摘要
空间结构特征是高分影像最显著的特征之一,高分影像的各种地物类型都表现出不同的结构特征,这些结构既包括纹理的、几何的,也包括空间关系的,有效地利用高分影像的结构特征可以弥补光谱特征进行分类的不足。本文以高分影像的空间结构特征建模与信息的提取为主题,主要研究了空间自相关统计量的计算和利用空间自相关统计量对图像进行处理,以及基于空间半变差函数高分影像样本的空间结构信息提取,最后以空间自相关统计量与空间半变差函数所提取得空间结构信息为特征进行神经网络分类。
Spatial structure feature is one of the most significant features of high-resolution image.All kinds of surface features of high-resolution image show different degrees of structural features,including texture,geometry and spatial relationship.Effective use of structural features of high-resolution image can make up for the lack of spectral features for classification.This paper focuses on the spatial structure feature modeling and information extraction of high-resolution image.It mainly studies the calculation of spatial autocorrelation statistics,the image processing using spatial autocorrelation statistics,and the spatial structure information extraction based on spatial semivariogram high-resolution image samples.Finally,the spatial structure information extracted by spatial autocorrelation statistics and spatial semivariogram function is used as the feature of neural network classification.
作者
李呈祥
LI Chengxiang(Heilongjiang Institute of Geomatics Engineering,Harbin 150081,China)
出处
《测绘与空间地理信息》
2021年第S01期147-150,154,共5页
Geomatics & Spatial Information Technology
关键词
高分影像
典型地物
结构特征
空间自相关统计量
空间半变差函数
神经网络
high resolution image
typical object
structure feature
spatial autocorrelation statistics
spatial semivariation function
neural network
