays ranges from 10 MeV to 10 GeV, with an intensity of approximately 105cm-2 · s-1.
Due to the radiation effect on the detector, star sensors are prone to performance degradation and abnormal operation. At present, the commonly used detectors for star sensors are CCD or APS. Although the detectors are diverse, fundamentally speaking, all detectors are integrated circuits built using semiconductor optoelectronic effects, with the main difference being the charge collection method. So different detectors will produce various abnormal phenomena after being affected by space radiation. From the analysis of the impact mechanism of space radiation on star sensors, the main effects brought by space radiation include displacement damage, total dose effect, and transient effect. Among them, displacement damage and total dose effect belong to cumulative effects, and transient noise caused by transient effect can easily cause abnormal operation of star sensors.
Transient effect refers to the phenomenon where charged particles enter the sensitive layer of an imaging element, and the energy of the charged particles is absorbed. Ionization effect can cause the generation of electron hole pairs (as electron hole pairs generally only appear briefly and only exist in the current cycle without affecting the next cycle, they can be referred to as transient effect). The emergence of transient effects often causes a series of noise and pseudo star interference in star sensor imaging systems. Due to the interference of noise and pseudo star points, the success rate of star sensor recognition will be greatly reduced, and the attitude effectiveness will decrease in tracking mode, even leading to exiting the tracking mode and conducting new star point recognition. The shape characteristics of transient noise can be obtained from ionization energy loss and track length at corresponding pixels. The ionization energy loss of charged particles can be obtained through the incident energy characteristics of the particles and the linear energy transfer density (LET, LinearEnergy Transfer). Generally, particles with an atomic weight greater than 1 exhibit a straight line characteristic; Due to the extremely small electron mass, the scattering of electrons during the incident process also needs to be considered. Usually, at room temperature, it takes approximately 3.65 eV to generate an electron hole pair through ionization. Therefore, the energy transferred, the electron track length at the corresponding pixel, and the charge collection efficiency can be used to determine the amount of charge generated by the transient effect of the particle. By analyzing the energy spectrum and particle flow rate of the current particles in the star sensor, the degree to which the star sensor is affected by transient effects can be quantitatively analyzed.
Figure 4.1 Noise and pseudostars caused by transient effects captured during strong irradiation experiments on a certain detector.
The noise caused by transient effects:
When a star sensor passes through the SAA region, it is susceptible to transient effects in imaging devices (CMOS detectors) due to strong spatial radiation. Transient effects refer to the phenomenon where high-energy particles enter the sensitive layer of the imaging device, absorb energy, and generate electron hole pairs, resulting in transient noise on the detector. The manifestation of transient noise on the detector target varies depending on the direction in which high-energy particles enter the detector. The more perpendicular the angle of penetration, the closer the noise shape is to the star point; The more inclined the angle of insertion, the longer the trailing edge of the noise. As shown in Figure 4.2, transient noise exhibits different shape features in the star map. Transient noise has the characteristics of randomness and transience. The emergence of transient noise brings great difficulties to the star recognition process of star sensors.
This algorithm can make the star sensor more effective in star point recognition under the interference of a large amount of transient noise. The FSI algorithm utilizes the randomness and transience of transient noise to eliminate diagonal noise in the process of single frame star point detection using scale information, and eliminates other similar star point noise by comparing the position and energy information of adjacent frame star points, thus achieving preliminary suppression of a large amount of noise; The FSI algorithm uses K-vector based multi star search technology to complete the final star recognition. The multi star search technology first constructs a custom star attribute table, then calculates the angular distance between the main star and neighboring stars, and uses the K-vector method to search for star pairs corresponding to the angular distance in the star attribute table. A fast repeated star search technology based on address index is used to achieve the initial positioning of the main star; Finally, the FSI algorithm determines the correct information of the main star in the field of view through a dual screening method of field of view and angular distance.
As shown in Figure 4.3, the main execution process of FSI includes two steps: noise suppression and multi star search.
The execution process of FSI algorithm:
Noise suppression
The shapes of transient noise generated by transient effects are various, with many shapes as large as stars and some appearing as diagonal lines. If the noise is not initially suppressed, a large amount of noise will reduce the success rate of star recognition, and may even cause the star sensor to stay in LIS mode for a long time. Therefore, the FSI algorithm proposes a noise suppression method based on the shape characteristics of noise, which mainly utilizes the scale and transient characteristics of noise for noise processing. Therefore, noise suppression can be divided into two steps: scale analysis and neighborhood comparison.
