基于动态阈值的HJ-1B图像云检测算法研究

Research on Algorithm of Cloud Detection for HJ-1B Image Based on Dynamical Thresholding

针对环境与灾害监测预报小卫星星座1B星(HJ-1B)的传感器波段特征,通过对云检测阈值的年度波动性分析,提出了一种基于波谱标准差异常的"动态阈值云检测算法"。该算法的实现主要是采用图像配准、波段运算、线性回归和误差分析等方法,获取云异常区域,进而剔除图像中的云像元。研究结果表明,该算法可以很好地检测出不同时相、不同类型下垫面上空的云像元,为有效利用HJ-1B数据并提高其图像分类精度发挥作用。

Through analyzing the annual volatility of cloud detection thresholds and integrating the band characteristics of HJ-1B,the authors proposed a new algorithm of cloud detection for HJ-1B image based on dynamical thresholding according to the spectral standard deviation anomaly.Using image registration,band math,linear regression and error analysis,the authors acquired the cloud abnormal regions which could be used to remove cloud pixels from the image.The results show that this approach can detect cloud pixels over different periods and in different scenes successfully,thus promoting the use of HJ-1B data and improving the precision of image classification.