|
Optical remote sensing observations often are the primary data source for many studies and applications in large scale,
even in tropical regions. Frequent clouds over moist tropical regions often cause difficulty in obtaining good-quality
high-resolution images. Different cloud styles and shadows make it hard to be masked on the images. The ZY-1 02C
satellite, where multi-spectral and panchromatic imagers were on-boarded, was launched on 22 Nov. 2011. The objective
of this satellite is to acquire data contributing for earth resources and environmental monitoring, as well as for other
applications such as land use and disaster reduction. The multi-spectral imager has three wavebands, centered at 0.55
micrometer (green), 0.66 micrometer (red) and 0.83 micrometer (near infrared), at a spatial resolution of 10 meters.
Panchromatic band, centered at 0.68 micrometer, has a spatial resolution of 5 m and is not used in this study. A twostage
algorithm is presented to detect cloud and cloud shadow for ZY-1 02C multi-spectral measurements in this study.
First, maximum and minimum filters with a moving window size of 5 by 5 pixels are operated on ZY-1 02C multispectral
measurements. Optimal thresholds are selected by spatial statistics and visual examination. Pixels in maximumfiltered
images with a grey level higher than the given thresholds (Tc) are labeled as potential cloud. In contrast, pixels in
minimum-filtered images with a grey value lower than the given thresholds (Ts) are considered as potential shadow.
Second, the contextual are used to mask out errors with potential cloud and shadow (e.g. vegetation canopy-cast shadow,
road and bare soil) in previous stage. A window size of 9 by 9 centered at each potential cloud position is searched. If no
potential shadow is found the potential cloud is rejected. Meanwhile, each potential shadow also is tested to get the final
cloud and cloud shadow mask. Two ZY-1 02C multi-spectral images acquired over Pearl River Delta, a subtropical
region in South China, are used to validate the two-stage algorithm. The research result indicates that, two-stage
algorithm developed in this study has successfully detected the cloud and shadow on ZY-1 02C multi-spectral images.
|
