文章基本信息

标题：CRF based Remote sensing image segmentation using Co-Sand algorithm
本地全文：下载
作者：I.G. Rowlandcy ; J. Suganthi
期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2016
卷号：5
期号：4
页码：1140-1146
出版社：Shri Pannalal Research Institute of Technolgy
摘要：In this study, we research the issue of multiclass pixel marking of very high-resolution (VHR) optical remote sensing pictures. For MR images, the panchromatic and multispectral segments are handled freely, separating both the edge maps and the morphological and phantom markers that are in the end intertwined at the most elevated determination, along these lines maintaining a strategic distance from any data misfortune incited by pansharpening. We propose a novel higher request potential capacity taking into account nonlocal shared limitations inside of the system of a conditional random field (CRF) with Co-sand model. The proposed approach consolidates grouping learning revelation from marked information with unsupervised division signals got from the cosegmentation of test information. The cosegmentation of unannotated test information joins nonlocal imperatives, which are encoded in a novel truncated powerful consistency potential capacity. The class names are then redesigned iteratively by exchanging between evaluating semantic divisions utilizing CRF and coordinating cosegmentation-inferred labels in higher request potential capacities to refine naming results. We tentatively exhibit the enhanced marking exactness of our methodology contrasted and best in class multilevel CRF approaches in light of quantitative and subjective results. We likewise demonstrate that our methodology can address the issue of lacking precisely named preparing information
关键词：Conditional random field; cosegmentation; ; high.