The Research Of Level Set Method In Radar Image Segmentation

Image segmentation is an important area of image processing research.It has been the rapid development in recent decades. It has been widely applied in medical,military, industrial, meteorological fields,and so on. As a preliminary step in image analysis, image segmentation is one of the basic problem in computer vision area. The quality of the segmentation result has a great influence on the next work. In this paper, level set method is used in radar image segmentation,and heavy rain clouds is segmented.Level set is a common method in image segmentation.This paper studies the theory of level set method,and specifically gives the two level set improved algorithm.Level set method is widely used in image segmentation, tracking and other fields.In this paper, the three kinds of ITK level set methods are tried to use in two dimensional radar image segmentation.The three kinds of ITK level set methods are Threshold Level Set Segmentation,Geodesic Active Contours Segmentation and Canny-Edge Level Set Segmentation.Threshold level set segmentation method combines threshold segmentation,Define a range of brightness values to type classification of the image.Find the spread factor through this brightness values based on the level set equation. Evoluting the curve with level set method.Geodesic Active Contours Segmentation is based on Shape Detection Segmentation and add the convection to the level set equation. Canny-Edge Level Set Segmentation is based on Canny-Edge and level set method.In this method,the convection coefficient is constructed by canny distance.This paper selects one of the typical examples as the experimental data from Beijing and Tianjin radar station.Firstly,the NVI method and eSnake external force model are used to treat radar data, This paper achieves the improvement and experiment of the method on the VC++ development platform, in the Windows environment. The results show that the three methods are effective.But Geodesic Active Contours Segmentation method is the best.