结合像元形状特征分割的高分辨率影像面向对象分类

针对高分辨率遥感影像空间分辨率高,结构形状、纹理、细节信息丰富等特点,提出一种新的融合特征的面向对象影像分类方法来提取城市空间信息。基本过程包含以下4个方面:①提取影像的几何纹理等结构;②融合几何与纹理特征的面向对象影像分割;③提取对象的形状、纹理和光谱特征,并优选最佳特征子集;④最后基于支持向量机(SVM)完成面向对象的影像分类。通过对福州IKONOS影像数据实验,结果表明融入影像特征后的分割效果明显优于原始影像的分割结果,而信息最大化(mRMR)的特征选择能够快速地获得较好的特征子集。通过与eCognition最邻近分类方法比较,表明本文方法的分类总体精度大约提高了6%,效果显著。

Object-oriented Classification of High Spatial Resolution Remote Sensing Imagery Based on Image Segmentation with Pixel Shape Feature

According to the high resolution and rich spatial information of high spatial resolution remote sensing imagery,this paper proposes to integrate geometric,shape,texture feature for high spatial resolution remote sensing imagery classification in urban area with object-oriented method.The proposed method is a four-step classification routine that involves the integration of：①extraction of geometric shape feature;②segmentation of high spatial resolution remote sensing imagery based on spectrum information and geometric shape feature that extracted;③extraction of object shape feature,texture feature,spectal feature and so on,then use mutual information minimum redundancy and maximum relevance（mRMR）criterion to select optimal subset features;④support vector machine（SVM） for classification.To validate the proposed method,a case study with IKONOS high spatial resolution remote sensing imagery in Fuzhou city is implempented.The experimental results demonstrate that fused pixel shape index（PSI）feature can improve the multiscale segmentation sigificantly,and feature selection can acquire optial feature subset.Moreover,the proposed method for high spatial resolution remote sensing imagery classification in urban area can increase classification accuracy by about 6% in terms of overall accuracy compared with the nearest neighborhood method.