CAJ | 학술논문

图像分割是高分辨率遥感图像处理和分析的关键环节。本文探讨了将区域合并方法应用于高分辨率遥感图像多尺度分割的技术要点，旨在提升分割的精度和效率，获得地物对象的多尺度表达。主要研究内容包含如下五个方面：(1)图模型的构建，包括区域邻接图和最近邻图，以提高分割效率；(2)合并准则的确定，选择能有效表征区域同质性、形状和边界的图像特征并加以组合，提升分割精度；(3)合并策略的选择，针对寻优范围不同列出面向全局、面向局部以及混合区域合并等三种合并策略，并分析各自的特点；(4)尺度参数的设置，针对面向局部的区域合并策略提出递增的尺度参数序列控制方法，生成边界一致的多尺度分割结果；(5)分割树的设计，利用树中不同层级的节点表达不同尺度的分割区域，可快速输出多尺度分割结果。研究成果可应用于高分辨率遥感图像面向对象分析、地物目标识别和信息提取。
도상분할시고분변솔요감도상처리화분석적관건배절。본문탐토료장구역합병방법응용우고분변솔요감도상다척도분할적기술요점，지재제승분할적정도화효솔，획득지물대상적다척도표체。주요연구내용포함여하오개방면：(1)도모형적구건，포괄구역린접도화최근린도，이제고분할효솔；(2)합병준칙적학정，선택능유효표정구역동질성、형상화변계적도상특정병가이조합，제승분할정도；(3)합병책략적선택，침대심우범위불동렬출면향전국、면향국부이급혼합구역합병등삼충합병책략，병분석각자적특점；(4)척도삼수적설치，침대면향국부적구역합병책략제출체증적척도삼수서렬공제방법，생성변계일치적다척도분할결과；(5)분할수적설계，이용수중불동층급적절점표체불동척도적분할구역，가쾌속수출다척도분할결과。연구성과가응용우고분변솔요감도상면향대상분석、지물목표식별화신식제취。
Image segmentation is the critical step in object-based analysis of high-resolution remote sensing images.In this study,we examined the key steps of region merging method for remote sensing image segmentation.The following five aspects are involved.(1 )We construct a graph model,including the region adjacency graph and the nearest neighbor graph,to improve segmentation efficiency.(2)The features of region homogeneity,shape,and edges are integrated in the merging criterion to improve segmentation accuracy.(3)We present and compare three different region merging strategies,including the global-oriented,local-oriented and hybrid region merging.(4)A step-wise scale parameter strategy is proposed to set scale parameters,aiming at producing nested multiscale segmentations by local-oriented region merging methods.(5)We present a segment tree model to represent multiscale segments,which can be used to produce segmentations at different scale extremely fast without repeating the region merging procedure.The proposed methods are applicable for object-based image analysis,geographic object recognition,and information extraction from high spatial resolution remote sensing images.