摘要
为了精确分割高分辨率全色遥感影像,提出一种基于多特征的区域化能量分割算法。利用曲波变换提取高分辨率全色遥感影像的边缘特征和纹理特征,结合原有的光谱特征构成特征集合;利用规则划分技术划分影像域,在划分的影像域上,利用能量函数和非约束Gibbs概率分布建立基于多特征的区域化能量分割模型;利用MCMC算法求解该分割模型,以实现高分辨率全色遥感影像分割。为了验证提出算法,以Worldview2全色遥感影像为实验数据进行分割,并对其分割结果进行定性及定量评价。
To accurately segment high resolution panchromatic remote sensing image,a regional energy segmentation algorithm based on multiple features is proposed.First,curvelet transform is used to extract the boundary and texture features of high resolution panchromatic remote sensing image,based on which a feature set is formed by combining with the original spectral features.Then the image domain is partitioned by regular tessellation,and on the partitioned image domain,the energy function and non-constrained Gibbs probability distribution are used to build a regional energy segmentation model based on multiple features.Finally,a MCMC algorithm is applied to solve the segmentation model.To verify the proposed algorithm,the Worldview2 panchromatic remote sensing images are used as test data for segmentation,and their segmentation results are evaluated quantitatively and qualitatively.
作者
马占瑛
王玉
尤号田
MA Zhanying;WANG Yu;YOU Haotian(Guangxi Key Laboratory of Spatial Information and Geomatics,Guilin University of Technology,Guilin 541004,China;College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541004,China)
出处
《无线电工程》
北大核心
2021年第8期732-737,共6页
Radio Engineering
基金
广西科技计划项目(2020GXNSFBA297096)
广西空间信息与测绘重点实验室基金项目(19-050-11-20)
桂林理工大学科研启动基金资助项目(GUTQDJJ2018065)。
关键词
多特征
区域化能量分割
高分辨率全色遥感影像
multiple features
regional energy segmentation
high resolution panchromatic remote sensing image
