南京大学学报（自然科学版）Issue(5)：1039-1048,10.DOI:10.13232/j.cnki.jnju.2015.05.015
基于协同分割的高分辨率遥感图像变化检测
Change detection from high-resolution remotely sensed images based on cosegmentation
摘要
Abstract
Due to the inconsistency of multi-temporal objects’boundaries for object-based change detection,this paper proposes a new change detection approach from multi-temporal high-resolution remotely sensed images based on the concept of cosegmentation in the filed of computer vision.First,multi-temporal remotely sensed images are co-processed to discover the change feature and a map of change intensity is obtained using the magnitude of spectral change between images.Then cosegmentation is performed under the guidance of the change intensity map, combined with each image features.Multi-temporal change objects with accurate boundaries and spatial correspondence are directly generated by energy function minimization finally.Experimental results obtained on multi-temporal aerial images show that multi-temporal change objects are preferably segmented and the change process can also be clearly acquired through establishing a correspondence between multi-temporal change objects. This novel method can provide a workable way for object-based change detection from high-resolution remotely sensed images.关键词
高分辨率遥感图像/多时相图像/变化检测/协同分割/最小割/最大流Key words
high-resolution remotely sensed images/multi-temporal images/change detection/cosegmentation/min-cut/max-flow分类
信息技术与安全科学引用本文复制引用
袁敏,肖鹏峰,冯学智,张学良,胡永月..基于协同分割的高分辨率遥感图像变化检测[J].南京大学学报（自然科学版）,2015,(5):1039-1048,10.基金项目
江苏高校“青蓝工程”(201423),浙江省科技计划(2014F50022) （201423）
