基于光谱信息的空间碎片材料种类数目估计方法仿真研究

发布时间：2018-05-19 00:38

  本文选题：空间碎片 + 端元数目估计　； 参考：《光谱学与光谱分析》2017年09期

【摘要】：在观测空间碎片时,受碎片结构紧凑、组成材料复杂,以及地基观测设备空间分辨率的限制,同一像元中通常会包含多种材料的信息,即产生"混合像元"。目前国内外对混合像元的研究主要集中在获取混合像元的纯物质光谱以及丰度上,往往忽略了高光谱数据中纯物质个数的确定对于没有任何先验信息的混合像元分析是至关重要的。如果估计的材料数目过少,将会导致解混出的材料光谱仍然是混合状态的像元;如果估计的材料数目过多,提取出的端元中将很有可能包含冗余噪声成分。基于光谱线性混合模型,提出一种改进的p范数纯像元辨识算法。主要利用光谱数据具有近似于低维流形的特性,首先采用正交投影的原理,将提取的端元扩充至正交投影算子中,然后分析投影后各个像元向量的p范数值,最终将p范数值高于阈值的向量个数作为材料种类数目。对实测碎片常用材料和美国地质勘测局数据库分别进行仿真实验,实验结果表明:提出的方法在估计材料种类数目的同时,还能提取出目标所包含的材料光谱,这在一定程度上提高混合光谱分解过程的自动化程度;相对于现有的一些主流算法,该方法有较强的鲁棒性,并且在信噪比不高的情况下仍能正确地估计空间碎片材料种类数目。
[Abstract]:When observing space debris, the structure of debris is compact, the composition of material is complex, and the spatial resolution of ground-based observation equipment is limited, the same pixel usually contains information of various materials, that is, "mixed pixel" is produced. At present, the research on mixed pixels at home and abroad is mainly focused on obtaining pure material spectrum and abundance of mixed pixels. It is often neglected that the determination of the number of pure matter in hyperspectral data is essential for mixed pixel analysis without any prior information. If the estimated number of materials is too small, the spectrum of the material will still be the pixel in the mixed state. If the estimated number of materials is too large, the extracted end elements will probably contain redundant noise components. Based on the spectral linear mixed model, an improved p-norm pure pixel identification algorithm is proposed. In this paper, the spectral data is used to approximate the low dimensional manifold. Firstly, by using the principle of orthogonal projection, the extracted endelements are extended to the orthogonal projection operator, and then the p-norm values of each pixel vector after projection are analyzed. Finally, the number of vectors whose p-norm value is higher than the threshold value is taken as the number of materials. Simulation experiments were carried out on the commonly used debris materials and the USGS database. The experimental results show that the proposed method not only estimates the number of materials, but also extracts the spectrum of materials contained in the target. This method can improve the automation of the mixed spectral decomposition process to some extent, and it has strong robustness compared with some existing mainstream algorithms. And the number of space debris materials can be estimated correctly under the condition of low signal-to-noise ratio (SNR).
【作者单位】： 北京航空航天大学仪器科学与光电工程学院精密光机电一体化技术教育部重点实验室;
【基金】：国家自然科学基金项目(61575015)资助
【分类号】：O657.3;V528
，

本文编号：1907954