首页 | 本学科首页   官方微博 | 高级检索  
     检索      

一种基于分割图斑的海岸带遥感图像分类方法
引用本文:吴均平,毛志华,陈建裕,潘德炉.一种基于分割图斑的海岸带遥感图像分类方法[J].海洋学研究,2006,24(2):70-78.
作者姓名:吴均平  毛志华  陈建裕  潘德炉
作者单位:国家海洋局,第二海洋研究所,国家海洋局,海洋动力过程与卫星海洋学重点实验室,浙江,杭州,310012
摘    要:传统的基于象元的分类方法在处理海岸带这类动态多变、混杂度大的地物时常出现“椒盐现象”,难以矢量化形成专题图。提出了一种新的遥感图像分类方法——基于分割图斑的分类法,即选择一个合适的尺度,利用光谱信息按照一定的策略将图像分割为一系列的图斑,并确保图斑内大多数象元的光谱特征相近,分别对图斑内诸象元进行统计,求出各个波段的均值,用该均值替换图斑内所有象元的原始亮度值,这样分类时可以确保同一图斑内的所有象元被分到同一个类别中,从而有效地避免了“椒盐现象”,使分类结果便于矢量化成图,也便于对特定地物的提取。实验结果表明,该方法还能在一定程度上提高分类精度。

关 键 词:海岸带  图像分割  分类  尺度
文章编号:1001-909X(2006)02-0070-09
修稿时间:2005年2月28日

A new classification method for coast remote sensing image
WU Jun-ping,MAO Zhi-hua,CHEN Jian-yu,PAN De-lu.A new classification method for coast remote sensing image[J].Journal of Marine Sciences,2006,24(2):70-78.
Authors:WU Jun-ping  MAO Zhi-hua  CHEN Jian-yu  PAN De-lu
Abstract:The "salt and pepper" is a serious problem of traditional pixel-based classification method when processing the coastal image with dynamic land types. A new classification approach based on plot segmentation is proposed in this paper.An image is segmented into a series of plots according to their spectral features in an appropriate scale to make sure majority pixels in one plot have similar spectrum.And then,within each plot,the intensity of all pixels in each band is averaged and replaced by the average value.In this case,all pixels will be classified into the same class,avoiding the "salt and pepper",and all of classes are totally independent.Experiments show that this new approach is not only easy for image vectorization and object extraction,but also able to improve the classification accuracy.
Keywords:coastal zone  image segmentation  classification  scale
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号