采用主成分分析的改进云检测算法

Advanced algorithm of PCA-based Fmask cloud detection

针对Fmask云检测算法难以区分Landsat遥感影像上云和冰这一技术难题,该文提出一种基于主成分分析的改进Fmask云检测算法。首先对Fmask中归一化植被指数和归一化雪指数固定阈值进行自适应阈值改进,并对影像进行主成分变换,然后对主成分分析变换后的组合波段进行改进的Fmask云检测,最后进行算法对比分析。以北极地区的TM影像进行实验,结果表明,对同时覆盖冰层和云层的Landsat遥感影像,该文提出的算法能够提高云检测精度。

Cloud detection via Fmask usually experiences technical difficulties while being applied in Landsat remote sensing image with coverage of icy sheet and clouds,an advanced Fmask cloud detection algorithm called PCA_Fmask was proposed based on Principal Component Analysis in the paper.Firstly the thresholds of NDVI and NDSI were improved adaptively.Then Principal Component Analysis was employed as transformation for TM image,and the improved Fmask method was used to complete cloud detection in the transformed image.Finally,the performances of diverse algorithms were compared and analyzed.Experiment was carried out on TM images from Arctic region,the result proved that algorithm in this study could effectively improve the accuracy of detecting cloud in Landsat remote sensing image with ice and cloud coverage.