Raft cultivation area extraction from high resolution remote sensing imagery by fusing multi-scale region-line primitive association features
Raft cultivation area extraction from high resolution remote sensing imagery by fusing multi-scale region-line primitive association features
复制标题
融合多尺度区域线原始关联特征的高分辨率遥感影像筏式耕作面积提取
DOI：
10.1016/j.isprsjprs.2016.10.008
复制
发表时间：
2017
影响因子：
12.7
通讯作者：
Lv Guonian
中科院分区：
文献类型：
--
作者：
Wang Min;Cui Qi;Wang Jie;Ming Dongping;Lv Guonian
本文首先在区域线基元关联框架（RLPAF）的基础上，提出了几个用于基于对象的图像分析的新概念，包括基于线的形状规则性、线密度和基于尺度的最佳特征值（SBV）。在此基础上，提出了一种基于多尺度特征融合和空间规则归纳的高空间分辨率遥感影像筏式养殖区域提取方法。该方法包括以下步骤：（1）利用HBC-SEG图像分割方法得到多尺度区域基元（线段），利用基于相位的直线检测方法得到直线基元（直线）。(2)基于RLPAF建立区域与直线之间的关联关系，提取多尺度RLPAF特征并选择SBV。(3)设计了几种空间规则来提取陆地和水分离后的海洋沃茨内的RCA。实验结果表明，该方法可以成功地从HR图像中提取不同形状的RCA，具有良好的性能。
In this paper, we first propose several novel concepts for object-based image analysis, which include line-based shape regularity, line density, and scale-based best feature value (SBV), based on the region-line primitive association framework (RLPAF). We then propose a raft cultivation area (RCA) extraction method for high spatial resolution (HSR) remote sensing imagery based on multi-scale feature fusion and spatial rule induction. The proposed method includes the following steps: (1) Multi-scale region primitives (segments) are obtained by image segmentation method HBC-SEG, and line primitives (straight lines) are obtained by phase-based line detection method. (2) Association relationships between regions and lines are built based on RLPAF, and then multi-scale RLPAF features are extracted and SBVs are selected. (3) Several spatial rules are designed to extract RCAs within sea waters after land and water separation. Experiments show that the proposed method can successfully extract different-shaped RCAs from HR images with good performance.