Star sensors are one of the most commonly used attitude determination instruments. Compared with other common attitude measurement equipment such as sun sensors, magnetometers, horizons, and gyroscopes, star sensors not only have higher attitude measurement accuracy, but also can It has the ability to realize autonomous navigation and has strong anti-interference ability. It is currently the most important attitude measurement instrument on satellites and other spacecrafts, and is also used on missiles, aircraft and ships.
Research on star sensor technology began in the 1950s. Up to now, many different types of star sensor products have been developed and successfully applied. Aerospace product R&D institutions in the United States, Germany, France, Denmark, Italy and other countries have developed many star sensor products for use in different environments, some of which have attitude positioning accuracy of 1″ or higher. Domestic star sensors Technology research began in the late 1980s. After years of accumulation and development, many domestic scientific research institutes and universities have also developed star sensor products that have been successfully used in aerospace and aviation. However, compared with star sensors developed abroad, they are still There is a big gap. This article analyzes the development trend of star sensor technology by sorting out the development history of star sensors at home and abroad and the research status of key technologies, and provides some reference ideas for domestic star sensor researchers.
It is an algorithm that uses the characteristics of the mutual position relationship between stars for identification. The triangle algorithm was first proposed by Junkins in the 1970s. This algorithm is relatively intuitive and is also the most commonly used star map recognition algorithm in current projects. The core idea of the algorithm is to use the triangle features composed of observed stars and the same ones in the navigation constellation database. Constructive triangle matching. Liebe, Quine and Douma designed an improved triangle star pattern recognition algorithm.
Liebe selects all stars that can form a triangle for identification based on the size of the field of view and the number of bright stars; Quine first selects the brightest star in the field of view as the main star, and then selects the two brightest stars in the circular area around the main star. The main star forms a star triangle; Douma’s method is similar to Liebe’s, but he considers the probability of stars in the field of view forming a triangle and only selects the triangle with the highest probability. Compared with the star map recognition algorithm that simply uses star angular distance as a feature, the advantage of the triangle algorithm is that it has more feature dimensions, reduces the probability of mismatching, facilitates the establishment of a navigation star library index, and shortens the search time of the navigation star library. , improve the speed of star map recognition.
The disadvantage of the triangle recognition algorithm is that when the number of star triangles is large, redundant matching or mismatching will occur, reducing the recognition success rate. Mortari proposed a Pyramid recognition algorithm based on the k-vector method. The k-vector method can quickly initialize positioning, reduce the number of navigation star catalog searches, and improve the speed of star map recognition. The Pyramid algorithm uses tetrahedron as the identification feature and selects 4 observation stars, with 1 star as the vertex and the remaining 3 stars as triangles to form a tetrahedron. The k-vector method is used as the navigation star library search algorithm, which can be implemented in Quickly identify navigation stars when there are a lot of noise and pseudo star points. The disadvantage of this algorithm is that as the star pair information table increases, the accuracy of the fitting curve decreases, and it cannot ensure that the best matching star pair falls within the angular distance error range. Zhang Guangjun used a linear database search method to modify Liebe’s algorithm and improve the speed of the algorithm.
Padgett et al proposed the grid algorithm. This algorithm maps star coordinates to a sparse matrix, providing a new idea for star map recognition. The grid algorithm has the advantages of small storage capacity, fast recognition speed, and the algorithm is not sensitive to the measurement error of the star sensor. However, when the star position error or magnitude error is relatively large, the recognition rate of the grid algorithm will decrease rapidly. Meng Na proposed an improved algorithm for the grid algorithm and proposed the “elastic grayscale grid algorithm”, which adds a virtual elastic template during the recognition process.
This algorithm significantly improves the error tolerance for magnitude and noise errors and improves the recognition rate. Hyunjae Lee also proposed an improved grid star map recognition algorithm. He used a circular grid instead of the square grid of the original algorithm. This not only overcomes the dependence of the original grid algorithm on the reference star, but also makes the improved algorithm more sensitive to image rotation environments. It has strong robustness and introduces virtual grid, which increases the space for selecting modes and greatly improves the success rate of star map recognition.
Li Baohua and others proposed the KMP star map recognition algorithm, which is another representation of the raster algorithm. After high-pass filtering the collected star maps, they directly generated a 0-1 string matching pattern, and then used the KMP string Search algorithm performs star map string identification. Since the storage capacity of the original image string is too large, an improved KMP algorithm based on wavelet transform is proposed.
was produced with the rapid development of artificial intelligence technology. Hong introduced neural networks into star map recognition and proposed a star map recognition method based on fuzzy neural networks. This algorithm is based on the three angular distance characteristics of triangles, conducts neural network learning on the selected navigation triangle library, and uses the learned neural network structure to identify star maps. The neural network recognition algorithm has the characteristics of high recognition rate and fast recognition speed. Its disadvantages are slow learning speed and certain probability of misrecognition. Compared with traditional algorithms, neural network algorithms have the advantages of low data storage, good real-time performance and robustness.
However, a large number of sample sets are required for training. The accuracy of recognition is affected by the size of the training set and training time, and the requirements for hardware are also relatively high. McClintock introduced genetic algorithms into star map recognition for the first time and conducted preliminary research on star map recognition methods based on genetic algorithms. Paladugu conducted in-depth research on the application of genetic algorithms in star map recognition and proposed an improved star map recognition method based on genetic algorithms. .
Select a main star, encode the star-to-star angular distance between the main star and the companion star and the angle between the stars, define the sum of the star angular distance error and the angle error corresponding to the two sets of star maps as the fitness function, and search It is divided into two stages: coarse positioning and fine positioning. In the coarse positioning stage, the variation factor is appropriately larger, and in the fine positioning, the variation factor is adjusted smaller. Quan Wei and others used the adaptive ant colony algorithm (AAC) to realize star map recognition.
Comparative analysis of three types of star map recognition algorithms, their advantages and disadvantages are summarized in Table 4. Type 1 refers to the star chart recognition algorithm based on constellation characteristics, Type 2 refers to the star chart recognition algorithm based on character patterns, and Type 3 refers to the recognition algorithm based on intelligent behavior. Currently, in actual engineering, type 1 is the most commonly used algorithm, while type 2 and type 3 algorithms have not been widely used.
In the vast expanse of space, where the stars shine as beacons of light and guidance, star tracker algorithms serve as pillars of precision and reliability. By harnessing celestial guidance and computational ingenuity, these remarkable algorithms enable spacecraft to navigate the cosmic seas with unparalleled accuracy and confidence. As technology advances and missions push the boundaries of exploration, star tracker algorithms will continue to play a crucial role in shaping the future of space navigation and exploration endeavors.
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.