(1) Scale analysis
The purpose of scale analysis is to eliminate diagonal noise, mainly achieved through the method of clustering within a single frame image. After clustering the targets, the scale information of the targets can be calculated. By analyzing the scale within a single frame image, some abnormal scale noise can be removed.
(2) Neighborhood comparison
The shape of transient noise is similar to that of a star point, and this type of noise cannot be filtered out through scale information within a single frame due to its close proximity to the star point. So it is necessary to utilize the randomness and transience of transient noise to determine the correlation of target position information detected in adjacent frame images.
Through scale analysis within a single frame and neighborhood comparison between multiple frames, a large amount of transient noise will be eliminated. The noise suppression method proposed for transient noise has good applicability, which can effectively suppress some transient noise and facilitate the subsequent main star search work.
Multi main star search technology
After noise suppression, star recognition is mainly achieved through subsequent multi star search techniques, which mainly rely on the statistics of repeated stars to achieve star recognition, which is different from the traditional triangle method of equal angular distance matching. In multi star search technology, the first step is to use k-vector technology to construct the k-vector function 𝑘(𝑥), which aims to achieve fast search of custom star attribute tables. In addition, a fast repeated star search technique has been proposed, which can quickly locate stars with a large number of duplicates in a large number of star pairs. By combining k-vector technology with fast repeated asterisk search technology, the overall execution speed of the algorithm can be improved. After locating the star attribute information for multiple main stars (matching with the star attribute table), two steps of verification are required to achieve accurate recognition of the main stars: comparison of field of view angles and verification of angular distance.
The multi star search technology mainly includes three parts:
Firstly, a custom star attribute table needs to be built, just like the star library built by the triangle method and matching group method. For the search algorithm of FSI, a suitable custom star attribute table needs to be built. The star library finally solidified in the memory mainly includes: star attribute table and declination table.
The process of searching for matching star pairs in the star attribute table using the k-vector function can be expressed as: ① Calculate the angular distance 𝐴𝑗 between the stars 𝑌𝑗 in the measurement coordinate system. ② Using the previously obtained straight line 𝑦 (𝑥), calculate the k index value 𝑥. ③ Substitute the values of 𝑥𝑑 and 𝑥𝑢 obtained in ② into the k-vector function 𝑘 (𝑥), and obtain the row numbers 𝑗𝑑 and 𝑗𝑢 in the star attribute table. ④ Considering the calculation error in the calculation process of star angular distance 𝐴𝑗, take the row numbers 𝑗𝑑 and 𝑗𝑢 as the center and move upwards, respectively. Spread an offset on the lower two sides, taking into account the four lines of information: (𝑗𝑑– 1), (𝑗𝑑+1), (𝑗𝑢– 1), and (𝑗𝑢+1), Use its star pairs for subsequent repeated star searches. Therefore, in the star attribute table, the corresponding information of the 6 planetary pairs is extracted for subsequent repeatability comparisons.
It can be mainly divided into the following three steps. The first step is to obtain the statistical values of duplicate star signs based on the star pair information obtained by the k-vector method mentioned above. Through the statistics of duplicate star signs, the preliminary positioning of the main star is achieved, that is, to preliminarily determine the star signs and declination and declination information of each main star. The second step is field of view angle analysis, mainly to eliminate stars that meet the statistical relationship of duplicate stars but are located outside the field of view. The third step is angular distance verification, which ultimately eliminates false stars and preserves the true stars by checking the error of angular distance between the measurement coordinate system and the inertial coordinate system. After searching and filtering through the above three steps, the above search and filtering conditions cannot be met, and accurate asterisk information cannot be obtained. It is considered as noise and can be directly removed. The remaining star signs 𝑆𝑛′ and their corresponding declination values 𝑆𝑛𝑟𝑎′ and declination values 𝑆𝑛𝑑𝑒′ are retained for subsequent attitude calculations.
The innovation of FSI algorithm lies in:
Compared to the triangle algorithm, matching group algorithm, and MPA algorithm,
The FSI algorithm has a higher recognition success rate, and the FSI algorithm has a faster execution speed with an average execution time of only 60.1 seconds, which is conducive to achieving fast recognition of star sensors.
Send us a message,we will answer your email shortly!
TY-Space Technology (Beijing) Ltd. is professional focusing on advanced attitude optical sensors, especially Star Trackers for space industry.
