Classification of clouds in the Japan Sea area using NOAA AVHRR satellite images and self-organizing map

摘要

This paper introduces a method for cloud classification using NOAA AVHRR satellite images. AVHRR (Advanced Very High Resolution Radiometer) data consists of five-channel multi-spectral images. To reduce the dimensionality of the data, principal component analysis (PCA) is calculated for each channel separately. The most significant principal component values are then composed into an image feature vector. Finally, the feature vectors are clustered using self-organizing map (SOM). This method is applied for the study of winter season clouds in the Japan Sea area.