Typhoon Locating and Reconstruction from the Infrared Satellite Cloud Image

Abstract

Typhoon inflicts terrible damage due to thunderstorms, violent winds, torrential rain, flooding and extreme high tides. Improving the early typhoon forecast capability is important for the disaster prevention. In recent years, many scholars have made efforts in typhoon center location, typhoon intensity estimation and moving path prediction from the satellite images. Moreover, it may be useful to transfer the useful information of typhoon to the audience in a low bandwidth environment. In this paper, we proposed a novel approach that partitions the satellite cloud image into slices to extract typhoon features. Then we used morphology operations and statistical image classification methods to locate the center and the contour of the typhoon. In addition, we applied the vector quantization method to encode typhoon satellite images owing to its fast image reconstruction capability in different resolutions and regions of interest. The infrared satellite images of 71 typhoons from mature period to decaying period occurred between 1995 and 2006 are tested in this paper. We used the proposed method to recognize, locate and reconstruct typhoon images from these collected satellite cloud images. Experimental results show that the main body of a typhoon image can be segmented effectively from the infrared satellite cloud image with complex background and reconstructed in regions of interest and resolution required.