期刊文献+

面向对象矿区复垦植被分类最优分割尺度研究 被引量:5

Study of optimal segmentation scale in object-based classification for rehabilitated vegetation in coal mining site
收藏 分享 导出
摘要 针对面向对象分类中直接影响最终分类精度的分割尺度问题,该文通过基于对象的精度评价,定量对比分析eCognition软件自带的ESP工具、欧几里得距离二指数(ED2)和分割误差(sE)3种分割尺度选择方法,研究不同尺度在矿区复垦植被面向对象分类的精度差异,为高分遥感数据分割尺度的选择提供定量的标准。得出结论:分割尺度的微小变化会导致分类结果的显著差异;ESP工具分类精度最低,SE和ED2全面考虑参照样本与分割对象之间的特征和几何差异,可显著提高面向对象分类精度;将SE和ED2作为最优分割尺度选择算法嵌入面向对象分类体系中,可以快速获取最优分割尺度。 As the most important part of OBIA, multi-resolution segmentation has direct impact on the final classification accuracy. The ESP tools in eCognition software, Euclidean distance 2 index (ED2) and segmentation error(SE)methods were chosen to select optimal segmentation scale in object-based clas- sification for rehabilitated vegetation in mining site. The quantitative accuracy assessment method based on the segments was applied to develop the segmentation scale and evaluate each method. The results showed that the relatively small variations in scale parameter could lead to significant differences of final classifica- tion results; The ESP tools carried with eCognition software had the lowest accuracy, while ED2 and SE significantly improved the classification precision of OBIA with feature and geometric discrepancies being considered; embedded SE and ED2 in OBIA system as new optimal segmentation scale selection would oh- rain the optimal segmentation scale efficiently and accurately.
作者 李娜 包妮沙 吴立新 刘善军 刘小翠 LI Na , BAO Nisha , WU Lixin, LIU Shanjun , L IU Xiaocui(1. College of Resource and Civ il Engineering, Northeastern University, Shenyang 110004, China; 2. IoT Perception Mine Research Center, China University of Mining and Technology, Xuzhou 221116, China; 3. Liaoyang Land Survey and Plan Institute, Liaoyang 111000, China)
出处 《测绘科学》 CSCD 北大核心 2016年第4期66-71,76共7页 Science of Surveying and Mapping
基金 中央高校基本科研业务经费项目(N120801002) 国家自然科学基金项目(4140010440)
关键词 复垦植被 Worldview-2影像 面向对象分类 分割尺度 rehabilitated vegetation Worldview-2 imagery object-based classification segmentation scale
作者简介 李娜(1989-),女,河南巩义人,硕士研究生,主要研究方向为矿区环境遥感。E-mail:sky_faye@126.com
  • 相关文献

参考文献16

二级参考文献101

共引文献332

同被引文献74

引证文献5

二级引证文献2

投稿分析

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部 意见反馈