基于WorldView-3影像的融合算法对比研究

Comparative Research on Fusion Algorithms Based on WorldView-3 Image

适宜的融合算法有利于遥感影像的质量提升与产品应用。本文基于WV-3影像,选用PCA、GSA、ATWT、BDSD、Indusion、PRACS算法对全色和多光谱影像进行融合实验,在采用定性和定量分析对融合结果进行直接评价的基础上,引入地物分类精度对融合结果进行间接评价。实验结果表明:不同场景下6种算法的融合效果不同,ATWT算法在定性与定量评价中表现最佳,并具有较高的分类精度;GSA的融合效果及分类精度表现较好,可作为WV-3影像融合的替补方法;BDSD在建筑物与道路中光谱保持性和空间细节表现最差,但在地物分类中的整体精度最高,适用于地物分类;PCA在植被与水体中的空间细节增强不足、分类精度差,不适用于WV-3数据的影像融合。

An appropriate fusion algorithm is conducive to the quality improvement of remote sensing images and product applications. In this paper,based on WV-3 images,the PCA,GSA,ATWT,BDSD,Induction,and PRACS algorithms were used to conduct fusion experiments on the panchromatic and multispectral images. On the basis of the direct evaluation of fusion results through the qualitative and quantitative analysis,the classification accuracy of ground objects was introduced to evaluate the fusion results.The experimental results showed that the fusion effects of the six algorithms were different in different scenes,the ATWT algorithm had the best performance in the qualitative and quantitative evaluation with a high classification accuracy. The GSA algorithm performed better in the fusion effect and classification accuracy,and could be used as a substitute for the WV-3 image fusion. The BDSD algorithm performed worst in the spectral retention and spatial details of buildings and roads. PCA had insufficient spatial detail enhancement and poor classification accuracy in the vegetation and water bodies,so it was not suitable for the image fusion of WV-3 data.