项目名称: 基于仿生视觉的高光谱遥感异常检测方法研究
项目编号: No.41301448
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 天文学、地球科学
项目作者: 李敏
作者单位: 河海大学
项目金额: 25万元
中文摘要: 高光谱遥感异常检测可以在无先验知识的条件下检测与周围背景存在光谱差异的目标,在诸多领域(如:国防、军事安全等)具有重要的应用价值。然而,大尺度空间观测下地物属性的精细表征导致的光谱干扰严重、复杂等问题,致使现有的异常目标检测方法存在高虚警率、低识别率等不足。为此,受蝇视觉系统快速、完备信息处理模式、高效压缩融合机理以及自适应敏感性调节机制等生物学机理的启发,提出并初步建立一套基于仿生视觉的高光谱遥感异常检测理论和方法。开展如下研究:多孔径"同构异步"分块信息仿生处理模式研究;仿生视觉异常感知机理建模;高光谱遥感异常检测算法以及"弹性"评价指标设计。重点解决:(1)高光谱遥感数据立方孔径尺寸及映射函数设计;(2)信息的高效压缩、超敏锐度融合以及有效增强与提取机理建模;(3)视神经纤维非线性自适应"异常"敏感性调节机制的研究与模拟。为该技术在实际中的应用打下扎实的理论基础和实验验证准备工作。
中文关键词: 高光谱遥感;异常检测;数据融合;仿生算法;特征提取
英文摘要: anomaly detection of hyperspectral remote sensing is a techonolgy that can detect spetral anomaly target refer to the neighborhood without any prior knowledge, and has important value in many application domain. However, finer description of ground content attribution under large scale measure of spatial domain induced badly and complex disturbtion of spectra that makes present anomaly detection method have higher false alarm rate,lower recognise rate and so on. So inspired by the advantages of fly vision such as fast and completed information procession, effective compression fusion mechanism, self-adaptive sensitivity modulation.this project try to proposed and construct a suit of theory and method that used to anomaly detection for hyperspectral data based on bionic vision. It included the research of block-based information process inspired by multi-apperture "same form and asynchronism" of fly vision, bionic anomaly perception model and bionic anomaly detection algorithm for hyperspectral remote sensing. foucsing on three problems: (1) design of aperture size and mapping function of hyperspectral data cube; (2)model of information compression, super acuity information fusion,enhancement and extraction mechanism;(3)research on the self-adaptive mechanism of sensitivity modulation of vision neuro fiber.
英文关键词: Hyperspectral remote sense;Anomaly detection;Data fusion;Bionic algorithm;Feature extraction