DEVELOPMENT OF IRREGULAR CLOUD CLUSTER ENCAPSULATING STRUCTURE FROM SATELLITE INFRARED IMAGES

摘要

A major task in flood prediction is the estimation of precipitation. Detection and tracking of convective cloud is an integral part of it. Convective cloud can be detected from thermal infrared (TIR) images because clouds are associated with very low temperature. They can be tracked in a series of TIR images, so that their position can be estimated in order to identify the place of precipitation. For tracking of cloud structures in a series of TIR images, the shape of these structures are required to be determined, so that each cluster can be identified as a single object. Additional features based on the shape can be determined for effective tracking of cloud clusters in a sequence of images. In the present paper, a cloud cluster encapsulating structure developing technique has been proposed that fits absolutely close to a cluster and does not include any non-cluster pixels at all. The enclosing structure takes exactly the form of the individual cluster and hence is unique for each cluster. It helps to treat a cluster of pixels as a single object and can be helpful in its tracking.