<abstract xmlns="http://www.w3.org/1999/xhtml"><p> While SIFT (Scale Invariant Feature Transform) features are used to match High-Resolution (HR) remote sensing urban images captured at different phases with large scale and view variations, feature points are few and the matching accuracy is low. Although replacing SIFT with fully affine invariant features ASIFT (Affine-SIFT) can increase the number of feature points, it results in matching inefficiency and a non-uniform distribution of matched feature point pairs. To address these problems, this paper proposes the novel matching method ICA-ASIFT, which matches HR remote sensing urban images captured at different phases by using an Independent Component Analysis algorithm (ICA) and ASIFT features jointly. First, all possible affine deformations are modeled for the image transform, extracting ASIFT features of remote sensing images captured at different times. The ICA algorithm reduces the dimensionality of ASIFT features and improves matching efficiency of subsequent ASIFT feature point pairs. Next, coarse matching is performed on ASIFT feature point pairs through the algorithms of Nearest Vector Angle Ratio (NVAR), Direction Difference Analysis (DDA) and RANdom SAmple Consensus (RANSAC), eliminating apparent mismatches. Then, fine matching is performed on rough matched point pairs using a Neighborhoodbased Feature Graph Matching algorithm (NFGM) to obtain final ASIFT matching point pairs of remote sensing images. Finally, final matching point pairs are used to compute the affine transform matrix. Matching HR remote sensing images captured at different phases is achieved through affine transform. Experiments are used to compare the performance of ICA-ASFIT and three other algorithms (i.e., Harris- SIFT, PCA-SIFT, TD-ASIFT) on HR remote sensing images captured at different times in different regions. Experimental results show that the proposed ICA-ASFIT algorithm effectively matches HR remote sensing urban images and outperforms other algorithms in terms of matching accuracy and efficiency.</p></abstract>

While SIFT (Scale Invariant Feature Transform) features are used to match High-Resolution (HR) remote sensing urban images captured at different phases with large scale and view variations, feature points are few and the matching accuracy is low. Although replacing SIFT with fully affine invariant features ASIFT (Affine-SIFT) can increase the number of feature points, it results in matching inefficiency and a non-uniform distribution of matched feature point pairs. To address these problems, this paper proposes the novel matching method ICA-ASIFT, which matches HR remote sensing urban images captured at different phases by using an Independent Component Analysis algorithm (ICA) and ASIFT features jointly. First, all possible affine deformations are modeled for the image transform, extracting ASIFT features of remote sensing images captured at different times. The ICA algorithm reduces the dimensionality of ASIFT features and improves matching efficiency of subsequent ASIFT feature point pairs. Next, coarse matching is performed on ASIFT feature point pairs through the algorithms of Nearest Vector Angle Ratio (NVAR), Direction Difference Analysis (DDA) and RANdom SAmple Consensus (RANSAC), eliminating apparent mismatches. Then, fine matching is performed on rough matched point pairs using a Neighborhoodbased Feature Graph Matching algorithm (NFGM) to obtain final ASIFT matching point pairs of remote sensing images. Finally, final matching point pairs are used to compute the affine transform matrix. Matching HR remote sensing images captured at different phases is achieved through affine transform. Experiments are used to compare the performance of ICA-ASFIT and three other algorithms (i.e., Harris- SIFT, PCA-SIFT, TD-ASIFT) on HR remote sensing images captured at different times in different regions. Experimental results show that the proposed ICA-ASFIT algorithm effectively matches HR remote sensing urban images and outperforms other algorithms in terms of matching accuracy and efficiency.

ICA-ASIFT-Based Multi-Temporal Matching of High-Resolution Remote Sensing Urban Images

Postdoctoral Programme of Electronic Science and Technology, Fuzhou University, Fuzhou 350116

Key Lab of Spatial Data Mining & Information Sharing of Ministry of Education, Fuzhou University, Fuzhou

National Engineering Research Centre of Geospatial Information Technology, Fuzhou University, Fuzhou

Spatial Information Engineering Research Centre of Fujian Province, Fuzhou University, Fuzhou

Spatial Information Engineering Research Centre of Fujian Province, Fuzhou University, Fuzhou,

Cybernetics and Information Technologies

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

While SIFT (Scale Invariant Feature Transform) features are used to match High-Resolution (HR) remote sensing urban images captured at different phases with...

ICA-ASIFT-Based Multi-Temporal Matching of High-Resolution Remote Sensing Urban Images

Publicado en línea: 20 oct 2016

Páginas: 34 - 49

DOI: https://doi.org/10.1515/cait-2016-0050

Palabras clave
Remote sensing image matching, Independent component analysis, SIFT, Affine transform

© by Fenghua Huang

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

ICA-ASIFT-Based Multi-Temporal Matching of High-Resolution Remote Sensing Urban Images

Publicado en línea: 20 oct 2016

Páginas: 34 - 49

DOI: https://doi.org/10.1515/cait-2016-0050

Palabras claveRemote sensing image matching, Independent component analysis, SIFT, Affine transform

© by Fenghua Huang

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Palabras clave
Remote sensing image matching, Independent component analysis, SIFT, Affine transform