Identifying and tracking storms in satellite images

The identification and tracking of individual storms visible in satellite weather imagery is a difficult task. Both the correlation methods and cell-identification methods currently available have drawbacks. In this paper, we use a segmentation-based approach to identify storms in images and place the storms in a heirarchial scheme. A genetic-algorithm method of matching these segmented regions across frames is introduced in this paper. We show how this method can easily handle splits and merges of storm cells. Since it deals with all scales, this method overcomes the drawbacks inherent in cross-correlation and cell-tracking methods.