基于波段子集特征融合的高光谱图像异常检测

Anomaly Detection in Hyperspectral Imagery Based on Feature Fusion of Band Subsets

高光谱图像分析中,对未知环境下伪装目标的检测识别具有较大难度,因为缺乏背景与目标的先验光谱信息.针对这一问题,提出一种高光谱图像异常检测算法.将高光谱图像分成波段子集进行特征提取,利用对图像中噪声程度及目标、背景之间可分性敏感的特征样本高阶统计量构造基本置信指派函数,通过DS证据推理实现特征层智能融合异常检测.理论分析及仿真实验结果表明了算法的有效性.

Detecting camouflaged targets in an unknown environment presents a great challenge in hyperspectral image analysis since the prior knowledge about targets and background is not available. A nomaly detection method for hyperspectral imagery was proposed for this problem. Features were extracted from subband sets of hyperspectral imagery,then fusion algorithm for detection was implemented by D-S evidence reasoning while basic belief assignment function was constructed involving high-order moments of features. Theoretical analysis and results of experiment verify the effectiveness of the algorithm.