快速星图识别算法设计与实现

发布时间：2018-05-19 15:45

本文选题：星图识别 + 三角形算法　；参考：《南京理工大学》2017年硕士论文

【摘要】：导航系统是航天器必不可少的设备,在航天器工作中具有重要的作用。准确的姿态信息是航天器导航系统运作的基础,星敏感器是一种航天领域中高精度的姿态信息测量仪器。研究星敏感器工作中的关键技术——星图识别,对于航天事业的发展具有重要的意义。星图识别一般包括星图采集及预处理、导航星数据库构建和星点特征匹配等过程。本文分别从这三个方面对星图识别进行了研究。在现有算法的基础上,探寻在星点提取和星图识别环节进一步提高精度的方法和途径。在星图采集及预处理方面,提出了一种用于星点检测的自适应变邻域背景预测算法。该算法在现有背景预测算法的基础上,当待预测像素点在星点边缘处时,仅用待预测像素点邻域内灰度小于邻域灰度中值的像素点进行预测计算,而在其他区域时,直接使用固定权值进行预测。同时,该算法还判断待预测像素点是否为单点噪声,自适应调整待预测像素点自身灰度值在背景预测计算中的权值。实验结果证明,改进算法较现有算法对星点的提取更清晰,对噪声的抑制能力更强,有利于后续星点检测的处理。在导航星数据库构建方面,主要研究了均匀化导航星表的构建。提出了一种基于正交网格的改进算法。该算法中在分析全天球恒星分布特点的基础上,引入一个"距离-星等"加权的导航星综合权值来筛选导航星。和现有正交网格算法相比,改进算法提高了导航星均匀化程度。另外利用等效星处理双星问题,提高数据库的完整性。在星点特征匹配方面,提出一种改进的星图识别算法。该算法在现有三角形算法及其相关改进算法的基础上,通过引入星三角形之外的辅助观测星,构造四边形,并引入一种统计次数法对星点进行识别,利用星三角形特征对识别结果进行验证。仿真证明,在保证识别速率的情况下,改进算法的抗干扰能力有所提高。最后,总结论文的主要内容和创新点,并对进一步工作进行展望。
[Abstract]:Navigation system is an indispensable equipment for spacecraft and plays an important role in spacecraft work. Accurate attitude information is the basis of spacecraft navigation system operation. Star sensor is a kind of high precision attitude information measurement instrument in space field. The study of star map recognition, the key technology of star sensor, is of great significance to the development of space industry. Star map recognition includes star map acquisition and preprocessing, navigation star database construction and star feature matching. In this paper, star map recognition is studied from these three aspects. Based on the existing algorithms, the methods and ways to improve the accuracy of star extraction and star map recognition are explored. In the aspect of star map acquisition and preprocessing, an adaptive variable neighborhood background prediction algorithm for star detection is proposed. On the basis of the existing background prediction algorithms, when the pixel is at the edge of the star, only the pixel in the neighborhood of the pixel to be predicted is calculated, but in other regions, the gray level of the pixel is less than the median of the gray value of the neighborhood. Use fixed weights directly to predict. At the same time, the algorithm also determines whether the pixel to be predicted is a single point noise, and adaptively adjusts the weight of the gray value of the pixel to be predicted in the calculation of background prediction. The experimental results show that the improved algorithm can extract stars more clearly than the existing algorithms, and has stronger ability to suppress noise, which is beneficial to the processing of subsequent star detection. In the aspect of navigation star database construction, the construction of homogenized navigation catalog is studied. An improved algorithm based on orthogonal mesh is proposed. In this algorithm, based on the analysis of the star distribution characteristics of the whole celestial sphere, a "range-star magnitude" weighted navigation star synthesis weight is introduced to screen the navigation star. Compared with the existing orthogonal mesh algorithm, the improved algorithm improves the homogenization of navigation stars. In addition, the equivalent star is used to deal with the binary star problem to improve the integrity of the database. In the aspect of star feature matching, an improved star map recognition algorithm is proposed. On the basis of the existing triangle algorithm and its related improved algorithm, this algorithm constructs a quadrilateral by introducing an auxiliary observation star in addition to the star triangle, and introduces a statistical number method to identify star points. The star triangle features are used to verify the recognition results. Simulation results show that the anti-jamming ability of the improved algorithm is improved under the condition of ensuring the recognition rate. Finally, the paper summarizes the main content and innovation, and prospects for further work.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V448.2;TP391.41

【相似文献】