基于变分水平集方法的遥感影像地物分割方法研究

Research of the Segmentation of Remote Sensing Image's Surface Features Based on Variation Level Set Method

影像信息提取自动化程度低是高分影像应用潜力得不到充分发挥的主要限制因素,水平集分割方法凭借其自由拓扑性及多信息共融性得到了学者们的广泛重视与研究。本文通过实验研究分析了变分水平集方法中内部能量项参数μ、阶跃函数Hε(z)和狄拉克函数δε(z)的组成参数ε对遥感影像地物分割的影响,并提出了两种结合最近邻插值法和高斯平滑的变分水平集方法,提高了分割准确性且提高了效率。

Low level of image information automatic extraction is a major factor that limits the full potential of high-resolution image.Level set segmentation method has received extensive attention and research by scholars for its free topology and rich information communion.In this paper,the influence of the internal energy term’s coeffect μ,the composition parameterεof the Heaviside function Hε(z) and Dirac function δε(z) on variational level set method for image segmentation is analyzed by experiments.And two variational level set methods which are combined with Nearest Neighbor Interpolation and Gauss Filter are proposed to optimize the accuracy of the segmentation and improve the efficiency.