摘要
针对当前TM影像中细小河流提取存在的不连续问题,提出一种基于结构相似区域搜索的细小河流提取方法。首先,利用地物间不同的反射特性,使用水体提取模型对细小河流和无关信息进行区分;其次,结合TM不同波段上水体灰度值差异,设定不同阈值去除无关噪声;然后,通过搜索评估不连续河流区域,确定河流待连接断点;最后,利用河流像元间结构相似度,实现不连续细小河流的启发式搜索连接。研究结果显示,相比传统的乘性Duda算子、线状特征增强算子以及区域生长等方法,基于结构相似区域搜索的细小河流提取方法,有效地解决了细小河流提取不连续的难题,准确地实现了细小河流的完整提取。该方法对基于遥感技术的完整水体信息提取,具有较大的实际应用价值。
The structural similarity region search algorithm is used to realize the automatic extraction of TM image narrow rivers,which is of great value for disaster assessment and soil and water resources management.The discontinuity of narrow river extraction is the main problem which causes the difficulty in accurate obtaining of information about rivers.Many experts have studied various characteristic properties of water bodies to avoid the phenomenon of river information leakage during extraction.However,due to the complex flow of narrow rivers and the vulnerability to environmental disturbances,it is difficult to achieve complete extraction of river information.Combining structural similarity and heuristic search algorithm,this paper proposes a new method for accurately connecting faulted rivers.The specific process of the method is as follows:Firstly,according to the reflection characteristics of the ground objects,the water body extraction model is used to distinguish the narrow rivers from the irrelevant information.Then,the difference between the gray values of the water bodies on different bands is used to set different thresholds for unrelated noise removal.Third,the discontinuous rivers are evaluated by searching.The area is used to determine the breakpoints to be connected to the river.Finally,the heuristic automatic search connection is realized by using the structural similarity between the 5,4,and 3 bands of river pixels in the TM image.A comparison with several algorithms shows that the proposed method can solve the problem of river extraction fracture of traditional algorithms and realize the precise connection of discontinuous narrow rivers.
作者
孙玉梅
王保云
张祝鸿
韩文科
孙显辰
张玲莉
SUN Yumei;WANG Baoyun;ZHANG Zhuhong;HAN Wenke;SUN Xianchen;ZHANG Lingli(School of Information Science and Technology, Yunnan Normal University, Kunming 650500, China)
出处
《国土资源遥感》
CSCD
北大核心
2020年第2期63-72,共10页
Remote Sensing for Land & Resources
基金
国家自然科学基金项目“基于深度迁移学习的遥感影像中泥石流孕灾沟谷识别——以云南省为例”(编号:61966040)
国家级大学生创新训练项目“高山峡谷泥石流沟危险性评价研究(以怒江流域为例)”(编号:201810681009)共同资助。
关键词
遥感影像
细小河流
启发式搜索
结构相似
remote sensing image
narrow rivers
heuristic search
structural similarity