欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2014, Vol. 31 ›› Issue (4): 484-491.DOI: 10.7523/j.issn.2095-6134.2014.04.007

• 环境科学与地理学 • 上一篇    下一篇

一种FNEA影像分割算法中初始对象的快速构建方法

邓富亮1, 范协裕1, 王刚1, 马娟2   

  1. 1. 中国科学院遥感与数字地球研究所 遥感科学国家重点实验室, 北京 100101;
    2. 中国地质环境监测院, 北京 100081
  • 收稿日期:2013-05-21 修回日期:2013-09-04 发布日期:2014-07-15
  • 通讯作者: 邓富亮,E-mail:fldeng8266@gmail.com
  • 基金资助:

    国家重大科技专项(02-Y30A04-9001-12/13,30-Y20A02-9003-12/13)资助

A fast method for creation of the initial objects for FNEA image segmentation

DENG Fuliang1, FAN Xieyu1, WANG Gang1, MA Juan2   

  1. 1. Institute of Remote Sensing and Digital Erath Chinese Academy of Sciences, State Key Laboratory of Remote Sensing Information Sciences, Beijing 100101, China;
    2. China institute of Geo-Environment Monitoring, Beijing 100081, China
  • Received:2013-05-21 Revised:2013-09-04 Published:2014-07-15

摘要:

针对分形网络演化法存在分割效率相对较低和区域合并准则无法适用于单像素区域对象2个问题,提出采用快速扫描法构建初始区域对象,进而采用基于异质性最小区域合并算法实现分形网络演化分割.实验证明,快速扫描法能快速构建初始区域对象.通过设置适当初始阈值参数,在不影响分割精度情况下,该方法较大程度上提高了整体分割效率.从可信度和通用性角度来看,具有较高的实用价值.

关键词: 分形网络演化法, 快速扫描法, 高分辨率遥感影像, 图像分割

Abstract:

The fractal net evolution approach (FNEA) is an object-oriented segmentation method with high precision for high spatial resolution remote sensing images. However, experiments indicate that this method has two problems. It has considerable low execution efficiency, and the region merging criteria cannot be applied to a region object with single pixel. A fast scanning image segmentation method is proposed to generate initial region objects, and the region merging approach based on the minimum heterogeneity principle is applied to the FNEA. Experimental results show that the proposed method significantly improves efficiency by setting appropriate threshold parameters without loss of segmentation accuracy. So, the method has highly practical value in universality and reliability.

Key words: fractal net evaluation approach(FNEA), fast scanning image segmentation, high resolution remote sensing image, image segmentation

中图分类号: