基于先验特征的台风云系影像分割方法研究

Research on the Typhoon Segmentation Method with Priori Features

传统的基于数学形态学的台风云系影像分割方法多采用圆形或菱形模板进行,这种方法的缺点是当运算模板较大或经多次腐蚀、膨胀运算后,不能很好地保留台风云系影像原始的形态信息。为了解决此问题,探讨了一种基于先验特征的台风云系影像分割原理与方法,即先提取台风云系影像的大致形态,并以此作为形态学模板;然后利用形态学方法提取台风云系影像;最后,以2008年7月28日18时的09号FY-2C气象卫星云图台风云顶温度数据为例进行分析,同时根据本次示例台风(200809号台风)形状自定义的模板来腐蚀分割台风云系影像,并详细对比和分析了不同形状、大小的模板对台风云系影像分割结果的影响。实验表明,基于先验特征的自定义台风云系结构模板,其分割结果不仅较好地保留了台风云系影像的形状特征,而且腐蚀后保留的台风云系信息最多。

The traditional segmentation of typhoon cloud based on mathematical morphology usually uses round or diamondshaped template to segment satellite images. This conventional approach fails to keep a good typhoon original form information after large or multiple erosion or dilation. In order to solve this problem, the authors studied the theory and method of typhoon segmentation with priori features in this paper. Concretely speaking, the authors first extract the general form of typhoon cloud as a morphological template, and then extract typhoon cloud by morphology methods. With TBB data of FY2C, NO. 200809 typhoon obtained at 6：00 p. m on July 28 as an example, the authors made a template according to the shape of the typhoons to erode clouds images, and then made a detailed comparison as well as analysis of different template shapes or sizes Of the cloud segmentation results. The results show that the priori features templates segmentation results can not only well retain the characteristics of typhoon cloud but also preserve most information after cloud erosion.