Studies on Cloud Detection of Atmospheric Remote Sensing Image Using ICA Algorithm

In atmospheric remote sensing science, the cloud detection is an essential basis for the inversion of the parameters in atmosphere such as temperature, humidity, sea surface temperature etc. At present, the main cloud detection technologies can be divided to two categories, the method based on the different spectral characteristics of clear and cloudy area and the method based texture features of cloud relative to the earth surface. The Independent Component Analysis (ICA) developed in recent years is a method which can obtain higher order statistics information in the observed data. It not only can remove the correlation, but also can obtain images that are mutual independent. In the paper, the cloud detection of the atmospheric remote sensing image of AVHRR is tested using the FastlCA. The results of the experiment shows that the FastlCA can separate the land, water, cloud, and high level ice cloud area in the image exactly, with the results of high level ice cloud detection more precise than the results of other cloud detection method. This provides another good approach for the cloud detection.