文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 自然科学 > 天文/地理 > CLASSIFICATION-ORIENTED HYPERSPECTRAL AND POLSAR IMAGES SYNERGIC PROCESSING
侵权投诉

CLASSIFICATION-ORIENTED HYPERSPECTRAL AND POLSAR IMAGES SYNERGIC PROCESSING

CLASSIFICATION-ORIENTED HYPERSPECTRAL AND POLSAR IMAGES SYNERGIC

PROCESSING

Tong Li, Junping Zhang, Honglei Zhao and Cuiping Shi

Dept. of Information Engineering, Harbin Institute of Technology, Harbin, China

ABSTRACT

Classification is one of the most important applications in the field of remote sensing. How to improve the accuracy of classification is the critical topic that has long obsessed the researchers. In this paper, a fusion method based on a synergic use of hyperspectral data and Polarimetric SAR (PolSAR) data is presented. This method consists of two main parts, feature-level fusion and decision-level fusion. In feature-level, parallel feature combination strategy is introduced to classification of remote sensing images. Results of feature-level fusion are used as inputs of decision-level fusion based on fuzzy set theory. The final results are compared with processing of single level and single data set, and it shows that the synergic method proposed in this paper has a superior performance in joint classification of hyperspectral and polarimetric SAR data.

Index Terms—Data fusion, classification, synergic processing, hyperspectral, Polarimetric SAR

1. INTRODUCTION

The goal of data fusion in remote sensing is to preserve the respective superiority and utilize the complementary information of the multisource data. For instance, spectra acquired from hyperspectral data provide information on radiance of the ground cover in different wavelength while PolSAR data contain scattering mechanism in the polarization signatures which reflect the geometric structure and physical characteristic of the surface. That means PolSAR data possess information that may not be acquired by optical data and the two data sets can complement each other in processing. Therefore, data fusion topic has always been drawing much attention in remote sensing.

Several fusion approaches have been proposed recently using hyperspectral and PolSAR data. For instance, Dempster-Shafer evidence theory is adopted to combine features extracted from each data source and handle the uncertainty in measurement for land use mapping in a system prototype called Intelligent Data Fusion System

This work was supported by the National Natural Science

Foundation of China under Grant 61271348.

(IDFS) as part of the Earth Observation Application Development Program sponsored by Canada Space Agency [1]. In [2], a novel feature selection algorithm is applied to the greedy modular eigenspaces (GME) to explore a multi-class classification technique. Moreover, lots of work has been done in fusion of multispectral data and SAR data. Neural Networks (NN) is used to combine the ERS-1 and Landsat-TM data [3]. Although the overall accuracy is not significantly improved by adding texture features of SAR data, it makes a great improvement in identifying a certain class. Markov random fields combining spatial context also improves the precision of classification [4][5]. Consensus Theory, Boosting and Bagging algorithm are also performed to train the classifiers, after which the comparison and analysis of the result are given [6]. Besides, Support Vector Machine (SVM) shows more advantages for not requiring any probabilistic assumption for class distribution. Therefore, SVM is used as the classifier for joint classification of Landsat TM and SAR imagery in [7][8].

In this paper, a fusion method is proposed, which combines the feature-level fusion and decision-level fusion with a synergic use of hyperspectral and PolSAR data. The results are compared to the classification outputs of single data source and single level fusion method.

The rest of this paper is organized as follows. The overview and specific processing of method are given in Section 2. Data sets and experimental results are shown in Section 3. Conclusions are drawn in Section 4.

2. SYNERGIC FUSION FOR HYPERSPECTRAL AND

POLSAR IMAGES

The methodology of synergic processing for hyperspectral and PolSAR data is addressed in this section. Fig. 1 shows the overview of the method. In order to get the best set of spectral features for classification and overcome the Hughes phenomenon, we research the double nearest proportion feature extract method which is especially used for hyperspectral images. For the PolSAR images, several decomposition methods are applied. Inspired by the general two levels of fusion methods, we propose a new synergic classification method which combines the feature level and decision level methods. The feature fusion classification algorithm is based on parallel feature combination, while the decision level classification is based on fuzzy set theory.

第1页

你可能喜欢

热门文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email