项目名称: 基于人工蜂群算法的高光谱遥感图像端元提取方法
项目编号: No.41201356
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 地理学
项目作者: 孙旭
作者单位: 中国科学院遥感与数字地球研究所
项目金额: 25万元
中文摘要: 由于成像光谱仪的空间分辨率限制,混合像元问题在高光谱遥感图像中普遍存在,端元提取是进行混合像元分解的关键步骤,对混合像元分解的精度有重要的影响,因此端元提取算法的研究一直是高光谱图像处理领域的重点研究内容。目前广泛使用的端元提取算法大部分是基于线性光谱混合模型的几何学描述,普遍存在端元评价标准简单、端元数量依赖于图像波段数量并且在端元提取过程中缺乏正反馈机制的不足,使得图像中的个别异常像元对端元提取结果产生重大的不利影响。群智能算法中的人工蜂群算法具有自组织学习、间接信息交流、并行分布式计算等特点,这些特点恰好能够克服现有端元提取算法的不足。本项目从混合像元的数学模型和蜂群算法的机理两个方面入手,重点研究蜂群算法中的各种规则与混合像元问题的联系,尝试设计一种以蜂群算法为基础的高光谱图像端元提取方法,利用蜂群算法的优势提高端元提取的精度,避免异常像元对端元提取的影响。
中文关键词: 高光谱图像;混合像元分解;端元提取;人工蜂群算法;
英文摘要: Due to the spatial resolution limitation of the imaging spectrometer, the problem of mixed pixel always exist in hyperspectral remote sensing image. Endmember extraction is a key step of mixed pixel unmixing, which infuleces a lot in its precision. therefore, the study of endmember extraction algorithms has been the focus of hyperspectral image processing research. Currently, the widespread used endmember extraction algorithms are based on linear spectral mixture model of geometry description, which lead to many limitations such as simple evaluation criteria of endmembers, the endmember number on image's bands quantity dependence, the lackage of positive feedback mechanism. This makes some individual abnormal pixels in image result to a significant adverse impact in endmember extraction. Artificial Bee Colony (ABC) algorithm, a new swarm intelligence algorithm, with the features of self-organized learning, the indirect exchange of information and parallel distributed computing, can effectively overcome the deficiencies of the existing endmember extraction algorithms. This research will start with the study of mathematical model of mixed pixels and the mechanism of ABC algorithm, mainly study the relationship between the rules of ABC and mixed pixels, and try to design an new endmember extraction method of hyper
英文关键词: Hyperspectral image;Spectral unmixing;Endmember extraction;Artificial bee colony ;