摘要
对高光谱影像的RX异常检测算法进行了研究。针对RX算法中对高维数据局部背景协方差矩阵估计存在较大误差的局限性,提出一种基于决策级融合的RX算子高光谱影像异常目标检测算法。首先,对同一场景下的可见近红外数据和短波红外数据分别运用经典的RX算子进行异常检测,得到初步异常检测的目标判决。在此基础上,利用传感器获取信息的冗余性和互补性特性,结合基于规则的决策级融合方法,得到最终的RX异常检测判决结果。在实测高光谱数据上进行了实验仿真,验证了本算法的有效性。
In this paper, an anomaly detection method based on RX algorithm for hyperspectral image is studied. Aiming at the error of estimating the local background covariance matrix for the RX algorithm, an anomaly detection algorithm based on decision-level fusion is proposed for hyperspectral RX detectors. Firstly, we apply the RX algorithm to calculate the preliminary result of anomaly detection for hyperspectral image on visible/near-infrared bands and short-wave infrared bands respectively. On this basis, considering the information redundancy and complementary characteristics of different sensors, we construct a decision-level fusion method to refine the result of the previous RX anomaly detection. The experimental results based on a real data set show that the proposed method achieves satisfactory improvement.
出处
《红外技术》
CSCD
北大核心
2013年第6期339-344,共6页
Infrared Technology
基金
国防基础科研计划
教育部留学回国人员启动基金
关键词
高光谱影像
异常检测
RX算法
决策级融合
hyperspectral image
anomaly detection
RX detector
decision-level fusion
