基于支持向量机的云相态分析

为了提高云相态的反演精度,结合MODIS多通道的光谱特性,首先提出了分别针对白天和夜间的MODIS云检测方法,得到了较高精度的云检测结果,为云相态反演奠定了基础。然后,依据不同相态云在不同光谱的辐射特性,选取了合适的特征参量,采用支持向量机方法反演云相态,在2009年7月7日的一次大暴雨过程中的云顶粒子相态反演试验中取得了较好结果。结果表明：精确的云检测结果有利于提高云相态反演精度，利用支持向量机的方法反演云相态是可行的,并且反演结果具有较高的精度。

Analysis of cloud phase based on support vector machine

To improve the inversion accuracy of cloud phase, the idea of moderate resolution imaging spectroradiometer (MODIS) cloud detection algorithm credibility and MODIS cloud detection methods were used, combining with the spectral characteristics of MODIS multicenter, and high pixel credibility for retrieval of cloud phase was provided. The differences of cloud phase’s spectrum were analyzed, and appropriate characteristic parameters selected, with the inversion of cloud phase based on the method of support vector machine.The cloud top particle phase of the rainstorm on July 7th 2009 was analyzed and significant results presented. The conelusions are that the good results of cloud detection help to improve the accuracy of cloud phase inversion,and that support vector machine approach to the inversion of cloud phase is feasible, and of with high accuracy the inversion results.