一种基于多通道卫星云图的台风分割方法

有效的台风主体分割能够提高台风预报的准确性.提出一种基于矢量图像的改进水平集方法进行台风分割,将C-V模型推广到矢量空间,加入了无需重新初始化的能量项;提出了一种迭代收敛条件并设计了针对云图特点的模型参数及分割方法,该方法利用了多通道的云图信息,避免了无谓的迭代计算,能够加快分割曲线的演化速度并提高分割的准确性.实验结果验证了所提出方法的有效性和可行性,其平均目标错误率与平均非目标错误率分别为1.21%与2.25%,可得到较理想的台风分割效果. 

Method base on multi-channel satellite cloud image for typhoon segmentation

A modified level set method for typhoon segmentation base on vector image was proposed to improve the accuracy of typhoon prediction.The C-V model was extended to the vector space and an energy term without re-initialization was introduced.An iteration convergence condition was proposed,and the parameters of the model and the method of segmentation were designed for the features of cloud image.Through the use of cloud image information and avoiding the unnecessary iterative calculation,this method can accel...