基于最小异质性区域生长法的多尺度城市地物影像分割  

Multi-scale Segmentation of Urban Land Image base on the Region Growth of Minimum-heterogeneity

作　　者：肖奥[1] 赵文吉[1] 

机构地区：[1]首都师范大学,北京100048

出　　处：《首都师范大学学报（自然科学版）》2010年第1期60-64,共5页Journal of Capital Normal University：Natural Science Edition

摘　　要：遥感影像分割是面向对象影像分类中的关键技术.针对地物景观的多尺度效应与高分辨率影像的高空间分辨率特征研究了多尺度分割与多特征最小异质性准则,分析了其有效性及可行性.基于区域生长的原理与方法,采用多尺度分割与最小异质性算法,对不同分辨率遥感影像和地物类型的样区进行分割技术的应用实践,分析总结了分割参数与影像特征参数的关系和各种参数的确定方法.Segmentation is the key of object-oriented image classification. Due to the multi-scale character of landscape and the high resolution character of remotely sensed image, we research the multi-scale segmentation and the minimum-heterogeneity rule. Based on the scale affect and minimum-heterogeneity rule, this paper presents the necessity and possibility of multi-scale affects analysis as well as the principle and practice of the region growing image segmentation. There are many sites of different resolution remotely sensed image and land types to test the segmentation process. The result show the contact between the segmentation parameter and image character parameter and the manner of choose all the parameter.

关 键 词：影像分割 面向对象分类 高分辨率 多尺度 区域生长. 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]