题目：Mapping vegetation-covered urban surfaces using seeded region growing in visible-NIR air photos
领域：REMOTE SENSING 三区
来源：IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
发表年代：2014年
作者：Jianhua Zhou*, Yan Huang, Bailang Yu.
摘要：Unreliability involved in the extraction of shadedvegetation-covered surfaces (VS) is a common problem in urbanvegetation mapping. Serving as a solution to it, a novel methodnamed Nonlinear Fitting-based Seeded Region Growing (NFSRG)is explored. With NFSRG, a series of classified results are organizedby a seeded-region-growing process. In order to adapt to thevariable separability between VS and background, the growingis limited in several weighted buffers defined by some nonlinearfitting relationships. When searching new VS members (membermeans both pixel and patch) within such a buffer, a graduallyreduced weight makes the buffer width continually narrowed asthe separability worsens. To avoid unexpected entrances of waterand smooth shaded background members, a during-growing constraint,named expansion rate, is proposed. Accuracy assessmentsreveal that more than 96% of VS members can be accuratelyextracted by the proposed method.
