基于台风无论是在生成期、成熟期, 还是在消亡期, 都具有螺旋性的特征, 非台风云系无此特征, 充分挖掘云系边界特点, 统计出单幅云图中各云系边界的旋转程度。运用Bezier直方图曲率曲线两次确定分割阈值, 迭代分割卫星云图, 结合台风的旋转与面积、形状等几何特性, 识别卫星云图中的台风云系。实验表明该方法对台风云系有很好的识别率。

Typhoon has different characteristics, such as texture, shape, and area, in different development stages. We could not automatically recognize the typhoon clouds in all stages based on these features. During different development stages, typhoon all has helicity and non-typhoon has no helicity. Based on this, we extract boundary information of clouds and statistics of the rotation degree of boundary clouds in single satellite image. In this paper, we use curvature curve of Bezier histogram to obtain two segmentation thresholds, iteratively segment the satellite image, and combine typhoon’s geometric features, such as rotation, area and shape to recognize the typhoon. Experimental results show that the typhoon can be recognized effectively in all different development stages.