典型卫星局部构件光学识别方法研究

发布时间：2018-03-11 19:24

本文选题：卫星识别　切入点：局部构件　出处：《哈尔滨工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：目前,针对以卫星为主的重要太空资源与设施的攻击、保护和在轨服务已成为世界各国航天技术的重要发展方向,而卫星识别技术是其中的关键环节。而飞行器抵近技术的日益成熟,使得利用天基平台对太空目标进行高分辨率光学成像成为可能,这对卫星识别特别是局部构件的精确识别也提出了更新更高的需求。然而,尽管国内外在天基目标探测识别方面的研究较多,但多集中在远距离情况下的点目标探测及在轨运动状态的识别方面,针对卫星本体、太阳帆板等重要构件的识别方面的研究罕有报道。本论文针对该问题,开展典型卫星局部构件光学特性及其聚类特征与参数化表征方法研究,在此基础上研究复杂太空环境下的目标局部构件识别方法与算法,并开展实验验证。主要研究工作与成果如下:(1)建立卫星典型局部构件的参数化表征方法与聚类模型。针对不同类型的典型卫星,分析了卫星局部构件的几何结构与光学特性,在此基础上提取了用于辨识卫星本体、太阳帆板等局部构件的聚类特征,并研究了这些特征的参数化表征方法,同时结合卫星先验知识,建立了本体、太阳帆板、天线构件的聚类参数集与聚类模型。利用该模型可实现局部构件特征的分类与识别,并支持后续目标识别方法与算法研究;(2)提出卫星典型局部构件的识别方法与算法。结合典型卫星先验知识与构件聚类模型,综合考虑卫星在在轨运行中成像系统离焦、电子学噪声、平台抖动等导致的图像模糊、噪声、对比度下降等像质退化问题,以及太空目标与天基平台之间复杂相对运动关系导致的目标几何变形以及遮挡等问题,提出了基于单帧与多帧序列图像的卫星局部构件识别方法及算法流程和策略,从而提出了基于单帧与多帧序列图像的卫星局部构件识别算法;(3)算法实验验证及适用性分析。结合数字仿真和地面半物理实验,获取不同像质退化程度及太空目标与天基平台间不同相对位姿关系导致局部构件自遮挡与互遮挡以及几何变形条件的序列帧图像数据。利用该数据,验证了基于单帧和多帧序列图像的目标聚类与识别算法的准确性与适用性。
[Abstract]:At present, protection and in-orbit services have become an important development direction of space technology around the world in response to attacks on important space resources and facilities dominated by satellites. The technology of satellite identification is one of the key links, and the increasingly mature technology of aircraft approach makes it possible to use space-based platform for high-resolution optical imaging of space targets. This also puts forward a higher demand for satellite recognition, especially for the precise identification of local components. However, although there are many researches on space-based target detection and recognition at home and abroad, However, most of them are focused on the detection of point targets and the recognition of the state of motion in orbit at a long distance. There are few reports on the identification of important components such as satellite body, solar canvas and so on. The optical characteristics, clustering characteristics and parameterized characterization of typical satellite local components are studied. Based on this, the methods and algorithms of target local component identification in complex space environment are studied. The main research work and results are as follows: 1) the parameterized characterization method and clustering model of typical local satellite components are established. The geometrical structure and optical properties of local satellite components are analyzed for different types of typical satellites. On this basis, the clustering features used to identify local components such as satellite body, solar canvas and so on are extracted, and the parameterized representation method of these features is studied. At the same time, combined with the prior knowledge of satellite, the ontology and solar canvas are established. The clustering parameter set and clustering model of antenna components can be used to classify and recognize local component features. The method and algorithm of satellite typical local component recognition are proposed. Combined with the prior knowledge of typical satellite and component clustering model, the defocusing of imaging system in orbit is considered synthetically. The image quality degradation problems such as electronic noise, platform jitter, noise, contrast decline, and the geometric distortion and occlusion caused by the complex relative motion relationship between the space target and the space-based platform are discussed. This paper presents a method of local satellite component recognition based on single-frame and multi-frame sequence images, as well as its algorithm flow and strategy. The experimental verification and applicability analysis of satellite local component recognition algorithm based on single-frame and multi-frame sequence images are presented, which are combined with digital simulation and ground semi-physical experiments. The serial frame image data with different degradation degree of image quality and different position and pose relationship between space target and space-based platform resulting in local component self-occlusion and mutual occlusion and geometric deformation condition are obtained. The accuracy and applicability of the target clustering and recognition algorithm based on single frame and multi frame sequence images are verified.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V52;TP391.41

【参考文献】