【作者】 柳毅；

【导师】 孙伟； 陈龙；

【作者基本信息】 西安电子科技大学 ， 工程硕士（专业学位）， 2022， 硕士

【摘要】 深空探测作为一种拓展人类生存空间,探索地外文明的重要手段,近年来随着人类对深空的不断探索,现有的一些导航技术已无法满足越来越高的导航需求。基于星敏感器的天文导航作为一种新兴的导航技术,以其对地面的依赖性低、使用宇宙中相对恒定不变的恒星星体作为其导航依据,可以保持长时间、高精度的导航性能,成为如今深空探测中不可或缺的一部分。其中星点的快速高精度定位、星图数字模拟技术以及星图识别算法作为星敏感器自主天文导航的关键技术,成为了国内外学者的研究重点。本文针对这三个方面进行了深入的研究,主要贡献与工作内容如下:1、构建了基于星敏感器下的星图全天区多情况星图环境模型。基于目前的天文理论,深入研究星图识别的先验基本星表信息以及星敏感器成像原理,对全天区多视轴下的星图进行计算机模拟生成,并针对星敏感器真实成像中遭受的噪声、伪星以及缺星干扰等多种情况进行星图模拟仿真实现,为后续星图识别算法提供了全面可靠的星图模拟仿真环境。2、提出了星点质心快速高精度定位算法。针对目前星点定位速度慢、定位精度较低等情况,对星图识别中的噪声处理、阈值分割、星点粗提取方法以及星点质心细分定位方法都进行了详细的分析。将基于灰度交叉投影的快速星点粗提取方法与带阈值的高精度星点质心细分定位方法相结合,在提高星点提取速度的同时保证了星点质心的高精度定位。3、提出了基于星间角距的奇异值分解可靠星图识别算法。针对目前星图识别算法中存在的误匹配率过高、识别速度较低以及算法鲁棒性较差的情况,对目前星图识别领域的几种典型星图识别算法进行分析。利用星间角距这一稳定的星图信息对特征库星组进行选取,同时对观测特征库进行结构优化,算法在星图无噪声干扰的情况下识别率可达99.95%;在识别时间与特征库容量上相较于传统奇异值分解算法分别提升了23%和10%,同时算法对于噪声、星等、缺星以及伪星等干扰有较好的鲁棒性。4、设计了一款星图模拟识别软件。根据本文中所涉及的星图模拟、星点质心定位、星图识别算法以及一些现有的算法,集成设计了一款星图模拟识别软件,实现了从星图模拟仿真到星图识别的整个过程。该软件共分为星图模拟、星点质心定位以及星图识别三个模块,通过此软件可以对星图识别算法进行高效的性能测验和算法分析。更多还原

【Abstract】 Deep space exploration is an important means to expand human living space and explore extraterrestrial civilization.In recent years,with the continuous exploration of deep space by human beings,some existing navigation technologies have been unable to meet the increasingly high navigation needs.As an emerging navigation technology,star sensor-based astronomical navigation has low dependence on the ground and uses relatively constant stars in the universe as its navigation basis,which can maintain long-term and high-precision navigation performance.become an integral part of deep space exploration today.Among them,fast and high-precision positioning of star points,star map digital simulation technology and star map recognition algorithm,as the key technologies of star sensor autonomous astronomical navigation,have become the research focus of scholars at home and abroad.This paper conducts in-depth research on these three aspects,and the main contributions and work are as follows:1.A multi-situation star map environment model of the star map all-sky area based on the star sensor is constructed.Based on the current astronomical theory,the a priori basic star table information of star map recognition and the imaging principle of star sensor are deeply studied,and the star map under the multi-horizontal view of the whole sky area is generated by computer simulation.The simulation of star map is carried out under various conditions such as noise,pseudo-star and star-missing interference,which provides a comprehensive and reliable star map simulation environment for the subsequent star map recognition algorithm.2.A fast and high-precision localization algorithm for the centroid of star points is proposed.Aiming at the slow speed of star point positioning and low positioning accuracy,the noise processing,threshold segmentation,star point coarse extraction method and star point centroid subdivision positioning method in star map recognition are analyzed in detail.Combining the fast star point coarse extraction method based on gray cross-projection and the high-precision star point centroid subdivision positioning method with threshold,the high-precision positioning of star point centroid is ensured while improving the star point extraction speed.3.A reliable star pattern recognition algorithm based on the angular distance between stars is proposed by singular value decomposition.Aiming at the high mismatch rate,low recognition speed and poor algorithm robustness in the current star map recognition algorithm,this paper analyzes several typical star map recognition algorithms in the current star map recognition field.Using the stable star map information of the inter-satellite angular distance to select the star group in the feature database,and optimize the structure of the observation feature database at the same time,the algorithm can achieve a recognition rate of 99.95% when the star map has no noise interference;Compared with the traditional singular value decomposition algorithm,the capacity of the feature library is increased by23% and 10% respectively.At the same time,the algorithm has better robustness to noise,magnitude,lack of stars and pseudo-stars.4.A star map simulation and recognition software is designed.According to the star map simulation,star point centroid positioning,star map recognition algorithm and some existing algorithms involved in this paper,a star map simulation and recognition software is integrated and designed,which realizes the whole process from star map simulation to star map recognition.The software is divided into three modules: star map simulation,star point centroid positioning and star map recognition.Through this software,efficient performance testing and algorithm analysis can be performed on the star map recognition algorithm.更多还原