An adaptive segmentation arithmetic adapted to intertwined irregular convective storm images

摘要

Storm cell is the fundamental element of forming varieties of strong convective disastrous weather. It is usually difficult to distinguish the cell because in an echo image the shapes of storm cells are complicated and the distributions are different inside while intertwined outside. This paper grasps the distribution that the storm cell's intensity decreases from inside to outside. Then get to start from the strong cell nuclear areas which are determinately independent of each other and design the algorithm of molting layer by layer to obtain a series of independent largest cell nuclear. This method can not only adaptively ascertain the sum of the subjects waiting to be divided, but also tremendously diminish the later division time. This paper also proposes expansion of collision avoidance algorithm based on nuclear in order to settle the adaptive division problem of the multi-cell subjects intertwined outside. The test result shows that the partly overlapped storm cells, which are divided by the method proposed by this paper, have integrated hierarchical structure but don't contain irrelevant information. This train of thought and arithmetic can also be adapted to other adaptive division of multi-subject which is intertwined.